TenthE.C.

Photovoltaic Solar Energy Conference

 Commision of the European Communities

***
**e**
* *** *

Tenth E.C.
Photovoltaic Solar
Energy Conference
Proceedings of the International Conference,
held at Lisbon, Portugal,
8-12 April 1991

Edited by

A. LUQUE
Universidad Politecnica, Madrid, Spain
G.SALA
Universidad Politecnica, Madrid, Spain
W.PALZ
Commission o/the European Communities, Brussels, Belgium
G. DOS SANTOS
Secretaria de Estado da Energia, Lisbon, Portugal
and
P.HELM
WIP, Munich, F.R. Germany

KLUWER ACADEMIC PUBLISHERS

DORDRECHT I BOSTON I LONDON
ISBN ~7923-1389-5

Publication arrangements by
Commission of the European Communities
Directorate-General Telecommunications, Information Industries and Innovation, Scientific and
Technical Communications Service, Luxembourg

EUR 13807
© 1991 ECSC, EEC, EAEC, Brussels and Luxembourg

LEGAL NOTICE
Neither the Commission of the European Communities nor any person acting on behalf of the
Commission is responsible for the use which might be made of the following information.

Published by Kluwer Academic Publishers,

P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

Kluwer Academic Publishers incorporates the publishing programmes of

D. Reidel, Martinus Nijhoff, Dr W. Junk and MTP Press.

Sold and distributed in the U.S.A. and Canada

by Kluwer Academic Publishers,
101 Philip Drive, Norwell, MA 02061, U.S.A.

In all other countries, sold and distributed

by Kluwer Academic Publishers Group,
P.O. Box 322, 3300 AH Dordrecht, The Netherlands.

Printed on acid-free paper

All Rights Reserved

No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner.
 INTRODUCTION

I have great pleasure in presenting the Proceedings of the 10th European Photovoltaic Solar Energy
Conference held in Lisbon from 8 to 12 April 1991. These Proceedings contain all the scientific papers delivered
at the Conference. The following is a short summary of the Conference activities.

The Conference was opened by the Minister of Industry and Energy of Portugal, Eng. Luis Mira do Amaral.
At the opening ceremony the Becquerel Prize, created by the Commission of the European Communities, was
awarded to Professor Werner Bloss of the University of Stuttgart, and presented by Professor Philippe Bourdeau,
Director at the Directorate-General for Science, Research and Development. The Becquerellecture delivered by
Professor Bloss constituted the scientific opening to the conference.

About 760 delegates from 53 countries presented around 350 contributions, 50 of them as plenary lectures;
the contributions were selected among the many papers submitted, this time more strictly than ever before. Also a
selected group of scientists were invited to deliver 15 review lectures, to provide an adequate context to the
contributions to the Conference.

A Symposium on Photovoltaics in Developing Countries, which was very well attended, took place as a
parallel event. The Symposium provided an opportunity to hear not only experts of the industrialized countries,
but also speakers from the countries where photovoltaics provides services of paramount value.

The now-traditional Panel Discussion, chaired by Professor Gerry Wrixon, dealt this time with the topic
"Are we prepared for mass production of PV electricity?". A lively discussion took place beween those who
considered that the growth of PV electricity must be left solely to market forces, and those who asked for a
"Manhattan project" (some delegates found the comparison objectionable owing to the warlike connotations of this
name).

The Exhibition of the Photovoltaic Industry brought together 43 companies, whose activities varied from
photovoltaic manufacturing equipment and the preparation of basic semiconductor material, through cell
fabrication, to companies involved in power systems and in measuring equipment.

As usual the work in the Conference Hall lobbies, where informal meetings take place, was as interesting as
that in the auditoria; to make it even more animated, a daily newspaper - The Conference News - containing
topical articles and interviews, was issued under the leadership of Alain Liebard.

At the Summaries session, the most important conclusions on every subject were presented by the session
organizers, and those concerning the Exhibition by Mr. Fred Treble. From this session and that on National
Programmes, it appeared clearly that we are witnessing today a stronger commitment of big corporations, in
particular European ones, to the development of photovoltaics as well as the industrialization of concepts
developed in recent years in the academic sector.

At the closing session the Mouchot Prize was awarded to Dr. Wolfgang Palz, head of the Renewable Energy
Programme of the Commission of the European Communities, the inspirer of this series of conferences, in
recognition of his outstanding contribution to the promotion of photovoltaics. The prize was presented by the
President of Eurosolar, Dr. Hermann Scheer, Member of the German Parliament. The laudatio of the awardee
was delivered by Mr. A. Liebard, Chairman of the "Comite d'Action pour Ie Solaire". The session was closed by
the Secretary of State for Energy of Portugal, Dr. Nuno Ribeiro da Silva.

The next European Photovoltaic Solar Energy Conference, the 11th in this series, will take place at
Montreux, Switzerland, on the 12 to 16 October 1992 under the chairmanship of Professor Leopoldo Guimariies.

Madrid, 15 May 1991

Antonio Luque, General Chairman

v
Sponsored by:

Commission of the European Communities

Ministerio da Industria e Energia

Secretaria de Estado da Energia

Co-sponsored by:

UNESCO

Academia das Ciencias

Direc~iio Geral de Energia

Petroleos de Portugal, Petrogal SA

Daimler-Benz AG

Neste Polimeros, SA

Caixa Geral de Depositos

Camara Municipal de Lisboa

Eurosolar

EPIA

Collaborating Entities:

Laboratorio Nacional de Engenharia e Tecnologia Industrial

Sociedade Portuguesa Acumuladores Tudor

Portsol

Universidade Nova de Lisboa

Direc~iio Geral da Comunica~iio Social

Instituto de Promo~iio Turistica

Sociedade Portuguesa de Energia Solar

Instituto Superior Tecnico

Conference Organiser:

WIP-Munich

vi
 CONFERENCE COMMmeE
General Chairman
Prof. A. Luque, Univ. Politecnica, Madrid, Spain
Technical Programme Chairmen
Prof. l. GuimarAes, Univ. Nova, Lisbon, Portugal
Prof. G. Wrixon, NMRC, Cork, Ireland
Conference Director
Dr. W. Palz, CEC, Brussels, Belgium
Conference Secretariat
Prof. G. Sala, Univ. Politecnica, Madrid, Spain
Mrs. A. Baubin, CEC, Brussels, Belgium
International Organisation, Administration I exhibition
Dipl. Ing. H. Ehmann, WIP, Munich, F. R. Germany
Mrs. H. Riegel, WIP, Munich, F. R. Germany
Local Organisation
Prof. R. Martins, Univ. Nova, Lisbon, Portugal
Eng. E. Fortunato, Univ. Nova, Lisbon, Portugal
Posters
Dr. H. W. Schock, Univ. Stuttgart, F. R. Germany
Prof. R. Martins, Univ. Nova, Lisbon, Portugal

STEERING COMMmeE
Prof. W. H. Bloss, Univ. Stuttgart, F. R. Germany
Prof.l.J.M. Guimaraes, Univ. Nova, Lisbon, Portugal
Dipl. Ing. P. Helm, WIP, Munich, F. R. Germany
Prof. R. Hill, Newcastle Polytechnic, UK
Prof. A. Luque, Univ. Politecnica, Madrid, Spain
Dr. W. Palz, CEC, Brussels, Belgium
Dr. M. Reis, CEC, Brussels, Belgium
Prof. G. Sala, Univ. Politecnica, Madrid, Spain
Eng. G. dos Santos, Secretaria de Estado da Energia, Lisbon,
Portugal
Dr. H. Scheer, Eurosolar, Bonn, F. R. Germany
Prof. R. Van Overstraeten, IMEC, Leuven, Belgium
Prof. G. T. Wrixon, NMRC, Cork, Ireland

COMMmeE SYMPOSIUM PV
IN DEVELDPING COUNTRIES
Chairman:
Prof. R. Hill, Newcastle Polytechnic, UK
Prof. W. Bloss, Univ. Stuttgart, F. R. Germany
Prof. l. GuimarAes, Univ. Nova, Lisbon, Portugal
Dipl. Ing. P. Helm, WIP, Munich, F. R. Germany
Prof. E. Lorenzo, Univ. Politecnica, Madrid, Spain
Dr. B. McNelis, IT Power, Eversley, UK

SESSION ORGANISERS
Subj. 1: Prof. G.l. Araujo, Univ. Politecnica, Madrid, Spain
Subj. 2: Prof. R. Mertens, IMEC, Leuven, Belgium
Subj. 3: Prof. I. Solomon, Ecole Polytech., Palaiseau, France
Subj. 4: Prof. W. Bloss, Univ. Stuttgart, F. R. Germany
Subj. 5: Dr. J. Schmid, Fraunhofer Inst., Freiburg, F. R. Germany
Subj. 6: Dr. S. McCarthy, Hyperion Energy Syst., Cork, Ireland
Subj. 7: Prof. R. Hill, Newcastle Polytechnic, UK

vii
NATIONAL COMMITTEE
Eng. Gonyalves dos Santos,
Adviser to the Secretary of State for Energy
Chairman of the National Committee
Prof. Rodrigo Martins, Universidade Nova, Usbon
Prof. Collares Pereira,
Portuguese Center for Energy Saving, Usbon
Prof. AntOnio Valera, Faculty of Sciences, Usbon
Prof. JoAo Bessa, University of Oporto
Dr. Pedro Paes, LNETI, Usbon
Eng. Elvira Fortunato, Universidade Nova, Usbon
Eng. Penaforte e Costa, DirecyAo Geral de Energia, Usbon
Prof. Traya de Almeida, University of Coimbra

HONORARY COMMITTEE
Prof. A. Cavaco Silva,
Prime Minister of Portugal,
President of the Honorary Committee

On. Filippo Maria Pandolfi,

Vice President of the CEC,
E. C. Commissioner for Science and Technology
Eng. A. Cardoso e Cunha,
E. C. Commissioner for Energy
Eng. Luis Mira Amaral,
Minister for Industry and Energy
Eng. Valente de Oliveira,
Minister for Planning and Regional Administration
Eng. Roberto Carneiro,
Minister for Education
Prof. Paolo Fasella,
General Director for Science, Research and Development,
DGXII,CEC
Dr. Nuno Ribeiro da Silva,
Secretary of State for Energy
Prof. Sucena Paiva,
Secretary of State for Science and Technology
Eng. Alves Monteiro,
Secretary of State for Industry
Prof. Veiga SimAo,
President of National Laboratory of Engineering and Industrial
Technology, LNETI
Eng. CustOdio Miguens,
General Director for Energy, DGE
Prof. Carlos Salema,
President of National Institute for Research, Science and
Technology, JNICT
Prof. Azeredo PerdigAo,
President of Calouste Gulbenkian Foundation
Prof. Leopoldo GuimarAes,
President of Uninova
Prof. Jacinto Nunes,
President of Portuguese Science Academy
Dr. Helena Vaz da Silva,
Portuguese Commissioner for UNESCO

viii
 SCIENTlFlC COMMITTEE
Prof. V. M. Andreev,lOFFE Physical Tech. Inst., Leningrad, USSR
Prof. G. L. AraOjo. Univ. Politecnica, Madrid. Spain
Dr. C. E. Backus, Arizona State Univ., Tempe, AZ., USA
Mr. C. Baraona, NASA Lewis Res. Center, Cleveland, OH, USA
Dr. A. M. Barnett, AstroPower, Newark, DE, USA
Dr. G. Beer, Con phoebus, Catania, Italy
Dr. B. Berkovski. Div. of Technological Res. and Higher
Education, UNESCO, Paris, France
Dr. B. Bini Smaghi, CEC, Brussels. Belgium
Prof. W. H. Bloss, Univ. Stuttgart, F. R. Germany
Dr. E. C. Boes, Sandia Nat. Lab .• Albuquerque. N.M., USA
Dr. K. P. Bogus, ESA-ESTEC, Noordwijk, The Netherlands
Mr. R. Buhs, Telefunken Systemtechnik, Wedel, F. R. Germany
Dr. J. Cap ilion, EDF, Chatou, France
Dr. D. E. Carlson, Solarex, Newton, PA, USA
Dr. A. M. de Andrade. Escola Politec.• Univ. Slio Paulo, Brazil
Prof. B. Equer, Ecole Polytechnique, Palaiseau, France
Dr. W. Freiesleben, Eurosolar, Brussels, Belgium
Prof. J. Furlan, Fac. Electrical Engineering, Ljubljana, Yugoslavia
Dr. R. Gislon. ENEA, Rome, Italy
Prof. A. Goetzberger, Fraunhofer Inst., Freiburg, F. R. Germany
Dr. J. G. Grabmaier, Siemens, Munich, F. R. Germany
Prof. M.A. Green, Univ. New South Wales, Australia
Prof. L.J.M. Guimarl1es, Univ. Nova, Usbon, Portugal
Prof. Y. Hamakawa, Univ. Osaka, Japan
Dr. D. Helmreich, Heliotronic, Burghausen, F. R. Germany
Prof. R. Hill, Newcastle Polytechnic, UK
Mr. M. S. Imamura, WIP, Munich, F. R. Germany
Mr. W. Kaut. CEC, Brussels. Belgium
Dr. L. L. Kazmerski. SERI. Golden, CO. USA
Dr. H. Kiess, Paul Scherrer Inst., Lab. RCA, Zurich. Switzerland
Prof. W. Kleinkauf, ISET, Kassel, F. R. Germany
Dr. K. Krebs, CEC, JRC, Ispra, Italy
Dr. Y. Kuwano, Sanyo ElectriC, Osaka, Japan
Prof. D. Lalas, CRES, Athens, Greece
Prof. E. Lorenzo, Univ. Politecnica, Madrid, Spain
Prof. R. Martins, Univ. Nova, Usbon, Portugal
Prof. H. Matsunami, Univ. Kyoto, Japan
Dr. S. McCarthy, Hyperion Energy Systems, Cork. Ireland
Mr. B. McNelis, IT Power. Eversley, UK
Prof. R. Mertens, IMEC, Leuven, Belgium
Dr. C. Messana, ENEA, Rome, Italy
Mr. G. Oelert, GTZ, Eschborn, F. R. Germany
Mr. A. Previ, ENEL, Milan, Italy
Dr. M. B. Prince, US Dept. of Energy, Washington, DC, USA
Prof. D. Rapakoulias, Univ. Patras, Greece
Prof. T. Saitoh, Univ. of Agric., Tokyo, Japan
Mr. F. Sanchez. CIEMAT-IER, Madrid, Spain
Mr. A. Sauze, CEN/CEA, Cadarache, France
Dr. H. Scheer, Eurosolar, Bonn, F. R. Germany
Dr. J. Schmid. Fraunhofer Inst. ISE, F. R. Germany
Dr. H. W. Schock, Univ. Stuttgart, F. R. Germany
Prof. A. V. Shah, Univ. Neuchatel, Switzerland
Prof. Y. Shapira, Univ. Tel-Aviv, Israel
Dr. P. Siffert, CRN-PHASE, Strasbourg, France
Prof. I. Solomon, Ecole Polytechnique, Palaiseau, France
Prof. D. S. Strebkov, VIESH, USSR
Mr. F. C. Treble, Farnborough, UK
Dr. M. Umeno, Dept. of Elec. and Computer Eng., Japan
Prof. W. F. van der Weg, Univ. Utrecht, The Netherlands
Prof. R. van Overstraeten, IMEC, Leuven, Belgium
Dr. C. VElrie, CNRS, Valbonne, France
Dr. G. Willeke, Univ. Konstanz, F. R. Germany
ix
 TABLE OF CONTENTS

Alphabetical List of Authors at the end of the book

Introduction v

Conference Sponsors vi

Conference Committee vii

Opening Session

Opening Address 3
Dr. Nuno Ribeiro da Silva
Secretary of State for Energy, Lisbon, Portugal

Speech on the Occasion of the Mouchot Prize Award

Dr. W.Palz 5
Commission of the European Communities, Brussels, Belgium

SESSION 01: Fundamentals, High Efficiency Cells and Novel Devices

01.02 Efficiency of a Combined Solar Concentrator Cell and Thennal Power Engine System 11
Goetzberger A., Bronner W., Wettling W.
Fraunhofer Inst. ISE, Freiburg, F.R Germany

01.03 Holographic Optical Elements for the Constrnction of Tandem Cells 15

Herz K., Corlatan D., Anders G., Schafer M.
ZSW, Stuttgart, F.R Germany

01.04 Light Trapping with Submicron Gratings in thin Crystalline Silicon Solar Cells 19
Kiess H., Epler J.E., Gale M.T., Krausbauer L., MorfR, Rehwald W.
Paul Scherrer Inst., Zurich, Switzerland

01.05 A Technology-Based Comparison Between Two-Sided and Back-Contact Silicon Solar Cells 23
Cuevas A.
Inst. Energia Solar, UPM, Madrid, Spain
Sinton RA.
Sun Power Corp., Mountain View, CA, USA
KingRR
Georgia Inst. Technology, Atlanta, GA, USA

01.06 DeVelopment of a Polysilicon Emitter Solar Cell 27

Papadopoulos G., Tarr N.G.
Carleton Univ., Ottawa, Ont., Canada
BoivinL.P.
Nat. Res. Council, Ottawa, Ont., Canada

xi
POSTER SESSION IA: Fundamentals, High Efficiency Cells and Novel Devices

IAOI Emitter-base Coupling in Highly Illuminated Solar Cells and Applications in OCVD
Technique 33
Pelanchon F.
CEDUST, Damascus, Syria
BoeglinE.
ISSAT, Damascus, Syria
Mialhe P.
Univ. Perpignan, France

IA02 Lifetime and Surface Recombination Detennination 36

Mialhe P.
Univ. Perpignan, France
Salagnon J.M., Pelanchon F.
ISSAT, Damascus, Syria
Sissoko G., Kane M.
Univ. Dakar, Senegal

IA03 Reduction of Reflection Losses by Structured Surfaces 39

Scheydecker A, Goetzberger A, Wittwer V.
Fraunhofer Inst. ISE, Freiburg, F.R. Germany

IA04 The I-V Silicon Solar Cell Characteristic Parameters Temperature Dependence -An
Experimental Study Using the Standard Deviation Method 43
Veissid N.
Inst. Pesquisas Espaciais, Sao Jose dos Campos, Brazil
De Andrade AM.
Univ. Sao Paulo, Brazil

IA05 Sizing of Light Confining Cavities for GaAs and Si Solar Cells 48
Tobias I., Luque A, Miftano J.C., Alonso J.
Inst. Energia Solar, UPM, Madrid, Spain

IA06 AIGaAs-GaAs Solar Cells with Increased Radiation Stability 52

Andreev V.M., Kalinovskii V.S., Sulima O.V.
AF.Ioffe Physico-Technical Inst., Leningrad, USSR

IA07 New Techniques in Solar Cell Theory 55

Markvart T.
Univ. Southampton, UK

IA08 Organic Solar Cells - A Survey 58

Bonnet D., Volkheimer J.
Battelle Inst., FrankfurtlMain, F.R. Germany

IA09 Optimization ofAntireflection Film Structures for Surface-Passivated Crystalline Silicon

Solar Cells Using Spectroscopic Ellipsometry 62
Saitoh T., Kamataki 0., lida S.
Univ. of Agriculture and Technology, Tokyo, Japan

IAll The Influence of Surface Recombination on the Limiting Efficiency and Optimum Thickness
of Silicon Solar Cells 66
Demesmaeker E., Symons J., Nijs J., Mertens R.
IMEC, Leuven, Belgium

xii
lA.12 Development of Cost-Effective, High-Efficiency Concentrator Solar Cells 70
Leboeuf C.M., Benner J.P.
SERI, Golden, CO, USA

lA.13 Enhancement of Sub-Bandgap Absorption in X-Si Solar Cells by Alloying with Germanium 73
Healy S.A., Taouk M., Chan B., Green M.A.
Univ. New South Wales, Kensington, Australia

lA.14 New Concepts in 2-Dimensional Solar Cell Modeling 75

Jimeno J.e., Uriarte S., Zamora J.J., Icaran e.
Univ. Pais Vasco, Bilbao, Spain

lA.16 Hydrogen Evolution on Catalyst-Coated Photoelectrodes p-InP and p-Si 79

Schefold J.
IPE, Univ. Stuttgart, F.R Germany
Specht M.
ZSW, Stuttgart, F.R Germany

lA.17 Experimental Determination of Field Profiles in Diodes Formed on Amorphous Silicon

Germanium Alloys 84
Longeaud e., KIeider J.P., Glodt 0., Mencaraglia D.
Univ. Paris VI et XI, Gif-sur-Yvette, France
Vanderhaghen R
LPICM, Ecole Polytechnique, Palaiseau, France

lA.18 Highly Transparent and Conductive ZnO:AI Thin Films for PVApplications 88
Mauch RH., Schock H.W.
IPE, Univ. Stuttgart, F.R Germany

lA.19 The Hydrogen-Carbon Complex: Build-up, Electronic Properties and Annealing Behavior 92
Endros A.L., Grabmaier J.
Siemens AG, Munich, F.R Germany

lA.20 Defect Chemistry and Homogeneity Range of Pyrite and Their Influence on the Opto-
electronic Behaviour 96
Birkholz M., Hartmann A., Fiechter S., Tributsch H.
Hahn-Meitner-Inst., Berlin, F.R Germany

lA.21 Comparison of Dark and Light I-V Curves of Solar Cells 100
Beier J., Bucher K.
Fraunhofer Inst. ISE, Freiburg, F.R Germany

lA.22 The LDS Solar Cell- A New Approach to Higher Efficiency 104
Barnham K.W J., Braun B., Nelson J., Paxman M.
Imperial College of Science Technology and Medicine, London, UK
Button e., Roberts J.S.
Univ. Sheffield, UK

lA.23 Spectral Response Measurements of Multi-junction Solar Cells with a Grating Mono-
chromator and a Fourier Spectrometer 107
Bucher K., SchOnecker A.
Fraunhofer Inst. ISE, Freiburg, F.R Germany

lA.24 Measurement of Multi-junction Solar Cells 111

Heidler K., Muller-BierI B.
Fraunhofer Inst. ISE, Freiburg, F.R Germany

xiii
lA.26 Contact-Free Luminescent Measuring of (n-p) GaAs-pAl GaAs Solar Cell's Parameters 115
Rumyantsev V.D.
AF. loffe Physico-Technical Inst., Leningrad, USSR

POSTER SESSION IB: Amorphous Silicon and Related Materials I

IB.OI A Comparison Between Fluorinated and Hydrogenated Amorphous Silicon Carbide

Prepared by Reactive Sputtering 121
Demichelis F., Pirri C.F., Tresso E.
Politecnico di Torino, Italy
Stapinski T.
Acc. Mining & Metallurgy, Krakow, Poland
BoarinoL.
lEN Galileo Ferraris, Torino, Italy
RavaP.
ElettroRava, Torino, Italy

lB.02 Influence of Silicon on Properties of Hydrogenated Amorphous Carbon 125

Demichelis F., Giachello G., Pirri C.F., Tagliaferro A
Politecnico di Torino, Italy
Spagnolo R.
lEN Galileo Ferraris, Torino, Italy

lB.03 A New Approach for Characterizing Surface Interface and Bulk Defects in a-Si:H 128
AmatoG.
lEN Galileo Ferraris, Torino, Italy

IB.04 Modelling and Characterization of PV Tandem Cells with a-Si:H, a-SiC:H and a-SiGe:H
Intrinsic Layers 132
Gramaccioni C., Grillo G., Galluzzi F.
Eniricerche, Rome, Italy
ConteG.
ENEA-CRIF, Portici, Italy

IB.05 A Photoemission Study ofgdJlc-Si 135

Willeke G., Fischer R., Bucher E.
Univ. Konstanz, F.R. Germany
Prasad K., Shah A
Inst. Microtechnology, Univ. Neuchatel, Switzerland

lB.06 Modelling ofAmorphous Thin Film Solar Cells 139

Block M., Bonnet D., Zetzsche F.
Battelle Inst., Frankfurt/Main, F.R. Germany

lB.07 CO 2 Laser CVD ofAmorphous Hydrogenated Silicon: Gas Phase Processes and Properties
of Thin Films 143
Golusda E., Luhmann K.D., Mollekopf G., Wacker M., Stafast H.
Battelle Inst., Frankfurt/Main, F.R. Germany

IB.08 Simulation and Analysis of the Bias Dependent Spectral Response of a-Si:H pin Solar Cells 147
Bruns J., Gall S., Wagemann H.G.
TU Berlin, F.R. Germany

lB.09 Influence of Deposition Temperature on the Stability of a-Si:H Solar Cells 151
Asensi I.M., Bertomeu I., Andreu J., Puigdollers I., Morenza J.L.
Univ. Barcelona, Spain

xiv
lB.1O Thickness Mapping of Thin Films by Low-contrast Optical Inteiferometry 154
Solomon I., Bhatnagar M., Rosso M.
PMC, Ecole Polytechnique, Palaiseau, France

lB.1l Doping Efficiency of Boron Trifluoride in the Preparation of P-Type Amorphous Silicon
Carbide Thin Films 157
Ganrua J.J., Gutierrez M.T., Carabe J.
CIEMAT-IER, Madrid, Spain

lB.12 Stationary Primary Photocurrents for the Characterization of a-Si:H PIN-Diodes 161
Abel C.D., Bauer G.H.
IPE, Univ. Stuttgart, F.R. Germany
PaesH.R.
Univ. Federal, Rio de Janeiro, Brazil

lB.13 Correlation Between Discharge Geometry and Film Properties in PE-CVD of a-Si:H 165
Mataras D., Rapakoulias D., Kounavis P., Mytilineou E.
Univ. Patras, Greece

lB.14 Diagnostics of Very High Frequency Silane Plasmas 169

Howling AA, HoUenstein c., Dorier J.L., Paris P.
CRPP/EPFL, Lausanne, Switzerland
FavreM.
IMT, Univ. Neuchiitel, Switzerland

lB.lS Electric Field Profile and Charge Collection in a-Si:H Solar Cells 173
Konenkamp R.
Hahn-Meitner Inst., Berlin, F.R. Germany
Muramatsu S., Itoh H., Matsubara S., Shimada T.
Hitachi Central Res. Lab., Tokyo, Japan

lB.16 Relation of Small Signal Transient Response to Density of States in a-Si 177
Furlan J., Basaneze E., Smole F.
Univ. Ljubljana, Yugoslavia

lB.17 Characterization of Highly Conductive Phosphorus Doped pc-Si:H Films 181

Dirani E.AT., de Andrade AM., Pereyra I.
Univ. Sao Paulo, Brazil

lB.18 A Model for the Direct Bias C- V Measurements on a-Si:H P-I-N Solar Cells 184
Caputo D., De Cesare G., Irrera F., Palma F.
Univ. La Sapienza, Rome, Italy

lB.19 Influence of the Buffer Layer of a-SiC:H/a-Si:H pin Solar Cells on the Spectral Response 188
Kopetzky W.J., Schwarz R.
TU Munich, Garching, F.R. Germany

IB.20 Open-Circuit Voltage Enhancement in MIS Solar Cells by Thin Plasma-Grown Tunneling
Oxides 192
Meier J., Kragler G., Willeke G., Bucher E.
Univ. Konstanz, F.R. Germany
Keppner H., Fischer D., Viret V., Shah AV.
Univ. Neuchiitel, Switzerland

xv
IB.21 Microstrncture and Stability of VHF-GD Amorphous Silicon under Illumination 196
VanecekM.
Inst. Physics, Prague, Czechoslovakia
Holoubek J.
Inst. Macromolecular Chemistry, Prague, Czechoslovakia
Shah A, Brechet H., Mettler A
Univ. Neuchiitel, Switzerland

lB.22 VHF Plasma Deposition for Low-Cost a-Si:H Solar Cells 201
Fischer D., Keppner H., Finger F., Prasad K., Shah A.
Univ. Neuchiitel, Switzerland

IB.23 Continuous Monitoring of VHF-GD Deposited p-i-n Solar Cells During Unifonn
Degradation Till Saturation 205
Keppner H., Chabloz P., Fischer D., Mettler A, Tscharner R, Shah A
Univ. Neuchiitel, Switzerland

IB.24 Reversal of Field and Photocu"ent in a-Si pin Solar Cells 208
Pfleiderer H.
Siemens AG, Munich, P.R. Germany

lB.25 Optical Modelling of a-Si:H and a-Ge:H P-I-N Solar Cells 212
Saeng-udom R, Bullemer B.
Federal Armed Forces Univ., Neubiberg, F.R Germany
Kusian W.
Siemens AG, Munich, F.R Germany

lB.26 An a-Ge:H pin Solar Cell with High Red Efficiency 216
Kusian W., Giinzel E., Plattner R.D.
Siemens AG, Munich, F.R Germany

IB.27 Deuteron Magnetic Resonance Spectral Components in a-Si:H,D 220

Santos-Filho P.B.
Inst. Tecnol6gico Pernambuco, Recife, Brazil
Bodart J., Norberg RE.
Washington Univ., st. Louis, MO, USA

lB.29 Detennination of the Density of States in a-Si:H Over a Broad Energy Range by the
Thennally Stimulated Conductivity Technique 224
von der Linden M.B., Schropp R.E.I., Bezemer J., Van der Weg W.F.
Univ. Utrecht, The Netherlands
Landweer G.
Univ. of Technology, Delft, The Netherlands

lB.30 Diode Recovery Measurement of a-Si:C:H Diode Using Different Silane-Propane Ratio for
Tandem Solar Cells 228
Ibrahim K.
Univ. of Science, Penang, Malaysia

SESSION 02: High Efficiency Crystalline Silicon Solar Cells

02.01 - Review lecture -

Review on High Efficiency Silicon Solar Cells 233
Luque A
Inst. Energia Solar, UPM, Madrid, Spain

xvi
02.02 Progress in the Manufacturing of Production-Type Crystalline Silicon Solar Cells 240
Mertens RP.
IMEC, Leuven, Belgium

02.03 High-Efficiency Backside Contact Solar Cells with a Self-Aligned Process and New Texturization
Technique for Silicon 246
Verlinden P.
Sun Power Corp., Mountain View, CA, USA
Lafontaine B., Evrard 0., Mazy E., Crahay A.
Univ. Catholique, Louvain-Ia-Neuve, Belgium

02.04 RecentAdvances in Silicon Solar Cell Peifonnance 250

GreenM.A.
Univ. New South Wales, Kensington, Australia

02.05 High-Efficiency and Low-Cost Option in the Development of Crystalline Silicon Solar Cells 254
Hayashi Y., Shimokawa R
Electrotechnical Lab., Ibaraki, Japan
Saitoh T.
Tokyo Univ. of Agric. and Technology, Tokyo, Japan
HaneK.
Keio Univ., Kanagawa, Japan

POSTER SESSION 2A: Crystalline Silicon - Materials and Devices

2A.Ol Elaboration of Photovoltaic Silicon by a Plasma Process and its Characterization 261
Humbert P., Combes R, Erin J., Madigou N., Morvan D., Amouroux J.
ENSCP, Paris, France

2A.02 Classical and Rapid Thennal Process Induced Gettering in Multicrystalline Silicon 267
Hartiti B., Muller J.e., Stuck R, Siffert P.
Centre de Recherches Nucleaires, Strasbourg, France
Sarti D.
Photowatt, Caen, France

2A.03 Comparative Efficiency of Hydrogen Passivation Procedures for Cost Effective Multi-
crystalline Silicon Solar Cells 270
Hartiti B., Schunck J.P., Muller J.e., Stoquert J.P., Hussian E., Siffert P.
Centre de Recherches Nucleaires, Strasbourg, France
Sarti D.
Photowatt, Caen, France

2A.04 Thennodynamics for Removal of Boron from Metallurgical Silicon by Flux Treatment 273
Suzuki K., Sano N.
Univ. Tokyo, Japan

2A.05 An Analysis of High Efficiency Si Processing 276

Hogan S., Darkazalli G., Wolfson R
Spire Corp., Bedford, MA, USA

2A.08 A High Efficiency Silicon Solar Cell Production Technology 280

Mason N.B., Jordan D.
BP Solar Espana, Madrid, Spain
Summers J.G.
BP Solar, Leatherhead, UK

xvii
2A09 Development of Solar Cell Manufacturing Technology under the Sunshine Project of Japan 284
Yokoyama e., Kuno K Sawadaishi S., Ohyu N., Kuroda T., Sugimoto H.
NEDO, Tokyo, Japan

2A.1O Removal of Boron from Molten Silicon by Argon-Plasma Mixed with Water Vapor 286
Baba H., Yuge N., Sakaguchi Y., Fukai M., Aratani F., Habu Y.
Kawasaki Steel Corp., Chiba, Japan

2A12 Passivated Semicrystalline Silicon Solar Cells: The Effect of Hydrogen Plasma Parameters 290
Soler M.AG., Andrade AM.
Univ. Sao Paulo, Brazil

2A15 Texture Etching of Multicrystalline Silicon 293

Kaiser U., Kaiser M., Schindler R
Fraunhofer Inst. ISE, Freiburg, F.R Germany

2A16 Localization of the Solid/Liquid Interface During Directional Solidification of Silicon by a

Pulse-Echo Ultrasonic Technique 295
Eyer A, Haas F., Schatzle P.
Fraunhofer lust. ISE, Freiburg, F.R. Germany
PaulM.
Fraunhofer Inst. fUr Zerstorungsfreie Prufverfahren, Saarbriicken, F.R Germany
KnobeiRM.
Heliotronic, Burghausen, F.R Germany

2A17 Development of High-Quality P-Type Si Thin-Films by a New SPC Method and its
Application to Solar Cells 298
Matsuyama T., Taguchi M., Tanaka M., Matsuoka T., Tsuda S., Ohnishi M., Kishi Y.,
Nakano S., Kuwano Y.
Sanyo Electric Co., Osaka, Japan

2A18 Silicon-Film"'''' Product I: Initial Production Perfonnance 302

Barnett AM.
Univ. Delaware, Newark, DE, USA
Bottenberg W.R., Bragagnolo I.A, Brooks D.S., Checchi J.e., Kendall e.L.,
Laswell P.G.,13rown I.E., Mulligan W.P., Rand J.A, Hall RB.
AstroPower, Newark, DE, USA

2A19 Silicon-Film Product II: Initial Light Trapping Results 306

Rand I.A, Ford D.H., Bacon e., Ingram AE., Ruffins T.R., Hall R.B.
AstroPower, Newark, DE, USA
Barnett AM.
Univ. Delaware, Newark, DE, USA

2A20 Characterization of Commercial Single Crystal Si Cell Technology 310

Mitchell KW., Aldrich D.L., Walle I.R, Gretlein D.G., Pauls KL., Probst R
Siemens Solar Industries, Camarillo, CA, USA
Gee I.M., McBrayer J.D.
Sandia Nat. Lab., Albuquerque, NM, USA

2A.21 Recrystallization Methods in Production of Photovoltaic Devices 313

Urli N.B., Vlahovic B., Persin M.
Rudjer Boskovic Inst., Zagreb, Yugoslavia

2A22 Polycrystalline Photovoltaic Silicon Ingot Production 317

Dorrity I.A, Garrard B.J., Hukin D.A
Crystalox, Wantage, UK

xviii
2A.24 Towards Low Cost Multicrystalline Silicon Wafers for High Efficiency Solar Cells 320
Martinuzzi S., Perichaud I., Gervais J.
Univ. Aix-Marseille, France
Sarti D.
Photowatt, Caen, France

2A.25 Thennal Environment and Process Optimization of Solarex Cast Polycrystalline Silicon 324
Narayanan S., Terry J., Brenneman R
Solarex Corp., Frederick, MD, USA

2A27 A Low-Cost Diffusion Process for the Fabrication of Solar Cells 327
Dias J.AS., Avilez O.V.
UNICAMP, Campinas, Brazil
SerranN.V.
Rober, Campinas, Brazil
Andrade AM.
Univ. Sao Paulo, Brazil

2A.28 Influence of Hydrogen Plasma in the Properties of a-Si:H/a-SiC:H Interfaces 329

Fantini M.C.A, Pereyra I., Carreno M.P., Andrade AM.
Univ. Sao Paulo, Brazil

POSTER SESSION 2B: Amorphous Silicon and Related Materials II

2B.01 Thin Metallic Interfacial Layers: A Means to Improve the Tin-Oxide/Boron Doped
Hydrogenated Amorphous Silicon Interface? 335
Roca i Cabarrocas P.
LPICM, Ecole Polytechnique, Palaiseau, France
EickerU.
Solems, Palaiseau, France

2B.02 Improved Thickness Unifonnity and Optoelectronic Properties of a-Si:H Thin Films 339
Meiling R., Schropp RE.I., van Sark W.G.J., Bezemer J., van der Weg W.P.
Univ. Utrecht, The Netherlands

2B.03 Low Temperature Short Time Response of Light Induced Defects in a-Si:H 343
Kasaneva Reinoso J.
Univ. Antofagasta, Chile
Andreu Batalle J.
Univ. Barcelona, Spain

2B.04 Investigation of Spectral Mismatch in a-SiGe:H Based Tandem Solar Cells by Optical
Modelling 346
Gorn M., Curran C.
Phototronics Solartechnik, Putzbrunn, F.R Germany
Antoine-Labouret A, Schmitt J.P.M.
Solems, Palaiseau, France

2B.05 Characterization of a-Si:H Based Pin Cells by Subbandgap Photocu"ent Spectroscopy 350
Rubel H., Frammelsberger W., Geyer R, Scheppat B., Lechner P., Gorn M., Kniffler N.
Phototronics Solartechnik, Putzbrunn, P.R Germany

2B.06 Stability Optimization of a-Si:H Based Solar Cells 354

Lechner P., Rubel H., Kniffler N.
Phototronics Solartechnik, Putzbrunn, P.R Germany

xix
2B.08 Accelerated Stability Testing ofAmorphous Modules 358
Eicker U.
Solems, Palaiseau, France

2B.09 Light Trapping Due to TCO Roughness in a-Si:H Solar Cells Measured by Spatially-
Resolved Photothennal Deflection 361
Leblanc F., Ollivier P., Antoine-Labouret A
Solems, Palaiseau, France
Perrin J.
LPICM, Ecole Polytechnique, Palaiseau, France

2B.1D Photoinduced Effects in Hydrogenated Amorphous Silicon 365

Fortunato E., Martins R, Vieira M., Guimaraes L.
New Univ. Lisbon, Portugal

2B.ll Influence of the RF Power Density on the electrical Properties of Glow-Discharge 368
Amorphous Silicon
Amaral A
Inst. Sup., Univ. Tecnica, Lisbon, Portugal
Rodrigues L., Guimaraes L., Martins R.
New Univ. Lisbon, Portugal
Mencaraglia D., Djebbour Z., Kleider J.P., Longeaud C.
LGEP, Univ. Paris VI et XI, Gif-sur-Yvette, France

2B.13 Temperature Dependence of the Perfonnances of a-Si:H Solar Cells 371

Schirone L., Vincenzoni R. .
Univ. La Sapienza, Rome, Italy

2B.14 Electronic Transport Properties of High Deposition Rate a-Si:H Material 375
Mohammed-Brahim T., Rahal A, Ababou N., Beldi N., Aoucher M.
U.S.T.H.B., Algiers, Algeria
Mencaraglia D., Longeaud c., Kleider J.P., Glodt 0., Djebbour Z.
Univ. Paris VI et XI, Gif-sur-Yvette, France

2B.15 a-Si:H Ambipolar Diffusion Length and Effective Lifetime Measured by Flying Spot
Technique (FST) 379
Vieira M., Martins R, Fortunato E., Guimaraes L.
New Univ. Lisbon, Portugal

2B.16 Thennal Stability of a-Si:H/a-SiC:H Superlattices Studied by Hydrogen Effusion, X-ray

Diffraction, IR Spectroscopy and Photoluminescence 383
Fischer T., Muschik T., Hanesch P., Kolodzey J., Schwarz R
TU Munich, Garching, F.R. Germany

2B.17 Light Induced Defect Influence on a-Si Field Test Perfonnance 387
Berry W.B., Hahn MJ.
Univ. Notre Dame, IN, USA

2B.18 Effect of Unusual Experimental Parameters in R.F. Plasma Deposition ofAmorphous-

Silicon Based Materials 391
Bruno G., Capezzuto P., Cicala G., Mandoro P.
Univ. Bari, Italy

2B.20 Intrinsic Limitations of Single p-i-n a-Si Solar Cells 395

RuizJ.M.
Inst. Energia Solar, UPM, Madrid, Spain

xx
2B.21 Outdoor Evaluation ofAmorphous Silicon Solar Cell Modules 399
Joshi J.e., Konar P.K.
Indian Inst. Technology, New Delhi, India

2B.22 Analysis of Perfonnance Evolution ofAmorphous Silicon Modules by Experimentation in

Indoor and Outdoor Conditions 403
Ragot P., Costa H.S., Desmettre D.
CEN Cadarache, Saint Paullez Durance, France
Chenevas-Paule A
CEN Grenoble, France
Rossi E., Ossenbrink H., Van Steenwinkel R
Joint Res. Centre, Ispra, Italy

2B.23 Analysis of Different Commercial Technologies a-Si Submodules: Characterization and

Shorl- Time Outdoor Degradation Test 408
Chenlo F., Vela N.
CIEMAT-IER, Madrid, Spain

2B.26 Evaluation of Two Theoretical Models in Simulating the Perfonnance ofAmorphous-Silicon

Solar Cells 412
Protogeropoulos e., Brinkworth B.J., Marshall RH., Cross B.M.
Univ. Wales College of Cardiff, UK

2B.27 A New P(a-SiN:H!a-Si:H)-I(a-Si:H)-N(a-SiC:H!a-Si:H) Superlattice Solar Cell 416

Varonides Ae.
Univ. Scranton, PA, USA
Rothwarf A
Drexel Univ., Philadelphia, PA, USA

2B.28 Photoelectric Properlies of a-Si:H with Low Gap-States Deposited by VUV Photo-CVD 420
Shirafuji T., Yoshimoto M., Fuyuki T., Matsunami H.
Univ. Kyoto, Japan

POSTER SESSION 2C: PV Systems Technology

2e.01 Neural Networks - a Proper Approach to the Energy Management Problem? 427
Stoll M.
Fraunhofer Inst. ISE, Freiburg, F.R Germany

2e.02 The Effect of Tilt Angle and Voltage Conditions on PV System Perfonnance - An
Experimental Investigation 431
Jantsch M., Stoll M., Roth W., Kaiser R, Schmidt H., Schmid J.
Fraunhofer Inst. ISE, Freiburg, F.R. Germany

2e.03 Comparison Between Measured and Computed Data for a PV System 435
Ambrosone G., Catalanotti S., Coscia U., Troise G.
Univ. Napoli, Italy
SamoA.
ENEA-CRIF, Portici, Italy

2e.04 ISEE - Renewable Energy Infonnation System 439

Grebe R, Koch H.
ISET, Kassel, F.R Germany

xxi
2C.05 Mathematical Models for the Constrnction of a Renewable Energy Hybrid Plant 442
Marchetti G.P., Piccolo M.
Post and Telecommunication Ministry, Rome, Italy

2C.06 Photovoltaics, Utilities, and the Environment in the 21st Century 446
Darkazalli G.
Spire Corp., Bedford, MA, USA

2C.0? Control Considerations for Photovoltaic Powered Water Pumping Systems 449
BaItas P.
Center for Renewable Energy Sources, Koropi, Greece
Russell P.E.
Arizona State Univ., Tempe, AZ, USA

2C.09 Spatial and Temporal Chatacteristics of Short Tenn Fluctuations in Solar Radiation for PV-
Plant Applications 453
Beyer H.G., Luther J., Steinberger-Willms R.
Univ. Oldenburg, F.R. Germany
Decker B.
Inst. for Solar Energy Res., Hannover, F.R. Germany

2C.1O INSEL - A Stimulation System for Renewable Electrical Energy Sypply Systems 457
Luther J., Schumacher-Grohn J.
Univ. Oldenburg, F.R. Germany

2C.U Sensivity Analysis of Energy Requirements for Thin Film Photovoltaic Module Production 461
Hynes K.M., Pearsall N.M., Hill R.
Newcastle Polytechnic, UK

2C.12 ITE-BOSS, A New Software Tool for Photovoltaic System Design 465
Bosch A, Hoenes H.P., Jossen A, Karl H., Lehner G., Saupe G., Zahir A
ITE, Univ. Stuttgart, F.R. Germany

2C.13 Behaviour Analysis of Experimental Photovoltaic Systems 469

Paes P., Rodrigues C.
LNETI - Dep. Energias Renovaveis, Lisbon, Portugal

2C.14 PV System Perfonnance as Computed with Synthetic Rudiution Dutu 473

AguiarR.
LNETI - Dep. Energias Renovaveis, Lisbon, Portugal
Collares-Pereira M.
Center for the Conservation of Energy, Amadora, Portugal

2C.15 Hydrogen Energy Storage for Photovoltaic Power Systems 476

Kauranen P.S., Lund P.D.
Helsinki Univ. Technology, Espoo, Finland
Leppanen J.R.
Neste Corp. R&D, Porvoo, Finland
Spiers D.J.
Neste Advanced Power Systems, Abingdon, UK

2C.16 Analysis of PV Energy Possible Applications for Supplying Drinking Water to Rural Zones 480
MichelE.
COSTle, Digne, France

xxii
2C.17 PV System without Charge Controller for Grid-Connected and Stand-Alone Operation 483
Reismayr D.
Ludwig-Bolkow-Systemtechnik, Ottobrunn, F.R. Germany

2C.18 Methodological Approaches for the Estimation of a Capacity Credit of a Grid-connected

Photovoltaic Electricity Generation 486
Kaltschmitt M., Voss A.
Univ. Stuttgart, F.R. Germany

2C.20 Fiability Coefficient Maps for PV System Design 490

Egido M.A., Lorenzo E.
Inst. Energia Solar, UPM, Madrid, Spain

2C.21 Stochastic Modelling of a PV System with Small Concentration 494

Fara V.L., Grigorescu R.
Polytechnical Inst., Bucharest, Romania

SESSION 03: Crystalline Silicon - Materials and Devices

03.01 - Invited review lecture -

The Development of the Wafer Cost andAvailability for the Photovoltaic Industry; Ribbon/Sheet
Approaches and its Comparison with Wafers: Efficiency-cost Trade-off 501
HerzerH.
Wacker-Chemitronic, Burghausen, Germany

03.02 Behaviour of Multicrystalline Silicon Materials under High Efficiency Process Conditions 507
Schmidt W., Rasch KD., Wahl G.
Telefunken Systemtechnik, Heilbronn, F.R. Germany
EbnerW.
Telefunken Systemtechnik, Wedel, F.R. Germany

03.03 Recent Advances in Silicon Inversion Layer Solar Cells and their Transfer to Industrial Pilot
Production 511
Hezel R.
Univ. Erlangen-Nurnberg, F.R. Germany
Hoffmann W., Jaeger K
Nukem, Hanau. F.R. Germany

03.04 Increasing Cell Efficiency of S-Web Silicon Ribbon Material Using Synthetic Silica Crucibles 515
Falckenberg R., Grabmaier J.G., Lerchenberger A.
Siemens AG, Munich, F.R. Germany
Schindler R.
Fraunhofer lost. ISE, Freiburg, F.R. Germany

03.05 On-Line Manufacturing of Si Solar Cells Compatible with High Speed Si-Ribbon Growth 519
Grabmaier J.G., Urbach H.P., Eisenrith KH.
Siemens AG, Munich, F.R. Germany
Munzer KA
Siemens Solar, Munich, F.R. Germany

03.06 Controlled Nucleation: A New Approach to Improve The Crystallinity of Raft-Foils 524
Reis I., Hurrle A, Rauber A
Fraunhofer Inst. ISE, Freiburg, F.R. Germany
Helmreich D., Beck A
Heliotronic, Burghausen, F.R. Germany

xxiii
POSTER SESSION 3A: Compound Semiconductor Solar Cells I

3A01 Stem-Stebic Technique for Defect Investigation in Semiconductor Photovoltaic Materials 529
Garozzo M., Parrett a A.
ENEA-CRIF, Portici, Italy
Vittori M.
ENEA-Casaccia, Rome, Italy
Camanzi A, Alessandrini P.
Eniricerche, Rome, Italy

3A02 Improvement of Device Efficiency by Time-Resolved Photoluminescence Measurements 533

Ahrenkiel RK., Keyes B.M., Dunlavy DJ., Kazmerski L.L.
SERI, Golden, CO, USA

3A03 Investigations of Emitter Characteristics of LPE GaAs Solar Cells 537

Welter H., Bett A, Ehrhardt A, Wettling W.
Fraunhofer lost. ISE, Freiburg, F.R Germany

3A04 InGaP, a Promising Material for Tandem Solar Cells 541

van Geelen A, Hageman P.R, Gabrielse, Giling L.J.
Univ. Nijmegen, The Netherlands

3A05 Design and Measurement ofAntireflection Coatings for AIGaAs/GaAs Solar Cells 545
Nell M.E., El-Ankah Z., Eschrich H., Lin D.D., Nischwitz G., Reinicke B.,
WagemannH.G.
TU Berlin, F.R Germany

3A06 Comparison of an Analytical and a Finite Differences Model Referring to GaAs Solar Cell
Performance 549
Eschrich H., Bruns J., Nell M.E., Reinicke B., Wagemann H.G.
TU Berlin, F.R Germany

3A07 Photovoltaic Activity in Thin Polycrystalline GaAs Films Produced by Ion-Assisted

Deposition 553
Bonnet D., Oelting S., Luke U.
Battelle Inst., Frankfurt/Main, F.R Germany

3A08 Graded Ga1Jn,!' Window on GaAs Solar Cells Grown by Isothermal Liquid Phase Epitaxy 556
01chowik J.M.
Univ. Lublin, Poland
Soumana H., Gavand M., Mayet L.
INSA, Villeurbanne, France

3A09 A TTC2 Device with an Al.32Ga.6w4s-p GaAs Heterojunction 559

Mayet L., Gavand M., Montegu B.
Univ. Lyon, France
Laugier A
INSA, Villeurbanne, France

3A1O Cost Perspectives of GaAs Thin-Film Solar Cells 563

Alsema E.A., Cuelenaere RF.A., Turkenburg w.e.
Univ. Utrecht, The Netherlands

3All Development and Evaluation of CdS/CdTe Thin Film PV Cells 567

Skarp J., Koskinen Y., Lindfors S., Rautiainen A, Suntola T.
Microchemistry, Espoo, Finland

xxiv
3A12 Photovoltaic Properties of an Epitaxial CdS/CdTe/ZnTe n-i-p Heterostmcture Grown by
MOVPE 570
Simmons M.Y., AI-Allak H.M., Durose K, Brinkman AW.
Univ. Durham, UK

3A14 Thin Films of Cadmium Telluride Produced Using Stacked Elemental Layer (SEL)
Processing 574
Bhatti M.T., Groarke E.P., Miles R.W., Carter M.J., Hill R
Newcastle Polytechnic, Newcastle upon Tyne, UK

3A15 Screenprinted CdS-CdTe Solar Cells 577

Clemminck I., Burgelman M., Vervaet A., De Poorter J.
State Univ. Gent, Belgium

3A16 Photovoltaic Properties of n-CdTe/p-Te Junctions 581

Yown I., Cadene M.
Univ. Montpellier, France
LaplazeD.
Univ. Cheikh Anta Diop, Dakar, Senegal
Lincot D.
ENSCP, Paris, France

3A17 Variations in the Resistivity of Sprayed Films of Cadmium Sulphide 584

Buckley R.W., Cotier B., Dixon K, Gallagher T, Goddard N.
Twyford CE High School, London, UK

3A18 The Development of CuInS2 Single Crystals and Thin Films for PVApplications 586
Fearheiley M., Dietz N., Scheer R, Kanis M., Fiechter S., Briissler M., Dzionk e.,
Metzner H., Migge H., Lewerenz H.J.
Hahn-Meitner-Inst., Berlin, F.R. Germany

3A19 Preparation of CdSiAsz-Homojunction Solar Cells and its Anisotropic Optical Properties 590
Baumgartner F.P., Schweikardt H.P., Bucher E.
Univ. Konstanz, F.R Germany

3A20 Thin Film Preparation of FeS2 (Pyrite) byArgon Sputtering, Plasma Reaction, MOCVD and
Spray Pyrolysis 594
Hopfner C., Ennaoui A., Lichtenberger D., Birkholz M., Smestad G., Fiechter S.,
Tributsch H.
Hahn-Meitner-Inst., Berlin, F.R. Germany

3A21 Thin Films of MoSe2 and WSe2 Prepared by Soft Selenization as Bulk Material for Solar
Cells 597
Jager-Waldau A., Lux-Steiner M., Bucher E.
Univ. Konstanz, P.R. Germany

3A22 Interface Recombination Velocity ofAI/WSe2 and ZnO/WSe2 Solar Cells Detennined by
Ebic and Light Induced Current Measurements 601
Vogt M., Friemelt K, Lux-Steiner M.e., Keil M., Reetz W., Bucher E.
Univ. Konstanz, Konstanz, P.R. Germany

3A23 ZnO Transparent Conducting Thin Films Deposited by Ultrasonic Spray Pyrolysis for Solar
Energy Conversion 605
Tiburcio Silver A
CINVESTAV-IPN, Mexico
Joubert J.e.
ENSPG-INPG, St. Martin d'Heres, France

xxv
3A24 Solvent Effect on 2nO Thin Films Prepared by Spray Pyrolysis 609
Guillemoles J.F., Lincot D., Cowache P. Vedel J.
ENSCP, Paris, France

3A25 Sprayed 2nO Thin Films as Optical Window in CUlnSez Based Solar Cells 613
Belghit K, Subhan M.A, Duchemin S., Bougnot J.
Univ. Montpellier II, France
RUhle U.
IPE, Univ. Stuttgart, P.R. Germany

POSTER SESSION 3B: High Efficiency Crystalline Silicon Solar Cells

3B.01 Analysis of the Operation of Laser Grooved Silicon Solar Cell 619
Strollo AG.M., Vitale G.
Univ. Napoli, Italy

3B.02 Progress on High Efficiency Silicon Solar Cells 623

Alonso J., Luque A, Tobias I., Sala G., Cuevas A
Inst. Energia Solar, UPM, Madrid, Spain

3B.03 The PV Eye; a High Efficiency Converter Based on Light Trapping and Spectrum Splitting 627
Luque A, Sala G., MmanoJ.C., Davies P., Tobias I., Alonso J., Algora C., Araujo G.L.,
Ruiz J.M., Cuevas A, Olivan J.
lost. Energia Solar, UPM, Madrid, Spain
Dunn P., Rice G.
Univ. Reading, UK
Knobloch J., Voss B.
Fraunhofer Inst. ISE, Freiburg, P.R. Germany
Flores C.
CISE, Milan, Italy

3B.04 SiOz-Passivated High Efficiency Silicon Solar Cells: Process Dependence of Si-SiO z
Interface Recombination 631
Aberle A, Glunz S., Warta W., Kopp J., Knobloch J.
Fraunhofer Inst. ISE, Freiburg, P.R. Germany

3B.05 The Use of Plasma Silicon Nitrides and Hydrogen Plasma on Polycrystaliine Silicon Solar
Cells 636
Coppye J., Ghannam M., De Schepper P., Nijs J., Mertens R.
IMEC, Leuven, Belgium

3B.07 EffeCt of Temperature and Light Intensity on the Perfonnance of MIS Inversion Layer Solar
Cells 640
Jaeger K, Hoffmann W.
Nukem, Hanau, F.R. Germany
HezelR.
Univ. Erlangen-Niirnberg, F.R. Germany

3B.09 Laser Scanning Analysis of Silicon Solar Cells 644

Hughes AE., Neville B.M., Heasman KC., Lesniak M.P.
BP Res. Centre, Sunbury on Thames, UK

xxvi
3B.1O Defects Induced in Bulk Silicon by Rapid Thermal Annealing 647
Susi E., Fabbri R, Poggi A.
CNR-Ist. LAMEL, Bologna, Italy
Passari L., Carotta M.e., Merli M.
1st. di Fisica dell'Univ., Ferrara, Italy

3B.ll Evaluation of Silicon Nitride SUiface Passivation of Crystalline Silicon by Means of Time
Resolved Microwave Conductivity 65'1
Bickl T.
Univ. Wiirzburg, F.R Germany
Creutzburg u., Silber D.
Univ. Bremen, F.R Germany
Ebner W., Eyckmans M., Wandel G.
Telefunken Systemtechnik, Wedel, F.R. Germany

3B.12 Photothermal Measurement of Minority Carrier Diffusion in Devices 653

Buchner B.
IPE, Univ. Stuttgart, F.R Germany
Cella N.
Inst. Politecnico, Rio de Janeiro, Brazil
CahenD.
Weizmann lost. of Science, Rehovot, Israel

3B.13 Non-Conventional Emitters for Polycrystalline Silicon Solar Cells 657

Coppye J., Demesmaeker E., Elgamel H.E., Szlufcik J., Ghannam M., Nijs J., Mertens R.
IMEC, Leuven, Belgium
Le Q.N., Rodot M., Laroche J.M.
CNRS/LPSB, Meudon, France
Sarti D., Loubly P.
Photowatt, Caen, France

3B.14 Improvements of Polycrystalline Silicon Wafers and Conventional Solar Cells by a Boron
Diffusion Step 661
Le Q.N., Rodot M.
CNRS/LPSB,Meudon, France
Sarti D., Loubly P.
Photowatt, Caen, France
Coppye J., Demesmaeker E., Nijs J.
IMEC, Leuven, Belgium

3B.15 An Approach to the Development of Efficient 31 cm 2 Area All-Implanted n +pp + Silicon

Solar Cells 664
Ruzinsky M., Baratka S.
Slovak Technical Univ., Bratislava, Czechoslovakia
Palaj J.
Tesla, Piestany, Czechoslovakia
Seidl P.
Czech Tech. Univ., Prague, Czechoslovakia
Kosek F., Cimpl Z.
Univ. of Chemical Technology, Pardubice, Czechoslovakia

3B.16 Maximising Minority Carrier Lifetime in High Efficiency Screen Printed Silicon BSF Cells 667
Bruton T.M., Mitchell A., Teale L.
BP Solar, Leatherhead, UK
KnoblochJ.
Fraunhofer lost. ISE, Freiburg, F.R Germany

xxvii
3B.17 About the Effect of Impurity Contaminations, Processing and Thennal Annealing on the PV
Properties of Polycrystalline Silicon 670
Pizzini S., Acciarri M., Binetti S., Acerboni S.
Dip. Chimica Fisica ed Elettrochimica, Milano, Italy
Falconieri M., Pirozzi L., Garozzo M.
ENEA-FORI, Rome, Italy

3B.18 Microstructure Investigations and Electrical Measurements in Polycrystalline Solar Silicon 674
Moller H.J., Rodak E., Huang L.
Case Western Reserve Univ., Cleveland, OH, USA
Griess M.
Univ. Gottingen, F.R. Germany

3B.19 Donor and Acceptor Neutralization in Multicrystalline Silicon 678

Margadonna D., Ferrazza F., Peruzzi R.
Italsolar, Rome, Italy
Pizzini S., Acerboni C., Tarchini L.
Univ. Milano, Italy
De Lillo A.
ENEA-Casaccia, Rome, Italy
Wei Xiwen
Univ. of Technology, Dalian, P.R. China

3B.20 Limiting Factors in the Emitter and Base Region of Diffused Solar Cells Made from 681
Crystalline Lower Purity Silicon
Schlosser V.
Inst. fUr Festkorperphysik, Univ. Vienna, Austria

3B.21 Transport Properties in the Emitter of the Silicon Solar Cells: A Comparison Between a New
Analytical Model and 17le Experimental Results 685
Abenante L., Arabito G., La Motta S., Pirozzi L.
ENEA-FORI, Rome, Italy
Di Francia G., Garozzo M.
ENEA-CRIF, Portici, Italy

3B.22 A Flexible Measurement System for R&D in Industry 689

Martinez V.E., Saenz MJ., Gutierrez J.R., Jimeno J.e.
Univ. Pafs Vasco, Bilbao, Spain
Perezagua E.
Isofot6n, Madrid, Spain

3B.23 663 m V Open Circuit Voltage Epitaxial Thin Film Silicon Solar Cells 692
Blakers A.W.
Univ. New South Wales, Kensington, Australia
Werner J.H., Bauser E., Queisser H.J.
Max Planck lnst. fUr Festkorperforschung, Stuttgart, F.R. Germany

POSTER SESSION 3C: Hybrid and Utility PV

3e.01 Line Photovoltaic Back-up System for Rural Nucleus with Low-quality Electric Supply 699
Gracia Navarro S.
Compania Sevillana de Electricidad, Seville, Spain

xxviii
3C.02 Scenario for Integration of PV Systems into Multiple-Carrier Grids with Integrated Energy
Storage 701
S(Ilrensen B.
COWIconsult, Lyngby, Denmark

3C.03 PVj Diesel Systems on Houseboats 705

van der Weiden T.CJ., Dunselman C.P.M., Hoekstra K, Blok K
Ecofys, Utrecht, The Netherlands

3C.05 A Photovoltaic-Hydrogen-Fuel Cell Energy System: Preliminary Operational Results 708

Lehman P.A., Chamberlin C.E.
Humboldt State Univ., Arcata, CA, USA

3C.06 Gridconnected 30-kWp Photovoltaic Power Plant in Finland 712

Nyman C.
Technical Res. Centre, Espoo, Finland

3C.07 Design Tool Photo for Sizing of Hybrid Power Systems: Program Verification 716
Manninen L.M., Lund P.D.
Helsinki Univ. of Technology, Espoo, Finland

3C.08 Development Potential of Small-Scale Grid-Connected Residential PV Systems 720

Peippo K, Lund P.D.
Helsinki Univ. of Technology, Espoo, Finland

3C.12 Eight Years Experience with the 300-kWp Solar Power Plant on the Island Pellwonn (FRG)
and the Extension to a 1.25-MW Hybrid System 724
Kruse C., Lawalt HJ., Maass K
Telefunken Systemtechnik, Wedel, F.R. Germany

3C.13 Autonomous Photovoltaic System with Auxiliary Source for Battery Charging 728
van Dijk V.A.P., Alsema E.A., Albers R.A.W.
Univ. Utrecht, The Netherlands
Degner T., Gabler H., Wiemken E.
Univ. Oldenburg, F.R. Germany

3C.14 Integration of Photovoltaics into Electric Utilities: Issues and Approaches 734
Annan R.H., Rannels J.E.
US Dept. of Energy, Washington, DC, USA
Cabraal R.A.
Meridian Corp., Alexandria, VA, USA

3C.15 lOO-kW Grid-Connected PV Insta//ationAlong Rail Infrastructure in Southem Switzerland

- A Feasibility Study 738
Hiichler R., Nordmann T.
TNC Consulting, Chur, Switzerland

3C.17 PV Electric Utility Interface Issues, Understanding the Problem of Hannonic Cu"ents in the
Grid System 742
Tortoreli M.
Center for Renewable Energy Systems, Koropi, Greece
Russell P.E.
Arizona State Univ., Tempe, AZ, USA

xxix
3C.18 Design of a Hybrid PV/Diesel System from the Analysis of the Present PV System
Peifonnance in a Dairy Fann in Pozoblanco, Spain 746
GonzaIezA
CIEMAT-IER, Madrid, Spain
MayerD.
Centre d'Energetique, Valbonne, France

3C.19 PV System Generation Analysis, 0 & M and Cost Evolution - Hidroelectrica Espanola SA.
Experience 750
Pantoja A, Garcia J., de Julian A
Hidroelectrica Espanola, Madrid, Spain

3C.20 Half a Dollar per kWh and Still Cheap Electricity 752
Daey Ouwens C.
Provincial Bureau Energie, Haarlem, The Netherlands

3C.21 Tiso: 4-kW ExperimentalAmorphous Silicon PV Power Plant 755

Chianese D., Camani M., Ceppi P., Iacobucci D.
Dip. dell'Ambiente, Bellinzona, Switzerland

3C.22 Influence of Different Revenue Stmctures on the Economy of PV Plants in Selected

Countries 759
Hille G., Staiss F., Steinborn F.
7BW, Stuttgart, F.R Germany

3C.23 The 4-kW PV Pilot Plant "Muttler" 763

SchadelJ.
Ascom Energy Systems, Bern, Switzerland

3C.24 30-kW Photovoltaic Plant in the Alps ofAustria 766

Nentwich A, Szeless A
Verbundgesellschaft, Vienna, Austria
Schneeberger M., Wilk H.
Oberosterreichische Kraftwerke (OKA), Linz, Austria

3C.25 Battery Research in the Terschelling PV System 771

van der Weiden T.C.J., Boumans J.H., Blok K
Ecofys, Utrecht, The Netherlands

SESSION 04: Compound Semiconductor Solar Cells

04.01 - Invited review lecture -

Thin Film Compound Semiconductor Solar Cells: An Option for Large Scale Applications? 777
SchockH.W.
IPE, Univ. Stuttgart, F.R Germany

04.02 RecentAdvances in Single-Junction and Tandem Thin Film Modules Based on CuInSe2 783
Eberspacher C., Ermer J., Fredric c., Jensen c., Gay R, Pier D., Willet D.
Siemens Solar Industries, Camarillo, CA, USA
Karg F.
Siemens AG, Munich, F.R Germany

04.03 Photocurrent Transport in Heterojunctions with Graded Cu(In,Ga)Se2Absorbers 787

Menner R, Walter T., Schock H.W.
IPE, Univ. Stuttgart, F.R Germany

xxx
04.04 Stable, High Efficiency Thin Film Solar Cells Based On Electrodeposited Cadmium Telluride 791
Turner AK., Woodcock J.M., Ozsan M.E., Summers J.G.
B.P. Solar, Leatherhead, UK

04.05 Analysis of the Diode Currents in Back Contacted Emitter GaAs Solar Cells 794
Marti A, Araujo G.L., Algora C., Maroto J.C.
Inst. Energia Solar, UPM, Madrid, Spain
Flores C.
CISE, Milan, Italy

04.06 GaAs on Si Solar Cells: Photovoltaic Characterization of GaAs Grown directly on Si and with
Intermediate Buffer 798
Vilela M.F., Leycuras A, Freundlich J.c., Grenet J.c., Strobl G., Leroux M., Neu G.,
Gibart P., Verie c.
CNRS, Valbonne, France
BremondG.
INSA, Villeurbanne, France

POSTER SESSION 4A: Stand Alone PV Systems

4A01 Photovoltaic Systems for Archaeological Areas 805

Barra L., Castello S., Messana C.
ENEA-FORI, Rome, Italy
Bricca E.
Italsolar, Rome, Italy

4A02 Lighting ofAirfield Landing Strips by Photovoltaic Systems 809

Barra L., Castello S., Messana C.
ENEA-FORI, Rome, Italy
Bricca L.
Italsolar, Rome, Italy

4A03 The Solar Boat "KORONA ": Two Years of Experience 813
Schaffrin C.
Fachhochschule Konstanz, F.R. Germany

4A04 Micropower Telemeasurement Device - Application on a Safety PV Telecom Relay for an

Anticipated Geological Disaster in the French Alps 816
AubreeM.
France Telecom, Issy Les Moulineaux, France

4A06 Design and Development of a Car Ventilation System Powered by a Photovoltaic Generator 819
Messana C., Parisi Presicce L.
ENEA-FORI, Rome, Italy
Palazzetti M., Petruccioli G., Salotti G.
Centro Ricerche Fiat, Turin, Italy

4A09 Zambelli Photovoltaic Pumping Station (70 kWp) Operating Experiences and Results 822
SorokinA
TEAM, Rome, Italy
ZamboniG.
AGSM, Verona, Italy

4AI0 Photovoltaic Systems for Railways in Italy 826

Trentini M.
Sistemi Energetici Integrati, Florence, Italy

xxxi
4A.ll Expanding Photovoltaic Energy Usage: The Competitive Energy Environment 830
Callaghan W.T.
Caltech-Jet Propulsion Lab., Pasadena, CA, USA

4A.12 A Methodological Framework for Calculating the External Costs of Energy Technologies 834
Baumann A., Hill R.
Newcastle Polytechnic, UK
Ferguson R., Fells I.
Univ. Newcastle, UK

4A.13 Preliminary Operation Data of the Improved PilofPlant ''Isola del Giglio" 838
Li Causi S., Castello S., Murzilli G.
ENEA-FORI, Rome, Italy

4A.14 Evaluation of the Photovoltaic Reverse Osmosis Desalination Plant in Almeria, Spain 842
Andujar Peral J.M, Contreras Gomez A.
Facultad de Ciencias Experimentales, Almeria, Spain
Trujillo Navarro J.M.
Gomar, Malaga, Spain

4A.15 Rural Electrification Experience in South Spain 847

Trujillo Navarro J.M., Navarro Sevilla J.L., Jimenez Aguilera F.
Gomar, Malaga, Spain

4A.16 Application of Photovoltaics for the Illumination of Infonnation and Advertising Systems 850
Steinhiiser A., Schmidt H., Roth W.
Fraunhofer Inst. ISE, Freiburg, F.R. Germany

4A.20 Photovoltaic Powered Cathodic Corrosion Protection Systems with Advanced System
Technologies 854
KoruppK.H.
SET, Wedel, F.R. Germany

4A.21 Experiences with a PV Power Plant for a Remote Greenhouse Cooling System 858
Laukamp H., Braun P.
Fraunhofer Inst. ISE, Freiburg,F.R. Germany
Ayyash S.
Inst. for Scientific Res., Safat, Kuwait

4A.22 Multi-purpose PV Plant in Lisbon, Portugal 862

Guimaraes L., Fortunato E., Martins R., Ribas A.
New Univ. Lisbon, Portugal
Helm P., Hanel A., Ehmann H., Imamura M.S.,
WIP, Munich, F.R. Germany

4A.23 Design, Development and Demonstration of Environmental Measuring Stations with

Photovoltaic Energy Supply 865
Wiesner W., Leisen P., Schwenke S.
TOV Rheinland, Cologne, F.R. Germany

4A.24 Lessons from a PV Rural Electrification Project at "Sierra de Segura" (Spain) 869
Eyras J .R., Lorenzo E.
Inst. Energfa Solar, UPM, Madrid, Spain

xxxii
POSTER SESSION 48: Compound Semiconductor Solar Cells II

4B.01 A New Coevaporation/Selenization Process for High Quality CuInSe2 Devices 875
Dimmler B., Content A, Schock H.W.
Univ. Stuttgart, F.R. Germany

4B.02 Low Pressure Vapor Phase Selenization of Cu-In Films Without HzSe 879
Kessler J., Dittrich H., Grunwald F., Schock H.W.
IPE, Univ. Stuttgart, F.R. Germany

4B.03 Scaleable Large Area Compatible Technique for the Production of CuInSe2 Based Solar
Cells 883
Badawi M.H., Hyland M., Carter MJ., Hill R.
Newcastle Polytechnic, UK
Knowles A
Pilkington, St. Asaph, UK

4B.04 Thin Films of CuInSe2 Produced by Thennal Annealing of Multilayers with Ultra-Thin
Stacked Elemental Layers 887
Nakada T., Yuda K., Kunioka A
Aoyama Gakuin Univ., Tokyo, Japan

4B.05 CuInSe2/CdS Thin Film Solar Cells by Selenization and Sputtering 891
Romeo N., Bosio A, Mussini P.
Univ. Parma, Italy

4B.06 One Step Electrodeposition of CuInSe2 893

Thouin L., Rouquette-Sanchez S., Vedel J.
ENSCP, Paris, France

4B.07 Effect ofAnnealing Temperature on the Optical Properties of Electrodeposited CuInSe2 Thin
Films 897
Guillen c., Herrero J.
CIEMAT-IER, Madrid, Spain

4B.08 Screen Printing of CIS Films for CIS-CdS Solar Cells 900
Vervaet A
Univ. Bujumbura, Burundi
Burgelman M., Clemminck I., Casteleyn M.
State Univ. Gent, Belgium

4B.09 Stmctural Properties of CuInSe2 Thin Films for High Efficiency Solar Cells 904
Fujisawa H., Ihara T., Sato H., Hama T., Ohsawa M., Ichikawa Y., Sakai H.
Fuji Electric Corp. Res. and Development, Kanagawa, Japan

4B.1O Atomic Level Studies: Nanoprocessing and Nanocharacterization of Defects in CuInSe2 908
Kazmerski L.L.
SERI, Golden, CO, USA

4B.ll Investigation of the Defect Chemistry of CulnSe2 Single Crystals by DLTS Measurements 913
Moller H.J., Rodak E.M.
Case Western Reserve Univ., Cleveland, OH, USA

4B.14 Epitaxial Effects and Diffusion in Polycrystalline Chalcopyrite Thin Films 917
Dittrich H.
ZBW, Stuttgart, F.R. Germany
Waiter T., Schock H.W.
IPE, Univ. Stuttgart, F.R. Germany
xxxiii
4B.15 Etching and DLTS Measurements in Bridgman-grown CuInSe2 921
Yip L.S., Weng W.S., Li L., Shih 1., Champness c.R.
McGill Univ., Montreal, Que., Canada

4B.16 CuInSe2/CdS Interface Fonnation Studied by XPS 924

NiemiE.
Royal Inst. Technology, Kista, Sweden
Stolt L.
Swedish Inst. of Microelectronics, Kista, Sweden

4B.17 Optimizing Thin Film, Chalcogenide-Based Solar Cells via Chemical Surface Treatments 927
Klenk R., Schock H.W.
IPE, Univ. Stuttgart, F.R. Germany
Cahen D., Engelhard T., Moons E.
Weizmann Inst. of Science, Rehovot, Israel

4B.18 Chemical Bath Deposition of Cadmium Sulfide Thin Films 931

Lincot D., Vedel J.
Ecole Nat. Sup. de Chimie, Paris, France

4B.20 Diffusion Length Measurement and Modeling of CuInSe2-(Zn,Cd)S Solar Cells 935
Schmid D., Jager-Waldau G.J., Schock H.W.
IPE, Univ. Stuttgart, F.R. Germany

4B.21 The Effect of Localized Defect Features on CuInSe2 Device Perfonnance 939
Willett D., Aldrich D., Chesarek W., Mitchell K.
Siemens Solar Industries, Camarillo, CA, USA

4B.22 SnOiF Transparent Conducting Films Prepared By Pyrosol Deposition 942

Yoon K.H., Song J.S.
Inst. Energy and Resources, Daejeon, Rep. of Korea

4B.23 The Effect of Hydrogen on The Plasma Deposition of a-SiGe:H Thin Films for Tandem
Solar Cell Applications 946
Zeman M., Tao G., Geerts M.J., Metselaar J.W.
Univ. of Technology, Delft, The Netherlands
Ferreira I.
New Univ. Lisbon, Portugal

4B.24 Anomalous X-Ray Scattering Study of an Amorphous Si.sGe.s:H Alloy 950

Bouchet-Fabre B.
URA-INSA, Mont St. Aignan, France
Elkaim P., Cuniot M., Dixmier J.
CNRS, Meudon, France

4B.26 The Mobility Edges ofAmorphous Silicon-Carbon Alloys 953

Wu Z.Y., Siefert J.M.
Solems, Palaiseau, France
EquerB.
LPICM, Ecole Polytechnique, Palaiseau, France

POSTER SESSION 4C: Component Performance and Development

4C.01 Considerations About the Design of PV Modules for Central Power Plants 959
RediP.
Univ. Florence, Italy

xxxiv
4e.02 Electrostatic Cover Glass Bonding to GaAs Solar Cells 963
Nowlan J.M., Darkazalli G.
Spire Corp., Bedford, MA, USA

4e.03 Modules Assembled with Diffuse Reflectors for Photovoltaic Bifacial Cells 967
Moehlecke A, Krenzinger A
PROMEC/UFRGS, Porto Alegre, Brazil

4e.04 Conceptual Development of Low-cost A"ay Strncture 971

Hanel A
WIP, Munich, F.R. Germany

4e.05 Constrnction and Measurement of a Prototype of PV Module with Static Concentration 975
Parada J., Mifiano J.e.
lust. Energia Solar, UPM, Madrid, Spain
Silva J.L.
Isofot6n, Madrid, Spain

4C.06 Photovoltaics on Light-Weight Strnctures - A Mounting Technology with Many Advantages 979
Wecker R.
MWB Messwandler-Bau, Bamberg, F.R. Germany

4C.07 Concentration Limits of Primaries for Cells Inmersed in Optically Dense Media 982
Vidal P.G., Almonacid G.
EU Politecnica Jaen, Granada, Spain
Luque A, Mifiano J.C.
Inst. Energia Solar, UPM, Madrid, Spain

4C.08 On The Energy Collected by Compound Parabolic Collectors 985

Macagnan M.H., Lorenzo E., Sanchez Martin J.1.
Inst. Energia Solar, UPM, Madrid, Spain

4C.09 The Concept-90 Photovoltaic Concentrator Module 988

Maish AB., Hund T.D., Quintana M.A
Sandia Nat. Lab., Albuquerque, NM, USA
Chiange.J.
General Electric Astrospace, Princeton, NJ, USA

4C.1O PV Concentrators Today and Tomo"ow 992

Maish AB., Chamberlin J.L.
Sandia Nat. Lab., Albuquerque, NW, USA

4C.ll A Reliable Photovoltaic Battery Charge Regulator with Ideal VB; I B vs t Profile 996
Pressas S.A., Makios V.
Univ. Patras, Greece

4C.12 Concerted Action on Battery Control and Management in PV Systems 1000

McCarthy S., Hill M.
Hyperion Energy Systems, Cork, Ireland

4e.13 Development and Evaluation of a 20-Watt MPP Tracking Charge Controller for PV Supplied
Remote Monitoring Stations 1004
Decker B., Gentz M.
lust. fiir Solarenergieforschung, Hannover, F.R. Germany

xxxv
4C.14 Equipment for Testing of Maximum Power Point Trackers Under Variable, Programmable
Conditions 1008
Veltman AT., Klockner R., Geers KKW., van Twisk J.
Netherlands Energy Res. Foundation, Petten, The Netherlands

4C.15 Battery Control Unit with State of Charge Indicator 1012

Jossen A, Bosch A, Hones H.P., Karl H., Lehner G., Saupe G., Zahir A
IPE, Univ. Stuttgart, F.R. Germany

4C.16 Prevention of Islanding Operation in Utility Interactive Photovoltaic Systems 1016

Wang Y., Russell P.E.
Arizona State Univ., Tempe, AZ, USA

4C.17 Active Filters in Autonomous Energy Supply Systems 1020

Caselitz P., Krengel u., Mevenkamp M., Petschenka J.
ISET, Kassel, F.R. Germany

4C.18 A Passive Cooling System for Remote PV Power Systems 1024

Pederson S., Selbagen L.
Neste Advanced Power Systems, Skarholmen, Sweden
Spiers D.J.
Neste Advanced Power Systems, Abingdon, UK

4C.19 Solar Inverter for Grid-Connected PV Systems 1027

Crastan V.
Eng. College of the State of Bern, Biel, Switzerland

4C.21 Perfonnance of the Power Conditioning Unit of the 300 kW Delphos Plant 1029
Li Causi S., Apicella F., Noviello G., Sarno A
ENEA-FORI, Rome, Italy
Bucci G., Parasiliti F.
Univ. degli Studi di L'Aquila, Italy

4C.22 Results of the Solar Operation of a 10-kW PV-Electrolysis System with Different Power
Matching Concepts 1033
Brinner A., Bussmann H., Steeb H.
DLR, Inst. for Technical Thermodynamics, Stuttgart, F.R. Germany

4C.24 A Graphic-Analytical Method to Study the Direct Coupling Photovoltaic Generators

-DC Motors 1037
Abete A, Barbisio E., Tommasini R.
Politecnico di Torino, Italy

4C.25 Electronic Ballasts - Important Components in Photovoltaically Powered Lighting Systems 1042
Pfanner N., Roth W.
Fraunhofer Inst. ISE, Freiburg, F.R. Germany
Gerhold V., Vaassen W.
TOV Rheinland, Cologne, F.R. Germany

4C.26 Testing PV Devices for the Community 1+ D and Demonstration Projects 1046
Chenlo F., Martin N., Copetti J., Mukadam K, Zaragoza J.M.
CIEMAT-IER, Madrid, Spain

xxxvi
4C.27 A Novel Method of Monitoring Total Nonnal Irradiance Data with a Stationary Sensor 1050
Imamura M., Helm P.
WIP, Munich, F.R Germany
Beer G., Guastella S., Chimento G.
Conphoebus, Piano D'Arci, Italy

4C.28 Errors in Current-Voltage Measurements of Photovoltaic Devices Introduced by Flash

Simulators 1055
Ossenbrink H., Zaaiman W.
Joint Res. Centre, Ispra, Italy
Drainer A.
Strathdyde Univ. Glasgow, UK

4C.29 Analysis of the Variation in the Spectral Distribution of Solar Irradiance 1058
Fabero F., Chenlo F.
CIEMAT-IER, Madrid, Spain

SESSION 05: Amorphous Silicon and Alloys

05.01 Interface Characterization During their Fonnation in Amorphous Silicon Solar Cell Stntctures by
In-Situ Transient Photoconductivity Measurements 1065
Neitzert H.C., Hirsch W., Swiatkowski C., Kunst M.
Hahn-Meiner Inst., Berlin, F.R Germany

05.02 The Role of the Reduction of ITO on the Electrical Properties of the p-i Junction: A Thin SiO
Layer as a Remedy 1069
Carvalho C., Martins R, Guimaraes L.
New Univ. Lisbon, Portugal
de Nijs J .M.M.
Univ. of Technology, Eindhoven, The Netherlands

05.03 Potassium Ion Implantation Doping of the n-Layer for p-i-n Amorphous Silicon Solar Cells 1072
Rizzoli R, Summonte C., Galloni R
CNR-Ist. Lamel, Bologna, Italy
Zignani F.
Dip. Chimica Applicata e Scienza dei Materiali, Bologna, Italy
Nylandsted Larsen A.
Univ. Aarhus, Denmark

05.04 Comparison of a-Ge:H Preparation Techniques in View of Optoelectronic Optimization 1075

Heintze M., Eberhardt K., Kessler F., Bauer G.H.
IPE, Univ. Stuttgart, F.R Germany

05.05 Stability of a-Si/Ge:H Device Material to Light and KeV-Stress Exposure 1079
Ebersberger B., Karg F., Kausche H., Kriihler W., Pliittner R
Siemens AG, Munich, F.R. Germany
Pierz K.
Univ. Marburg, F.R. Germany
Schneider U., SchrOder B.
Univ. Kaiserslautern, F.R. Germany

xxxvii
05.06 Helium Dilution as a Means to Obtain a Low Defect Density Hydrogenated Amorphous Silicon
at High Deposition Rates in RF Glow-Discharge Systems 1083
Roca i Cabarrocas P., Vanderhaghen R
LPICM, Ecole Polytechnique, Palaiseau, France
Bouizem Y., Theye M.L.
Univ. Pierre et Marie Curie, Paris, France
Mencaraglia D., Djebbour Z., Sib J., KIeider J.P., Longeaud C, Glodt o.
Ecole Superieure d'Electricite, Gif-sur-Yvette, France

05.07 Absorption Optimization ofAmorphous Silicon Solar Cells 1087

Schropp RE.I., Meiling H., van Sark W.G.J., Stammeijer J., Bezemer J., van der Weg W.F.
Univ. Utrecht, The Netherlands

05.08 Explanation of the !J-;-Problem in Undoped a-Si:H 1091

KockaJ.
Inst. of Physics, Prague, Czechoslovakia
Abel CD., Nebel CE.
IPE, Univ. Stuttgart, F.R. Germany

05.09 Electric Field in a-Si Solar Cells Measured by Electron-Beam Testing 1095
Mayer A, Jank A, Schmoranzer H.
Univ. Kaiserslautern, F.R Germany

POSTER SESSION SA: PV in Developing Countries I

5A02 Tin Zaouatine - A Full Solar Energy Village in the Algerian Sahara 1101
Labed S., Yaici B.
Centre de Developpement des Energies Renouvelables, Algiers, Algeria

5A03 PV Electrification of a Village in a Hot-AridArea ofAlgeria 1105

Yaici B., Labed S.
Centre de Developpement des Energies Renouvelables, Algiers, Algeria
LorenzoE.
Inst. Energia Solar, UPM, Madrid, Spain

5A05 The 2000 km Algerian Desert Tracks Beaconed with Solar Energy 1108
Aillane A, Touti A
Centre de Developpement des Energies Renouvelables, Algiers, Algeria

5A08 Cost Analysis of Photovoltaic Water Pumping Systems Used in Arid and Semi-Arid Zones in
Algeria 1111
HamoudaC.
Univ. Batna, Algeria
Wagemann H.G., Hanitsch R, Siekmann H.E.
TU Berlin, F.R. Germany

5Al0 Algeria's Experience with Photovoltaic Module Production 1115

Mesbahi Y.
Haut Commissariat ala Recherche, Algiers, Algeria

5All Electro-Osmotic Subsoil Irrigation Using PV Electricity 1118

KamelF.M.
Suez Canal Univ., Port Said, Egypt
ShawkiE.
Nat. Res. Center, Cairo, Egypt

xxxviii
SA 13 Study of the Overall Efficiency of a 7 kWp Solar PV Pumping System at Mugombwa
(Rwanda) 1122
Mpawenayo P.
Inst. Recherche Scientifique et Technologique, Butare, Rwanda

SAl4 Solar Refrigeration for the Storage of Vaccines in the Expanded Programme on
Immunization in Developing Countries 1126
ZaffranM.
World Health Organization, Geneva, Switzerland

SA 16 PV Systems Development in Romania 1130

Tanasescu F.T., Olariu N., Popescu c.1.
Res. Inst. for Electrical Engineering, Bucharest, Romania

SAl8 Qualification Labels for Refrigerators Destined to the Developing Countries 1134
Michel E.
COSTlC, Digne, France

5Al9 Production and Experimentation ofAdapted Photovoltaic Refrigerator for Vaccine Storage 1137
MichelE.
COSTlC, Digne, France

POSTER SESSION 5B: PV in Developing Countries II

5B.OI Status and Experience of Solar PV Pumping in Developing Countries 1143

Barlow R, McNelis B., Derrick A
IT Power, Eversley, UK

5B.02 PV-Powered Data Collecting Platfonns for Data Acquisition in Remote Areas 1147
Reiniger K.D., Ebke W.
DLR, Oberpfaffenhofen, F.R Germany

5B.03 PV Pumping System Optimization: Tasks Perfonned in Laboratory and Field Tests 1151
BucherW.
German Aerospace Res. Establishment, Cologne, F.R Germany

5B.05 Status of Solar Energy Conversion and Applications in Yemen 1155

Bahaj AS.
Univ. Southampton, UK
Bin Gadhi S.M.
Univ. Aden, Yemen
Al-Ashwal A.
Univ. Sanaa, Yemen
Shamsan M.H.
Local Government Administration, Sanaa, Yemen

SB.07 Photovoltaic-Powered Water Pumping in Mali 1158

Cabraal R.A, Kennedy T.G., Hoelscher J.F.
Meridian Corp., Alexandria, VA, USA
Hart T., Sylla L., Dicko M.
IT Power, Bamako, Mali
AnnanRH.
US Dept. of Energy, Washington DC, USA

xxxix
SB.08 Present Status & Prospects of PVApplications in Western China 1162
Mao. Y., Wang A
Natural Energy Res. Inst., Gansu, P.R.China

SB.12 Photovoltaic-powered Television Relay Station 1165

Wiesner W., Pitzer H.L., Adrian M.
TOV Rheinland, Cologne, F.R. Germany

SB.14 Four Years of Photovoltaic Energy in Notto (Senegal) 1169

Montero Bartolome M.
Inst. Energfa Solar, UPM, Madrid, Spain

SB.1S Electrification of Rural Primary Schools by Photovoltaics in Tunisia 1173

Chaouch I., Ullerich D.
Nat. Energy Agency (AME), Tunis, Tunisia
NjaimiM.
CNMA, El Kef, Tunisia
Ben Mabrouk C.
ASSAD, Negrine, Tunisia

SESSION 06: Amorphous Silicon and Solar Cells

06.01 - Invited review lecture -

Amorphous Silicon Solar Cells 1179
WagnerS.
Univ. Princeton, NJ, USA
Carlson D .E.
Solarex, Newtown, PA, USA

06.02 Progress in Manufacturing at the Megawatt Level of a-Si Based PV Technology 1184
Ricaud AM., Schmitt J.P.M., Siefert J.M., Meot J.
Solems, Palaiseau, France
Roelen E., Bubenzer A, Kiimmerle W., Haussler W., Bottger M.
Phototronics Solartechnik, Putzbrunn, F.R. Germany

06.04 Recent Improvements in Very Large Area a-Si PV Module ManUfacturing 1189
Macneil J., Eser E., Kampas P., Xi J., Delahoy A, Ellis JT. P., Liu C.H.
APS, Princeton, NJ, USA

06.0S High Efficiency Double-Junction Tandem Solar Cells and Their Reliability Tests 1193
Yoshida T., Fujikake S., Fujisawa H., Saito S., Sasaki T., Ichikawa Y., Sakai H.
Fuji Electric Corp. Res. and Development, Kanagawa, Japan

06.06 Can We Improve the Stability ofAmorphous Silicon-Based Solar Cells Through the Use of
"More Stable" i-Layer Materials? 1197
von Roedern B.
SERI, Golden, CO, USA

SESSION 07: Component Performance and Development

07.01 - Invited review lecture -

Developments in Battery Storage for PV 1203
McCarthyS.
Hyperion Energy Systems, Cork, Ireland

xl
07.02 Use of the Modeling of Lead-Acid Battery Operation for the Development of a State of charge
Meter 1209
Mayer D., Biscaglia S.
Ecole de Mines de Paris, Centre d'Energetique, Valbonne, France

07.03 Perfonnance of Different Types of Battery Charge Regulators in Stand-Alone PV Systems 1214
Iliceto A., Previ A.
ENEL-CREL, Cologno Monzese, Italy
Guastella S., Pappalardo A.
Con phoebus, Piano D' Arci, Italy

07.04 Specification 503 - Implementation of PV Module Qualification Tests at ESTI 1219

Ossenbrink H., Rossi E., Bishop J.
Joint Res. Centre, Ispra, Italy

07.05 Comparative Assessment of Silicon Cell Pyranometers 1223

Grottke M., Imamura M.
WIP, Munich, F.R. Germany
Bechteler W., Mayer O.
Univ. der Bundeswehr Munich, Neubiberg, F.R. Germany

07.06 Holographic Soft Concentrators for PVApplications 1229

Anders G., Corlatan D., Herz K., Schmid D.
ZI:)W, Stuttgart, F.R. Germany

SESSION 08: Systems and Subsystems

08.01 - Invited review lecture -

Photovoltaic Power Conditioning/Inverter Technology 1235
Kleinkauf W.
ISET, Kassel, F.R. Germany

08.02 On-Site Power Measurements on Large PVArrays 1240

Blaesser G., Zaaiman W.
ESTI, Joint Res. Centre, Ispra, Italy

08.03 Detennination of Optimum Tilt Angle of Panels in Large Scale Photovoltaic Arrays 1244
Iliceto A., Previ A.
ENEL-CREL, Cologno Monzese, Italy
Chimento G., Guastella S.
Conphoebus, Piano D'Arci, Italy

08.04 Electrical Degradation of the Carrisa Plains Power Plant 1248

Schaefer J.
EPRI, Palo Alto, CA, USA
Rosenthal A.
Southwest Technology Development Inst., Las Cruces, NM, USA
Schlueter L.
Siemens Solar Industries, Camarillo, CA, USA
WengerH.
Pacific Gas and Electric Co, San Ramon, CA, USA

08.05 Experimental Comparison of Different Concepts for Photovoltaic Power Plants 1254
Knaupp W., Bleil A., Klotz F.H., Stellbogen D., Weeber V., Pfisterer F.
ZI:)W, Stuttgart, F.R. Germany

xli
08.06 Concerted Action on Computer Modelling and Simulation 1259
Keating L., McCarthy S., Wrixon G.T. ,
NMRC, Cork, Ireland
MayerD.
Ecole Des Mines, Sophia Antipolis, Valbonne, France

SESSION 09: PV Systems Technology

09.01 - Invited review lecture -

PVfor Bulk Power Applications: CostlPerfonnance Targets and Technology Prospects 1269
DeMeoE.A.
EPRI, Palo Alto, CA, USA

09.02 ENEL's 3-MW PV Power Station Preliminary Design 1277

Corvi c., Vigotti R.
ENEL-DSR, Rome, Italy
Iliceto A., Previ A.
ENEL-CRE1, Cologno Monzese, Italy

09.03 Operating Results of the 340 kWp Plant "Kobem-Gondorf'. Design and Constrnction of the
330 kWp Plant "Neurather See" 1281
Beyer U., Dietrich B., Hotopp R. Pottbrock R
RWE Energie, Essen, F.R Germany

09.04 Design and Development of a Standard 100 kW Photovoltaic System 1285

Barra L., Castello S., Messana C.
ENEA-FORI, Rome, Italy

09.05 Description and Performance ofAustin's Two 300-kW Photo voltaic Plants 1289
Hoffner J.E.
City of Austin Electric Utility Dept., Austin, TX, USA

SESSION 10: PV Systems Programmes

10.01 - Invited review lecture -

Grid-Connected Solar Houses 1295
Bloss W.H., Pfisterer F.
ZSW, Stuttgart, F.R Germany
Kleinkauf W., Landau M.
ISET, Kassel, F.R Germany
Weber H., Hullmann H.
Inst. fiiI Industrialisierung des Bauens, Hannover, F.R Germany

10.02 Results and Lessons Learned from the DG XVII PV Demonstration and THERMIE 1301
Programmes
KautW.
CEC, DG XVII, Brussels, Belgium
Blaesser G., Riesch G.
ESTI Lab., JRC, Ispra, Italy
Gillett W.B., Hacker RJ.
Halcrow Gilbert Associates, Swindon, UK

10.03 Project Megawatt: 333 Residential Grid-Connected PV houses Spread over Switzerland 1305
RealM.
Alpha Real, Zurich, Switzerland

xlii
10.04 Technical Requirements and Consultation Concept for Small Grid-Connected PV Plants within
the Gennan "1000-Roof PV Programme" 1308
HotoppR.
RWE Energie, Essen, F.R. Germany

10.05 The Influence of the Maintenance on the Operation: Lesson Learned by Two Years
Management of the Delphos Plant 1312
Sarno A
ENEA-CRIF, Rome, Italy
Cordisco S., Guerra M.
Area Sperimentale di Monte Aquilone, Manfredonia, Italy

10.06 The Full Cost of Electricity Generation - How Does PV Compare Today? 1316
HohmeyerO.
Fraunhofer-Inst. fUr Systemtechnik und Innovationsforschung, Karlsruhe, F.R. Germany

10.07 DetailedAnalysis of the Fotavoltaic Project after 7 Years of Operation 1321

O'Riordan A, Rudden M., McCarthy S., Wrixon G.T.
NMRC, Cork, Ireland

10.08 The Development of Photovoltaic Power Generation System under the Sunshine Project of
Japan 1326
Morishita H., Agawa T., Morishige T., Miyazato M.
NEDO, Tokyo, Japan
Hashimoto E., Takigawa K., Kobayashi K.
CRIEPI, Tokyo, Japan

SESSION 11: PV Technology for Successful Deployment in Developing Countries - The Lessons of the past
20 Years

No paper for this session.

SESSION 12: The Successful Introduction of PV Technologies into the Communities in Developing Countries

12.01 Photovoltaics Manufacturing in Developing Countries 1335

Darkazalli G., Hogan S.
Spire Corp., Bedford, MA, USA

12.02 Export Promotion Support of Photovoltaic Technology Manufactured in Developing Countries 1337
SeminN.
Internat. Trade Centre, Geneva, Switzerland
Derrick A
IT Power, Eversley, UK

12.03 Integration of Solar Electrification: Pacific Islands Example 1339

Jourde P.
GENEC, CEN Cadarache, France

12.04 PV Rural Electrification in the Bolivian High Plateau 1342

Aguilera J., Lorenzo E.
Inst. Energia Solar, UPM, Madrid, Spain

xliii
12.05 Sale of Excess Solar Energy: Contribution to the Recurrent Costs of Immunization
Programmes? 1344
DurandJ.M.
IT Power Central Africa, Kinshasa, Zaire
ZaffranM.
World Health Organization, Geneva, Switzerland

12.06 Household Photovoltaic Systems in Developing Countries: Experience, Issues, and Outlook 1348
Crosetti M., Durand P.
The World Bank, Washington, DC, USA

SESSION 13: Transforming International Aid into PV Services to the Communities

13.01 - Invited -
From Aid Projects to the Market Introduction of PV Systems - Experiences and Strategies 1353
Posorski R.
GTZ, Eschborn, F.R. Germany

13.02 - Invited -
Experiences with Village PV Systems in Pakistan 1356
Lovejoy D.R.
United Nations, New York, NY, USA

SESSION 14: Organisation and Training Needs for the Development of Endogenous Capabilities in
Photovoltaics

14.01 Organization and Training Needed for the Development of Endogenous Capabilities in
Photovoltaics 1361
HillR.
Newcastle Polytechnic, UK
Guimariies L.
New Univ. Lisbon, Portugal
LorenzoE.
Inst. Energia Solar, UPM, Madrid, Spain

National Programmes

NPI - Invited -
The European Community R&D Programme on Photovoltaics 1369
PalzW.
CEC, Brussels, Belgium

NP2 -Invited-
U.SA. Photovoltaics Program Plan FY 1991-FY 1995 1371
Annan R.H., Rannels J.E.
US Dept. of Energy, Washington, DC, USA

NP3 - Invited -
Sunshine Project and Recent Progress in Photovoltaic Technology in Japan 1375
Hamakawa Y.
Univ. Osaka, Japan

xliv
NP4 - Invited -
Photovoltaic R&D in the Federal Republic of Germany 1381
Wollin K.
BMFf, Bonn, F.R. Germany
BatschJ.
Res. Centre Jiilich, F.R. Germany

NP5 - Invited -
The Italian Photovoltaic National Programme 1385
Farinelli U., Ambrosini G.
ENEA, Rome, Italy
Vigotti R.
ENEL, Rome, Italy

NP6 - Invited -
The Netherlands Photovoltaic National Programme 1390
Muradin-Szweykowska M., ter Horst E.W.
NOVEM, Utrecht, The Netherlands

NP7 - Invited -
Portuguese National Effort in the Area of Photovoltaics 1394
Collares Pereira M.
Center for the Conservation of Energy, Amadora, Portugal

NP8 - Invited -
International Photovoltaic Markets, Developments and Trends 1396
Maycock P.D.
PV Energy Systems, Casanova, VA, USA

NP9 The Stmcture of Solar R&D - Is it adequate? 1401

Einsenbeiss G., Hertlein H.
Forschungsbund Sonnenenergie, DLR, Cologne, F.R. Germany

Late News

LNl The Shockley-Queisser Argument as the Limiting Case of a More Accurate Theory 1407
Landsberg P.T.
Univ. Southampton, UK
De VosA.
State Univ. Gent, Belgium
BaruchP.
Univ. Paris 7, France

LN2 Tino - A Solar Car for Daily Use 1409

Baumgartner F.P., Simon M., Burkhardt R.
Auto-Sol, Konstanz, F.R. Germany

LN3 P-type Doping in Heteroepitaxial CdTe and Lift-off Technique for Making Thin Film Single
Crystal CdTe Solar Cells 1411
Tiwari A.N., Blunier S., Kessler K., Zogg H.
AFIF, Swiss Fed. lost. of Technology, Zurich, Switzerland

LN4 17le Photovoltaic Megawatt Project Berlin 1413

Reismayr D.
Ludwig-Bolkow-Systemtechnik, Ottobrunn, F.R. Germany

xlv
LN5 High-Efficiency ZnO-CdS-Cu!nSe2 Solar Cells - Recent Results of Eurocis Collaboration 1415
Mauch R.H., Ruckh M., Kessler J., Klinger R., Schock H.W.
IPE, Univ. Stuttgart, F.R. Germany
Hedstrom J., Stolt L.
SIM, Kista, Sweden
Lincot D., Vedel J.
ENSCP, Paris, France

LN6 GaAs/Si Monolithic 2-Tenninal Tandem Solar Cell Grown by MOCVD 1418
Umeno M., Shimizu H., Egawa T., Soga T., Jimbo T.
Inst. of Technology, Nagoya, Japan

LN7 Slicing, Cleaning and Etching thin 30 x 30 cm Wafers of Multicrystalline Silicon 1420
Lievens P.P., de Villers Grandchamps T., Smekens E.A., Smekens G.R.
ENE, Brussels, Belgium

Closing Session

Closing Address 1423

Dr. Nuno Ribeiro da Silva
Secretary of State for Energy, Lisbon, Portugal

Conference Summary 1427

Alphabetical List of Authors 1429

xlvi
Prof. A. Luque, general chairman of the Conference

The launch of this outstanding international event

 us ra
•

The Minister for Industry and Energy of the host country

Prof. P. Bourdeau, CEC, awarding the Becquerel Prize to Prof. Werner Bloss
One of several press conferences during the event

Very animated poster session

 ACKNOWLEDGEMENT

It is a particular concern to the Commission of the European Communities and to the general
organiser to express thanks for the extraordinary support offered by

Eng. Luis Mira Amaral

Minister for Industry and Energy

Through the support and participation of the Ministry for Industry and Energy an outstanding
environment was found in Lisbon in both professional and human relation respect.

Eng. Luis Mira Amaral, Minister for Industry and Energy during the opening session.
Opening Session
 MINISTRY OF INDUSTRY AND ENERGY
OFFICE OF THE SECRETARY OF STATE FOR ENERGY

SPEECH BY HIS EXCELLENCY THE SECRETARY OF STATE FOR ENERGY

Dr. NUNO RIBEIRO da SILVA

OFFICIAL OPENING CEREMONY OF THE 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY

CONFERENCE

It is particulary gratifying for me to be making the final speech in this official opening ceremony of the 10th
European Photovoltaic Solar Energy Conference, which was organized by the Commission of the European
Communities and was wholeheartedly welcomed from the start by the Ministry of Industry and Energy.

The unequivocal support by Portugal for the holding of this international forum should be seen in the
context of the philosophy and strategy of the Portuguese government when it took the clear decision to back the
development and promotion of so-called renewable energy sources, such as biomass, small hydroelectric plants,
and thermal and photovoltaic solar energy.

Whilst not relegating any of the stages of scientific and technological progress to second place, be they
research into, development or demonstration of energy technologies, the government, through the Secretariat of
State for Energy, has pursued a one-speed national energy policy, where 'traditional' and 'renewable' energies have
co-existed side-by-side.

It was in this broad sense that, through supporting this conference, it was hoped that scientific and technical
specialists, as well as public and private bodies - coming from the most diverse countries, but all sharing the same
fundamental concerns about energy - would be able to impart their knowledge and explain their points of view.

In the present post-Gulf War scenario, the Portuguese government does not intend to boost its own use of
renewable sources of energy any further, in the light of considerations of economic viability and its philosophy of
energy diversification. Nevertheless, it should be a global concern to get the maximum use out of the natural and
locally available resources of each country.

For this reason, the efforts which the European Community has been making in this sphere should be
underlined, especially since these efforts have been directed at a very early stage in the production process, that of
research. This is important, given that the period of time necded, to determine the scientific and technical viability
of projects is naturally enormous, and a more or less long-term perspective is almost always required.

Portugal has also not refrained from asserting its rights on the various boards and decision centres where it
is represented, on the Councils of European Ministers, in the European Parliament and the International Energy
Agency, at the World Energy Conference or even at other international or Community forums.

I am sure that this conference will facilitate the exchange of knowledge about photovoltaic solar energy, and
that this exchange will lead in the medium term to the development and optimisation of technologies capablc of
turning local, regional, Community and international projects into economically viable propositions.

The sectors with the greatest potential for using photovoltaic solar energy are vast, including the aerospace
industry, telecommunications, airports, road and rail transport, energy and water, agriculture, forestry and tourism.

In Portugal, a number of extremely interesting projects have been supported and financed by the
Portuguese government and by the Commission of the European Communities through the various official
programmes and incentive schemes, to wit STURE (the Specific Programme for the Development of Portuguese
Industry), the Valoren Programme, the Demonstration Project Programme and the Thermic Programme, the Joule
Programme and the Sprint Programme.

3
 I would like to mention some of the most important projects:

- the New University's photovoltaic power plant, which is located in the Technology Park and is to be
inaugurated tomorrow;
- the provision of electricity to rural communities in the Algarve;
- water pumping systems in the Vale da Rosa;
- small photovoltaic units installed in houses, schools, social centres, protected areas, reserves and rural
areas;
- the energy-generating plant on Madeira;
- the provision of electricity to Berlenga Island;
- the building of a factory for making photovoltaic cells.

Here I would like to highlight the work carried out at a number of Portuguese research centres and by
Portuguese scientists and researchers who have devoted their attention to this field in a most commendable way
and who, by their knowledge and intelligence, have contributed to the success of important projects in both
fundamental and applied research.

My hope is that one day this commendable work will also be recognized internationally through the
awarding of a "Nobel Prize for Photovoltaics", in the same way that just a moment ago we were privileged to
witness the awarding of the Edmond Becquerel Prize to the winner, the scientist and professor Werner Bloss.

Such a prize would represent recognition of the work being carried out by everyone who is striving to bring
about a quality of life and a state of social well-bring which is compatible with the future needs of the new social
and economic orders which will be responsible for the balanced functioning of energy systems.

Finally, let me finish by saying that Portugal, as a member state of the European Community, will do
everything in conjunction with the other member states to ensure that greater funds are made available for
supporting applied research projects, demonstration projects and projects for disseminating energy technologies,
particularly those connected with renewable energy sources, and thus also for measures to promote technology
transfer, both within and beyond the Community.

4
 Speech given on the occasion of the Mouchot Prize Award
at the 10th European Photovoltaic Solar Energy Conference
in Lisbon, 12 April 1991

by w. Palz
Commission of the European Communities

The time of Mouchot, the 1860s and 70s, was particularly important for present-day
society. It was a time when man first became conscious of the importance of energy. It
was a time when man started to realise what industrial energy is, to appreciate its
potential, and at the same time recognise the limitations of the fossil energy resources
available on Earth.

It is interesting to note that, simultaneously with the start of oil consumption, there arose
a keen interest in solar energy. Mouchot's solar experiments, shown for instance at the
universal exhibition in Paris in 1879, attracted enormous public interest. Other important
technologists of that time, such as Ericsson, the American inventor of the ship's screw,
developed solar devices. Werner von Siemens, the founder of an industrial empire,
worked on photovoltaic cells in 1873 and was the first to recognise their importance as
a potential energy source.

Mouchot's lifetime was the start of a particularly fruitful period of new social ideas and
technical innovations. It is a striking fact that the new technical ideas were much more
successful than the new social concepts of that time. Nationalism and communism had
their roots in that period and they eventually led the world to disaster. A social concept
which proved more durable is a much older one, namely the American revolution of a
century earlier providing the world of today with the "pax americana". An interesting
element in this social concept is the high rank which American society accords to the
individual.

As regards the new technical ideas, they were manyfold. Most of the technical
innovations from which we benefit today, such as the automobile, the telephone, radio
and TV to name just a few originated from Mouchot's time onwards. Later on came
propulsion in space and the idea to go to the moon, and finally nuclear energy emerged.
By the time of World War II, most of these technologies, though not all, had matured
but had not been used on a very significant scale. Implementation came after World War
II. The tremendous increase of Gross National Products and individual wealth which
occurred in many countries since the last war was associated with the mass production
and mass consumption of technical goods which had been invented during the very
fruitful period from the 1860's to the 1940's.

In some ways, our generation seems to be less imaginative and less fertile in new ideas
than our fathers and grandfathers; on the other hand, there is no doubt that we are very
strong in comprehensive implementation. Informatics, one of the important new
developments of our generation, came at the right time to help us manage a complex
cycle of mass production, mass consumption and finally mass production of waste which
is characteristic of our society.

5
There is of course the temptation to apply modern technologies "ad absurdum". For
example, it cannot make sense to attempt to draw our energy supplies exclusively from
nuclear fission by building reactors all over the world, or to build up tremendous space
programmes whose usefulness is most questionable. Already at the beginning of the
1970s, the realisation drawned that progress must proceed in harmony with the natural
limitations of the Earth and its resources. Nevertheless the race went on; mass
production and consumption have brought us a world whose environment is seriously
disturbed and with a widening gap between North and South in terms of wealth and
progress.

What we need again today are new ideas and visions like those of Mouchot and the
people of his time.

It seems essential to recognise the importance of solar energy for us and for future
generations. What would the solar age, as proposed by EUROSOLAR, imply for society
and individuals ? It need not be stressed that the principles of society as it exists today,
the role of the "res publica" to develop wealth, protect the poor and save the natural
environment would be unchanged by solar energy, whilst personal opportunities and
individual responsibility would be strengthened. Solar energy does not mean that we must
do away with current economic concepts and mass production and go back to hand-made
energy systems. Photovoltaics offers a good demonstration that renewable energy devices
should not be produced in a "do-it-yourself' way: nobody could imagine PV modules
being produced in a backyard workshop. Industrial companies, large and small, are the
key to the future of renewable energies.

However, there is no reason why in future, electricity should not be produced more and
more at the site of consumption. This is particularly true for photovoltaics which lends
itself very well to integration in large buildings and family dwellings. In comparison, wind
energy, the other important renewable electricity source, is more suited to concentrated
implementation in parks outside the cities. Heat is currently produced locally in factories,
large buildings and houses. So, why should people not be able to manage their individual
power systems, just as they learned to operate a much more complex decentralised power
system in their cars ? The total capacity installed in 110 million automobiles in the
European Community is more than 10 times the power capacity of all central electricity
plants.

It would be technically easy to add an engine and a generator to the existing heating
systems in buildings. Decentralised electricity production in small "combined heat and
power" systems of this type would work with a very high overall efficiency, hence they are
economically and environmentally desirable and, furthermore, they would stimulate
individual responsibility. In Europe, roof- and facade- integrated photovoltaic generators
could complement in an ideal way these combined heat and power systems because in
the European climate solar electricity is produced mostly in the summer when the heating
load is low. In such a scheme, solar electricity would find the widespread and
de centralised application it deserves.

In conclusion, I firmly believe that solar energy as we understand it today will fit perfectly
well with the wealth of the individual and progress of the overall economic systems. The
vision of widespread utilisation of natural energy is an old one which has emerged again
strongly and is widely supported by the general public. The paradox is that this scheme

6
still encounters little recognition from the established energy system, although it offers
promise to improve the overall energy situation to the benefit of our economy. Let us
therefore share the vision of the first solar pioneers and revive their ideas. They were
never more up to date than now.

I would like to thank Eurosolar for having honoured me with the Mouchot Prize and I
thank in particular its President, Hermann Scheer, for whom I have the highest esteem
because I recognise in him one of these visionaries who have become so rare today.

7
 SESSION 01

Fundamentals, High Efficiency Cells

and Novel Devices
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EFFICIENCY OF A COMBINED SOLAR CONCENTRATOR CELL AND THERMAL

POWER ENGINE SYSTEM

A. Goetzberger, W. Bronner, W. Wettling

Fraunhofer-Institut fUr Solare Energiesysteme, Oltmannsstrasse 22, D-7800 Freiburg FRG
Phone: -49-761-4014-0; FAX: -49-761-4014-100

ABSTRACT. The combination of a concentrator solar cell and a thermal power engine is
investigated theoretically. The excess heat of the PV concentrator cell can be used to drive a
thermal power engine such that the thermal engine acts as a cooling device for the solar cell.
The combined efficiency of the solar cell and the thermal engine is calculated as a function of
cell band-gap energy and temperature. The individual efficiency of the thermal engine was
assumed to be one half of that of a Carnot engine.
The maximum of the combined efficiency lies between 35 and 40 % for concentration
factors between 100 and 1000. It can be increased to values above 40 % when a tandem
configuration for the solar cells is used.
By introducing the 'double tandem' arrangement, in which the temperature of the cooling
medium is raised in stages by first cooling the cell with lowest band-gap and then the higher
gap cells, even higher efficiencies of values between 45 and 50 % can be achieved theoretically.

1. INTRODUCTION concentrator cells have reached, on a laboratory scale,

values close to 30 % AM1.5 and might exceed these
High efficiency concentrator solar cells have a values in the not too distant future.
large potential for application in future power (2) New solar cell systems with various band-gaps have
generating systems because the costs of energy been investigated in more detail, as e.g. AI"Gal_xAs,
production are drastically reduced due to improved InxGal_xP or InxGal_xAsyPl_X which can be 'tailored"
module efficiency and increased concentration factor II, in wide band-gap ranges. With these cells numerous
21. The efficiency can be further increased if a tandem solar cell combinations have been developed on
tandem or multijunction configuration is used. With laboratory scale during the past years with promising
high-efficiency tandem concentrator cells efficiencies results for future developments. The idea of a
close to 40 % can be achieved theoretically provided the combination of tandem solar cells and solar thermal
cell temperature can be held at 300 K. Efficiencies power engines has, however, not been discussed.
above 35 % have already been realized 131. The The present paper addresses the efficiency of a
remainder of the solar energy cannot, however, be combined system of concentrator solar cells (single
converted directly to electricity but is dissipated as junction, double and triple junction) and thermal power
thermal energy. engine. The thermal engine extracts heat from the solar
An alternative concept of solar energy conversion is cells thus acting as an active cooling device for the
the thermal power engine which uses the solar heat to concentrator cells. It will be shown that the combined
drive electric generators by steam turbines, Sterling or efficiency of conversion of solar energy to electricity
Rankine engines. is distinctly higher than with solar cells alone. Using
In the present paper a combined use of photovoltaic realistic values for solar cell and thermal conversion
and thermal power conversion of solar energy is efficiencies, a two cell tandem configuration can reach
considered theoretically. The idea to use the excess efficiencies above 40 %. It will be further shown in
heat of a photovoitaic concentrator array to drive a this paper that with the new 'double tandem' principle
thermal engine is not new. A number of papers and still higher efficiencies can be realized.
reports treated this problem at the end of the seventies
and the beginning of the eighties with respect to space
and I or terrestrial applications 14-121. Several 2. THEORY
advantages are associated with this concept:
Because it is not possible to operate the photovoltaic 2.1 Theoretical model
device under high concentration at ambient temperature, In this section a simple model is presented to
the theoretical efficiencies cannot be realized in calculate the combined efficiencies of a photovoltaic
practice. Converting, in addition, the thermal energy into (PV) and thermal engine (TE) power generator. We
electricity overcomes this difficulty. assume a PV concentrator cell or module in perfect
The lowest band-gap cell in a tandem structure thermal contact to a thermal engine by either a direct
shows the strongest dependence of efficiency on bond to the hot side of the engine or by contact to a
temperature. One can consider the combined photovoltaic liquid heat exchange system that transports the excess
and thermal cycle as a set-up in which this lower gap heat of the PV system to the TE. In this section the
cell is replaced by a Carnot engine. temperature of the TE and the PV are assumed to be
As will be shown in this paper, the optimal the same.
operating temperature is in an intermediate range For the PV system a one-diode model is used. This
around SOO K. Therefore materials problems for the model will give an upper limit of efficiencies that can
heat engine are relaxed. be reached only if the technology of solar cells is
Although to our knowledge no experimental developed to a state that is not yet possible on
verification of this principle has been published, it production scale but that may be achieved in not too far
seems worthwhile to reconsider this idea for the a future with some semiconductor materials, as e.g.
following reasons: with Si and GaAs.
(1) The efficiencies of solar cells have steadily been Reflection losses and Ohmic losses are neglected.
improved during the last decade; GaAs and Si The band-gap energy, Eg , of the PV system is assumed

II
for the moment to be a variable parameter. concentrations C = 100 (dotted) and C = WOO (solid
Furthermore the temperature T [K] and the light lines) and for various temperatures between 300 and
concentration factor C are used as free parameters. 800 K. These curves show that 'I7(PV) decreases
For a "real" system there exists, of course, a drastically with increasing T, in particular for sma1l
dependence of T on C but this can only be calculated if band-gap energies as is expected. The numerical values
the thermodynamic properties of the system are known. must be taken with some precaution, since each
For a realistic set-up it can be assumed that a well individual semiconductor has a band-gap that is
defined temperature of the PV device can be estahlished temperature dependent, e.g. for GaAs one has 1161
by choosing a concentration factor, e.g. between C= 100 (9) Eg(GaAs) = 1.519 - (5.4 10-4 T2)1 (T+204)
and C= 1000. One must therefore use the E -values that are
The following four equations describe the appropriate for a given temperature. 'rhe T -dependence
characteristics of the solar cells: of EJ! was used in calculations of Wysocki et a1. for
current density: C =I 114/.
(1) 1 = 10 exp(qV/kT) - 1L The efficiency of a thermal engine depends very
dark saturation current density: much on the system that is used. As a rough estimate,
(2) 10 = q ni2 (OeAe/Le NA + 0hAh/Lh NO) however, one may assume an efficiency amounting to
one half of that of a Carnot engine which works at
intrinsic carrier concentration: temperature T against To = 300 K. Thus
(3) lli2 = Nc Nv exp (-Eg/kT) (10) 'I7(TE) = 0.5 (1- 300 IT)
density of states in conduction and valence band: is a realistic figure for thermal engines of the
(4) Nc = 2 (2mn/ kT/h2 )3/2 Rankine, the Sterling or the Brayton type 117/.
Nv = 2 (21Tmh* kT Ih2 ) 3/2 The combined efficiency of a PV and TE system is
Here, 0 and L are diffusion constant and diffusion given by
length and A is a geometry factor with e and h (11) '17 = 'l7c(PV) + '17 (TE) (1 - 'l7c(PV»
referring to electrons and holes, respectively, NA and 40 ,-------------------------r=====~
NO are the numbers of acceptors and donors,
respectively.
Inserting (3) and (4) into (2) and collecting all T-
dependent terms o~e finds 30
(5) 10 = K T exp (-Eg/kT)
The factor K contains all T -independent terms. If
one choses K = 0.2 one finds for Si and GaAs values E
of 10 that are close to measured values: g 20
.!
(6) 10(SO = 2 10- 12 [A/cm2] ~
10(GaAs) = 8.5 10- 18 [A/cm2] i:;

The light current density 1L for the AMl.5 solar

10
spectrum can be approximated to very good accuracy by
(7) 1L = C 8.46 10-3 (3.38- Eg)1.96 [Eg in eV]
The efficiency of the solar cell is given by
(8) 'I7(PV) = Pm/Pss O+--------.--------.-------~r-------~
where Pm is the power at maximum power point and M W ~ M ~

P ss is the integrated power ~ensity of the AM1.5 solar E_ rev]

spectrum, P ss = 98.4 mWIcm . Figure 1: Calculated 1 sun solar cell efficiencies
In order to model a weaker dependence of 'I7(PV) at T = 300 K as a function of semiconductor band-gap
on C a heuristic correction factor is used for 'I7(PV): energies E; comparison of different models: Green:
(8a) 'l7c(PV) = 'I7(PV) (1- 0.Q25 10g(C» AMfs-effi!iency 1131; Queisser: Shoc~ley and QU~I~ser:
It accounts for losses at larger C which may be detailed balance limit 1151; WysockI: AMO-efflclency
due to high-injection effects andlor not optimized series lUI; ISE: this paper
resistances. 40
~.::.~.--
Using (1), (5), and (7) to calculate Pm and inserting ~
this into (8), resp. (Ba), 'l7c(PV) as a function of Eg
and T is calculated. 30
It should be emphasized at this point that the use of
K-values independent of specific semiconductor
materials is a great simplification and not justified by :it
theory. A more detailed calculation has to insert values >-
of K that are specific for individual semiconductors. .,g 20
Our method gives, however, reasonably good values :;;;
t:S
compared to other calculations in the literature. This is
demonstrated in figure 1 where 'l7c(PV)=f(Eg )
according to our model is compared to models by Green 10
(calculated for AM1.5) 1131, Wysocki et a1. (for AMO)
1141 and by Shockley and Queisser (detailed balance
limit) /151. Our calculated efficiency values for Si
(1.1 eV) and GaAs (1.42 eV) are close to experimental
values that have been achieved for solar cells on w ~ M U
E... [ev]
laboratory scale. At large band-gaps our values may be
too high, but since no large band-gap material has been Figure 2: Calculated AMI.5 efficiencies as a
developed to show efficiencies close to the theoretical function of Ed. for concentration factors C=100 and C=lOOO
limit the decision which curve 'I7(Eg) is the best cannot and for temperatures between 300 and 800 K
be made to date.
In figure 2 efficiencies 'l7c(PV)=f(Eg ) as calculated
according to the above model are presented for

12
 60 60 ~----------------------------------~
C ~..Ipq .. c." IC!Q ..
C = 1000 C" 1000
50 50

..
(PV+Ccmol!2)-[(f;dency

40 (pv+Ccm<>I/2)-[fnc;oncy 40
'7
.!....
g 30 g
.S'
.!!
30
~ ~
W W
20 20

10 10

O~--.--,,--.---.--.---.---.--,---.--~
o ~--~--~--~--~--~--~--~--~--~--
500 350 '00 450 500 550 600 650 700 750 800 300 350 .00 450 500 550 600 650 700 750 800
Te"'4"!",lure [K) Temperoture [K]
Figure 3: Calculated AMl.5 efficiencies for
concentrator solar cells as a function of T. The cells are Figure 4: Calculated AMl.5 efficiencies of a
assumed to be optimized with respect to Eg for every T. concentrator two·cell configuration as a function of T.
For C=lOOO ED varies linearly with T from !-1 e~ at 300 The cells are assumed to be optimized with respect to Eg
K to 1.9 eV ar800 K. Also shown are the effICIenCIes c;f a for every T. For C=lOOO Ep) varies linearly with T from
TE (Carnotl2) according to equ.(JO) and the combtned 1.05 eV (300 K) to 1.65 eV (800 K) and ED2 from 1.85 eV
PV+TE efficiencies according to equ.(J1) (300 K) to 2.25 eV (800 K). Also cs.hown are the
efficiencies of a TE (CarnotI2) accordtng to equ.(JO)
and the combined PV+TE efficiencies according to
2.2 Results of Modelling equ.(J1)
Figure 3 shows the evaluation of (8a), (10) and 60.-------------------------~------_,

(11) as a function of T. Here it was assumed that for

every T a solar cell with optimal band-gap is available.
It is seen that the decrease in PV-efficiency at large T 50
is overcompensated by the TE. Values of the combined
efficiency between 35 and 40 % for values of C 40
(GaAs+Cornotj2)-Efficiency
between 100 and 1000 and optimized band energies g:
'I.... = 35.32 l';
between 1.5 and 2 eV are obtained. It is obvious that >-
u
for larger T a higher Eg would be optimal. I: 30
Our model can be mOdified to include a tandem cell "
~
GoAs-Efficiency
CarnoV2-Efficiency
configuration in the PV system. If one assumes that the
wide gap cell 1 absorbs light with energy > Eg1 without 20
losses one can use equ.(l1) by just inserting
(12) '17c(PV) = '17c(PV)l + '17c(PV)2 10
One thus finds for a tandem configuration
efficiencies as shown in figure 4 with optimal cell
band-gaps EgI(T) and Eg2(T). In this case the TE O~-,---'--r--'--'---'--'--'-~~~
approximately compensates the temperature losses of 300 350 400 450 500 550 600 650 700 750 800
the PV system leading to an almost T - independent total Temperature [K]
efficiency above 40 % AM1.5. Figure 5: CalCUlated efficiencies of a GaAs cell
In a realistic system one has to choose available under concentrated light as a function of T. Also shown is
semiconductors even if their value of E,g is not optimal the efficiency of the TE and of the combined PV+TE
for air temperatures. As an example tor one specific system
semiconductor the combined PV and TE performance
using an GaAs cell is shown in figure 5. This figure
demonstrates that with GaAs cells efficiencies between Reference is made to the spectrum splitting approach
33 and 35% for temperatures of about 500 K are /18/ in which parts of the solar spectrum are directed
possible. to the appropriate solar cells by spectrally sensitive
mirrors.
2.3 The Double Tandem Principle In figure 6 it is shown how the heat sink of the
The results presented above demonstrate the low gap cell Cl is used to heat the transfer medium
interdependence of operating temperature and optimal fr.om TL ~o an intermediate tempe~ature T 1, At cell C2
energy gap of the solar cell(s). The higher the wlth the hlgher gap Eg2 >Egl' WhlCh tolerates a higher
temperature the higher the band-gap of the cell must be temperature, the output temperature T2 is reached.
(see figure 2). This is expected because the Subsequently the heat transfer medium is heated by 13
temperature dependence of ni. decreases with increasing to temperature TH and then transferred to the TK
gap. It is now proposed to eliminate this correlation by Thus it is possible to operate low gap cells at reduced
applying a new principle which we call the 'double. temperatures and still attain the sum of heat energy at
tandem' principle: The tandem principle is not only the highest output temperature TH.
applied to the PV-part but also to the thermal With this approach the highest overall conversion
conversion. A Carnot engine operates between two efficiencies from sunlight to electricity are possible.
temperatures TL and TH (L and H stand for low and As is evident from figure 7, a two cell, two
high, respectiveTy). It is possible to heat up the heat temperature arrangement with achievable solar cell
transfer medium in separate stages from TL to TH parameters and half the Carnot efficiency yields a total
such that stages with lower T are associated with low efficiency between 43 and 45%. Even higher efficiencies
band-gap solar cells and vice versa. Different are possible with the three cell, three temperature
embodiments of this principle are possible, and only one configuration that was shown schematically in figure 6.
of them can be described here to illustrate the concept. In this case our theoretical model indicates an

13
efficiency close to 50% (see figure 8). If it is further 3. CONCLUSIONS
assumed that the solar cells are not at the final
temperature, e.g. at T 1 for stage 1 but at the average An improved model for the estimation of the
temperature (Tt+TL)j2 the uppermost curve in fig. 8 performance of a combined photovoltaic plus thermal
(calculated for C=1000) is obtained. Arrangements other conversion has been developed. It permits to evaluate
than the spectrum splitting approach for the double ultimate efficiencies of solar energy conversion to
tandem principle are currently under investigation and electricity assuming ideal parameters but also to
will be published at a later date. estimate them for realistic conditions. A new concept,
the double-tandem principle, has been proposed. It
permits separate optimization of PV and TE conversion
and predicts highest conversion efficiencies under
realistic conditions. A double tandem cell arrangement
can yield a total efficiency of 45 % which rises to
almost 50 % for a triple solar cell configuration.

REFERENCES
111 J .C.C. Fan "Promises of III/V solar cells", Tech.
Digest of the Int. PVSEC-5, Kyoto 1990, p. (fJ7
121 J .M. Gee "Status of concentrator collector and
high-efficiency concentrator cell development",
Tech. Digest of the Int. PVSEC-5, Kyoto 1990, p.
Figure 6: Schematic layout of the 'double .tande!f1' 95
principle. shown here for a three solar cell conftguratIon 131 L.M. Fraas, J.E. Avery, P.E. Gruenbaum, R.J.
(see text) Ballantyne, W.E. Daniels "Over 30 % efficient
60 .----------------------------------, GaAs cell assemblies for use with concentrated
C. =100
sunlight", Tech. Digest of the Int. PVSEC-5, Kyoto
C = 1000 1990, p. 89
141 D.F. Gluck "A combined Qhotovoltaic/thermal
50 (PV+Com o!/2)-(W";ency electric central power plant", Proc. 14. Intersoc.
12_· 4'.~J:.
Energy Convers. Eng. Conf., Boston, Mass. 1979, p.
...... -_....................... .............................. . 255
40 151 R.C. Neville "Study of combined (photovoltaic-
~ thermic) solar energy systems", Proc. 2nd Miami
g 30
Intern. Conf. on alt. enery sources, in Altern. En.
Sources 3, 1267 (1980)
;g" 161 S.M. Moite, J .A. Castle, R.L. Lessley, J .M. Kallis,
W E. Aerni "Simulation of a hybrid solar photovoltaicl
20 solar thermal electric power system", Proc. 14th
IEEE-PVSC ,(1980), p.1036
l.
171 .A. Castle, S.M. Moite, E. Aerni, J .M. Kallis
10 'Performance evaluation of conceptual hybrid solar
photovoltaicl solar thermal electric power systems",
Proc. 14th IEEE-PVSC, (1980), p_ 1106
0 4L--r--.r-~------.---.---r---------~ 181 "Conceptual _design and system analysis study for a
300 350 400 450 500 550 600 650 700 750 800 hybrid solar photovoltaicl solar thermal electric
Temp.rohJ", [KJ power sytem", Report Sandia National Labs., 1981
Figure 7: Calculated AMi.S efficiencies of a two- 191 A. Bar-Lev, S. Waks, G_ Grossman "Analysis of a
cells. two-temperatures 'double tandem' arrangement combined thermal-photovoltaic solar system based on
with optimal band-gap parameters the spherical reflector Itracking absorber
concentrator", Proc. ISES- Solar World Congress
60 J (PV+Corn0t/2)-(froelency ,_ • mo. I
1983, Pergamon Press, p. 1783
/lO/T.J. Lambarski "Electrical design guidelines for
photovoltaic/thermal systems", Proc. 17th IEEE-
50 - - -- ~ ~~~~---- - --­ PVSC, (1984), p. 1045
~-
....................
111/0. Chubb "Performance Characteristics of a
40 combined solar photovoltaic heat engine energy
converter", Proc. 22. Intersociety Energy Conv.
E,.. Engineer. Conf. (1987), p. 1
~ 30 1121 A. Goetzberger, W. Wettling "Kombinierter
;g'" Wirkungsgrad von
Wll.rmekraftmaschine",
PV -Konzentratorzelle
Proc. 7.
und
Intern_
W
20 Sonnenforum, Frankfurt, 1990, VoU, p. 1335
113/M.A. Green "Solar Cells", Prentice Hall Inc.
Englewood, NY, 1982
10 C = I~ .•. 1141 J .J. Wysocki, P. Rappaport "Effect of temperature
C = 1000 on Qhotovoltaic solar energy conversion", J.
App!.Phys. 31, 571 (1960)
o ~-,------,--,---r--,--,---r--,-~ 115/W. Shockley, H.J. Queisser "Detailed balance limit
300 350 400 450 500 550 600 650 700 750 800 of efficiency of p-n junction solar cells", J. App!.
Temperolure lK) Phys. 32, 510 (1%1)
Figure 8: Calculated AM 1.5 efficiencies of a f16/S.M. Sze "Physics of semiconductor devices", John
three-cells. three-temperatures . double tandem' Wiley and Sons, New York, 2nd ed. 1981, p. 15
configuration with optimal band-gap parameters f17lS. Kaneff "On the practical and potential viability
of paraboloidal dish solar thermal electric power
systems", Proc. ISES- Solar World Congress 1983,
Pergamon Press, p. 2767
118/N.S. Alvi, C.E. Backus, G.W. Madsen "The
potential for increasing the efficiency of
photovoltaic systems by using multiple cell
concepts", Proc_ 12th IEEE-PVSC, 1976, p. 948

14
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

HOLOGRAPHIC OPTICAL ELEMENTS FOR THE CONSTRUCTION OF

TANDEM CELLS

K. Herz, O. Corlatan, G. Anders, M. Schafer

ZSW
(Centre for Solar Energy - and Hydrogen - Research
Stuttgart and Ulm )
HeBbriihlstr. 61
D - 7000 Stuttgart 80

ABSTRACT. Reflection holograms were developed as spectral

beamsplitters for use in solar tandem cell configurations. The
holograms have a high reflectance and transmittance and a variable
spectral bandwidth between 20 and 200 nm. The central wavelength of
reflection/transmission can be shifted in a wide range. The swelling
behaviour of the gelatine was modelled and i ts influence on the
hologram properties was studied by computer simulations . Efficiency
calculations considering various solar cell combinations showed how
the hologram properties have to be specified to get maximal tandem
efficiencies. A simple tandem concept was tested in an experimental
setup using a poly-Si and a GaInP cell.

1. INTRODUCTION (transmission cell). We exami ned nonconcentrating

systems like a) and b). Example c) using a
We investigated on dispersive tandem systems, concentrating reflection hologram is discussed in
which use a holographic optical element (HOE) for (2).
splitting white light into two spectral ranges,
Fig. 1. Each wavelength range is directed to a
so 1ar cell with apropri ate band gap . System 2. REFLECTION HOLOGRAMS IN DCG TECHNIQUES
performance depends on the spectral selectivity of
the opical element and the type of solar cells. The 2.1 Principles of holographic techniques
conversion efficiency of such tandem systems can Volume phase holograms are recorded by
achieve higher values compared to a single cell superposing two laser beams in a recording
(1) . material. The resulting interference pattern is
Holographic optical elements offer a lot of determined by the properties of the beams
features, which are well suited for this (wavelength, wave fronts) and the geometry of the
application: opt i ca 1 set up. After the exposure of the. 1i ght
- their spectral selectivity can be adjusted to sensitive layer and a proper developing process the
the spectral sensivity of the solar cells, intensity distribution of the interference field is
- they have a high diffraction efficiency, converted into a periodic modulation of the
- they are flat, light, variabel in size, refractive index over the whole depth of the
- they can be copied and therefore are inexpensive. medium: a refractive index grating is formed. At
There exist two kinds of volume phase reconstruction the refractive index modulation acts
holograms: transmission and reflection holograms . like a stack of Bragg planes in a crystal
Because of the steeper filter slopes of reflection diffracting a spectral range of the incident light .
holograms and because their spectral selectivity is Such volume phase holograms have a theoretical
not as strongly influenced by a variation of the diffraction efficiency of 100% (3).
reconstruction angle our current work concentrates Using dichromated gelatine (OeG) as recording
on this type . Attempts were successful to broaden material, which is sensitve for blue and green
their spectral selectivity and to shift the central light, we influenced the swelling and shrinkage of
wavelength. A physical model was developed and the layer by varying process parameters in order to
evaluated in a program to calculate these two change the distance between Bragg planes. DCG
parameters. Efficiency calculations showed how to offers the advantages of high diffraction
select bandwidth and band position to attain efficiencies, high resolution , low scattering and
optimal tandem efficiencies. high transparency. Reflection holograms in DCG
The tandem concepts shown in Fig. 1 are based behave like dichroitic optical filters: they have a
on the idea of a reflection hologram grating us~ ng bandpass characteristic in reflection and a
one solar cell for the diffracted light (reflection stopband in transmission. The transmittance/
cell) and one for the nondiffracted light reflectance is the wavelength dependent ratio of
incident light to transmitted or reflected light. A
a) b)
optical filter is characterized by the bandwidth
~ lambda at half-maximum, the center wavelength lc
Rericcllon at the midpoint of the Alambda intervall ana
Cell the steepness of the slopes.

~~ 2. 2 Experimental results
Fig.2 presents the transmission curves of a
Ilologl'am :!5~~~
narrow band (lc=550nm, Alambda- 20nm) and a broad
band reflection hologram (lc=670nm, Alambda-200nm).
Tran mis ion Cell The maximum transmittance IS over 80% in the case
of the narrow band and about 70% in case of the
broad band hologram. Light scattering and
Fig.l: Solar tandem systems using a reflection absorption in the gelatine layer are the reason for
hologram transmittance drop in the shortwave range . The

15
transmittance of the glass plates we used as ~ 100r-....................-------------------------,
substrates have a maximum transmission of 90%, no ~ - - O' --- 30'
antireflection coatings were used. Subtracting ....... 10' - - 40'
glass absorption and Fresnel losses the
transmittance of the gelatine is 80% to 95 %.
oC 80 ···..··..···· 20'
'00
Scattered light is lost in transmittance (J)
measurement (small aperture) although a fraction of E
up to 50% of it may contribute to the illumination (J)
of the transmission cell, if the hologram is c
laminated directly onto the cell.
~100~---------------------,
.-ro
'-

cf?
I0...--Il
····7·......................._..... - . _ _ _ ...... - . , .

c 80 3
o
(J) 500 600 700 800 900
(J)

E Wavelength [nm]
~ 40
.=ro 20
Fig.3: Wavelength dependence on angle of incidence
of a reflection hologranm
the exposing light beam. Inhom0genious swelling
leads to a variation of the Bragg plane spacings.
~oo 500 600 700 800 900
3.2 Numerical Simulation.
Wavelength [nm] Therefore a numeri ca 1 s i.mul at i on was
implemented using a one dime.nsional refractive
Fig. 2. A narrow (1) and a broad band (2) index grid to describe the hologram. The evolution
reflection hologram, (3)transmittance of in time of an incident wavefront undergoing Fresnel
the glass plate reflection and diffraction at each discreet
gridpoint is calculated until a steady-state
Table I shows the experimental results in a brief condition of the reflected and transmitted beam is
illustration: reached.
The refractive index and the local distance
Table I: Characteristics of reflection holograms between Bragg planes is calculated according to the
model of swelling. Basically any shape of an index
curve can be used for simulation.
Central wavelength lc 520 nm - 750 nm
Bandwidth A 1ambda 20 nm - 200 nm 3.3 Results.
Maximum transmittance Tmax 80 % Thicker layers lead to higher refraction
Maximum reflectance Rmax 90 % efficiencies and steeper filter slopes but do not
increase the bandwidth. Broadband holograms as are
obtained at the ZSW can be simulated with an
It is a common feature of reflective type inhomogenious swelling between front and back side
holographic gratings that both their transmittance of about 30 %, Fig.4. The index modulation is
and reflectance spectra shift to shorter related to the smallest obtainable bandwidths and
wavelengths as they are tilted from normal to has a value between .08 and 0.12. Thicker layers
oblique incidence, Fig.3. This shift to shorter lead to higher refraction efficiencies and steeper
wavelengths at oblique incidence is very useful in filter slopes but do not increase the bandwidth.
tuning passband or stopband from one wavelength to
another. This shifting behaves according to the '#- 90 Layer Thickness: 16 jJm
Bragg condi1ion. -; 80
..g 70
3. SIMULATION OF HOLOGRAPHIC GRATINGS :rl 60
3.1 Swelling behaviour ~
a: 50
The actual properties of a holographic 40
optical element in dichromated gelatine technique 30
are not only defined by the optical setup and the
wavelength of exposure. Process parameters during 20
the depOSition of the gelatine layer, its drying 10
conditions and parameters for the chemical
sensitization and developement are crucial factors 0 490 540 590 640 690 740 790 840 890
for the determination of the bandwidth and central Wavelength [nm]
wavelength of the hologram.
The most simple assumption causing a shift of Fig.4: Reflection of a broadband hologram simulated
the central wavelength is a homogenious swelling with inhomogenious swelling 30%, thickness
process of the gelatine layer. In consequence the 16/.1m
distance between the inscribed Bragg planes is
increased. This process is apt to shift the central
wavelength while the relative bandwidth stays 4. EFFICIENCY ASSESSMENT
constant.
To achieve larger bandwidths an inhomogenious 4.1 Modelling of a tandem system
swelling must be considered. The gelatine is For the determination of required hologram
supposed to exhibit a different swelling behaviour properties a tandem system consisting of a
due to its local density and local absorption of holographic diffraction grating and two types of

16
solar cells as shown in Fig.Ia) or b) was modelled. efficiency to maximal 31% corresponding to eta/etaT
The following setup was examined: • 1.32.
- a hologram which reflects light in a certain
wavelength range Alambda and transmits the rest 1.6 ~
..........
of the spectrum .;:;
- a reflection cell with a bandgap EQR T 1.4 -:!
- a transmission cell with a bandgap EQ
Calculations of the conversion efficiency of the
cells are based on a one diode model proposed by 1.2
(1). The short curcuit currents jsc of the single
cells and the cells in tandem configuration have
been calculated at a temperature of 27·C using an 20 1.0
AMO spectrum. For simplicity a rectangular
transmission and reflection characteristic of the
hologram was used. The bandwidth and band position 15 L _ _---iL..-_ _---l_...:.-_ _ _'---_--' 0.8
as well as the degree of transmission and ~ ~ ~ ~ ~ ~ ~ m 1~
reflection could be chosen freely.
The open circuit voltage Vo of the cells is D.A ... A [nm]
determined by application of the aiode equation (1) [nm] ~ ~ ~ ~ 100 0
with a diode factor of one. jo is the dark
saturation current and can be expressed by equation Fig.5: Tandem efficiency eta and ratios eta/etaR,
(2). eta/etaT as a function of hologram bandwidth
A'lambda for EgT = 1.85 eV, EgR • 1.44 eV
Voc = kT/e*ln(jsc/jO + 1) (1)
~40 T
jo = c * exp(-Eg/kT) (2) ~ 11- l.85eV
R
1I- l.lev
>0.35 1
The constant c was fitted from one of the best o
silicon cells (5). Voc of a GaAs and a GaInP cell c:
Q) 30
10--_ ~ 1
was calculated and showed to be in good agreement
with experimental results.
"0
!i:
W25
-~ ~ r- ~ ~L
1.25
~
4.2 Results of calculation 20
~ rt/'L 1.0
Tandem efficiencies were calculated 1- rt/rtt
considering cell combinations with the bandgaps 1.1 15 0.75
eV(Si), 1.44 eV (GaAs) and 1.85 eV (GaInP). From
six possible cell combinations two will be w I I i
m
u
discussed here to demonstrate some basic ideas: ~ 300 ~ ~ ~ 700 800 1~ 1100 1~

f:::.A ... I I I I I Wavelength [nm]

4.2.1 E T = 1.85 eV, E R = 1.44 eV [nm) ~ 300 ~ 100 0
In ~ractice this arraHgement can be realized
by a GaInP and a GaAs cell. As GaInP can only Fig. 6: Tandem efficiency and ratios eta/etaR,
converse light with wavelengths below its bandgap eta/etaT as a function of hologram
(670 nm) the spectral range above 670 nm can be bandwidth Alambda for EgT = 1.1 eV, EgR =
reflected to the lower bandgap cell (GaAs) without 1.85 eV
any loss for GalnP. So the long-wave bandedge of
thR reflection hologram is optimally positioned at 5. EXPERIMENTAL TANDEM SETUP
E (860 nm), Fig.5. If the fi lter bandwidth
AY~mbda is increased beginning from Alambda = 0 at The concept of a nonconcentrating tandem was
Ea the tandem efficiency reaches a maximum at EgT tested in an experimental setup under laboratory
(070 nm). A bandwidth of 190 nm is sufficient to conditions, Fig 7. For illumination an irradiation
realize a tandem efficiency of 32 %, ideal cells of 100 mW/cm 2 from an ELH lamp was applied. USing a
and holograms (~=T=I) assumed. Further increase of reflection hologram in combination with a GalnP
the bandwidth has a negative effect because short cell in transmission and a polycrystalline Si cell
wave energy is then reflected to the lower bandgap we obtained tandem efficiencies increased by a
cell. factor of 1.06 to 1.08, Table II, as compared to
The quality of the tandem configuration must single cell efficiencies. The hologram bandwidth
further be examined comparing the tandem efficiency used would be ideal for a cell combination GalnP
eta with the efficiencies etaT and etaR of the and GaAs as explained in Fig . 5. But the
single cells, Fig. 5. The fraction eta/etaT and efficiencies of our GaInP and GaAs cells differed
eta/etaR show maxima of 1.41 and 1. 25 at EgT. too much to get an efficiency gain in a tandem
Because of the higher cell efficiency of GaAs arrangement.
eta/etaR is smaller than eta/etaT. Using realistic
reflection and transmission properties (R=T=0.9),
the maximum tandem efficiency decreases to 29% ELH Lamp
exceeding the efficiency of the best single cell by
a factor of 1.12.

4.2.2 E T = 1.1 eV , EaR = 1.85 eV

Thi~ cell combination can be realized by a Si R f1. Hologram
and a GalnP cell. In this configuration the best C 11
position of the reflector band edge is the bandgap
wavelength of the reflection cell (670 nm). As
GaInP has a higher conversion efficiency for the
blue spectrum than silicon the tandem efficiency mi ion
increases up to 35 % at a bandwidth of 300 nm (R =
T = 1), FIG. 6, exceeding the single cell II
efficiency of silicon by a factor of 1.47. A
reflection/transmission of 0.9 decreases the tandem Fig. 7: Experimental tandem setup

17
 A cell combination Si in transmission and
GalnP in reflection needs a hologram bandwidth in
the short wave range, Fig. 6. In this range our
holograms are not yet sufficient so we us~d a cold
mirror to demonstrate the practicability of the
described tandem concept. The tandem efficiency
exceeds the best single cell efficiency.by a factor
of 1.21, Table II. This is in ~ood agreement with
calculated values assuming R= T = 0.9.

TABLE II: Experimental result~ from a tandem setup

Irradiation 100 W/m (ELH lamp)
Transm. Refl. a= Bandwidth eta/etaT eta/etaR
cell cell B (nm)
GaInP poly-Si 30° 650 - 830 1.08 1.06
(hologram)
poly-Si GalnP 45° 420 - 680 1.21 1.23
(cold mirror)

6. CONCLUSIONS
Experimental results demonstrate a simple
concept of a nonconcentrating tandem module using a
spectral beam splitter Reflection hologram
gratings are an alternative to conventional
beamsplitters like dichroitic filters. They are
inexpensive, are easily reproducible by copying
techniques and their spectral selectivity can be
adjusted to the spectral sensitivity of the solar
cell.
Cell combinations of a low and a high bandgap
cell (Si/GalnP) demand a bandwidth of about 300 nm
between 370 and 670 nm. This spectral range covers
the optimal potential for a high tandem efficiency
of theoretically up to 35%, ideal cells and
holograms postulated. Therefore the reflection
wavelength has to be extended into this range. This
is possibl~ by shrinking of the gelatine layer
after exposure (6,7). The principal mechanisms of
gelatine behaviour are understood and experimental
results can be interpreted by means of computer
simulations using a swelling model. On the basis of
these experiences further steps forward seem to be
possible.

ACKNOWLEDGEMENT
The authors thank W. Wettling from ISES,
Freiburg and S. Kurtz from Seri, Colorado for some
GaAs resp. GalnP cells used for realizing an
experimental tandem setup. Further we acknowledge
valuable discussions with H.W. Schock, T. Walter,
IPE Stuttgart and S. Nann, ZSW Stuttgart.

REFERENCES
(1) J.C.C.Fan, B-V.Tsaur and B.J.Palm, Conference
Record 16th IEEE PV Spec. Conf., San Diego,
692, (1982).
(2) G. Anders, D. Corlatan, K. Herz, D. Schmid,
Conference Record lOth PV Sol.En.Conf., Lisbon
(1991).
(3) H. Kogelnik, Bell Syst. Techn. Journ. 48, 2909,
(1969).
(4) W.H. Bloss, M. Griesinger, E.R. Reinhardt,
Conference Record 16th IEEE PV Spec. Conf., San
Diego, 463, (1982).
(S) J.Zhao, A.Wang and M.A.Green, Conference Record
21st IEEE PV Spec. Conf., 333, (1990).
(6) J.Jannson, T.Jannson and K.H.Vu, Sol.Energy
Mat. 14, 289, (1986).
(7) J.A.Quintana, P.G.Boj, A.Bonmati, J.Crespo,
M.Pardo and C.Pastor, Pract. Hologr. III, Vol.
1051, 160, (1989).

18
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Light Trapping with Submicron Gratings in Thin Crystalline

Silicon Solar Cells

H. Kiess, J. E. Epler, M. T. Gale, L. Krausbauer, R. Morf

and W. Rehwald
Paul Scherrer Institut zurich, CH 8048 Zurich, switzerland

Abstract: Light trapping in silicon solar cells, fabrication of cells with and without
grating are investigated. First results on the impact of gratings on the electrical
properties of cells are reported.

1. Introduction less precision without loss in the effi-

ciency of the light trapping and are,
therefore, better suited for rough Si
The efficiency of crystalline silicon surfaces.
solar cellls can be improved by use of
thin substrates [1] if (i) the light In addition to these losses also the
absorption is enhanced by means of light reflection losses from the front side of
trapping and if (ii) recombination losses the silicon have to be minimized. Since in
at the cell surfaces are kept very low. practice solar cells are protected by a
Indeed, the cell thickness t should at glass cover we investigated optical coup-
best be small on the scale of the minority ling layers between the front surface of
carrier diffusion length L, which in si and glass covers. They were prepared
polycrystalline silicon (solar grade) either of silicon nitride layers with
material is of the order 100 to 200 varying refraction indices or of TiOZlSi02
microns, such that cell thicknesses of multilayer structures. The mult11ayer
less than 50 microns would be desirable. structures were slightly superior since
Since silicon has an indirect bandgap, the silicon nitride layers tend to have a
light with wavelengths at and slightly weak absorption at wavelengths below 450
below the bandgap is only weakly absorbed nm. We find that the average reflection of
and, therefore, partially lost for the silicon behind glass in the wavelength
photogeneration of charge carriers in thin range between 400 and 1100 nm can be
wafers. In order to reduce these losses, reduced to approximately 8 % in agreement
means have to be developed to increase the with our theoretical simUlations. This
optical path length of the weakly absorbed value refers to a glass cover without
light by appropriate light trapping. antirefler.tion coating. If the glass is
covered with a broad band antireflection
We have divided the task of making an coating we estimate to reduce the ref-
efficient thin solar cell of lower grade lection from the front face down to about
silicon into three subtasks: Firstly, we 5 % (Fig. 3).
investigated the light trapping in thin
silicon wafers and then, secondly, we
attacked the problem of making a silicon
solar cell with a reasonable efficiency
and, thirdly, we tried to attach our light 3. Solar Cell Fabrication
trapping scheme to these cells. In the
following we will report on these investi-
gations. We have made solar cells of 400, 250 and
40 micron thick silicon wafers. The 400
and 250 micron thick wafers were used as
obtained from the supplier, whereas the 40
2. Light. trapping in thin crystalline micron thin wafers were prepared by etch-
silicon wafers ing original wafers in KOH down to the
required thickness. Cells of 80 micron
thickness were also investigated, however,
The light trapping is achieved, as already obtained from an outside source. These
discussed previously [2], by attaching a cells had a back surface field in order to
metallized submicron grating to the rear reduce recombination at the rear side of
side of the silicon cell as shown in Fig. the cell.
lb. The dimensions, depth and period, are The cell fabrication itself followed the
the result of a numerical optimization by procedure as described in detail by Bla-
excact solution of Maxwell's equations. kers et al. [4]. Modifications of the pro-
The gratings of 0.31 or of 0.62 micron cedure were required if the grating had to
period and of about 0.07 micron groove be applied to the rear side of the cell.
depth were prepared by holographic tech- In contrast to Blakers et al. we neither
niques and dry etching [3]. The efficiency microstructured the front face of the
of the light trapping is best given in cells nor did we try to make an antire-
terms of an equivalent thickness, 1. e. flection coating directly on the silicon.
the thickness which a silicon wafer with- The reflection losses of the front side of
out grating but with a planar reflector on the cells were, therefore, determined by
the rear side must have in order to the thickness of the passi-vation oxide.
achieve the same reflectivity (see Fig. 1a We usually investigated cells of 20x20 mm 2
and 1b). 40 micron thick silicon wafers size. The best cells we were able to make
with grating;J.re approximately equivalent had an efficiency of 18 %. If the reflec-
to 200 micron thick silicon wafers without tion losses are taken into account, the
grating. This is demonstrated in Fig. 2. efficiency would have reached values close
Gratinqs with 0.62 micron period require to 20 %. A typical I/V-characteristic of a

19
250 micron thick cell under 0.3 AM 1.5 4.1 Cells of 250 micron thickness
illumination and the spectral reflec~ion
curve of this cell are shown in Fig 4
variations of the efficiency of cells were
quite pronounced (+ - 1 %) even if the Though the effect of the light trapping in
cells were prepared on silicon wafers of cells of this thickness should be minor we
nominally the same quality and under the have prepared 3 cells each with and with-
same conditions. The reason for this could out grating under' identical conditions. If
not yet be found and is subject of a con- the diffusion length is not large compared
tinuing investigation. to the thickness, then the effect of
recombination at the rear face of the cell
should be less pronounced. The short cir-
cuit current of the cells with grating was
in all cases greater ,than that of the
4. Solar Cells with Grating cells without grating. The average of the
I~c of the cells without grating was 304
A/ml, with grating 333 A/m2. The increase
Light trapping using a grating can only be of I,sc by approximately 10 % applying a
successful if at the same time the elec- grat~ng to the cells is quite remarkable.
trical properties of the cell are not The efficiency of the cells, however, did
deteriorated by fabricating the grating. not increase since the cells with grating
The grating 0an be prepared directly into had lower fill factors than those without
the rear side of the silicon or alterna- grating which practically offsets the gain
tively into a layer of a weakly absorbing in current. The reason for the lower fill
material of a high refractive index of n > factors is unknown.
3. For passivating the rear surface of the
silicon is covered by an oxide layer, i.
e. a low index material. For optical rea-
sons this layer has to have a thickness of 4.2 Cells of 80 micron thickness
about 10 nm or less otherwise the
diffracted light would only weakly couple
into the silicon and the optical effi- In cells of only 80 micron thickness, the
ciency of the grating would be reduced. light trapping should be more pronounced
We made the grating only directly into the than in the 250 micron thick cells. HoW-
silicon in the case of the 80 micron thick ever, if the passivation of the rear side
cells. In all other experiments we used is not sufficient, already the influence
the second method assuming that a thin of an inadequate surface recombination
si0 2 -layer of about 10 nm between grating velocity might affect the electrical pro-
and rear side of the cell is sufficient to perties adversely. As already mentioned,
leave the electrical properties unaf- we investigated cells of this thickness,
fected. The cells were prepared as usual. which were prepared differently: the cells
Since the thickness of the passivation had a back surface field and the grating
oxide was about 100 nm it had to be was made directly into the rear face of
thinned down on the backside to about 6 to the cell. We assumed the depth of the
12 nm by etching or else it was completely grating grooves to be significantly
removed and regrown to a thickness of 10 smaller than the thickness of the p+-
nm. Subsequently an SiO x layer of 150 nm region so that its effectiveness should
was deposited and the grating fabricated. remain intact. An evaporated large area Al
The refractive index of Sio can be tay- contact served simultaneously as an elec-
lored to the required value ~y varying its trical back contact and as a reflector for
oxygen contents. Aluminium point contacts the grating. We observed that the cells
were evaporated on the rear side of the without grating had a short circuit cur-
cell. This process served simultaneously rent of 246.4 A/m2 and with grating of
to metallize'the grating. 255.3 A/m2. The efficiency had also
slightly increased from 9.29 % to 9.63 %.
Since the short circuit current Isc is The average reflection losses of these
directly x'elated to the light trapping
efficiency we will in the following pre- cells were high and about 35 %; for a fair
dominantly consider this quantity and the comparison with other results the effi-
external and internal quantum efficiencies ciency would have to be corrected for
which are measured spectrally. The exter- these losses. Since the open circuit volt-
nal quantum efficiency is defined as the ages (0.592 V, 0.591 V) and the fill fac-
number of carriers collected per incident tors (0.639, 0.637) of the cells were
photon, the internal efficiency as the practically identical we believe that the
number of carriers collected per photon increase in efficiency can be traced back
absorbed by the silicon. Thus the external to the light trapping.
efficiency reflects directly the light
trapping, whereas the internal efficiency
gives information on the diffusion length
and the surface recombination. Experimen- 4.3 Cells ot 40 micron thickness
tally the internal efficiency becomes
difficult to determine at long wavelengths These cells had again a passivating oxide
where the si is only weakly absorbing and and Sio x layer on the rear side. In cells
also due to the imperfect reflectivity of of only 40 micron thickness, the light
the metallJ,c back reflector. trapping should be quite pronounced and
hence also the increase of the short cir-
cuit current compared' 'to cells without
light trapping. However, this is only true
if the surface recombination velocity in
both types of cells can be kept equally
low. In an attempt to investigate this

20
question and to reduce the inaccuracy
inherent in our processing we prepared
cells whose rear sides had only on one
half a grating. The short circuit currents
of such cell halves with and without gra-
ting were nearly identical. The question
remained whether the dry etching process,
the only processing step which was omitted
on the cell half without grating, causes a
degradation of the surface passivation
such as to destroy the expected increase
of current by the light trapping. Measure-
ments of the external and internal quantum
efficiency shed some light on this ques-
tion (Fig. 5). The external efficiency is
for both cell halves identical so that the
light trapping is seemingly ineffective. Fig. 1. Cross sections of solar cells
The internal efficiency' is in the long without (a) and with ,(b) grating. The
wavelength range greater for the cell area reflection losses of cells with grating
without grating which indicates that the are reduced due to light trapping
surface recombination velocity must be
smaller than in the area with grating. The
electrical degradation due to the fabrica-
tion of the grating is obviously as severe
as to veil the improved optical properties
of the cell.

5. Conclusions

We have shown that the optical absorption

of light in a thin solar cell can be sig- 90r------------------------------,
nificantly improved by appropriate light
trapping. Light trapping was observed to
increase the short circuit currents of
cells with thicknesses above 80 micron.
However, the grating fabrication has led
to strong degradations of the electrical 70
properties of the rear surface, so that
the improved absorption of the long wave-
length light could not yet be converted
into an improved short circuit current in '#. 60
cells of 40 micron thickness. Cells of
this thickness are more sensitive to c
changes of the surface quality than 0 50
:;:::;
thicker cells. Further investigations are U
Q)
under way for solving this problem. ;:;:::
Q) 40
a:
Acknowledgments: We have benefitted from
the help of many colleagues in the 30
laboratory, who supported us in every
respect and to whom we wish to express our
gratitude. Financial support by the
Bundesamt fur Energiewirtschaft is
gratefully acknowledged.
10 ~~~~~~~_r~~rr,_~_r,_ro~
400 600 800 1000 1200 1400
Wavelength [nm]
References:
1. H. J . Hovel, Solar Cells, in Semicon- Fig. 2. Reflection of a 40 micron thick
ductors and Semimetals, Vol 11, Acad. solar cell covered with a passivation
Press, 1975
oxide and supplied with a planar reflector
2. R. H. Merf and H. Kiess, Proc. of the and a metallized grating on the rear side
Nineth EC Photovoltaic Solar Energy resp.. The grating shifts the reflection
Conference, Freiburg, 1989 curve to longer wavelengths.
3. H. Kie~s, R.H. Morf, M.T. Gale and
A.E. Widmer, Techn. Digest of the Intern.
PVSEC-5, Kyoto, 1990
4. A. W. Blakers, A. Wang, A. M. Milne, J.
Zhao, X. Dai, M. A. Green, 4Ih lntern.Photov.Sci.
and Eng. Conf., Sidney, 1984

21
 Ankopplung Si - Glas 90 ~-----------------------------,
i
~.

\ 60+-----------------~~~--~

J\\
16 i

-- --
c
o
'x
Ql
ti 8
~ -----._ IS13N4 1
.......... !-'-. .....

a:
lTI0 2lS I0 2J
4
401+-----~--~~--~------------_1

400 600 800 1000

301+-------~~~~~------------_1
Wellenlange [nm)
Internal .

Fig. 3. Reflection losses of silicon

cells covered with glass. The optical external
10 +-------------~~~--------~
coupling layers between glass and silicon
are multilayers of silicon nitride or of
Ti02/Si02 • No antireflection coating is O ~~~~~~~nTnTMT~~~TMT
applied to the glass surface. 600 900 1000 1100 1200
Wavelength [nm]
0
- - without grating
- O,Q1

Fig. 5. Internal, external efficiency and

<
reflection of a solar cell with and with-
.........
- O. CI
out grating. The curves marked with an
...
z:
a:
arrow refer to results without grating.
~ - O.CD
u

- 0 . 04

-O, DS
0. 1 0.2 0.' 0.4 O. S 0.' 0.1

VOLTAGE/V

Fig. 4. Typical I/V characteristic of a

solar cell measured at a reduced light
intensity of 0.3 AM 1.5.

22
10TH EUROPEAN PHOTOVOLTAIC SOLAR·ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A TECHNOLOGY-BASED COMPARISON BETWEEN TWO-SIDED AND

BACK-CONTACT SILICON SOLAR CELLS

Andres Cuevas F" Ronald A. Sinton' and Richard R, King"

Universidad Politecnica de Madrid. ETSI Telecomunicacion, 28040 Madrid, Spain

• SunPower Corp., 625 Ellis St., Mountain View, CA 94043, USA
,. Georgia Institute of Technology, School of Electrical Engineering, Atlanta, GA 30332-0250, USA

Abstract
A modeling comparison between Two-Side Contacted and Back-Contact cells is made to determine which
cell structure offers the best performance potential fo~ concentrated light. The comparison includes both
low-resistivity and nearly intrinsic silicon substrates. The model parameters are carefully selected to reflect
the state-of-the-art of crystalline silicon technology. Optical and ohmic losses due to the front metal grid
,constitute the most significant difference between the Front-Contact and Back-Contact cell designs. On
the other hand, recombination at the front surface emitter is the main factor that limits the internal cell
performance. If the fabrication process is kept reasonably simple, the Back-Contact cell is clearly advan-
tageous, with an achievable efficiency of 26% in the range of 100 to 300 suns. Possible advancements such
as polysilicon emitters are also discussed and predicted to give an efficiency of 27.5%.

1. INTRODUCTION the efficiency vs. irradiance curve, and this masks to some
extent the internal behaviour of the device. We prefer to
Several silicon cell designs have recently demonstrated assume a constant lens size and a variable cell sfze. The
conversion efficiencies in the range of 26% under concen: lens provides a fixed amount of power, for eXample, 8 W
trated light. These are the simplified version of the Back (a lOx10 cm 2 lens). To achieve a desired irradiance the cell
Point-Contact Cell (1, 2), a Front-Gridded Point-Junction area is adjusted, for example a 0.8 cm 2 cell would operate
Cell made on a high resistivity substrate (3), and a n+-p-p+ at 10 W /cm 2 , or 100 suns~ Since the cell size decreases as
PESC cell made on a low-resistivity substrate (4). The the irradiance increa(les, th~ output current is always about
purpose of this work is to model and compare these three 3.2 A. ,The metal grids that correspond to the different cell
cell structures. Its scope goes beyond a mere theoretical sizes should all h.ave the same ~eries resistance and produce
exercise: it intends to be a design optimization of feasible, the same relative power .loss. The design of the internal cell
practical cells. For this study, issues like the need for long structure is thus decoupled from the design of the metal
term stability (5), the new available data on recombination grid.
in highly doped regions and surfaces (6, 7) and the com- The optimization of the front grid is described in (8).
plexity of the fabrication process are taken into account. The parameters that determine this optimization are the
The final architecture of the cells will not be fully de- sheet resistivity of the metal, Pm, and the cell output cur-
fined until their design is completed. As an starting point rent, 1m, and voltage, Vm. For a chevron pattern, Pm =8xlO- 3
we shall assume the structures shown in Fig. 1 for the two- n (3.5 {lm of aluminium), 1m ~3.2 A and Vm ~ 0.7 V we
sided cells and for the Simplified BCC. Both types of cells obtain a shadowing loss of 5.5% and a total power loss of
can be fabricated using four photolithographic masks. Nev- 10%.
ertheless, we shall discuss the implications of more complex
structures having point diffusions rather than planar ones
at the back surface and also polysilicon emitters instead of
the conventional diffused ones. Undc~ JubsUale

n·typo dopant·dlflu.lpn
p· typ. p·tYPI
2. CELL DESIGN CONSIDERATIONS

The design of a concentrator cell must consider a num- AI-H '·Au

ber of different factors that can be placed in three main

METAL
categories: the design of the metal grid, the trapping of the
light and the design of the emitter regions. These factors
are interrelated and synergistically determine the perfor-
mance of the cell. Still, a final design has to take into
account the trade-off that exists between the cell architec-
ture and the fabrication complexity.

2.1. Cell size. External losses

In two sided cells the losses produced by the front metal grid
are strongly dependent on the cell size. To study the perfor-
mance under concentrated light it is customary to assume a
constant cell size and a variable lens size to produce a vari-
able irradiance. This leads to the necessity of redesigning
ME TAL
the cell metal grid for every irradiance. The series resis-
tance and shadowing loss are different for every point of Fig,1. Two-Sided and Back-Contact cell structure.

23
 Series resistance losses are also present in the Back- oxide/poly silicon/oxide stack (5). The transmittance of the
Contact Cell, although they are not subject to optical con- front surface is somewhat decreased because of absorption
straints. One metal grid is needed for the n+ regions, the in the poly silicon layer. Experimentally, the short circuit
p+ being vertically soldered to a metal plate. It typically current of a BPC with poly is 0.975 times lower than that
covers nearly a 25% of the back surface, which should pro- of a cell with Si0 2 , both textured.
duce a series resistance four times lower than that of the
front grid of a two-sided cell. We shall conservatively as- 2.3.2. Metal-semiconductor contact resistance
sume a total loss of 2% for the BCC. Given the small contact areas required to obtain a low
J o and the high current densities flowing in a concentrator
2.2. Light trapping cell, the voltage drop at the metal-silicon contact can be-
An efficient light trapping can be achieved by faceting the come important unless precautions are adopted to avoid it.
front surface of the cell and forming a Back-Surface-Reflector. The contact resistivity is dependent on the dopant concen-
Usually the back surface is flat and the BSR consists of a sil- tration at the semiconductor surface. For aluminum and
icon dioxide layer coated with aluminium. The reflectivity p+ silicon with NA • 2: 1019 cm- 3 the contact resistivity
of the BSR, Pb, the internal reflectivity of the front surface is pc ::; 1O- 6 fkm 2 , which permits one not to worry much
(for escaping light), Pi, and the substrate thickness deter- about the p+ contact. Focusing on the n+ contact, the volt-
mine the quality of the light-trapping structure. The cell age drop in it should be kept below a 1% of the maximum
output current is further affected by nonlinear phenomema, power voltage. For an irradiance of 10 W /cm2 and a frac-
which also depend on the substrate thickness. We describe tional contact area of 1%, the contact resistivity should be
them in Section 4. Pc ::; 2x10- 5 ncm 2 • Experimental data on the contact resis-
From an analysis of available experimental data (3, 9, tivity between aluminum and n-type silicon indicate that
10) similar to that suggested in (11, 12), we have extracted the surface phosphorus concentration should be higher than
Pb=0.92 and pi=0.87. These values are representative of 5xlO19 cm- 3 (13).
a textured/polished structure with inverted pyramids (3,
10), and also of a textured/textured structure with upright 2.3.3. Compensated regions
pyramids (9). They can also be approximately applied to The simplified Back Contact Cell is based on allowing
the simpler textured/polished structure with random up- the p+ and n+ emitters to touch and form a compensated
right pyramids. The measured J Be for a 130 p.m thick sim- region between them. It is necessary that the reverse break-
plified BCC having the latter light-trapping structure is down voltage be higher than a few volts and this puts a
41. 7 rnA, while the calculated value (for inverted pyramids, limit on the maximum dopant concentration. The tun-
using PC-1D) is 42.05 mA/cm 2 • neling mechanism becomes increasingly probable and the
reverse breakdown voltage decreases as both NAB and NDs
2.3. Emitter regions increase. Typical sheet resistivities are 18 n and 6 n for
The selection of the dopant diffusion regions involves the the p+ and n+ diffusions, respectively (1), which satisfies
consideration of a number of interrelated factors. Some the contact resistance requirements.
of them are quite well known. The emitter recombination
current (the Jo) should be as small as possible. Usually 2.3.4. Emitter area. 2-Dimensional current flow
there is a trade-off between the passivated and the metal- Restricting the n+ or p+ emitters to nearly lines makes
contacted areas that determines whether the emitter should the flow of current in the base 2-dimensional. It is neces-
be shallow or deep. The front emitter, if present, must have sary to place the contacts close enough and in a number
greater than 90% collection efficiency for short-wavelength high enough as to avoid significant voltage drops because
photons and it should preferably have a low sheet resistivity. of the crowding of current in their vicinity. The theoretical
Besides these conventional trade-offs we have to consider studies of the Back Point Contact Cell (14) determine how
some additional ones specific to concentrator cells. much the p+ and n+ -regions can be trimmed down: the
fractional area covered by each the p+ and n+ diffusions
2.3.1. Stability constraints should approximately be 5% for a 100 suns cell.
Very intense UV light has been found to degrade the
recombination properties of the silicon/silicon dioxide in-
terface. There are two ways of minimizing this effect when 3. MODELING ASSUMPTIONS
the surface is textured (5): a) creating a diffused region
in the front surface; b) depositing a thin layer of polysili- We consider two technology scenarios: thermal dopant
con. The first, most common approach gives a fairly stable diffusion and polysilicon deposition.
front surface if the phosphorus dopant concentration in it
is about 1019 cm- 3 or higher. For a pyramidally-textured 3.1. Diffused emitters
surface, the best UV-stablilized J o that has been experi- Having in mind a design for 100 suns, all the factors dis-
mentally demonstrated is 0.7xlO- 13 Acm- 2 (7). cussed above and a four-mask process complexity, it is now
The research on boron diffusions for stabilization pur- time to use the Jo plots provided in (6,7). In Fig. 1 we note
poses has yet to be done systematically. Preliminary results that:
indicate that all boron diffusions give a stable passivation.
• The front surface is textured with random pyramids
Unfortunately, the achieved Jo for a p+ diffusion on a tex-
and has a phosphorus diffusion that covers 95% of the
tured surface is relatively high, 2.9xlO- 13 Acm- 2 • Further
surface in the case of two-sided cells.
investigations might be able to improve this result, which
is considerably worse than the one for n+ emitters. • The heavy diffusions occupy 5% of the front surface
It is possible to use polysilicon to passivate and stabi- of two-sided cells, matching the metal grid lines. 4%
lize the illuminated surface. Some preliminary experiments is passivated and 1% is metal-contacted.
have already shown great promise but the technique still
needs further refinements. An stable J o of 0.4xlO- 13 Acm- 2 • Most of the back surface is passivated. In a four-
has been obtained for a textured surface covered with an mask process no extra diffusions are used to form n++

24
 and p++ regions in the 1% area to be contacted with performance. Silicon cells made with high resistivity mate-
metal. Because of the conflicting requirements of the rial frequently show a sublinear responsivity. The amount
passivated and unpassivated areas the design of the of sublinearity is approximately determined by the Jo of the
back emitters is quite delicate. A shallow diffusion front emitter, by the minority carrier diffusivity and by the
with a relatively high surface dopant concentration cell thickness (18). The effect can be very important in the
gives the best compromise. For the simplified Back cases of the Back-Contact cell and the n - i - p cell, and
Contact Cell the fractional areas are assumed to be appears to be far less severe for the p-i-n cell based on a
75% and 25% for the p+ and n+ regions, respectively. one-dimensional model. However, for the cell structure in
Fig. 1 there are transverse carrier concentration gradients
In Table I we summarize the saturation currents used in at the front surface that can also produce a significant sub-
the modeling calculations. The estimated values have been linearity. Because of this our modeling of p - i - n cells is
rounded to give the same total (emitter) saturation current optimistic.
for all the cell designs considered. If the cells were to be Low-resistivity cells show a superlinear responsivity even
designed for a higher irradiance then the surface dopant if the front emitter is n-type. The effect is attributable
concentration should be increased to keep the contact re- to the ohmic electric field in the base. It compresses the
sistance low, especially for the n+ emitter. This would lead minority carriers towards the collecting junction and thus
to slightly higher J 0 values. decreases their chance to recombine (19). It is remarkable
It is interesting to compare the Jo values chosen through only when the lifetime is relatively low, as in a 0.2 ncm
our design exercise and the experimentally demonstrated substrate, and for relatively thick ce\ls. This superlinearity
ones. For the simplified Back Contact Cell a total J o of makes their responsivity nearly as high as that of high-
2.44xlO- 13 Acm- 2 has been measured by PCD (2). Two- resistivity cells at high irradiances in spite of their lower
sided cells with a p+(point)/n+(floating) front diffusion struc- one-sun responsivity. Nevertheless, the ohmic voltage drop
ture and a n+(point) back diffusion have J o between 1.8xlO-13 produced in the base region at high currents makes the use
Acm- 2 and 2.7x10- 13 Acm- 2 (15). The values in Table I of thin wafers also necessary even with doped.substrates.
are therefore representative of the state-of-the-art. Considering the trade-off between light-trapping and
sublinearity (or ohmic voltage drop), we have selected a
3.3 Polysilicon emitters thickness of 75 J.Lm for the study (150 J.Lm for a 0.2 ncm
Polysilicon emitters present a very low J o even when they substrate). This is is the optimum thickness for the range
are metal contacted and, simultaneo1.!sly, a low contact re- of 100-300 suns. At the lower end of this range the require-
sistivity. Demonstrated values for polysilicon emitters are ment of a thin cell can be relaxed.
J o =O.82xlO- 13 Acm- 2 , Pc=·3.8xlO- 6 ncm 2 for p-type and
J o =0.45x10- 13 Acm- 2 , Pc=6xlO- 6 ncm 2 for n-type (16). 4.2. State-of-the-art technology
Using the fractional areas described above for the p+ and The performance for cells with diffused emitters is shown
n+ emitters and assuming that an oxide/polysilicon stack in Fig. 2. The efficiency of the Back-Contact cell is about
is used to passivate the front surface, a total J o of approxi- 1-2 absolute percent higher than that of the Two-Sided
mately 1x10- 13 Acm- 2 is estimated for all the cell designs.
cells in the range of irradiances 10-30 W /cm 2 • The main
reason for this is the 10% relative power loss caused in these
cells by the front metal grid, compared with the 2% loss
Emitters Diffused Polysilicon that occurs in the BCC. Actually the internal efficiencies
of all the cell designs are nearly the same at 100 suns. The
n+pp+ n+pp+ modeled efficiency of about 26% at 100 suns is in very good
p+in+ BCC p+in+ agreement with the experimental results obtained for the
Cell BCC
n+ip+ n+ip+ Back-Contact Cell. Efficiencies between 24.5% and 25.2%
are achievable with the p+/n-i-n+ structure and with the
n+-p-p+ cell in the range 10-20 W/cm 2 , and about half
JOlfront 0.7 1.0 0.4 0.4
2l!
JOlback 1.8 1.5 0.6 0.6
'Z7
JOltotol 2.5 2.5 1.0 1.0

Table I. Emitter saturation current densities representative l 2S

of thermal diffusion and polysilicon. All Jo values refer to >-

<..>
900 K and are to be multiplied by 10- 13 Acm- 2 . z ~
w
u
4. RESULTS H:W 24

The modeling has been done with the numerical pro- 23

gram PC-1D (17) for the two-sided cell structures and
with a quasi-three dimensional computer code for the Back
Point-Contact cell (14). The effect of the silicon resistivity 22 10" 10' 10'
has been explored for the n+-p-p+ cell structure.
IRRADIANCE 0N /cm~
4.1. Non-linearities in J. c Fig. 2. Comparison between p+/n-i-n+, n+/n-i-p+,
The behaviour of the short-circuit current with the incident low resistivity (0.2 ncm and 0.5 ncm) n+/n---p--p+, and sim-
power density determines to a great extent the overall cell plified Back-Contact Cells. The temperature is 25 °c. All
the cells have the same total Jo , 2.5x10- 13 A/cm2 at 900 K.

25
an absolute per cent point less for a n+/n-i-p+ structure. 5. CONCLUSIONS
Thanks to their nearly linear responsivity the p-i -n cell
and the low resistivity n+ -p-p+ cell get close to the BCC We have reviewed the state-of-the-art for the differ-
at higher irradiances. ent technological parameters that determine the design and
the performance of silicon solar cells. Based on this we
4.3. Possible improvements have modeled both Back-Contact and Two-Sided cells and
• Recombination at the front surface is what limits the BCC found for the first design an efficiency of about 26%, ap-
cell performance most. Because of this, the added complex- proximately the same that has already been achieved ex-
ity of using point diffusions at the back surface provides perimentally. Two-Sided cells are handicapped by a 10%
only a marginal improvement compared to the simplified loss produced by the front metal grid. This causes the pre-
cell structure. At high irradiances the three dimensional dicted efficiencies to be less than 25%. The main barrier to
effects thus introduced oppose the improved back J o ' achieving higher efficiencies for BCC cells is the front sur-
• Using an oxide/polysilicon stack at the front can re- face emitter that is required for a stable passivation. Ox-
duce the front J o • In principle there is an associated opti- ide/polysilicon stacks are a promising alternative for this
cal loss that negates the benefit of this lower J o • Further passivation. Using the parameters for polysilicon emitters
research should hopefully lead to a minimization of this that have already been demonstrated, efficiencies in the
loss, then the expected efficiency for a BCC with polysil- range 27%-28% should be attainable.
icon emitters is about 27.5% in the range 10-30 W / cm 2
(Fig.3). REFERENCES
• A better light-trapping than that assumed here can
significantly boost the performance. For example, the re- 1. R. A. Sinton and R. Swanson, IEEE Trans. Electron
sponsivity would increase by a factor 1.02 if the back- and Dev., Vol. ED-37, pp. 348-352, Feb. 1990.
front- surface reflectances increase to 0.95 and 0.92, respec- 2. R. A. Sinton et al., in Proc. Twenty-First IEEE Pho-
tively. The efficiencies would thus augment half an absolute tovoltaic Specialists Conf., Kissimmee, May 1990, pp.
per cent. 302-306.
• The front metal grid offers a substantial room for the 3. A. Cuevas, R. A. Sinton, N. Midkiff and R.M.Swanson,
improvement of Two-Sided Cells. Some record-efficiency IEEE Electron Devices Lett., Vol. EDL-ll, pp. 6-8,
cells have utilized special techniques to reduce the shadow- Jan. 1990.
ing loss: either a sculptured surface with sloped gridlines
4. J. Zhao, A. Wang, A. W. Blakers, and M. A. Green, in
(3) or a prismatic cell cover (4). We estimate that these Proc. Twentieth IEEE Photovoltaic Specialists Conf.,
advanced concepts can possibly reduce the total power loss Las Vegas, NV, Sept. 1988, pp. 583-585.
to 5%. p+/n-i-n+ or low resistivity n+/n-p-p+ cells
5. P. E. Gruenbaum, J. Y. Gan, R. R. King and R. M.
could then reach the same efficiency as the BCC and even Swanson, in Ref.(2), pp. 317-322.
surpass it at very high irradiances.
• Other options exist (20), for example the front surface 6. R. R. King, R. A. Sinton and R. M. Swanson l Trans.
can be flat instead of textured thus permitting an stable Electron Dev., Vo1.37, No.2, pp. 365-371, Feb. 1990.
diffusion with a very low J o (0.3x10- 13 Acm- 2 ). The back 7. R. R. King, P. E. Gruenbaum, R. A. Sinton and R. M.
surface would then be textured to trap the light within the Swanson, in Ref.(2), pp.227-232.
silicon, and front-surface reflectance would be reduced by 8. P. A. Basore, Solar Cells, Vo1.14, pp.249-260, 1985.
the use of multi-layer AR coatings and/or cell mounting
configurations which cause light reflected from one cell to 9. S. P. Tobin, C. J. Keavney, L. M. Geoffroy, M. M.
be incident on another. The trapping of the light and the Sanfacon, in Ref.(4), pp. 545-548.
reduction of reflection losses could also be achieved with an 10. M. A. Green, A. W. Blakers, J. Zhao, A.M.Milne, A.
optical cavity (21). The internal cell performance would be Wang and X. Dai, IEEE Trans. Electron Dev., Vol.
similar to that achievable with polysilicon emitters, since ED-37, pp. 331-336, Feb. 1990.
both give a similar J o • 11. P. A. Basore, IEEE Trans. Electron Dev., Vol. ED-
37. pp. 337-343, Feb. 1990.
28 12. j. M. Gee, in Ref. ( 4), pp. 549-554.
13. G.P.Carver. D.B.Novotny and J.J.Kopanski, in Proc.
27
Fourth IEEE VLSI Multilevel Interconnection Conf.,
June 1987, pp.337-343.
14. R. A. Sinton and R. M. Swanson, IEEE Trans. Elec-
i. tron Dev., Vol. ED-34, pp. 2116-2123, Oct. 1987.
>
()
15. A. Cuevas, R. A. Sinton and R. M. Swanson, in Ref.(2),
Z pp. 327-332.
W
(3 16. J. Y. Gan and R. M. Swanson, in Ref.(2), pp. 245-250.
ii:
u. 17. P. A. Basore, D. T. Rover and A. W. Smith, in Ref. (4),
w sec, poly pp. 389-396.
n-i-p. poly
23 0.2 Oem 18. M. A. Green, J. Zhao, and D.L.King, Appl. Phys.Lett.,
Vol.52(16), pp. 1361-1363, April 1988.
22 1a"I,;---......-----'~~10~'----------'--~.......-::!
lO> 19. M. Cid and J. M. Ruiz, IEEE Trans. Electron Dev.,
Vo1.36, No.3, pp.507-513, March 1989.
IRRADIANCE 0N/cm~ 20. R. R. King, Ph.D. Disertation, Stanford Univ., 1990.
Fig.3. Cells with polysilicon emitters. The temperature 21. R. A. Sinton and R. M. Swanson, IEEE Electron Dev.
is 25°C. The total Jo is 1.0xl0- 13 A/cm2 at 900 K. Lett. , Vol. EDL-8, pp. 547-549, Nov. 1987.

26
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

DEVEUlPMENl' OF A FOLYSILIa:N l'MITl'm SOLAR CELL

G. Papadopoulos, L.P. Boivin* and N.G. Tarr

Deparbnent of Electronics, Carletcn University
OttaloJa., O'1tario, Canada IUS SB6
*Division of Physics, Natiooal Research council
OttaloJa., cntario, Canada KlA OR6

ABS'I'IU\CT
SOlar cells with anitters consisting of a very thin «20 nm) layer of
polysilioon overlaid with a thicker (70-130 nm) layer of recrystallized
~ silioc:n have been fabricated and characterized. Both layers were
fomed by LPCVD with heavy in-situ fbosli!orus dcpng. 'nle highest tenpmlture
used in processing was the 6S0oC recrystallization anneal for the am:>J:Iiloos
material. Short circuit cun:ent densities of 30 mAan-" canbined with fill
factors close to 0 . 8 have been achieved uOOer sinulated 100 1lWalr2 AMl. . S
illumination. <:pm-circuit voltages as high as 669 mVat 28°C were obtained
for cells fomed on high quality 0.1 Gan float zone substrates when J eo was
set to 32 mAan-". 'lbtal-cu:ea AMl..S efficiencies were up to 14%.

l.INTRODUCTION 2 . CELL FABRICATION

State-of-the art solar cells formed on high- Since (5] had established that the interface
quality monocrystalline silicon substrates can between a monocrystalline substrate and material
provide short-circuit photocurrent densities J. e and deposited in the polycrystalline phase gives
fill factors close to theoretical maximum values favorable junction properties, recrystallized
under AMl illumination (1]. However, open-circuit material was first incorporated into the devices
voltages Voe still fall significantly below reported here through the use of a bilayer structure
theoretical limits (2]. Major losses in Voe often (Fig. 1). A very thin (~20 nm) layer of material
result from recombination along the emitter surface deposited in the polycrystalline phase establishes
and at metal contacts to the emitter (3]. the polysilicon/substrate interface, and is overlaid
with a thicker (70-130 nm) recrystallized amorphous
One possible approach to suppressing front layer to reduce emitter sheet resistance. Later
surface recombination is the use of a polysilicon experiments showed that high-quality solar cells
emitter (4]. Extrapolation of results obtained with could also be produced with emitters consisting
polysilicon emitter bipolar transistors (5] entirely of recrystallized amorphous material, but
indicates that a polysilicon emitter solar cell open-circuit voltages were consistently 10 mV lower
should give an open-circuit voltage of approximately than those obtained with a bilayer emitter on
730 mV at 28°C under AMl. 5 illumination in the identical substrates.
Fronl Metal Conwa Gnd
absence of significant recombination in the base.

Apart from suppression of front surface

recombination, in-situ doped polysilicon emitters
offer a means of forming high-quality pn junctions
at very low temperatures (~650°C). This may prove
beneficial in preserving long minority carrier
lifetimes in very high quality substrates. It may
also be advantageous in fabricating solar cells on
polycrystalline substrates, where conventional
furnace cycles can lead to dopant diffusion down
grain boundaries, shorting the cell.

In previous work the authors reported the Fig. 1 Bilayer emitter solar cell structure.
fabrication of solar cells with in-situ
phosphorus-doped polysilicon emitters (6]. When A. Polysilicon Material Development
formed on high-quality 0.1 Ocm float-zone substrates
these devices gave very high open-circuit voltages Preliminary experiments were carried out to
(up to 650 mV at 28°C when illuminated to give establish the dependence of polysilicon and
Joe - 30 mAcm- 2 ) and reasonable AMI photocurrent recrystallized amorphous silicon sheet resistance on
densities provided the polysilicon thickness was film thickness in the range suitable for
less than approximately 150 nm. Unfortunately, it photovoltaic applications. Films were formed using
was found that such thin polysilicon layers had low pressure chemical vapor depOSition (LPeVD) from
unacceptably high sheet resistances (typically silane/phosphine mixtures at a pressure of
several kG/D), giving poor curve fill factors (~0.6) approximately 0.4 Torr. Polycrystalline depositions
and correspondingly low efficiencies (~10% at best). were carried out at 627°C with a Silane/phosphine
ratio of 185/1, while amorphous depositions were
In 1986 it was demonstrated that the sheet done at 560 0 e with a flow ratio of 530/1. Amorphous
resistance of in-situ doped polysilicon films could films were subsequently recrystallized by annealing
be reduced by up to a factor of five if the films at 650 0 e for 60 minutes in a nitrogen ambient.
were deposited in the amorphous phase and Annealing at temperatures of 900 0 e or greater
subsequently recrystallized at low temperature [7]. following polysilicon deposition and doping is
This reduction in sheet resistance was attributed in common in the fabrication of bipolar transistors and
part to a greater incorporation of phosphorus into is known to significantly reduce the sheet
the amorphous films during deposition, as resistance of polysilicon films, but was avoided
demonstrated by SIMS, and in part to the relatively here for two reasons. First, 900 0 e anneals have been
large grain sizes obtained in recrystallized shown to lead to a slight increase in transistor
material [8] . This paper explores the application back injection current [9], which would translate to
of recrystallized amorphous silicon in solar cell a lower Voe in a solar cell. Secondly, high-
emitters. temperature annealing would lead to dopant diffusion
from the polysilicon into the substrate, deepening

27
the emitterfbase junction. This can lead to a loss minutes in hydrogen, polysilicon emitter diodes
in blue response in a solar cell. using an Al/TiW contact scheme had near-ideal dark
current-voltage characteristics over almost eight
The sheet resistance of in-situ doped decades of current and leakage currents on the order
polysilicon, recrystallized amorphous material and of 1 nAcm- 2 .
the bilayer structure are plotted as functions of
film thickness in Fig. 2. As expected, the sheet A comb-like grid with finger width and spacing
resistance of recrystallized amorphous material is of 19.5 and 711 ~m respectively was designed for use
roughly four times lower than that of material with 2 cm square cells having 5 ~m thick top metal.
deposited in the polycrystalline phase. The The total shadowing loss, including the busbar, was
resistivity of recrystallized amorphous silicon only 3.5%, while the resistive power loss was
increases very rapidly as the film thickness drops estimated to be 3%.
below roughly 80 nm, probably because the grains in
such thin films are very small. C. Complete Cells

Test cells were fabricated on a variety of

5000 r-----------------, substrates, ranging from lightly doped <",10 Ocm)
0 • poly-Sl/a-SI (recrystallized)
Czochralski material to long-lifetime 0.1 Ocm Wacker

--a
• a-SI (recrystallized)
• poly-SI float zone material. The latter material has been
4000 used extensively in previous work on high-efficiency
cells [1,3]. Polysilicon bilayer emitter

-
~ thicknesses in the range. 75-150 nm were
c 3000
CU investigated. The highest temperature process to
I II which the cells were exposed was the 650°C
'iii recrystallization anneal.

-
2000
~
Most test devices were left with bare front
I I)
1000 surfaces to ensure that measurements of ppotocurrent
II)
.c response would not be obscured by differences in
en antireflection coating properties. However, a few
0 cells were given a simple 75 nm thermally evaporated
0 500 1000 1500 2000 2500 SiO antireflection coating.

Thickness (A) To provide a reference with which to compare

the performance of the polysilicon emitter cells, a
Fig. 2 Sheet resistance of polysilicon as conventional cell with a shallow monocrystalline
deposited, recrystallized amorphous and emitter was fabricated using a 90 second diffusion
bilayer films. from a POC1 3 source at 850°C [2]. This very light
diffusion should give a junction depth of only
In order to select an appropriate emitter 0.15 ~m with virtually no "dead layer", allowing for
thickness for solar cell use from the data of very high blue response. The control cell received
Fig. 2, cell power loss as a function of emitter the same TiW/Al front contact scheme as the
sheet resistance was calculated following Green polysilicon emitter devices.
[10]. It was found that even with a grid spacing as
fine as 700 ~m, the power loss increases rapidly for 3. ELECTRICAL CHARACTERIZATION
emitter sheet resistances greater than roughly
1 kn/O. More closely spaced grid lines are not Absolute spectral response measurements of
practical due to rapidly increasing shadowing selected polysilicon emitter cells and the control
losses. From Fig. 2, this implies that the cell were carried out using a thermopile to
polysilicon emitter can not be made thinner than determine incident irradiance. Typical results are
approximately 80 nm. Conversely, an upper limit is shown in Fig. 3. As expected, the response of the
placed on emitter thickness by photo current losses. polysilicon emitter devices in the ultraviolet and
If one assumes that all carriers photogenerated in blue is considerably poorer than that of the
the polysilicon are lost to recombination at grain control.
boundaries, and that the absorption coeficients of
~ 0.5

.
polysilicon and monocrystalline material are the E
same, then approximately 10% of the photo current ;( c-SI (1500A)
theoretically available in the AMl.5 spectrum is §. poly (1450A)
discarded if the emitter thickness exceeds 150 nm 0.4 poly (900A)
[6] . This indicates that there is a relatively 51C
narrow window of polysilicon emitter thickness- 0
Q.
roughly 80 to 150 nm- that is useful for solar cell III 0.3
II)
applications. a:
B. Front Contact Grid ...
(ij
t)
0.2

Forming low-resistance ohmic contacts to 8.

-
(/)
shallow polysilicon emitters has proven difficult. 0.1
II)
The use of pure aluminum contacts lead to an almost
::J
complete loss of junction rectifying properties when
annealing was carried out at temperatures greater
'0
III
.Q 0.0
than approximately 350°C. Much better results were
obtained using a 0.1 ~m thick contact layer of
« 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2

Ti o. 1WO. 9 alloy overlaid with aluminum or plated gold. ').. (~m)

The TiW alloy was deposited by RF sputtering and the Fig. 3 Absolute spectral response for polysilicon
aluminum bye-beam evaporation. The titanium emitter cells and diffused junction control.
provides good adhesion and a low resistance contact
to the silicon, while TiW alloy is known to be a Once the spectral response had been determined,
very effective barrier to aluminum and gold the photocurrent which should be available under
diffusion [11]. Even after annealing at 450°C for 30 AMl.S illumination was calculated for the

28
 Table 3: Illuminated output paramelers of uncoated 0.SxO.5cm poly-Sila-Si'
ultraviolet and visible regions of the spectrwn; emitter solar cells measured using a AM1.S (100mW/cm 2) xenon
results are given in Table 1. The diffused junction arc solar simulator.
cell was found to have very small photocurrent
losses «0.5 mAcm- 2 ) for A < 0.8 ~m. Compared to the
diffused junction control, a 145 nm thick Solar Subs.rate Emiller Emitter
polysilicon emitter discards approximately 15% of Cell Resistivity Thlckne•• Re.l.tance Joe Voc FF Pmax
the available photocurrent in the ultra-violet and
blue regions, and a 92 nm thick emitter 8%. •
Control
In.em)
....
(A)
..... ... ...
(n/a)
. ....
mA/cm 2
35.5
Voltol
0.595 ....
(mW/em 2
.
......
c-SI 1.8-2.6 1500 400 23.4 0.514 0.1186 8.26
Table 1: AM1.5 short circuit current densities as calculated from 270 23.1 0.584 0.800 10.80
1 1.8-2.6 1,440
absolute spectral response measurements of uncoated
2 1.8-2.6 1080 900 22.0 0.583 0.767 9.84
poly-Slla-SI emitter cells. 3A .. 1.8-2.6 1,550 330 30.0 0.588 0.795 14.00
38 .. 1.8-2.6 1 100 900 29.6 0.590 0.772 13.48
4 0.1 (FZ) 1 400 360 19.2 0.648 0.790 9.83
Emitter (0.30-0.5"m) (0.5-0.8"m) (0.8-1.12"m) (0.29-1.12"m)
Thickness Juv.blue L\Juv+bl. Jvl1lbl. 4.1,10. J" 4.1" Joe lIJoe
(i) (mAlcm2 1(Lo•• ) mAlcm2 (Lo.. mA/cm2) (Lou) mAlcm2 Loss) - ASEC conlrol cell with doubl•• ntlrenectlon coating
Control 2.43 1-) 11.53 1,-) 6.22 1(-) 20.18 1(---) ... These cells are co.'ed with 8 single SIOx antireflection coating
1,450 0.95 (1.48) 9.97 (1.56) 6.18 (0.04) 17.10 (3.08)
925 1.31 1(1.12) 10.99 0.53 6.41 -0.19 18.71 1.47) Measurements of open-circuit voltage were also
made on uncoated cells with the incident light
J se was also measured in natural sunlight in intensity adjusted to give a photocurrent density of
Ottawa near solar noon on clear days within three 32 mAcm- 2 , which should be achi'evable with a well-
weeks of the summer solstice. Values obtained were designed two-layer antireflection coating. At 28°C
normalized to the photocurrent measured in a high- a cell fabricated on 0.1 Ocm Wacker float zone
efficiency Applied Solar Energy Corporation space material gave a very high open circuit voltage of
cell which had been calibrated against a secondary 669 mV (Table 4) . This confirms that the
standard. Results ~re given in Table 2, and agree polysilicon emitter structure is very effective in
well with Table 1. suppressing hole back injection. Other cells
fabricated on the same material gave comparable Vo•
Table 2: Estimated AM 1 photocurrent density Jsc for 2 x 2 em values. Open-circuit voltages as high as 696 mV
poly-51/a-51 em Iller cells, based on the measured have been reported at 25°C for state-of-the-art
photoeurrent Ise of uncoated cells In natural sunlight. silicon cells giving photocurrent densities of
40 mAcm- 2 [1], but the lower test temperature and
higher photocurrent should account for a difference
Poly-SI Melsured Conver.lon E.tlm.lad AMI Tol.1 Lo.. of of at least 10 mV in Voe '
Em Iller Photocurrent AMl Factor Pholocurrenl Density Pholocurrenl Table 4: Measured open circuit voltage (Voe) and fill factor (FF)
Thlckne.s lac lac** n.c· J.c=lsc{llc"lIlc')/4 aJsc (AM1)
for cells illuminated to give a photocurrent density of
(A) (mA) --- (mA/cm2) ImA/cm2)
32mA/cm2 at a controlled temperatura of 28°C.
1,500 (e-SI) 88.8 142/132 23.88 -. --
1,440 75.9 142/132 20.41 3.47
925 86.5 142/138 22.25 1.63 Substrata Emitter Emitter
Resistivity Thickness Sheet Resistance Voc FF
In.em) tAl In/a) (Volts)
Isc' : control cell's (ASEC) measured pholocurrenl.
Ilc": control ce"'. (ASEC) AMl rated photocurrent. 1.8-2.6 1 440 270 0.595 0_808
0.5 (FZ) 1 360 450 0.625 0.800
The complete illwninated current-voltage 0.1 (FZ) 1 400 360 0.669 0.798
characteristics of the test cells were recorded
using a Kratos Xenon arc lamp solar simulator set up 4. CONCLUSIONS
to duplicate 100 mWcm- 2 AMl. 5 illwnination. A
simulator was used in preference to natural sunlight The use of recrystallized amorphous silicon in
for these measurements to ensure that the chuck a polysilicon emitter solar cell has been found to
temperature would be well-controlled. Results for lower the emitter sheet resistance sufficiently to
both coated and uncoated cells are swnmarized in allow good fill factors (close to 0.8) to be
Table 3, while a representative J-V characteristic obtained in combination with reasonable short-
is given in Fig. 4. The highest total-area circuit current densities (~30 mAcm- 2 with a single-
efficiency (14%) was obtained for a cell with a layer SiO antireflection coating), allowing an
150 nm thick emitter. A 110 nm thick emitter gave overall total-area efficiency of 14% to be reached.
a slightly lower fill factor and marginally lower A bilayer emitter structure conSisting of material
efficiency (13.5%). deposited in the polycrystalline phase overlaid with
a recrystallized amorphous layer was shown capable
50 .of providing very high Voe's- up to 669 mV at 28°C
N FF,,0.795
with J sc set at 32 mAcm- 2 • Emitters consisting only
< Jsc=30mAlcm~2
of recrystallized amorphous silicon gave slightly
E 40
lower open-circuit voltages, perhaps because there
U Voc=590mV Elfie lency=14,O"Io
~ is less disorder near the polysilicon!silicon
.s 30 interface and therefore less blocking of hole back
injection.
c
......
GI

::I
20 Although the above results are encouraging, it
U should be noted that the thinnest useful polysilicon
0 emitters produced here still discard approximately
0
.c
10 8% of the available photocurrent in the visible
portion of the AMl.5 spectrwn compared to a cell
Il.
\ with a shallow diffused monocrystalline emitter. It
°° 100 200 300 400 SOD 600 remains to be seen whether the simple, low-
temperature polysilicon emitter process and very
Voltage (mV) ~i~ open-circuit voltages this structure can
Fig. 4 Illwninated J-V characteristic for poly proyide will ol.ltweigh this photocurrent loss in
emitter cell with SiD A.R. coating. commercial applications.

29
ACKNOWLEDGEMENT

The authors wish to thank L.P. Berndt, A. Ruhmland

and C.A. Adams for their expert assistance in
processing. Financial support for this work was
provided by the Canadian Natural Sciences and
Engineering Research Council (NSERC).

REFERENCES:

[1] M.A. Green, A.W. Blakers, J. Zhao, A.M. Milne,

A. Wang and X. Dai, IEEE Trans. Electron Dev.,
ED-37, 331 (1990).

[2] M.A. Green, IEEE Trans. Electron Dev., En-3l,

671 (1984).

[3] M.A. Green, A. Blakers, E. Gauja. M. Willison

and T. Szpitalak, Proc. 15th IEEE Photovoltaic
Specialists Conf .• 1982. p. 1219.

[4] F.A. Lindholm, A. Neugroschel, M. Arienzo and

P.A. lIes. IEEE Electron Dev. Lett. EDL-~, 363
(1985).

[5] E.P. Keyes and N.G. Tarr. IEEE Electron Dev.

Lett .• EDL-~. 312 (1987).

[6] N.G. Tarr. IEEE Electron Dev. Lett., EDL-~, 655

(1985).

[7] D.F. Waechter and N.G. Tarr. Can J. Phys. 65,

1030 (1987).

[8] G. Harbeke. L. Krausbauer, E.F. Steigmeier,

A.E. Widmer, H.F. Kappert and G. Neugebauer,
RCA Rev. 44, 287 (1983).

[9] E.. P. Keyes and N.G. Tarr, Can. J. Phys. fl, 179
(1989).

[10] M.A. Green, Solar Cells: Operating Principles

and Technology (Prentice-Hall, Englewood
Cliffs) 1982.

[11] P. Ghate, C. Blair, C. Fuller and G. McGuire,

Thin Solid Films 53, 117 (1978).

30
 POSTER SESSION lA

Fundamentals, High Efficiency Cells

and Novel Devices
10TH EUROPEAN PHOmVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EMITTER-BASE COUPLING IN HIGHLY ILLUMINATED SOLAR CELLS

AND APPLICATIONS IN OCVD TECHNIQUE.

F. PELANCHON, E. BOEGLIN' and P. MIALHE"

CEDUST, Ambassade de France a Damas, B.P. 3929, Damas ~Syrie)
'ISSAT, B.P. 7028, Damascus (Syrie)
"Laboratoire de Physique du Solide,Universite de Perpignan,
Ave de Villeneuve, 66025 Perpignan (France).

ABSTRACT. The effect of emitter-base coupling hetween the excess carrier

concentrations at the junction during open-circuit voltage decay in a n+-p
silicon solar cell when including recombination processes in both emitter and
base under low or high injection condition, is considered. A general
resolution of the continuity differential equation is performed and manageable
expressions of minority carrier densities, for any cell structure, are
derived. The theory yields a description of time variation of the junction
voltage at the very beginning of the OCVD. Measurements of carrier lifetimes
with the OCVD method have to take into account this result.

1. INTRODUCTION

The ultimate efficiency of solar cells y

is expected to be limited by so-called (

high injection effects. They appear with

large minority carrier concentrations
which are related to an increase of the -x H
stored carriers at the junction.
Recombination in the space charge layer is
increased. The majority carrier Figure 1: Representation of a solar cell.
concentration is disturbed and a reverse X (H) is the emitter (base) width.
current flows into the emitter. As a
result, the excess carrier concentrations
in the emitter and in the base are not 2.1 Equations
independent of each other. Consequently
the study of the p-n coupling between the We consider an abrupt junction with
excess carrier concentrations at the uniform doping in the quasi neutral
junction on both of its sides is of regions of the cell. The densities n(x,t)
importance. and p(y,t) verify the differential
The aim of this work is to give a continuity equations which read
description of the effect of the p-n during the initial steady state (t,O):
coupling on the decay of the minority
carrier density in the base of a solar 6 2n(x,O) n(x,O)
On. = -C.G(x) (1)
cell, in the first stage of an open 6x 2 'n
circuit voltage transient event. Until
now, descriptions have been given /1-3/ 6 2 p(y,O) p(y,O)
Dp • = -C ..G(y) (2 )
based on'the assumption that the minority 6y2 'p
carrier profile in the emitter remains
quasistatic /4/. This assumption may not
be applied for very short time values where Dn and the diffusion
because of the fast decay rate of carriers coefficients for electron and hole, 'n and
at the junction dn the emitter side. 'p are the electron and hole lifetimes,
G(x) is the photon generation rate. A
first order approximation of function G(x)
2. THEORY has been given 15/.

A n+-p silicon solar cell is During the transient event (t20+), the
considered in open circuit operating time variation of carrier concentration
conditions. The carrier injection is appears and the generation rate vanishes,
obtained with an applied steady photon consequentLy the continuity equations
flux (illumination level: C suns). reads:
At time t=O, the illumination is
abruptly terminated and we study the 6 2 n(x,t) n(x,t) 6n(x,t)
variation vs time (t~O+') of the excess Dn.---- = (3)
minority carrier densities n(x,t) in the 'n 6t
base and p(Y,t) in the emitter (Fig. 1). 6 2 p(y,t) p(y,t) 6p(y,t)
Dp .---- = (4)
6y2 6t

33
 This study considers a conventional solar
cell with a back ohmic contact and an
ideal front surface,
boundary conditions:
leading to the
J i
N2 x,t)

e
-Tn.U
=
rl r2
. [sl. [erfc(-) + erfcC--)]
2 • .[U 2 . .[U

n(R,t) =0 ( 5)
o
1'3
- SI. [ r3.erfc{--)
6p(-X,t) 2.JU
6y
=0 ( 6)

du
We now need to specify two boundary
( 11)
conditions at the p-n junction. One of
and,
them is the current condition. In open- -(c,X)
circuit conditions, the total current di=bi .e (i=I,3)
through the junction is equal to zero:
cl=6630 em-I, c2=1000 cm- I , c3=130 cm- 1
6n(0,t) 6p(0,t)
Dn. - Dp • = 0 (7)
bl=6.13 10 20 cm- 3 s- I ,
6x 6y
b2=0.54 10 20 cm- 3 s- I ,
Equation (7) relates the gradients of the b3=0.0991 10 20 cm- 3 s- 1

excess carrier profiles at the junction on

R-x R+x 2H-x x
both of its sides. The other boundary 1'1=--, r2=--, r3=--, 1'4=--
condition is a voltage condition: it lIln /Iln Wn ,fTJn
relates the excess carrier concentrations
at the junction to the junction voltage 3 -(ciR) 6n -
:E di.e , S2= -/Iln.-(O,O),
VItI in the device. From classical solid i ' I 6x
state physics theory: 3
n(O,t) z (ni 2 /NA).exp(V(t)/VT) where ni, S3= :E Ci .d;
i • I
VT are usual notations and NA is the base
doping level.
We assume that the junction voltage is
described by Fletcher's condition 12,3/.
3. RESULTS
We obtained:

n(O,t)/p(O,t)=n(O,O)/p(O,O)=a.No/NA Vt (8) Figures (2) and (3) display the

variations of the relative minority
where a=l in low injection condition and carrier densities in the base
a=1/2 in high injection condition, and No n(x,t)/n(O,O) at different values of the
is the emitter doping level. time t. A low injection level is
considered in Fig. (2) with AMI illumina-
2.2 Resolution tion (Vo.,=620 mV). The profiles show a
maximum which decreases with the time. The
Equations (3-4) with the conditions initial large reverse base current becomes
(5-8) form a coupled system which relates practically zero in a time of the order of
the excess carrier concentrations on the the minority carrier lifetime in the base.
two sides of the junction. A way to obtain A high injection level is obtained with a
the general solution is to take the 50 suns illumination (Fig. 3). A larger
Laplace transforms of equation (3-8) with base reverse current is pointed out with
regards to t. This converts the coupled an increase of the maximum, the quantity
differential equations to ordinary of stored carriers is larger. In this
equations which can be easily solved. In case, also, the number of recombinations
this general case, the inverse Laplace in the emitter decreases as time
transform does not exist. We avoid this progresses and a "normal" decay rate
difficulty when using the following appears at a time of the order of the base
approximations. First, we assume the lifetime.
doping condition No ~ 100 NA and we sim-
plify the Laplace transform solutions
(N(x,u), P(y,u» for the time range
[0+, 10- 5 s] (that is 10 5 < u). The in-
verse Laplace transform is then available.
We obtained the excess minority carrier
concentration in the base as:
n(x,t) = n(x,O) + Nl(x,t) + N2(x,t) (9)
with:
-Tn.t 3 dl -Ci.X
Nl (x, t) = (e 1).:E .e (10)
i::l Tn

34
 Transient excess
junction voltage (V)
0.01 T

~.O~ jk-----------',---.
"'-",-
"'-"'-,

j
~.02 "

~.03 "'-"

~.05 L
~.04 . "'-
o 150 300
Xillm) I "'-"'-,

o 2 3 4 5
Figure 2·: Relative minority carrier
densitie.s in the base n(x,t)/n(Q,O)at
different values of time t, with AMI
ini tid ill ulIlina t ion. Figure 4: VariatioJ;:ts of the trans-l,ent ex-
NA=10 16 cm- 3 , ND=10 19 cm- 3 , X=0.2 11m, cess junction voltage Vao(t)=V(t)-V(t~O).
H=300 11m. NA=10 16 em-a, ND=10 19 cm- 3 , X=0.2 11m,
H=300 J,lIIl, AMI initial illumination.

4. CONCLUSION
Manageable expressions of the minority
carrier densitiee in the emitter and in
the base of a n+-p silicon solar cell
during OCVD have been derived. The
emitter-base coupling and !'ec;ombination
processes in the quasi neutral region, of
the cell have been considered. The res'Jlts
show an i.ncrease of the voltage at the
junction d'.lring the beginning of the
decay. The ef.fect of toe coupling on OCVD
disappears atter a time of the order of
minority carrler base lifetime. This
result imposes cOild;i.tions for experiments
in OCVD to d.-rive meaeu.remente of carrier
lifetime.
Figure 3: Relative minority carrier
densities in the base n(x,t)/n(O,O) at
different values of time t, with an REFERENCES
initial illumination of 50 suns.
NA=10 16 cm a ND=10 19 cm- a , X=0.2 11m, III S.C. JAIN and U.C. RAY, J. Appl. Phys.
H=300 11m. ~~, 2079-85 (1983).
12/ V.K. TEWARY and S.C. JAIN, Adv. in
Electron. and Elect. Phys. .!i.?, 329-4'14
With operating conditions either at low (1986).
or high injection levels, the 131 S.C. JAIN, U.C. RAY, R. MURALIDHARAN
recombination rate in the base is and V.K. TEWARY, S'olid-State Electron .~.9.,
increased by the emitter-base coupling. 561-70 (1986).
Figure (4) shows the increase of the 141 S.C. JAIN and R. MURALIDHARAN, Solid-
initial junction voltage V(t), since the State Electron. ?~, 1147-54 (1981).
transient excess junction voltage 151 S.N. MOHAMMAD, Journal of Appl. Phys.
voc(tt=V(t)-V(tSO) is positive in the §L 767 (1987).
first stage of the OCVD curve. The voltage
decay depends on the emitter-base
coupl ing. Nega ti ve values of V'o c ( t ) show
that this effect disappears at the time
t)'tn.

35
l(JfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

LIFETIME AND SURFACE RECOMBINATION DETERMINATION

P. MIALHE", J.M. SALAGNON and F. PELANCHON

ISSAT, B.P. 70 28, Damascus (Syrie)
'Universite de Perpignan, Ave de Villeneuve, 66025tPERPIGNAN
G. SISSOKO and M. KANE
Universite de Dakar, Dakar-Fann, Senegal.

ABSTRACT. An analysis of the theory and a new experimental practice of the

transient short circuit current decay of solar cells has been performed. The
injection level was produced by illumination and preserved during the
experiments. All recombination processes have been conside~ed. An accurate
method for determining the minority carrier lifetime and the back surface
recombination velocity has been derived. Sensitivity and precision of the
method are shown to be related to the knowledge of structural parameters and
the injection level value.

1. INTRODUCTION 2. THEORY

Experimental methods have been used to The transient short ci.rcuit current
determine the carrier lifetime in the base (I.e) can be computed from the analytical
of a solar cell. For these methods a expression of the excess minority carrier
steady state is abruptly terminated I I I concentration, n(x,t) in the base of the
and the transient response is measured. solar cell:
The discrepencies among results available
with presently used techniques for I.e = q.D.6n(x,t)/6x, (1 )
measuring 121 carrier lifetime are well ~=o
known, They come from theoritical
interpretations which differ from both, The junction plane is situated at x=O and
experimental conditions 13,41 and cell the back surface at x=H, D is the carrier
structures 15,6/. Additionally, diffusion coefficient. n(x,t) is a
experiments do not consider actual solution of the time dependent diffusion
operating conditions and are performed by equation in the base:
using an applied bias or a variation of
carrier generation rate. This work D.6 2 n(x,t)16x 2 + G(x)
presents a new method for accurate - n(x,t)/~ - tUi = 6n(x,t)16x ( 2)
measurements. The method is based on a
transient measurement of the short circuit including carrier generation rate G(x),
current of a solar cell operating under linear recombination rates used to define
real conditions, refered to as Constant carrier lifetime ~ and any recombination
Illumination Short Circuit Current Decay processes Ui. No analytical solution of
(CISCCD) method. equation (2) is available with G(x) and Ui
A solar cell is illuminated with a functions of x or n(x,t).
steady photon flux (Fig. 1), operating During the transient event, and because
point O. At time t=O, a fast switch of the CISCCD procedure, the modification
removes ~he load R and a small external of the minority carrier density profile in
resistor r, used to monitor the transient the base is expected to be negligible
current, nearly provides a short circuit since the output current value remains
(point S). The transient terminal voltage close to the same value (Isc). A carrier
is monitored by using a data acquisition flow follows the junction barrier re-
system including a computer. adjustement imposed by the fast switching
at time t=o within a dielectric relaxation
time shorter than 10- 5 T, The carrier flow
I l/R+r is produced by an excess carrier
concentration q(x,t) which may be refered
o to the final steady state carrier density
ns (x):

n(x,t) = ns(x) + q(x,t). (3 )

o -.--~-
Since operating conditions (illumination
... ~ Acquisi tion and injection level) are not varied, it
can be assumed that 6Ui/6n is not

~ q]
significant during the transient event,
ce~ therefore, function q(x,t)
of the differential equation:
is a solution

D.6 2 q(x,t)16x 2 -q(x,t)/T : 6q(x,t)/6t. (4)

Figure 1: Experimental set up and
operating points 0, S. A detailed analysis of the solutions of
equation (4), with short circuit boundary

36
conditions, has been discussed 171; it The value of 0 is obtained from the
leads to the time variation of the intercept of F'.e(O) graph with the line
transient current, at time t>., in the i(O)/no(O). This representation enables a
form: determination of the error AO coming from
experimental uncertainties in the
itt) = no(O).FTae(O).exp(-tlToe) (5) measurements of the junction voltage VJO,
of the decay time ,Be and of the value
(02.D)/Tae itO): we get respectively AO=(10, 13, 18)
F'se(O)=e. (6) S-1/2 for A'Be=0.2 us, Ai(O)=O.l rnA and
H/2-(1/4)./D/O.sin(2(01ID)H) AVJo=5 mY. Therefore, the accuracy of VJO
measurement is of prime importance: the
1/'oe = 0 2 + 1/. ( 7) junction voltage differs from the cell
output voltage by the series resistance
tg(O.HIID) = -/D.O/S, .12 < O.HIID S x (8) voltage drop. For this experiment Rs=O.5
Ocm 2 , 100=10 mAcm- 2 and the voltage drop
where S is the back surface recombination is 5mV, that we have considered in figure
velocity and H the base width. 2. This practice of the CISCCD method need
A numerical analysis of the registered an accurate measurement of the cell series
itt) transient curve leads to the resistance and a low injection levels.
determination of the values 'oe and itO).
The initial and final steady states define S(Q) T(Q)
nolO) and ns(O). Parameter 0 may be cm/s I.1S
T(Q) :
computed from equation (5), consequently,
equations (7) and (8) yield the values of
lifetime. and recombination velocity S.

3. DISCUSSION 11

From these results, 0 appears as the

key parameter of the method, accordingly,
the accuracy of its determination is of
importance: it is illustrated in figure
(2) with a graphical representation.
Fe.o (Q)
a.u. Q parameter
290 580
Figure 3: Plot of minority carrier
lifetime and back surface recombination
velocity as a function of parameter Q. The
experimental value of 0 is indicated.
Figure (3) illustrates the
determination of , and S. The large
variation of T vs 0 displays a high
sensibility of this determination. It
shows that an error ll.Q=20 S-1/2 in the
measurement of parameter Q leads to ~he
experimental errors iH= 1. 5 ~fS and AS=300
il parameter cm/s. At very low or high values of Q, an
580 indetermination appears for the
290
measurements of S or T: this is the ca~e
Figure 2: Parameter 0 is determined from of high performances BSF type solar cells.
the intercept of the graph (dashed lines) Results are given in table 1 for a
of function F,sc(O) with the straight line commercial solar cell.
1(0)/no(0). Experimental values are 3.4 us
and 1.9 rnA, the upper line considers a 5
mV error in the determination of the
junction voltage for operating point O.

37
 R. A Ir 'toe i(O) 't S
W) ( itA) (itS) (mA/cmll) (S) (cm/s)

0.17 1.6 1.5 3.4 1.9 9.5 2300

(3 ) (350)
Table 1 : CISCCD measurements (precision
of the data) . (A: quality factor; Ir:
saturation current of the cell )
As the experiment is automatized, the
precision is obtained from a computer
process taking into account incertainty in
the base doping level.

4. CONCLUSION
We have given an analysis of a new
practice of the short circuit current
decay method of lifetime and surface
recombination velocity measurements.
Compared with previous methods, a constant
illumination level is used which allows
for experiments with an operating carrier
distribution. Accurate values are obtained
and a good preC1S10n is shown to require
low injection levels or a proper knowledge
of structural parameters.

RgFERENCES
/1/ S.R. DHARIWAL and N.K. VASU, Solid
State Electron. ~4, 915-27 (1981).
/2/ A. VISHINOI, R. GOPAL, R. DWIVEDI and
S. K. SRIVASTAVA, Solid-State Electron . .3. .:2.,
17-24 (1989).
/3/ P. MIALHE, G. SISSOKO and M. KANE, J.
Appl. Phys. ~Q, 762-5 (1987).
/4/ M.A. GREEN, Solar Cells ll, 147-61
(1984)
/5/ JUNG T., LINDHOLM F.A. and NEUGROSHEL
A., Solar Cells ~?, 81-96 (1987).
/6/ A. ZONDERVAN, VERHOEF L., LINDHOLM
F. A. and NEUGROSHEL A., J. Appl. Phys. 1>3,
5563-70 (1988).
17/ J.M. SALAGNON, S. MOUHAMMAD, P.
MIALHE, F. PELANCHON, Solar Cells (to be
published)..

38
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

REDUCTION OF REFLECTION LOSSES BY STRUCTURED SURFACES

A. Scheydecker, A. Goetzberger and V. Wittwer

Fraunhofer-Institut flir Solare Energiesysteme

Oltmannstr. 22, D-7800 Freiburg
Federal Republic of Germany

ABSTRACT: Structuring the transparent cover of solar cell modules reduces

reflection losses particularly at large angles of incidence. Relevant aspects are good
transmission efficiency independent of wavelength and a low sensitivity to pollution.
The macroscopic, linearly grooved structure proposed in this paper shows good
performance only in combination with a textured cell, because large angles are
likely to occur inside the structured cover.
A classification is made with the concept of annual averaged transmission
efficiency for the climatic zone of Freiburg. Calculations and measurements for
different combinations of smooth and structured covers and solar cells are
presented. From the calculated 97.8% entering the structured glass cover, a
measured 93.2% can be coupled into a pyramidal textured monocrystalline solar
cell. This is an absolute improvement of 17% compared to a smooth, uncoated
solar cell with a smooth glass cover.

1. INTRODUCTION reduce reflection. A glass COver (n = 1.5), for instance, not

only protects the solar cell against damage, but, assuming
A smooth, uncoated silicon surface reflects about one that there is good optical contact to the cell, also divides the
third of the incoming sunlight at normal incidence. This large change of refractive index into two smaller ones, and
reflected light cannot take part in the conversion process reflection losses are reduced to about 20 % (without an
inside the solar cell and so reduces the solar cell efficiency anti-reflection coating). Now, extending this simple model, a
dramatically. Reduction of reflection losses is thus a very structured surface has an average refractive index, which
important task in solar cell technology. rises continuously with depth and so further reduces
Transparent anti-reflection coatings are commonly used reflection losses.
in present-day solar cell fabrication to reduce reflection In this paper, both the structuring of the surface of the
losses. By choosing the right refractive index and thickness, solar cell and that of the transparent cover is examined.
reflection losses can be totally suppressed for one
wavelength (usually 600nm for silicon solar cells) and one
direction of the incoming light (usually normal to the solar 2. BASIC PRINCIPLE
cell surface). The unwanted dependence on wavelength and
direction can be reduced but not avoided combining two or On a structured surface, incoming light rays are
more anti-reflection coatings /1/. Reflection losses of 4% reflected several times, giving them several chances of being
for usable sunlight at normal incidence can be obtained. refracted into the material. The simplest case to think of
Light reaching the solar cell surface is reflected due to would be a randomly roughened surface. Such a surface is
a sudden change of the optical density or refractive index. not suitable for a cover of solar modules as pollution, e.g.
Smoothing this step in the refractive index function will dust particles, would give rise to an inverse effect, i.e. an
100

trensporent cover 95

90

T/% 85

80

75

70~U-~~~~~~~-+---+---+--~~~
-90 -70 -SO -30 -10 10 30 50 70 90
Angle of Incidence in degrees
Figure 1: Grooved structure on a transparent cover of a solar Figure 2: Transmission vs angle of incidence for a structured
cell surface (perpendicular (1) and parallel (2) to the grooves)
and for a smooth surface (3).

39
increase of reflection and absorption losses. Possible angles
of incidence of direct solar radiation onto a stationary
surface tilted towards the south by the geographic latitude
angle are _90° < a < 90° in the east-west and
-23.45° < a < 23.45° in the north-south direction respectively
(a being the angle of incoming light rays with respect to the
normal to the surface).
It is desirable to find a structure that transmits light
falling within these angles and is not sensitive to dirt. We
suggest the structure shown in figure 1. It is a structure of
arbitrary but macroscopic scale,so that geometric optics is
valid. This ensures - ignoring dispersion effects -
independence from the wavelength.
The aperture angle y should on the one hand be as
small as possible to ensure several reflections for incoming
rays from all directions. On the other hand, light that has
entered the glass should not be refracted outwards again.
So, a light ray coming from the most extreme direction of
incidence, i.e. a =90°, must undergo total internal reflection Figure 3: Angular distribution behind the structure, summed
after being refracted into the glass (figure 1). This leads to over all possible angles of incidence. u and v are the
the following formula for the aperture angle y: direction-cosines with the v-axis perpendicular to the
grooves.
l/n = sin ( 2 • y - arcsin ( sin y / n»

For a refractive index of n = 1.5, y = 32° is found. Three different solar cells were useo, all being
mono-crystalline silicon cells, 2·2 cm2 in area. One has a
smooth surface while the others are textured cells fabricated
3. PERFORMANCE in our institute. Selective etching of < 1,0,0> surfaces with
dilute KOH after photolithographic treatment leads to
Using a ray-tracing procedure, we calculated the < 1,1,1> planes tilted 55° to the surface /3f. One of the two
transmission of such a structured surface in dependence of textured cells has a structure similar to that of the cover.
the angle of incidence of incoming light rays. The results are The microgrooves are about 9j.1m apart and oriented
presented in figure 2. The curves parallel and perpendicular perpendicular to the lines of the top metal contact grid. The
to the grooves of the structure and that of a smooth surface other has square based pyramidal holes, again about 9j.1m
are plotted. A noticeable increase of 3.7% at normal apart.
incidence to almost 100% (99.7%) can be obtained in To ensure good optical contact of the acrylic glass
theory. As the curve perpendicular to the grooves, plates (n= 1.49) to the solar cells, and to be able to remove
corresponding to east-west directions, does not fall below the cover from the cell, silicon oil (n=l.4) was used.
95% even for large angles, direct solar radiation will be The measurements were made using light from a 150W
transmitted through the surface effectively throughout any halogen light source, l.4m away from a sample holder,
day of the year. which can be rotated in two directions with stepping motors.
One has to bear in mind that all transmission values Accuracies of 2° in direction and of 2% in homogeneity over
presented in this paper have to be multiplied with the cosine the cell area were obtained. Reference measurements with
of the angle of incidence to obtain realistic values, because a photodiode ensure equal conditions for all measurements.
only that part of the solar cen area is seen by the incoming As the short circuit current is proportional to first
light ray. order to the light transmitted into the semiconductor
The angular distribution beyond a smooth surface, material, it is a good measure for transmission values.
inside the cover, is restricted to angles inside a cone with an Absolute values were obtained by comparing the short
opening half-angle of 42° due to Snell's law of refraction. circuit current with reflection measurements made with a
This is not the case for a structured surface. Figure 3 shows goniometer in our institute.
the angular distribution behind the structure (y =43°)
integrated over all possible angles of incidence.
From this it can be seen that solar cells with absorption 5. SOLAR RADIATION DATA
coefficients almost independent of the incoming direction
are needed, to ensure good transmission through the For comparison and classification of different
structured cover into the solar cell. This could be· achieved configurations of the solar cell and the cover, it is important
with anti-reflection coated cells, but a more convenient way to take realistic solar radiation data into account. This was
is to structure the solar cell surface too, i.e. to use textured done for solar radiation data of a typical year in Freiburg.
cells. Figure 4 shows the solar radiation (in W/m2/st)
(st=steradian) summed over one year on a surface tilted by
48° (the geographical latitude) towards south. Again u and
4. MEASUREMENTS v are the direction cosines. Each point in the u,v-plane
represents one direction of incoming light (or point of the
An acrylic glass plate with the above described sky dome) to which a value of yearly summed radiation is
structure, manufactured by 3M /2/, was used for associated.The sum over all angles gives a total annual
measurements. It was originally designed for light trapping irradiance of 1156 kWh/m2/a. This curve was calculated
in wave guides. The grooves are 0.3 to 0.4 mm apart. The using data from the Test Reference Year (TRY) /4/. One
aperture angle of y =43° is not ideal in the sense described can easily distinguish the direct radiation from the two
above, but is still close enough to give reasonable results. plateaus of diffuse radiation and ground reflection.

40
 c:::!
~
w
c:::! N
$
r.:$c:::!
~~
~
"pc:::!
~
E
c:::!
0

Figure 4: Annual global radiation for Freiburg in dependence on the angle of incidence on a south-oriented surface, tilted 48°.

Convoluting this curve with measured or calculated A comparison of figure Sa and figure Sb shows that the
transmission versus angle of incidence data leads to a annual measured curves behave as predicted. The textured cell
average transmission efficiency. This value allows a realistic succeeds in collecting all rays up to large angles of incidence
classification of the transmission performance of any solar compared to figure 2. The accuracy of the measurements
module to be used in Freiburg. decreases with increasing angle, because the absolute signal
is lowered proportional to the cosine of the angle of
incidence, as mentioned above.
6. RESULTS Table I shows calculated and measured annual average
transmission efficiencies. The first line gives the value for
Figure 5 gives three measured curves (fig Sa) and the the structured acrylic surface alone, i.e. with ideal coupling
corresponding calculated curves (fig. Sb). Curves 1 show the into the solar cell. This value is approached more and more
transmission through a smooth cover into a smooth cell. closely as the structure of the solar cell surface is improved.
Curves 2 and 3 show the transmission through a structured In the end an absolute improvement of about 17%
surface into a microgrooved, textured cell, perpendicular and compared to the unstructured and untextured (and
parallel to the grooves of the structured cover. The grooves uncoated) case is obtained. It is notable that a textured cell
of cover and cell are parallel to each other. in combination with a smooth glass cover shows good

tOO
tOO
3 3
90
90 2 2
BO
60
70
70
60
60

T/% 50 T/% 50
40
40
30
30
20
20
to
to
0
0 -90 -70 -50 -30 -to to 30 50 70 90
-90 -70 - 50 -30 -10 to 30 50 70 90
Angle of incidence in Degrees Angle of incidence in Degrees

a b

Figure 5: Measured (a) and calculated (b) transmission vs. angle of incidence for a smooth cover with a smooth cell (1) and
a structured cover (perpendicular (2) and parallel (3) to the grooves) with a microgrooved textured cell.

41
 calculated measured
entering through the structure into the glass 97.8
smooth glass with smooth solar cell 77.1 75.6
structured glass with smooth solar cell 79.2 81.3
smooth glass with microgrooved soar cell 91.5 86.9
smooth glas with pyramidal textured solar cell 89.0
structured glass with microgrooved solar cell
grooves parallel 90.5 90.8
grooves perpendicular 92.4 92.7
structured glass with pyramidal textured solar cell 93.2

Table I: Annual average transmission efficiencies in % for different configurations.

performance, despite the fact that at large angles of

incidence a great amount of radiation is reflected at the
smooth surface of the <;over.

7. CONCLUSION

Reduction of reflection losses was achieved with

structured surfaces over a wide range of angles and
independent on wavelength. The concept of annual average
transmission efficiency was used for classification. The
structure of the transparent cover was designed to be kept
clean by rain-water and to collect direct solar radiation
effectively. A combination of the structured cover and
pyramidal textured solar cell yields the best results.
Calculating the annual average transmission efficiency for
other radiation conditions may lead to different results, but
will not alter the basic conclusions. Outdoor measurements
are to be carried out to confirm the results reported here
and to answer the question on how well the self cleaning of
this macroscopic structure by rain-water works.

REFERENCE

/1/ M. A. Green, Solar Cells, UNSW, 1986

/2/ S. G. Saxe, D. L. Wortman, 3M Optics Technology
Center, Minnesota USA
/3/ A. W. Blakers, 22.6% Efficient Solar Cell, 4th IPS&E
Conf. Sydney, feb. 1989
/4/ K Blumel et.al., Entwicklung von TRY rur
Klimaregionen der BRD, BMFT FB-T86-0S1

42
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THE I-V SILICON SOLAR CELL CHARACTERISTIC PARAMETERS TEMPERATURE

DEPENDENCE.AN EXPERIMENTAL STUDY USING THE STANDARD DEVIATION METHOD
N. VEISSID
Instituto de Pesquisas Espaciais-Laborat6rio Associado de Sensores
e Materiais, Phone:(123)41-8977, ext. 308, Fax:(123),
P.O. Box 515, 12201, Sao Jose dos Campos, SP-Brazil
and A. M. DE ANDRADE
Laborat6rio de Microeletronica-Escola Politecnica da Universidade
de Sao Paulo, Phone:(ll)8l4-6404, Fax:(11)211-4308,
P.O. Box 8174, 01051, Sao Paulo, SP-Brazil
ABSTRACT. This paper describes an experimental study of the
fundamental parameters (I L, 101 , 1 02 , A, Rs and l\)Of the
space qualified silicon solar cells with respect to the
temperature on I-V curves obtained in the dark and under
illumination. The values of those parameters and respective
standard deviations were obtained with the use of the
least-squares method, where chi-square minimization was
chosen as a criterion for the experimental curves fit.

1. Introduction 3 q.Vc
Two methods are commonly used to treat Cl·T .exp( - k.T ) (3)
the single crystal solar cells
current-voltage (I-V) curve
characteristics variations with the (4)
temperature. One method makes use of
linear coefficients relating the I-V curve
parameters with its variation with the where Cl and C2 are constants. The
temperature. The output I-V curve relations, obtained from equations (3) and
parameters Isc' Voc ' Imax and Vmax are (4), used to fit a linear regression and,
then, considering, linearly dependent consequently, calculating the fit
functions of the temperature [1]. constants (COl' CO2 ' VCl and VC2 ) are the
The second method uses a physical followings.
model, represented by an analytical
function, to describe the variation of the J OI q. VCl
current as a function of the voltage for In( T3 ) = In( COl) - k.T (5)
dark and illuminated solar cells. Equation
1 shows this function for the two-diode
model [2]. Where the parameters have their J 02 q.V C2
usual meanings. In ( ) = In ( CO2 ) - - - (6)
T3/2 2.k.T
I "" IL - I01·{exp[k~T(V+I.Rs)]-1} - NOW, we have saturation currents
densities and the new constants are Cl and
C2 divided by solar cell area.
(1)
For each temperature, the fitting of
the experimental I-V curve was done with
In this second method, the I-V curve the use of the method of the least-squares
temperature dependence is a function of in equation 1, where chi-square
the parameters, which are temperature minimization was chosen as a criterion for
dependent, of the characteristic curve the experimental curves fit. with this
(I L, 1 01 , 1 02 , A, Rs and 1\). Wolf et al. method it is possible to determine the
standard deviations of the obtained
[3] proposed in their work that saturation parameters values [5].
currents 101 and 102 are n~ and n 1
proportional, respectively. The n 1 is the 2.Experimental results
intrinsic concentration of the The I-V curves were obtained for any
semiconductor material and is temperature space qualified silicon solar cells at the
temperatures around 5'C, I5'C, 25'C, 35'C,
dependent. 45'C, 55'C, 65'C and 75'C under
Slotboom [4] proposed an expression for
illumination and in the dark condition. In
the variation of n: with temperature order to make the study of this article,
four space qualified silicon solar cells
3 q.Vc were removed at random from a lot produced
n: C.T .exp(- k.T ) (2) at the Microleletronic Laboratory of Sao
Paulo Uni versi ty . These lot was
with the value of q.v c "" 1.206 eV. manufactured to provide the Solar Cell
In this work, we will adopt equations 3 Experiment of the first satellite of the
Brazilian Full Space Mission [6] and the
and 4 for the variation of 101 and 102 devices have n+-p structure single crystal
with the temperature silicon s,?lar cell with a total area of
3 • 94 cm • They were prepared by

43
 -12. 0 .-------r'-----r----~--__,

conventional phosphorus thermal diffusion, • - dark o - illuminated

Ti/Pd/Ag front metallization, Al/Ti/Pd/Ag
back metallization and with a layer of
Ta 20 s like antireflection coating. The
four solar cells number 18, 41, 62 and 69
were used in this work. By way of
illustration the figures refers on sample -18.0
number 69. Figure 1 shows the temperature
dependent values of the photogenerated ....
current density. The saturation current ~ -20.0
density due to the diffusion of minority .....
carriers is showed in the figure 2. Figure
3 shows the saturation current density due ~ -22.°1~5 17 IU 20 21.5
recombination and generation of carriers 1/(2.k.Temperature) (lIeU]
through defect centers.Figure 4 shows the
temperature dependent values of the diode Figure 3 Saturation current d'lnllity due
quality factor obtained in the fitting. recombination and generation of carriers
Figures 5 and 6 show the values of series through defect centers, under illumination
resistance and parallel resistance and dark condition.
logarithm temperature dependence with 2.ror----r------~----~----~
their standard deviations.
e • - dark ° -illuminated

t--f-£-f_
HO
~
u..

-
k
•- 1=---1_
~ 2.10
t-r--j
. ~

-~
tU
~ 1.80
;1.0 §-f-f-~
° - illuminated 1.50
;0.0 i:5
N 1.20 0
e, 39.0
~.-!-o-~-.~
20 +0
Temperature (Oe)
60 80
i
'-'
38.0 0- Figure 4 : Diode quality factor
........J temperature dependence, under illumination
~ and dark condition.
37.0
0.20
36.0 0 ;0 • - dark ° -illuminated
~
20 60 80
Temper ature [OC] 0.16
<II

1--1--
Figure 1 : Photogenerated current density
...~
1--,--
temperature dependence, under illumination 0.12
'"

!f::::!:::r='-Y-Y-f-f-y
using Oriel solar simulator (ANO).
£ 0.08
x:
i 0.0;

0.00 0 20 ;0 60 80
Temperature rOC)
Figure 5 : Series resistance temperature
-22.0 dependence, under illumination and dark
~ • - dark o - illuminated condition.
M'"
~ -26.0
•
('o.j'

I I I I
U mate J
1.1
-30.0 25.0
'(
! 20.0 -
• - dark ° - 1-llIJm! -
M -34.0

" 5
..... -38. 0
~
I~O

~I - -I - ! =f ~r ::;:: .
-
10.0 -I "iO
-;2.0 31 3; 37 ;0 ;3
1/<k. Temperature) [lIeU) r- -
Figure 2 : saturation current density due I 1 I
to the diffusion of minority carriers, 0.0 0 20 ;0 60 80
under illumination and dark condition. Temperature [OC)
Figure 6 : Parallel resistance logarithm,
under illumination and dark condition.

44
 The values of current-voltage this work a linear function is proposed
parameters (I L, 1 01 , 1 02 , A, Rs and Rp) for the series resistance temperature
were obtained with the use of the dependence, the equation B. Similarly,
least-squares method, where chi-square equation 9 is proposed for the diode
minimization was chosen as a criterion for quality factor. Also, the logarithm of the
the experimental curves fit, see equation parallel resistance shows to be a linear

It {
(7) • function of the temperature, the equation
I ml - I(V I ) }2 (7)
10.
A = Aamb • [ 1 - CA. (T-300) ] (B)
X2 =
51
Rs = Rs • [ 1 + CRs' (T-300) ] (9)
1m I
is the experimental current value ....b
measured at the VI voltage, I is the value
In(Rp) = In(Rp ). [ 1 - CRp ' (T-300) ] (10)
of the calculated current from equation amb
(1), N is the number of points taken from
the I-V experimental curve and 51 is the The equation 11 shows the equation used
accuracy of the measured current. to fit the variation of the photogenerated
The values of standard deviations of current with the temperature.
these parameters were obtained using the IL = IL • [ 1 + CI • (T-300) ] (11)
method from the article already published amb L
[7]. In this method, a Taylor series
expansion of the parameters, around their Table 1 to 4 shows the values of the
best values, is made resulting in linear parameters obtained in the fit of
functions which permit the determination equations 1, 4, 5, 6, 7 and B for the
of the standard deviations with the solar cells number IB, 41, 62 and 69,
least-squares method. respectively.
This routine to obtain the parameters By the comparison of the value of
and respectively standard deviations was tables 1 to 4, it is seen that the
applied at I-V characteristic under parallel resistance values is obtained
illumination and dark condition. with a better accuracy from dark condition
I-V curve measurements rather than from
3. Discussion illuminated ones.
The saturation currents showed a
temperature dependence which can be
described by the equations 3 and 4. In

Table 1 Values of the fundamental parameters for cell number 18.

Equation Parameter Illuminated Dark Condition

IL [mAl 137.67±0.22 ---

11 ....b
CIL[%/K] 0.0859±0.0067 ---
X2 15.7 ---
In(C01 [mAjK3/cm2] ) 9.9BtO.51 11.60±0.14
5
VG1 [V] 1.162±0.014 1.1913tO.0037
2
X 4.19 25.2
In (C02 [mA/r/2/cm2] ) -0.67±2.26 -1.92±l.B2
6
VG2 [V] O.B5±O.12 O.825±O.099
2
X 3.54 0.105
Aamb I.B79tO.038 2.201±0.045
8
CAE %jK] -0.471±0.091 -0.26±0.10
2
X 2.21 0.173

Rs [Ohm] 0.OB47±0.0067 0.0661±0.0075

9 amb
CRs[%jK] 0.61±0.32 1. 76±0. 39
X2 9.83 2.86
In(Rp [Ohm]) 10.4±5.2 11.5l±0.47
10 amb
CRp[%jK] -0.66±1.66 -0.31tO.14
X<l 0.0515 0.825

45
Table 2 Values of the fundamental parameters for cell number 41.

Equation Parameter Illuminated Dark Condition

IL
amb
[mAl 143.50±0.22 ---
11
CIL[%jK] 0.0514±0.0065 ---
xi! 17.3 ---
In(COl [mA/r/cm2] ) 7.46±0.71 11.36±0.13
5
VCl [V] 1.103±0.018 1.1905±0.0034
2 0.627
X 6.99

In(C02[mA/K312/cm2] ) -0. 94±1. 41 -4.26±0.74

6
VC2 [V] 0.840±0.075 0.706±0.039
X2 9.80 1.37

Aamb 1. 902±0. 022 2.208±0.022

8
CA[%jK] -0.315±0.054 -0.398±0.042
xi! 4.98 0.505

Rs [Ohm] 0.0558:!:0.0049 0.0708:!:0.0083

amb
9
CRs[%!K] 0.72:!:0.42 1.8l±0.48
xi! 3.20 4.26

In(Rp [Ohm] ) 8.4±1.9 8.661±0.012

amb
10
CRp[%jK] 0.17±1.18 0.0068±0.0085
xi! 0.149 5.04

Table 3 Values of the fundamental parameters for cell number 62.

Equation Parameter Illuminated Dark Condition

IL
amb
[mAl 147.59:!:0.13 ---
11 CIL[%jK] 0.0468:!:0.0041 ---
2
X 35.2 ---
In(Col [mA/K3/cm2]) 4.65±0.49 10.75:!:0.14

5 VCl [V] 1.054±0.013 1.1822±0.0037

2 4.96
X 24.2
In(C [mA/K3/2jcm2 ]) 2.41±0.90 -2.44±0.79
02
6 VG2 [V] 2 0.982:!:0.050 0.807±0.043
X 26.4 18.7

Aamb 2.030:!:0.022 2.174±0.020

8 CA[%/K] -0.183:!:0.035 -0.413:!:0.044

xi! 23.1 6.94

Rs [Ohm] 0.0683:!:0.0045 0.0674±0.0088

amb
9 CRs[%jKj 0.85±0.28 2.11±0.54
X2 2.59 1.04

In(Rp [Ohm] ) 8.6±2.4 9.936:!:0.026

amb
10 CRp[%~~] 0.52:!:1.90 -0.105:!:0.025
0.0574 2.71

46
 Table 4 Values of the fundamental parameters for cell number 69.
Equation Parameter Illuminated Dark Condition
IL
....b
[mA] 150.38±0.26 --
11 CI L [%/!] 0.0435±0.0067 --
;t' 15.7 --
In(C 01 [mA/K3/cm 2 ] ) 7.90±0.45 10.85±0.10
5 VG1 [V] 1.109±0.012 1.1797±0.0027
2
;t' 1.11 34.3
In (C02 [mA/K3/2/cm2] ) 0.27±2.36 -2.08±0.81
6 VG2 [V] 0.96±0.13 0.854±0.042
;t'2 6.08 6.80
Aaab 1.827±0.046 2.109±0.016
8 CA[%jK] -0.224±0.088 -0.230±0.045
2
;t' 2.65 4.67

Rs [Ohm] 0.0532±0.0042 0.0697±0.0084

amb
9 CRs[%/K] 0.70±0.30 1. 31±0.51
X2 0.897 2.25

In(Rp~ [Ohm]) 10.4±6.9 12.69±0.47

10 CRp[%jK] 0.03±3.39 -0.39±0.12
2
;t' 0.0108 0.256

By the comparison of the value of A = exp(liE

tables 1 to 4, it is seen that the and 2.k.T (13)
parallel resistance values is obtained
with a better accuracy from dark condition for the future work we are intended to use
I-V curve measurements rather than from the equation 12 to fit the experimental
illuminated ones. I-V curve and compare the results with the
traditional equation.
4. Conclusions We conclude from this work that the
The more significant results of the temperature dependence of 101 and 102
work follows. The series resistance value follows the Wolf et al. equations [3] with
obtained from the illuminated I-V curves the modifications which we include taking
has shown to be more precise than its into consideration the work of Slotboom
counterpart obtained from dark I-V curves. [4]. As a new contribution of this work,
In some of the studied samples, it was empirical equations are proposed for the
observed that the differences in the temperature dependence of the diode
values of the mentioned parameters quality factor and series resistance.
obtained from the fit, does not imply,
necessarily, in a non-validity of the I-V
curve superposition principle. This can be References
1. H. S. Rauschenbach, Solar Cell Array
understood based on statistical Design Handbook, Van Nostrand Reinhold
considerations. In other samples, however, Company, New York, 1980.
the values of the 101 (from the 2. S. J. Fonash, Solar Cell Device
double-exponential model) obtained from Physics, Academic Press, New York, 1981.
fits of illuminated and dark I-V curves 3. M. Wolf, G. T. Noel and R. J. Stirn,
showed a difference many times bigger than IEEE Trans. on Electron Devices, vol.
that which could be statistically ED-24, p.419, 1977.
explained. For those samples it can be 4. J. W. Slotboom, Solid-State
said that the superposition principle is Electronics, vol. 20, p.279, 1977.
not valid since the probability of 5. P. R. Bevington, Data Reduction and
occurrence of that event is much smaller Error Analysis for the Physical
than the m1n1mum acceptable limit in Sciences, McGraw-Hill Book Company, New
experiments of this nature. This validity York, 1969.
may be related with considerations to 6. N. Veissid, P. Nubile, A. F. Beloto and
deduce the equation adopted. For example, A. M. Andrade. Twenty-First Specialists
Cerofolini and polignano [8] suggest Photovoltaic Conference, Kissimmee,
another analytical function for the second Florida-USA, p. 1184, 1990.
term of the I-V equation, which is 7. N. Veissid, M. T. F. da Cruz and A. M.
de Andrade. Solar Cells, vol. 28,
q. (V-I.Rs ) p.351, 1990.
exp[ k.T ] - 1 } 8. G. F. Cerofolini and M. L. Polignano,
I (12) J. Appl. Phys. 64(11), p. 6349, 1988.
r-g q. (V-I.Rs )
exp[ 2.k.T ] + A

47
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

SIZING OF LIGHT CONFINING CAVITIES FOR GaAs AND Si SOLAR CELLS

I.Tobias, A.Luque ,J.C.Mifiano and J.Alonso

Instituto de Energia Solar - U.P.Madrid
Ciudad Universitaria - 28040 MADRID

ABSTR~CT.- A family of light confining cavities containing GaAs and Si cells for spectrum
- splltting operation is analyzed. The cavity uses both spatial and angular confinement of
li~ht and hence the entry aperture can be large. In addition, the luminous power flux per
unlt.of entry area can be kept small. A potential cavity efficiency over 34% with good but
reallstic cells is calculated. The analysis, although approximate, allows for an
assessment of the potential losses to be done and identifies the most critical parameters.

1.- INTRODUCTION.
It is the property of the system that only
The operation of solar cells inside light rays coming directly from the bottom cell can
confining cavities presents interesting escape through the entry (these are the reverse
properties(1) allowing new photovoltaic systems paths of the entering rays). See figure 2 for a
wi.th high efficiency while adding no excessive description of the rays inside the cavity. Also, a
technological complexi ties. Inside a confining fraction of the rays reflected at the top cover
cavi ty the light reflected back by a cell is not returns back to it (there is an internal cavity for
definitely lost but gets further chances of being the cell L); the other ones reach the cell H. To
absorbed at the cells in the cavity, so that the our purpose this is enough about the optical
optical properties of the semiconductor become less features of the system. A more detailed description
critical and, i f properly reflecting, denser top is given in reference 2.
metal grids are tolerated hence minimizing series The bandgap energy of the cell H divides the
resistance effects. Spectrum - splitting operation solar spectrum into two regions with different
is easily achieved by placing into the cavity cells behaviour of the system, whereas the "fine"
with different bandgaps. variations of the spectral quantities with
When the rays escape the cavity with no wavelength are of minor importance: every
angular restriction the amount of achieved wavelength-dependant magnitude can be assumed to
confinement is directly related to the absorber - remain constant within each spectral region without
aperture area ratio: strong confinement requires loss of physical content. Superscript V (R) will
small apertures. If the cells are intended to denote the spectral range with photon energies
operate under reasonable irradiances high greater (lower) than the cell H bandgap.
concentration levels at the entry aperture are Note that i f the ellipsoid is fixed varying
needed. By using both spatial and angular the size of the cell H implies to modify the
confinement the family of cavities analyzed here angular distribution of the incoming bundle and
circumvents this technological problem. hence the previous concentrating stages. We will
consider that two concentrators are equivalent from
2.- CAVITIES WITH ANGULAR-SPATIAL CONFINEMENT

The elliptic cavity (fig. 1) consists of half fncoming light

a mirror-reflecting ellipsoid of revolution closed
with a cover through which a circular hole,
bun,dle
def tned by. the rotation of the foci F and F' is
made: the entry aperture. The cavity could be
filled with a dielectric medium. A cell with high
bandgap energy (cell H) is placed opposite to the white paint
entry so as to exactly intercept the incoming
bundle. An annulus-shaped cell with a narrower
energy gap (cell L), attached to the cover,
surrounds the entry. Should not it extend over the
whole cover a diffusive reflector (white paint)
Ice~ F' F
rmiiiII
_ _ _q

occupies the remaining area. In fact, according to

the following treatment, the shape of this cell is
irrelevant and we have adopted that guaranteeing
the rotational symmetry of the system.
The light strikes first the cell H and only
after reflection there it reaches the other cell.
So the cell H must behave as a good reflector for
photon energies lower than its gap, which can be
accomplished by putting a dichroic filter on top of
i t or by implementing a mirror at its back face.
Besides the reflected light must not be directed
back towards the entry: diffusive reflection will
be assumed hereafter to take place at the cells and
the top cover. The possibility exists also of using
another dichroic filter to prevent the cell L from
absorbing high energy photons that would be more Fig.1.- Cross section of the elliptic cavity; the
efficiently employed by the cell H. system has rotational symmetry.

48
 A B c o

Fig.2.- Families of rays inside the elliptic cavity : a: entering - escaping bundle (EH); b: rays
linking the cell "H and the top cover: some rays undergo reflection at the walls (TH); c: rays
linking the cover with itself (TT) ; d: rays that do not exist: the outlined arrow Is a non -
illuminated path; the solid arrows are impossible paths.

the point of view of technology if they present the spectrum in the same regions which, when impinging
same radiance (w'cm- z 'srad- 1 ) at their outputs, on th~ celJ' produce the current densities denoted
whichever the angular spread of the output bundle by JI , JI, respectively. The second equali ty is
could be (the etendue (cm z , srad); see references the defini tion of the spectral concentration
3 , 4). A family of concentrators with the same factors ~, Xv.
technological requirements is then characterized by The sta~~ars~ AM-l . 5 Direc t spectrum normalized
the value of the radiance BE at the cavity entry, to 100 mW'cm is adopted for calculations.
or, equivalently, by the irradiance concentration 22 % one-sun efficient GaAs cells and 19 % Si
Xiso for isotropic output . The following cells have been developed at our Laboratory. Table
relationship holds then for XE (irradiance gain of I collects til s(';} of representative parameters for
the concentrator) and gEH (etendue of the rays the devices ' . Note that the absorbance is not
entering the cavity) : included : we consider it as a parameter of the
gEH cavity rather than one associated to the' cells.
XE = Xiso IlAE (1)
TABLE I
AE is the area illuminated by the concentrator (the ELECTRICAL PARAMETERS OF PROTOTYPE CELLS
cavi ty aperture) . nAE equals the etendue in the GaAs cell Si cell
isotropic case. A value Xlso" 10000 will be
considered as representative of the available Jo (A/cm- 2 ) 4 . 65 ' 10- 19 4.50'10- 13
technology .
JI V(A/cm- 2 ) 0 . 0285 0 . 030
3 . - THE CELLS .
JIR(A/cm- Z ) 0.0 0.010
As prototype cells to operate inside the
cavity we will adopt the best devices, both Si rB (Ocm 2 ) 1. 26 ' 10- 4 2 . 80'10- 3
(cell L) and GaAs (cell H), already manufactured at
this Laboratory. Their electrical behaviour is RK (Qcm2)t 2.53 ' 10- 4 2.00 '10- 4
described with reasonable accuracy by the simple
one-exponential, low injection model : rc (Ocm 2 ) 10- 5 10- 5

J = JF - Jo (exp(V +V~ rS)_l) (2) rE (Ocm 2 ) 9.34'10- 7 4.17'10- 6

t One parameter, crossed gl'1d

JF and Jo (A·cm 2 ) are the photogenerated and dark metallization structure with 10ltlR
saturation current ~ensities , Vt (V) the thermal wide , 3ltm thick fingers .
vol tage and rs (0. cm) is a function of the front
grid transparenT~ factor F and in its simplest form 4 . - SIMPLIFIED ANALYSIS OF THE CAVITY OPERATION.
is expressed as' :
A RK+re rE In equation 4 the effect of the cavity is
rs = rB + - - - + - - - (3) included in the terms ~, Xv, so we must relate
1-F (1-F)2 them to the optical parameters of the system .
where ro, re , rE and RK are parameters that, Explicit reference to the angular distribution of
corresponding to the different regions of the light is to be done, hence we need to know the
device where ohmic losses occur, depend on the following etendues :
technological details of the device. The term A RK - To tal etendue of the entry .
(A being the device area) stands for the resistance 15E = nAE (5)
of the grid and , being proportional to the area of - ttendue of the entering-escaping bundle (rays
the cell, introduces the only dependence on an linking the entry an~ the cell H without reflection
"absolute" dimension in the equations of the at the cavity walls(2 ) :
cavity . gEH = IlAEAHlAc (6)
The photo-generated current density JF for - Total etendue of the cell H.
each cell is made up with the contributions from gH = nAH (7)
the R and V regions. This can be put in the more - Etendue of the rays linking the cell H and the
explicit, familiar form to be applied to both top cover (entry excluded) .
cells: grH = gH-gEH = nArAH/Ac (8)
- Total etendue of the top cover (entry excluded) .
JF = F(--
E ~R JI R+--
ER ~ V JI v) = F(~~RJIR+XV~VJIV) (4) gr = lIAr (9)
- Etendue of the rays I inking the top cover wi th
EIR El v
itself (through reflection at the walls) .
where ~R, ~v are the absorbances (the complement of grr = gr - grH = nAr (1-AH/ Ac) (10)
the reflectances) of the cell in the R and V Ar denotes the area of the top cover excluding the
spectral regions, ~ , EV the irradiances (w.cm-2~ entry, AH the area of the cell Hand Ac the total
on the cell for both spectral regions, and EI R, EI top cover area ( = Ar+AE ).
the irradiance content of the standard one-sun The preceding magnitudes are not spectral ones .

49
The reflectivities of both covers, on the contrary, reflector, 0% reflection at every absorber. This
depend on wavelength and so two values (for V and R does not apply to the white paint on the top cover,
bands) are to be given. The irradiance is assumed which has been assigned very small (0.2%)
to be uniform within each suriace AT and AH and the absorption losses.
ref lection perfectly diffusing. For the top cover Figure 3 collects the main results. The
(entry excluded) the reflectivity is written: abscissa is the ratio AEIAG, and the four remaining
variables are optimized at each point to give
PT AL (FL(l-aL)+(l-FL) (l-amL) ) +~(l-aB)
= AT AT-AL (11) maximum cavity efficiency.
The maximum efficiency is 34.4%, around 28%
with aL, amL, aB being the absorbances of the contributed by the GaAs cell and the remaining 6%
uncovered zones of cell L, the metal fingers and by the 5i cell. The entry diameter has a little
the white paint, respectively; AL is the cell L influence.
area. And the reflectivity of the bottom cover, The GaAs cell works at a concentration around
i.e., of the cell H, is expressed as: 700X with 6% grid shadowing and the cell diameter
is near 0.3 cm. These results do not differ very
pH = FH(l-aH)+(l-FH) (l-amH) (12)
much from those obtained by optimizing the cell
Three distinguishable families of rays exist without cavity. In fact the effect of the cavity is
into the elliptic cavity (subscripts EH, TH, TT). small for the GaAs device.
Each one is assigned a wavelength dependant The irradiance level at the 5i cell amounts
coefficient (L oS 1) describing the attenuation - 100 (small entry) and 200X (large entry) in the R
reflection losses undergone by the light when band of the spectrum, corresponding to around 30
travelling between the two regions that define the and 70X full spectrum. So it generates low current
bundle. densities. The shadowing factors (10 and 12%)
Let BT and BH be the radiances associated with resul t larger than those required outside cavi ty
the top cover and the cell H so that BTgT and BHgH showing the effect of the external confinement on
denote the power reflected by them (as diffusive this cell. Note that a great portion of the top
reflection is assumed, the total etendues apply cover is occupied by the white paint.
here), and ET and EH be the irradiances associated The concentration at the entry is low, 242X
with the same regions, so that ETAT and%aH denote for the small entry and 81X for the large one.
the incident power on them. In the steady state, we
can write balance equations of luminous energy for
each cover and each spectral region, expressing
that the power emitted by a surface is due to
reflection of the incident power:

BT gT = pT ET AT = pT(BH gTH LTH+BT gTT LTT) (13)

for the light on the top cover comes either from

the cell H or from itself through reflection at the
mirrors.

BH gH = pH EH AH = pH(BE gEl! LEH+BT gTH LTH) (14)

lOr
because the cell H is directly illuminated from the E
u
E
entry and receives also light from the top cover.
The last term in 13 is responsible for the
effect of internal cavity for the 51 cell. By
solving for BT and BH, and according to (1), we
find the concentrations on the cells:
~
XH = Xlso ~ LEH V- 1
AG
(15) 1.0 :2
u
- -- GaAs <{
x = XlsoAEAHLEH pH LTH(l- PT LTT l-AHI AG)-lV -1 (16)
LL

L AG2 l-AEI AG >-

U
where: - ... - ___ -- c._~
a:
<{
AH LTH 2 ( l-pT LTT 1-AHI
V = 1 - pHpT AG
AG)-l
l-AEIAG (17)
a..
U1
z
30 .8 g
f
The denominator V represents the influence of the
cavi ty in enhancing the concentration. The nested
04
configuration comes from the internal cavity effect
for the cell L.
~03f
GaAs
5.- EFFICIENCY OPTIMIZATION. t}02t
c;:
The cavity is a four terminal device and so the bOI . .
maximum power point for each cell is to be
Si
separately determined. For this the analytical,
°oc.OLrC-l~~~~'~"~'I~~~~~~~~--~'~'W'~'~"I
approximate expressions in reference 3 are 0.0' 0.1
applied. Two entry diameters (0.5 and 1.0 cm) are ENTRY AREA I L- tWE~ A~EA RAPe
considered in the calculations. The design
variables are the areas (AL, AHl and transparency Fig.3.- Optimization of the elliptic cavity: (from
factors (FL, FH) of both cells, and the cover area bottom to top) efficiency, optimum
(AG). The influence of the dichroic filter onto the transparency factors, optimum diameters,
5i cell has revealed very weak and the results and concentration factors at the cells ( in
refer to the case where no such a filter is the V-band for the cell H, in the R band
for the cell L ) and at the entry. Solid
present.
The optical parameters of the cavity have been lines: entry diameter 0. Scm; dashed: 1. °
given their ideal values: 0% absorption at every em. Full circles at the optimum points.

50
 Af!I$OlUTE SCNSITIVITY ,~
A=EL=A=
TI~
VE
~ 8E
=NS
~IT~ IT~Y______________-,
IV~

Cl CltL.l" m:53 CfU L

0.'
0.'
0.26

... 0. •

0 .16
0.'
..,
0.'

R r
... E

Fig.4.- a: Absolute sensitivity factors of the system to the cavity (optical) parameters. b: Relative
sensitivity factors of the system to the cells (electrical) parameters.
6.- SENSITIVITY ANALYSIS. expectancies to meet high efficiency.
The cavities analyzed use both spatial and
The derivative of the system efficiency n with angular confinement of light. Although spatial
respect to a certain parameter p is useful to confinement p::~one can lead to similar theoretical
quantify how sensitive the system is to the value efficiencies 2, practical considerations favour
of p. The parameters of the cavity belong to two the use of angular confinement: the system
classes. First, the optical magnitudes ranging from tolerates large apertures illuminated at low
o to 1 that describe the luminous losses and were irradiances. The advantage is double - fold, for
assigned their ideal values up to now; these will besides relaxing the technological problem of the
be characterized by the "absolute" sensitivity aperture the concentrator itself becomes
defined as a~/ap. Second, the electrical parameters simplified, perhaps avoiding the need of a
of the cells which were given values representative secondary.
of the state of the art and will be characterized We should highlight also two drawbacks of the
by the "relative" sensitivity a(to.q~)/ap. The configuration:
derivatives are calculated for the optimized - A large surface of white paint is needed and its
configuration described in the preceding section. reflectivity appears as a very critical parameter.
In this way additional information is obtained with - The GaAs cell cannot be textured and some
respect to which are the critical points in the arrangement is to be devised to prevent the mirror
performance of the cavity and what the efficiency reflection at its surface from escaping through the
could be when different values of the parameters entry.
are given. Both aspects can be improved with minor
The results of the analysis are represented in changes in the structure of the cavity without
fig. 4 for an entry diameter equal to 0.5 cm. Among affecting its basic properties. By-pass
the optical parameters, the large influence of the consequences of these modifications are that the
parameter LEH in the V region is obvious. More "internal cavity" effect is weakened while the
interesting information is contained in the high confinement of the light for the GaAs cell becomes
sensitivity factors exhibited by LTT and aB in the stronger. It is to be noted that the study of these
R-band: this points out the importance of the modified versions of the elliptic cavity does not
internal cavity effect for the Si cell. admit the simplified treatment adopted here but
Regarding the cells the results express that requires the aid of ray tracing.
the GaAs device gives the greatest contribution to Some of the parameters used here, in
the efficiency. The series resistance parameters particular those referring to the cells, can
show but little influence according to the change. The sensitivity analysis allows for an easy
principles of operation inside confining cavities. assessment of their influence that justifies the
With the sensitivity factors it is easy to ideal analysis we have undertaken.
perform efficiency extrapolations to different
system parameters. For example, when stronger
optical losses are considered (2 % parasitic
absorption at every reflector, 5 % reflection at
the absorbers), the efficiency drops from 34.4 to REFERENCES.
around 33 %.
At Spire Corporation 24.8 % :r;!~. 5G efficient 1.- R.A.Sinton and R.M.Swanson, IEEE Electron.
GaAs devices have been developed 8 . After some Devices Letters, EDL-8, 11 (1987).
rough modelling of these devices, we make use of 2.- J.C.Mifiano et al. Submitted for publication
the sensitivity factors to predict that with them (1990) .
the efficiency of the elliptic cavity increases to 3.- A.Luque, "Solar Cells and Optics for
35.5 %. The UNSW group has demonstr~\ed 24.2 % Photovoltaic Concentration", Adam Hilger,
AM1.5G efficiency with a Silicon cell 9 . Fitting Bristol (1989).
the reported data to the one-exponential model it 4.- W.T.Welford and R.Winston, "The Optics of Non -
is concluded that if this Si cell were placed into Imaging Concentrators", Academic, N.Y (1978).
the cavity instead of ours the efficiency would 5.- R.J.Matson et al., Solar Cells 11 (1984)
reach 36.5 %. And with both Spire and UNSW cells 105-145.
the system would exceed 37.5 % efficiency. 6.- C.Algora. Doctoral Thesis. Universidad
Complutense de Madrid (1990).
7.- CONCLUSIONS. 7.- A.Cuevas and M.A. Balbuena, Proc. 20 th IEEE
PVSC (Las Vegas 1988), IEEE£ N.Y. (1989).
The operation of light confining cavities with 8.- S.P.Tobin et al., Proc. 21 s IEEE PVSC (Orlando
GaAs and Si solar cells has been analyzed. The 1990), IEEE, N. Y. (1990). st
interesting properties displayed allow to conceive 9.- M.A.Green et al., Proc. 21 IEEE PVSC (Orlando
of novel photovol taic systems wi th good 1990). IEEE, N. Y. (1990).

51
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

AlGaAs-GaAs SOLAR CELLS WITH INCREASED RADIATION STABILITY

V.M Andreev, V.S.Kalinovskii, O.V Sulima

A.F.Ioffe Physico-Technical Institute
Polytechnicheskaya 26, Leningrad 194021, USSR

ABSTRACT. It is shown that the radiation resistance of the AIGaAs-GaAs

heteroface solar cells (HSC) can be considerably improved due to optimum
design. pAlO SGaO 2As-pGaAs-nAI Ga1_yAs (0<y<0.15) -nGaAs (substrate)
HSC fabricatea by 'combination 01 LPE and pseudo-closed-box diffusion of
zinc from vapor phase have been investigated. An improvement of the
radiation resistance of HSC exposed to 6.7 MeV proton irradiation was
achieved.

1. INTRODUCTION possible to achieve a high 1~cept03s

concentration (up to NA-S'10 cm-)
Solar cells (SC) for use in space with large NA gradien~ and. a low
have anti radiation coatings, but average resistIvity for pGaAs layer
nevertheless are considerably affected (down to p_2'10- 3 O·cm). Large gradient
by bombardment with high-energy of NA induces strong drift built-in
particles. GaAs-based SC usually have electric field (up to 4 kV/cm) thus
better radiation stabil1 ty than Si-SC facilitating the process of carriers
because of less influence of life-time separation. The HSC were bombarded with
decay in GaAs on SC parameters. monoenergeti8 bea~2 of pr~tons (E p -6. 7
However, that is true not for all GaAs- MeV, ~_101 -6'10 cm-) at room
based SC and is dependent on SC design. temperature during the time from
Earlier investigations of the influence several seconds to several hours.
of irradiation on photoelectric
parameters of the pAIGaAs-pGaAs-nGaAs
heterostructures have revealed an 3. EXPERIMENTAL RESULTS AND DISCUSSION
increase in the radiation stability
with reduction in the thickness of The life-time of nonequilibrium
pGaAs layer in the case of irradiation electrons and holes (T e , Th) is
with Ee -1 MeV electrons [1], Ep =52 MeV strongly influenced oy proton
[2] and E -6.7 MeV [3] protons. In irradiation. The drop of T values
order to Pimprove the SC radiation during irradiation causes deterioration
stabili ty we have investigated the of the carriers separation and
influence of wide bandgap pAIGaAs corresponding decrease of photocurrent.
window layer as well as drift built-in In the case of existence of drift
electric fields in p- and n-type built-in electric fields in the
regions on heteroface SC spectral photoactive layer both drift and
response and photocurrent. diffusion processes take part in the
separation of nonequilibrium carriers.
Figure 2 shows the calculated life-time
dependence of built-in electric field
2. EXPERIMENTAL DETAILS.
E=(kT/q)1/2'(~e'Te)-1/2
The investigated 12.6 mm 2 heteroface
solar cells (HSC) were based on in pGaAs layer, which provides equal
pAlO SGaO 2As-pGaAs-nAlvGa1_vAs-nGaAs values p'f diffusion length
(0<y<0.15) heterostructure~ schemati- le=(De'Te)1/2 and drift length
cally shown in Figure 1. The p-Iayers' Le=~e'E'Te of ~lectrons in this layer
thicknesses were H=0.01-2.0 ~. The for ~e=1000 cm IV' s (typical value of
graded bandgap nAlvGal_yAs layer was electr!?R mogllity for GaAs with
grown in order t6 induce the back p>5'10 cm-). It can be seen from
surface field (BSF) which is Figure 2 that the value of E~4 kV/cm is
particularly important in the HSC with large enough to make drift mechanism
small h values because of the predominant in the process of carriers
considerable light absorption in n- separation.
layer in this case. Thus existence of strong electric
As the first step in HSC fabrication field in pGaAs layer provided optimum
n-type heterostructure consisting of condi tions for separation of nonequi-
the BSF layer and the window layer was librium carriers generated in this
grown during one liquid phase epitaxial layer. This led to conservation of
process in graphi te boat with short wavelength photosensitivity of
compulsory squeezing out of melts [4]. HSC after irradiation (Figure 3). In
The next step was p-n junction order to increase the role of stable to
formation by simple from the point of radiation short wavelength photosensi-
view of technology process of pseudo- tivity in photocurrent conservation the
closed box diffusion of zinc from the window layer thickness (H) was
vapor phase. It has been shown [5,6] decreased to the minimum possible
that such diffusion method makes values of SO-100 A which guaranteed the

52
 p. on- 3 AlAs. mole %
lack of the electrons tunneling from
pGaAs layer to the surface (curves 1, 1020 D -:Lao

I
1', Figure 3). Figure 3 also shows that
the electric field in the graded
bandgap nAIGaAs layer reduced the 1019
influence of irradiaion on electrons 20
and holes in this layer. JOI8 ]
Figure 4 shows the influence of HSC 10
design on radiation resistance of
photocurren t the parameter whi ch is JOI7 o
the most sensitive to irradiation. In
HSC of mentioned above optimum design
the drop of initial photocurrent
density (1-30 mA/cm- , AMO) waf2as \ow
as 8~ for proton fluence ~=6'10 cm-.

4. CONCLUSIONS
The experimental results described
in this paper show that the radiation Figure 1_ Sketch of investigated AlGaAs
stability of the AIGaAs-GaAs HSC can be HSC structure.
considerably improved due to optimum
design. An improvement of the radiation
resistance of the pAlO 8GaO 2As-pGaAs-
nAlyGal_yAs(0<y<0.15)-n~aAs(suDstrate)
HSC exposed to 6.7 MeV proton E. V/cm
irradiation was achieved due to 500 .•••
following:
a. decreasing of the wide bandgap "
pAIGaAs window layer and the ",
",
photoactive pGaAs layer thicknesses
down to 100 A and 0.3 ~, respectively;
b. using of drift built-in electric
fields (up to 4 kV/cm) induced by
acceptors concentration gradient in 100 "
diffused p-region and by chemical
composi tion gradient in photoacti ve n- "
region. Though the thickness of
diffused pGaAs layer of HSC is small
the hi8h hf~e cOfcentration in this
layer (p>10 cm-) and corresponding
low sheet resistance makes possible to 7'. S
convert the concentrated sunlight. This
opens new possibilities in further
improvement of HSC radiation resistance
due to high-density current flow- and Figure 2. Calculated life-time de-
photoexitation- caused reduction of the pendence of built-in electric field
radiation-induced defects during the which provides equal values of
concentrated sunlight conversion. diffusion and drif~ lengths of
electrons (~e-1000 cm /V's) in pGaAs
layer.
5. ACKNOWLEDGEMENTS
We thank Alexander von Humboldt
Foundation for financial support
concerning presentation of this paper
in EPSEC 10. We express our 1 o~c=ocu==a~l=o~N~Ern~C=l~~~~__________________________--,
appreciation to Dr. V.D.Rumiantsev for
helpful discussions. We are also
grateful to Andreas Bett and Angela
Habich for their assistance during
preparation of this paper.

6. REFERENCES
(l)M.Yamaguchi and C.Amano,
J.Appl.Phys. 57, 537 (1985)
(2)S.Yoshida, K.Mitsui, T.Oda, and
Y.Yukimoto, Proc.Third Photovoltaic
Science and Engineering Conf., Kyoto,
1982, in: Jpn.J.Appl.Phys. 1, Suppl.,
27 (1982)
(3)V.M.Andreev, G.M.Gusinskii, V.S.Ka-
linovskii, O.K.Salieva, V.A.Solov'ev,
O.V.Sulima, and A.M.Khammedov, WAVELENGTH. 1,m
Sov.Phys.Semicond. 22, 88, (1988)
(4)Zh.I.Alferov, V.M.Andreev, S.G.Kon-
nikov, V.R.Larionov, and B.V.Pushny, Figure 3. Spectral distribution of the
Kristall und Technik, 11, 1013, (1976) collection efficiency of HSC with Wide
(5)V.M.Andreev and O.V~ulima, bandgap layer thicknesses 100 A (curves
Sov.Tech.Phys.Lett. 8, 187 (1982) 1 .1' ) and 2 ~ (2,2' ) before
(6)R.Romero and O.V.Sulima, irradiation (1,2) and after 6.7 MeV
Crystal Research & Technology, 18, 1053 proton irradiation with dose
(1983) 6'10 12 cm- 2 (1') and 4'10 12 cm- 2 (2').

53
 • 1 o 2 • 3

1.0

0.9
"' 8 ..
".
..
/"'0
'~

0.8

100 A

0.7

0.6
pip In
I ~ O.6/lm

-2
q,p, em
Figure 4. Ratio of final photocurrent (lip) to initial photocurrent
(11"0=30 mAl cm 2 • AMO) as a function of 6.7 MeV proton fluence for
different types of HSe.

54
 8-12 APRIL 1991 LISBON. PORTUGAL
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE

NEW TECHNIQUES IN SOLAR CELL TI-IEORY

Tomas Markvart
Department of Engineering Materials
University of Southampton, U.K.

Present address: Instituto de Energia Solar,

ETSI Telecomunicacion, Universidad Politecnica de Madrid

ABSTRACT It is shown that the solution of the transport equation f~r the I?ino~ity
carriers in a solar cell can be aided by the use of Green's. functions - solutIO~s whIch gIv~
the cell response to carriers generated in a single pomt. The use of this concept IS
illustrated by considering the difference between the effect of surface and bulk
recombination in silicon emitters.

No attempt will be made here to present a systematic

1. INTRODUCTION theory of the subject which requires a more extensive
treatment than is possible in this note. Instead, a plau-
There has been a continuing interest in the analy-tical or sibility argument will be given for the relevant results
semi-analytical solution of the carrier transport equations which will then be illustrated by specific examples. A
in the solar cell. Particularly in diffused emitters where the rigorous derivation of the results will be published else-
dopant distribution is not constant, a realistic solution of where [8].
these equations must take into account factors such as the
dopant concentration dependence of the minority carrier
diffusion constant and diffusion length. Since the doping is 2. THE GREEN'S FUNCTION
usually quite high, degeneracy and bandgap narrowing
must also be included. These points imply that, for a Suppose that the external surface of a planar solar cell lies
realistic solar cell, an approximation or a numerical at x = W (surface recombination velocity S) and x = 0
approach is usually necessary at some stage of the calcul- denotes the edge of the depletion region of the p-n
ation. junction. Let us consider the minority carrier current
collected at the junction which is produced if carriers are
A particularly elegant formalism for the minority carrier generated at point x with a unit rate. The carriers reaching
transport in the solar cell was developed by del Alamo and the junction can be divided into two groups. Some carriers
Swanson [1]. Approximate solutions were developed by reach the junction from x (with probability ur(x), say)
Park et al [2] and Bisschop et al [3], and the subject was without encountering the external surface at x = W.
reviewed by Cuevas and Balbuena [4]. The method can Clearly, u, is independent of S but depends on the bulk
also be used to discuss the photovoltage decay, as shown recombination lifetime. Other carriers reach the junction
by Landberg and Abrahams [5]. Green and Blakers [6] via x = W where they may recombine by surface recombi-
considered a relationship between solar-cell parameters in nation. The probability that these carriers are eventually
the dark and under illumination. This issue is also the collected can be written as as u,(x), where u,(x) is the
subject of this note where we show how the Green's probability that the carriers reach the external surface
functions can be used to provide a new insight into the after generation at point x without recombining in the
carrier transport in electronic devices. bulk, and as is the probability of reaching the junction
from x = W, without recombining either at the surface or
To a physicists, the Green's functions (which are, in this in the bulk. This probability, in turn, can be decomposed
context, often known as propagators) are closely associ- into two factors,
ated with the many-body problem of interacting particles
and the Feynman diagrams [7]. To a mathematician, the
Green's function (after George Green, 1793-1841, a British
mathematician) provides a means of expressing the where ps is the probability that a carrier at x = W recom-
solution of a linear differential equation with an bines via surface recombination, and a, is the probability
inhomogeneous ('source' or 'driving') term, via the sol- that a carrier reaches the junction from x = W without
utions of the homogeneous equation. recombining in the bulk. The current u(x) collected at the
junction is thus equal to
Translated to the solar-cell language, the method makes it
possible to express the solutions of the minority-carrier
transport equations in the space-charge regions of a cell u(x) = u,(x) + a,(1-ps) u,(x) (1)
under illumination in terms of the solutions in the dark.
The Green's function - in this context, the response of the
system to carriers generated in a single point - then The first term in (1) is the current at the junction under
becomes a useful tool for the analysis of the carrier carrier generation at x if the surface recombination velo-
transport, and was already introduced for transparent city were infinite. The second term in (1) gives the extra
emitters in references [1) and [6), and for the case of contribution to the current u(x) produced by the carriers
uniform field in [5]. Here, this problem is considered in a reflected from the external surface at x = W on account
rather more general setting. of the finite surface recombination velocity S.

55
The function u(x) is a particular form of the Green's
3. SURFACE AND BULK RECOMBINATION IN
function. Since the carrier transport is one-dimensional, it THEEMITIER
describes the response of the cell to carriers generated in
a thin slab with coordinate x rather than at a single point
The results that can be obtained from Eqs. (1)-(3) will be
as would be the case in two or three dimensional prob-
illustrated by a comparison between the effects of surface
lems. The usefulness of u(x) can now be seen since all the
and bulk recombination in the emitters of silicon solar
important characteristics of minority carrier transport
cells. We shall consider three different situations:
follow from u(x) by simple algebra:
(i) No surface recombination. This corresponds to an
(i) The light-generated current
emitter with good oxide passivation and small contact
area.
W

r u(x) g(x) dx (2) (ii) Infinite surface recombination velocity, approximating

J a continuous metal contact.
o
(iii) Emitter with S = 0 but short bulk minority-carrier
where g(x) is the optical generation rate at point x; and lifetime. In practice, a reduction of minority-carrier
lifetime occurs in space cells which have been irradiated
(ii) The dark saturation current jo by high-energy particles. The conditions chosen here
correspond to irradiation by the fluence tP = 9x1013 cm·2 1
Me V electrons. This irradiation was found to produce a
jo = 0(0) coCO) u'(O) = similar light generated current as in the case (ii), thus
= 0(0) coCO) {uc'(O) + a,(l-ps) u:(O)} (3) providing a good case for the comparison of the surface
and bulk recombination effects.
where 0(0) and coCO) are the minority-carrier diffusion
constant and the 'effective' equilibrium minority carrier The material parameters (diffusion constant, diffusion
concentration at x=O. It follows from the definition of u, length and bandgap narrowing) and their dependence on
that the second term in the brackets in (3) is negative. the doping concentration were chosen the same as those
This is expected since carriers reflected from the front sur- of [1]. The irradiation was modelled in the standard
face (rather than absorbed by it) reduce the dark current. fashion, by employing the Messenger-Spratt equation for
the minority-carrier diffusion length:
In addition to the physical insight, Eqs. (1) and (3) provide
a useful formalism for numerical solution once a recipe is 1 1
obtained how the functions u, and u, can be determined + K tP (7)
from the detailed theory. It turns out, in fact, that u~ u, L' L'
o
and a, are simply the same quantities introduced by del
Alamo and Swanson: they are the excess minority carrier where Land L., are the diffusion lengths before and after
concentrations, in units of the 'effective' equilibrium irradiation, and K is the damage constant which, in
concentration, which satisfy the boundary conditions general, depends on the dopant concentration and was
here taken in accord with the results [9].
u,(0) = 1 ; u,(W) = 0
(4) Before proceeding to discuss the spatial distribution of
u,(O) = 0 ; u,(W) = 1 minority carriers in the three instances, it is worth pointing
out that, although the light-generated current is identical
for a solar cell without illumination. The quantities a, and (6.3 mA/cm2), the dark currents of emitters (ii) and (iii)
Ps are given by are different: 1.8xlO·u Ncm 2 for (ii) and 4.5xlOo12 Ncm 2 for
(iii).

Uf ' (W) Other differences appear when the emitter characteristics

u r ' (W)
are examined as a function of distance of generation
(5) wavelength. Figure 1 shows the function u(x) for the three
emitters which demonstrate the difference between the
s bulk and surface recombination in terms of the collection
Ps efficiency as a function of distance of the point of gener-
D u r ' (W) + S ation from the junction. Figure 2 shows this difference in
terms of the spectral response; in addition to the three
emitters which have been considered so far, the spectral
Using (4) and (5), the theory presented here can be shown response of an ideal (Iossless) emitter is also shown.
to give identical results to those obtained by del Alamo
and Swanson [1] by virtue of the identity

D(O) coCO) u:(O) + O(W) co(W) uc'(W) = 0 (6)

which is proved in [8].

56
4. CONCLUSION REFERENCES

We have demonstrated that the Green's functions can 1. J.A del Alamo and R.M. Swanson, IEEE Trans.
provide an important insight into the solar cell operation, Electron Dev. ED-31 1878 (1984).
as well as a tool for the numerical solution of the transport 2. J.S. Park, AS. Neugroschel and F.A Lindholm, IEEE
equations. This technique is of particular value for the Trans. Electron Dev. ED-33 240 (1986).
design and diagnostics of solar cells, analysis of radiation 3. F.J. Bisschop, L.A Verhoef and W.C. Sinke, IEEE
damage and other situations where the spectral response Trans. Electron Dev. ED-37 358 (1990).
data and characteristics in the dark are needed to obtain 4. A Cuevas and M. Balbuena, IEEE Trans. Electron
information about the spatial distnbution of the solar cell Dev. ED-36 553 (1989).
parameters. 5. P.T. Landsberg and M.S. Abrahams, Proc. 16th IEEE
Photovoltaic Specialists Conf., San Diego (1982), p.
490.
ACKNO~DGEMENTS 6. M.AGreen and A W.Blakers, IEEE Trans. Electron
Dev. ED-30 1360 (1983).
This work was carried out with the support of the Space 7. See, for example, L.P. Kadanoff and G. Baym, Quan-
Department, Royal Airspace Establishment. It has tum Statistical Mechanics, Benjamin, N.Y. (1963).
benefitted greatly from discussions with B. Cocking, C. 8. T. Markvart, to be published.
Goodbody, AF.W. Willoughby and J.W. Peters. Particular 9. M. Yamaguchi and K. Ando, J. AppJ. Phys. 63 5555
thanks are due to P.T. Landsberg for his critical review of (1988).
the manuscript and for bringing my attention to the
photovoltage decay application of the present method.

Fig. 1. The probability u(x) that a carrier generated at x

is collected at x-O for low and high surface recombination
velocity (curves a and b, respectively), and for low surface
recombination but short bulk lifetime (curve c).

' tU(~X~)==========::========::~~~~----------l
a
0 .8

0 .6
b

0.2 0.3 0.4 0 .5

x, microns

Fig. 2. The spectral response for an ideal emitter (a), for

a realistic emitter with low and high surface recombination
velocity (b and c, respectively), and for an emitter with
low surface recombination but short bulk lifetime Cd).

spectral response, mA/mW

0 .4 0 .5 0.6 0 .7
wavelength, microns

57
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

ORGANIC SOLAR CElLS - A SURVEY

Dieter Bonnet, Jiirgen Volkheimer

Battelle Institut e.V.
D-6000 Frankfurt am Main 90

ABSTRACf. The objective of this study was to examine in some detail the question of efficient organic solar cells. The scientific-
technical situation concerning organic solar cells is highly confusing and unsatisfactory. Isolated and incomplete pieces of information have
been found on a great number of different materials. Especially the physical aspects of solar cells have been treated only inadequately
by most authors, and many open questions and unsolved problems have been left unconsidered. Only very few candidate materials - such
as the phthalocyanines and perylene derivatives - fulfill essential requirements such as those concerning stability, feasibility of p-n diodes,
etc. The highest uncorrected efficiencies under sunlight that have so far been measured on organic cells lie around 1%, while typical
efficiency values are well below 0.1 %. It is recommended that future work be focused on the most promising of those materials in order
to make a realistic assessment of the real potential of organic solar cells.

1 Background charge carriers.

Photoelectric effects and photoconduction of organic materials Maximum possible quantum efficien.cy over a broad
have long been known and are in fact utilised technically, in range of the solar spectrum.
photography and xerography. The first attempts in the 1960s at Doping and creation of depiction layers (As far as is
making use of the photovoltaic effect of organic photoconduc- known, the doping of organic semiconductors is critically
tors were discontinued because the "classical" solar cells based influenced by the ambient atmosphere.)
on inorganic materials were so successful. Ohmic contacts for the light-generated current. Especial-
Numerous minor R&D projects on organic solar cells ly with the organic semiconductors (which tend to be
were performed during the past 10 years or so, most of them at weakly doped) these contacts are difficult to achieve.
universities, but none of those projects resulted in continuous Stable properties over extended periods of time.
development programmes. There is an immense number of Inexpensive materials and simple methods of producing
organic semiconductors that could be suitable for making solar cell structures.
cells, because any change in the structure of a given organic The first brief survey already revealed considerable weaknesses
molecule will yield a "new" material. Dyes or pigments and of many of the articles that were reviewed. Especially the physical
conductive polymers are the best known of these materials. aspects of the matter are not given enough attention by most of
Up to this day, however, it is still unclear whether all the authors, and many problems and questions have not been
the necessary material properties can actually be combined in a understood.
single system, i.e. whether it is possible to "tailor" an organic
semiconductor system for solar cells.
It therefore appears advisable to survey all the available 3 Existing Data and their Interpretation
data and infornlation on organic solar cells before embarking on
further experiments - and that is what has been done with the 3.1 Measuring Techniques and Results
study that is outlined in the following. It was found that even the most essential basic data are often
lacking in the literature: Very often no information is given about
incident light intensity and spectrum, or about parameters such
2 Objectives as the wavelength of monochromatic light, the transparency of
The objectives of the study were as follows: the contacts, the size of the exposed area on the solar cell, the
• Describe the fundamentals and the present state of the art of rectification ratio, the mobility of the charge carriers, the doping
organic solar cells, and analyse the potential of those cells. level, the stability of the cells in the dark and in particular when
• Identify scientific-technical uncertainties and obstacles. illuminated, the limiting current density, the quality factor, and
• Indicate what further research and development is necessary. the surrounding atmosphere (vacuum, air, inert gas). However,
Roughly 1 500 scientific articles were identified which the surrounding conditions are extremely important in the case
deal with the photoelectric effects in organic semiconductors; of organic solar cells because oxygen, for example, has a doping
the large majority of those articles was examined, and between effect on organic p-photoconductors and plays a decisive role in
200 and 300 of them, which were directly related to the photo- barrier formation. (In organic n-semiconductors on the other
voltaic effect, were evaluated in detail. hand, oxygen seems to hinder the generation of charge carriers.)
The major problem areas associated with organic Many publications include no current-voltage characteristics
materials are basically the same as with inorganic materials: (although it is those characteristics that provide the most
Broad light absorption adapted to solar radiation. valuable information about the quality of a photodiode and the
Unlike inorganic semiconductors, many organiC materi- problems involved). Only monochromatic light or white light of
als have no absorption edge, but are characterised by extremely low intensity (compared with AMI) was frequently
absorption bands. used for illumination and - provided it is at all given - the
Mobile charge carriers. As their molecular orbitals efficiency is corrected with a view to the transparency of the
overlap only slightly, organic semiconductors usually contacts which is often extremely low (between 0.1 and 3%). This
have narrow energy bands with low mobility of the results in misleadingly high efficiencies of up to 6%, because

58
almost all the organic solar cells that have been fabricated so far 4.2 Stability
are characterised by a very high series resistance. At high light As the stability of solar cells under light is of particular impor-
intensities, however, such a high series resistance reduces the tance, one would have expected that the researchers experi-
efficiency of the solar cell significantly. Besides, the great menting with organic materials would first test those materials or
majority of organic solar cells cannot be compared with each materials modifications to see whether they are stable under light
other (for instance, as far as their efficiency or their fill-factor before they start building solar cells. However, not a single article
is concerned), beca\lse they were tested under different lighting mentions such stability investigations. Instead, the cells were
conditions. When stating rectification ratios, many authors fail fabricated right away - probably in the hope that the semicon-
to give the voltages that go with those ratios. Others provide ductors used "will stand the test". This procedure is particularly
figures for series resistances, short-circuit currents and satur- unreasonable when the solar cells are tested in an air atmosphere
ation currents without at the same time defining the areas (which was probably done in most cases): Decomposition
considered. Efficiencies are stated, but neither the intensity, processes that should have been identified beforehand, then
spectrum and wavelength of the light is given, nor is there any become obvious only on the final product. Stability tests of the
information about what side of the cell was exposed to light. individual photoelectric elements were not usually carried out
Some researchers even illuminated only a small point on the either. Some authors report a conspicuous instability, even in
active area, or gave dark and light characteristics that related to vacuum, of exposed solar cells containing specific cyanine dyes,
different areas. In many cases nothing is said about the absorp· chlorophyll-a and doped polyacetylene.
tion spectra and absorption coefficients that the organic
semiconductors have in the visible spectral region. Only in a few 43 Efficiency
exceptional cases was the reproducibility of the various photo- The survey has shown that all organic solar cells that have so far
cells mentioned. been tested under sunlight are characterised by low and in part
even by extremely low efficiencies (1 % to 10.8 %), and that their
3.2 Interpretation of Results fill-factors and short-circuit current densities are correspondingly
Current-voltage curves, if any, are found only in few publications low. Efficiency values, however, that relate to monochromatic
- and in most cases their interpretation is incorrect. The IU light (which corresponds to the absorption maximum of the
characteristics often clearly indicate that back diodes, back-diode photoelectric semiconductor), or that relate to light intensities
breakthroughs,ligh t -dependent series resistances (photo-doping) well below the intensity of sunlight, are absolutely uninformative:
and shunts exist, or that there is no rectification. However, none Unlike "good" inorganic photovoltaic cells, the efficiency (and the
of the authors was able to read those signs correctly. Instead, fill-factor) of organic solar cells usually decreases dramatically as
many of them speculate about space-charge-limited currents, the light intensity increases. The drop in efficiency often amounts
photosensitisation, detrapping processes and other phenomena to some thousand per cent within a light intensity region of just
which they have not fully understood apparently. For example, a few mW/cm 2 . This is due, among other things, to the frequently
dark currents in short-circuited photoelectric cells (which are in observed very high series resistances which result mainly from
fact battery effects or decomposition currents) are falsely inadequ~te doping of the organic materials.
identified as "emptying of charge carrier traps", 'and low :Only few authors report that they investigated the
collecting efficiencies (when diodes are illuminated and a relation~hip between the efficiency of otganic solar cells and the
backward current is applied) are termed "space-charge-limited conditions under which the organic semiconductor were deposit-
currents". Some of the investigated cells had no rectifying effect ed, for example, the substrate temperature during sublimation.
at all (i.e. they were in fact no photoelectric diodes), or had Rarely have investigations been made to identify the influence of
back diode breakthroughs between 3 and 5 V, which the authors crystallinity or the effect of modifications in the crystal structure
mistook for rectifying effects. Organic solar cells have been compared with the amorphOUS state. Anyhow, many publications
prepared whose IU curves clearly indicate the presence of back give no information at all about the modification the dyes.
diodes and substantial series resistances. Instead of trying to
remedy these obvious faults and optimise the cells, the responsi- 4.4 FJ.!m Thickness and Doping
ble people fabricated dozens of additional photodiodes with new Many of the organic photoelectric cells have their maximum
organic semiconductors, always in the hope to hit on the right efficiency at a semiconductor film thickness which, when taking
material sometime. the corresponding absorption coefficient into account, are too
small to absorb enough light and which are in addition suscepti-
ble to shunts and shorts. This phenomenon too is typical of very
4 Details high series resistances, as is the frequently observed efficiency
increase at elevated temperatures. Accordingly, all specific
4.1 Materials, Purity conductivity values that are given in the publications are ex-
Only in exceptional cases was the connection investigated that tremely low (or the specific resistance values are extremely high).
may exist between the purity of the organic semiconductors and This means that the doping levels were apparently much too low,
their efficiency, and those who made such investigations arrived or that the doping efficiency was inadequate. Besides, doping
at completely divergent conclusions: Some authors have it that often resulted in battery effects, for example when iodine vapours
the photoelectric behaviour of dyes is not influenced either by served as doping agent. Again, the authors usually did not
chromatographic cleaning or by vacuum sublimation, while recognize the problems involved.
others complain about unsufficient purities. Attempts that have
been made at purifying the organic materials (in most cases 45 FJ.!m Deposition
none such attempts were made, however) were based on Apart from sublimation, four different film deposition methods
sublimation processes. As any sublimation may involve the were frequently used: spin coating (or solvent casting), the
simultaneous sublimation of impurities or a partial decomposi- Langmuir-Blodgett (LB)technique,electrochemicalprecipitation,
tion of the dyes, a control of the purification-effect, e.g. by and film depOSition from polymer dispersion.
chromotography, would have been mandatory. No reports on The spin coating (solvent casting) method and elec-
such controls have been found. trochemical precipitation are problematic in so far as a part of
the solvent cannot be removed and crystallisation may be non-
ur:ifcrm.

59
 The LB technique also leads to solvent residues, and it 1 per cent under sunlight as well as relatively high fill-factors (up
has the added disadvantage that voids may be produced which to 0.65). However, those cells too were investigated in air. Future
may act as shunts in the solar cells. Absolutely all solar cells that research ought to be performed under more suitable and better
have so far been fabricated by this method are characterised by defined conditions, in particular with controlled doping levels.
extremely low efficiencies even when exposed to very weak Efficient doping agents for organic p-type conductors, and
polychromatic or monochromatic light. especially n-type photoconductors (which have been rather
When the film is produced from a suspension of organic neglected so far), will contribute significantly to reducing the
semiconductor particles in a polymer solution there is again the series resistances and increase barrier heights. The main objec-
problem of solvent residues. What is more, the particles form tive in this connection must be to achieve sufficient and equal
point contacts which lead to very high series resistances. The doping levels for p and n semiconductors, as well as electron
diode structures can thus only be- generated on point-shaped work function of p and n materials that differ as much as
contacts, so that only very narrow pathways are available for the possible. It will be important that the right ohmic contacts are
charge carriers. chosen (to avoid back diodes). P-n solar cells made of two
Considering all these problems, it appears that sublima- different organic p-n dyes seem to be particularly advantageous:
tion in high vacuum is the only really promising way of making Due to the possibility of complementary absorption the sunlight
the films. However, many authors whose publications were spectrum could be utilized more extensively. In contrast to
analysed fail to give any information on the quality of the perylene/phthalocyanine p-n solar cells, p-n cells with polymeric
vacuum during sublimation. In particular, nothing is said about semiconductors such as polyacetylene so far have only very low
the residual gas pressure, which should lie in the region of 10.6 technical efficiencies (approx. 10.3 %).
mbar. Many researchers worked with considerably less MS and MIS photocells are more problematic than p-n
favourable pressures, though. photocells. On the one hand this is due to the absorption range
Some organic compounds appear to be stable enough of a single organic material, which is too small in most cases, and
to withstand the vapourisation temperatures unharmed and be on the other hand, it is attributable to the barrier contacts which
able to form an intact layer on the substrate. Contrary to what often have a transparency of only a few per cent .. (II1uminating
is done with inorganic semiconductors, however, the substrate through a transparent ohmic back contacts so far only resulted
usually was not heated, but was coated at room temperature. in lower efficiencies compared with those achieved when
illuminating through the semitransparent barrier electrode.)
4.6 Structure Especially those MS (or MIS) cells in which organic semiconduc-
Some authors report that a high fluorescence yield, planarity of tor particles are embedded in a polymer matrix are no real
the molecules and the presence of only few degrees of freedom candidates for developing efficient organic solar cells. (The
for rotation and oscillation, as well as a special crystal structure reasons for this have been given before.) MS and MIS photoelec-
(J-aggregates) play an important role in exciton generation and tric cells with polymeric semiconductors (of which at least
exciton lifetime. Molecular planarity seems in fact to influence polyacetylene is capable of absorbing light over the whole visible
the lifetime of the excitons, because the most successful of all spectral region) are not promising; they too are characterised by
organic p-n photovoltaic cells did contain planar molecules. A very low technical efficiencies of only some hundredths of a
corresponding influence of the fluorescence yield and J-aggre- percent, low fill-factors and - as far as the relevant figures are
gates on exciton lifetime has to be considered as speculation. given in the literature - insufficient stabilities (in particular
when doped). Another reason lies in the basically unfavourable
4.7 Ambient Conditions fabrication method: The polymer films that are produced from
Again, only few authors have recognised how important the solution are very thick and contain solvent residues, and the
ambient conditions are for organic solar cells. Most publications series resistances are very high. In 1978 a specific merocyanine-
contain no information at all on this aspect, but it may be based MS or MIS solar cell attracted general attention, because
assumed (and has in fact been confirmed by some authors that it was said to have an efficiency of 0.7% under AM! light.
were interviewed on this question) that an air atmosphere was However, it appears that this efficiency is irreproducible, so that
used in most cases, whose oxygen molecules have a doping effect others considered it to be coincidental. It should be noted in this
on organic p-semiconductors. This means that data on the exact connection that merocyanines are known for their instability
doping level are also lacking. The presence of oxygen is particu- under sunlight (even in an inert gas atmosphere) and should
larly critical in the case of organic p-n cells, because charge therefore be viewed rather critically. However, practically all
carrier generation in organic n-type semiconductors is probably those who reported on merocyanine solar cells concealed the
significantly reduced by oxygen. Organic p-n junctions that have problem. Nevertheless, a few merocyanine cells had technical
been produced in air are therefore often characterised by efficiencies under sunlight between 0.1 and 0.2 per cent. The
considerable differences in the doping levels of the partner other MS and MIS photocells based on chlorophyll a (which is
materials. unstable under sunlight), porphyrin derivatives, phthalocyanines
(with or without polymer matrix), metal phthalocyanines and the
4.8 Cost Consideration like are reported to have had low to extremely low efficiency
The highest technical efficiency values that have so far been values (10.2 to 10.8 0/0) even when exposed to very weak and
achieved with organic solar cells exposed to sunlight lie at about monochromatic light, as well as high series resistances and small
1% . In view of the low manufacturing cost, some authors think fill-factors (0.15 - 0.3).
therefore that the cells are competitive and can be used for As far as can be seen from the literature, heterodiodes
electric power generation. However, such assumptions are far incorporating an organic plus an inorganic semiconductor
from realistic. likewise had high series resistances, small fill-factors and very low
technical efficiencies (under sunlight). A relatively good value
(0.4%) was achieved with a photoelectric cell made of cadmium
5 The Present Situation sulfide and polyacetylene. It appears, however, that the major
The most promising of all organic solar cells investigated so far part of the charge carriers were generated in the cadmium sulfide
are p-n cells with a planar perylene derivative as the n-type in this cell. Contrary to what some authors think, heterocells with
conductor and a phthalocyanine as the p-semiconducting an n-conducting silicon wafer and a p-conducting organic
material. They were found in the tests to have efficiencies up to photoconductor film do ~ open the avenue to low-cost solar

60
cell production, because the silicon would have to be of the A Physical Processes
highest quality standard (which is relatively expensive). What is Generation of cbaIge carriers_ It is generally assumed in the
more, the photoelectric effect occurs mainly in the silicon. literature that charge carriers result from the generation and
Of basic importance is the question of quantum yield, subsequent dissociation of excitons. The exciton dissociation is
which appears to be determined by the· field-dependent dis- believed to take place in the electric field of a diode, and to
sociation of light-generated excitons. As far as the known increase as the field grows. This mechanism should be investiga-
organic photoelectric semiconductors are concerned, effective ted in detail.
exciton dissociation requires fields that ar:e impossible to achieve
in solar cells (10 5 - 106 V/cm). In one of the expert interviews, CJlluge carrier transport properties (mobility, life). Due to the
however, the parallel was drawn to xerography, which operates narrow bands, the charge carriers in organic materials are less
with higher field strengths: There is some hope that certain mobile than those in inorganic materials. Considerable discrepan-
modified organic materials exist that might have quantum yields cies exist between the mobility values measured on films and on
around 100 per cent, even given the lower fields in solar cells crystals, which may be taken as a sign of insufficient film quality.
(approx. 104 V/cm). Detailed investigations will therefore be necessary to clarify
Diverging statements are made about char:ge carrier whether and to what extent the charge carrier mobility in films
mobility. Whereas low values (for example, 10.5 cm2Ns) have may be approached to that in single crystals (which amounts to
been measured on films and solar cell structures, relatively high as much as 100 cm 2Ns in copper-phthalocyanine, for example).
values have been recorded on crystals. (The few mobility
measurements that have been performed on organic films seem Doping mechanisms. Many more variants of doping mechanisms
to be rather unreliable however.) appear to exist in organic materials than in inorganic covalent or
As the literature survey has shown, the properties even ionic semiconductors which need clarification ..
of the most efficient organic solar cells are such that they would
satisfy not even the minimum requirements on commercial solar B MateriaJs
cells. The scientific-technical situation concerning organic solar Selection of materials. The spectrum of candidate materials is too
cells is rather unclear at present. Single aspects have been broad at present. In our opinion, theoretical and experimental
investigated on a large number of different cells, but no in- work ought to concentrate on a few materials; a promising
depth investigation has been made. Research efforts so far have combination appears to be that of phthalocyanines and perylene
not focused on a specific material class whose physical parame- derivatives.
ters might be optimised with a view to solar cells. However, the
study has shown that many materials can be left out of consider- Preparation and purification. Material preparation and purifica-
ation right away, because their light absorption or their stability tion need quantitative studies. These should be orientated on the
is insufficient. Only few organic materials, if any, fulfill the goals of classical semiconductor technology, although purities of
essential requirements concerning stability, doping behaviour, 99.999 per cent, which are often required, will certainly be out of
feasibility of p-n diodes, etc. These are specially the phthalo- the question.
cyanines and the perylene derivatives which so far yielded the
best res\llts when combined in p-n diodes. (However, nothing Film deposition. For each material the best possible method
was found in the literature about any purposeful optimisation of must be found to produce films. Many candidate materials may
those materials.) Compared with research and development in be vapour-deposited, in which case the questions concerning
the field of inorganic solar cells, little effort is spent on indivi- purity (quality of the vacuum), modification change, and decom-
dual organic compounds or cell types. It appears that the position must be answered. Besides, after-treatments must be
organic solar cells are still at the stage where the amorphous sought for optimising the lattice with a view to the electronic
silicon cells were before 1975. properties of the films produced. The morphology and structure
of the films are important points to be investigated.
To sum up, it can be said that the problems posed by organic
materials have been identified in past efforts only in part so far, C Solar Cell Technology
and that rea\Jy promising approaches to organic solar cells are Doping, Contacting, Interfaces. Doping is a fundamental
stiJ1 lacking. However, ooncluding from this that efficient organic problem, as it is more difficult with organic semiconductor
solar cells are not fC3Slble would certainly be premature at this materials than with inorganic ones. Organic materials appear to
stage. be doped even by the ambient air. Selective chemical doping, e.g.
by additives, is of course preferable to uncontrolled doping. The
question of contacts is closely connected with that of doping, and
6 Recommendations this is one of the major problems of organic solar cells. In
The 'authors of this study believe that the investigations on heterodiodes an interface exists at the most critical point of the
organic solar cells ought to be continued, because the potential potential diode, i.e. where the first semiconductor contacts the
of this class of materials has not been fully probed yet. Some second semiconductor. Because of the (generally assumed)
materials and combinations of materials have already been mechanism of field-dependent charge carrier generation, this
identified that might be useful. Particularly promising are interface is of particularly great importance and needs more
combinations of phthalocyanines and perylene derivatives, which intensive consideration ..
can be used to make p-n heterodiodes and on which the highest
efficiencies under solar-intensity light have been measured so Stability of photoelectric properties. Similar as with amorphous
far. We therefore recommend that research and development silicon, it may be assumed that the question of stability or
should concentrate on these material groups in the future. On po~sible degradation will grow the more problematic the higher
the one hand, basic physical information should be gained in the efficiency values are that will eventually be achieved. Howe-
such R&D, and on the other hand it would start from a ver, this question ought to be paid sufficient attention right from
relatively sound basis in terms of material technology. the start.
The following specific aspects ought to be investigated:
NOTE: This study was supported by the German Fede-
ral Ministry for Research and Technology.

61
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON,PORTUGAL

OPTIKIZATION OF ANTIREFLECTION FILK STRUCTURES

FOR SURFACE-PASSIVATED CRYSTALLINE SILICON SOLAR CELLS
USING SPECTROSCOPIC ELLIPSOKETRY

Tadashi Saitoh, Osamu Kamataki and Shigeo lida

Tokyo University of Agriculture and Technology
Koganei, Tokyo 184
JAPAN

ABSTRACT

Antireflection film structures for surface-passivated silicon solar cells have been OPtimized
using complex refractive indices measured by spectroscopic ellipsometry. A maximum short-circuit
current density of 44.8mA/cm Z is calculated for a V-grooved surface with an MgFZ/SiNx:H/Si02
triple layer, while 43.1 mA/cm 2 for a flat surface wi th an KgF2/ZnS/Si02 layer. A maximum ef-
ficiency of 27.8% is predicted for a completely light-trapped, thin cell with a low surface recom-
bination velocity of 10 cm/s.

1. I NTRODUCT I ON tion wi th photolithography. The V-grooved and flat

surfaces were oxidized in dry oxygen at 850·C. As
Photovoltaic power systems have attracted much for antireflection films, three films of Ti02'
attention due to electricity shortage in mid-summer SiNx:H and ZnS were employed due to its practical
and the global environmental problems. Substantial importance. The Ti02 films were deposited on the
technological advances have been realized in high- oxidized silicon substrates at 300 . C using atmos-
efficiency crystalline silicon solar cells(1~3). pheric chemical vapor deposition. The hydrolysis
To realize a theoretical efficiency, researches on reaction for Ti02 is as follows.
surface recombination and surface passivation are
most important. In addition, antireflection film
structures appropriate to passivated emitters have
to be optimized for attaining a limit short-circuit The optical properties of the Ti02 films were
current density and conversion efficiencY. To cal- reported to change slightly by annealing at tem-
culate the limit current density, precise complex peratures more than 500"(; (4).
refractive indices of practical antireflection SiN x :" films were prepared on oxidized silicon
films have to be measured. substrates at 300"(; using a conventional plasma-CVD
Using spectroscopic ellipsometry, complex method. The OPtical properties did not change wi th
refractive indices of Ti02 and SiNx:H films have annealing up to 900"(; (5). ZnS and MgF2 films can be
been measured(4,5). The ellipsometry is superior to deposi ted easily in vacuum evaporation. The refrac-
conventional methods due to a lot of the measure- tive index of the ZnS films is higher as compared
ment points and the sophisticated computer programs with other two films.
for mul ti laser structures (6). The former papers
reported that the complex refractive indices for 2.2 Refractive index measurement
Ti02 and SiN x :" films tended to increase with the
decrease in wavelength. In addition, limit values Spectroscopic ellipsometry was applied to
for short-circuit current density and conversion measure real refractive indices, n, and extinction
efficiency were calculated for oxide-passivated, coefficient, k. The measurement was conducted using
flat and V-grooved cell structures. an ellipsometer developed by SOPRA. The measured
In this paper, the spectroscopic ellipsometry va 1 ues are tan", and cos /':,. as a function of
has been utilized to measure wide-wavelength-range wavelength, where tan", is the amplitude ratio of
indices for practical Ti02, SiN x :" and ZnS an- refraction and /':,. the phase difference. The
tireflection (AR) films. Short-circuit current den- measurement was carried out at 240 wavelenghs in
sities were calculated for both the V-grooved and the wavelength range of 250 to 850 nm. Linear
flat surfaces wi th triple MgF2/AR/Si02 layers. In regression analysis was used to calculate the
addition, a limit conversion efficiency was also precise indices by minimizing the standard devia-
investigated for surface-passivated solar cells. tion between measured and calculated data for tan '"
and cos /':,. •
2. EXPER I MENT AL
2.3 Calculation of Cell Characteristics
2.1 Film preparation
Using the complex refractive indices measured
A V-grooved structure was fabricated using a by the spectroscopic ellipsometry, short-circuit
selective etching in alkaline solution in combina- current density was calculated for both flat and V-

62
grooved surfaces with a triple layer of Si02' Afi
and KgF2 on aim-thick silicon substrate. The cal-
culation was carried out especially as functions of 3.0
Si02 and AR thicknesses under an AMI.5, global --ZnS
><
cu ---Ti0 2
solar spectrum. It is postulated that almost all
the carriers generated by light absorption can '"
~
·n --------- SiN
reach at a pn junction. cu '.
\
:>
Prediction of conversion efficiency for ·n
..... 2.5 \,
surface-passivated silicon solar cells was per- u ,,
~ ,
formed using a one-dimensional simulator, PC-ID(7). 4-<
cu '.
" ,, -~-~---
The calculation was conducted as a function of ~

--------.-~---.-~-.---
' - , - ... --
primary parameters including surface recombination ------------
2. 0
velocity, substrate thickness and back-surface- ..... O. 6
field. c:::
cu
·n
.,.,U
4-<
4-<
3. RESULTS AND DISCUSSION <I.J
0
U
~
o. 3
3.1 Reflectance of Surface-Passivated Structures 0
·n
.....
U
~
Firstly, real refractive index, n, and extinc- ·n
.....
tion coefficient, k, for Ti02' SiN x :" and ZnS films ~
>< SiN
were obtained by analyzing the measured ellip- 0
sometric data. As shown in Fig. I, the real refrac- 300 500 600 700 800
tive indices shown in the upper curve were found to Wavelength ( nm )
increase with decreasing wavelength. The index of
the ZnS film is the highest of the three. From the Fig.1 Variation of refractive index and extinction
lower curve on extinction coefficient, all the coefficient with wavelength for Ti02, SiN x :"
films are found to be transparent up to a and ZnS films.
wavelength of about 350 nm, but begin to absorb at
wavelenghs less than the wavelength for the Ti02 20r------------------------------------,
and ZnS films. The extinction coefficient of the
- - - ZnS
SiN x :" film is much smaller than those for the
other films.
Using the complex refractive indices, surface
reflectance spectra were calculated for flat and V-
grooved structures. The reflectance spectra for
AR/Si02 double-layer, V-grooved structures with an
Si02 thickness of 10 nm are shown in Fig.2. The
reflectance was calculated at the OPtimized thick-
ness of the AR films. The reflectance for a struc-
ture with a higher refractive index is smaller. If
the KgF2 film is applied for the V-grooved struc-
tures with the AR/Si02 films, the reflectance be-
Wavelength (run)
comes smaller and similar irrespective of varying
the antireflection film.
Fig.2 Comparison of calculated reflectance spectra
3.2 Short-Circuit Current Density for V-grooved, passivated surfaces with a
single antireflection film. The antireflec-
Short-circuit current density, J sc ' was calcu- tion film is Ti02, SiNx:H or ZnS.
lated for both the flat and V-grooved substrates
with MgF2/AR/Si02 films. First, the relation be- A relation between J sc and Si02 thickness was
tween the J sc value and Afi thickness was inves- investigated for flat and V-grooved surfaces with
tigated at an Si02 thickness of 10nm. The relation Si02 and HgF2/Afi/Si02 films. As shown in lower two
was found to have a maximum value at a respective curves of Fig.4, the J sc value for the passivated
Afi thickness as shown in Fig.3. The maximum values surface without an antireflection film
for the flat surfaces indicated as a dotted line tends to increase with the increase in Si02 thick-
were obtained at the ZnS, Ti02 and SiN x :" thick- ness and have a maximum at a Si02 thickness of 100
nesses of 45, 50 and 55 nm, respectively. The maxi- nm. The maximum value was 43.6 IA/cm 2 , a slightly
mum value of 43.1 mA/cm 2 was obtained for a struc- higher than the observed value of 42.9 mA/cm 2
ture wi th the ZnS fi 1m. As for V-grooved surfaces reported by Green's group(8).
indicated as a solid line, the maximum J sc values On the other hand, for ihe surface-passivated
were obtained to be about 44 mA/cm 2 at the Ti02 and specimen with HgF2/AR/Si02 films, the J sc values
SiNx:H thickness of 50 nm and ZnS thickness of 40 were higher than that for the surface without AR
nm. films and did not vary with Si02 thickness up to

63
 Using simulated data on the highest efficiency
46~------------------------------------, crystalline silicon solar cell(8). limit conversion

Table 1 Comparison of short-circuit current density

= for flat and V-grooved cell structures with
various antireflection films.
:;' 43
...E
Surface AR structure AR film
.s< ZnS Ti02 SiNx:H
~
Flat AR/ Si0 2 39.5 39.9 41. 0
40
KgF2/AR/Si02 43.1 42.7 42.5
- - - V-grooved
-------- HOI
V-grooved AR/ Si 02 43.8 43.9 44.1
KgF2/AR/Si02 44.7 44.7 44.8
370~--~2~0----~~~---ro~---~80------1~00
Thicknesses of Si02 and KgF2 are 10 and 100 nm.
AR film thickness (nm)

Fig.3 Variation of calculated short-circuit current

density with antireflection film thickness efficiency was estimated using PC-ID as a function
for flat and V-grooved surfaces. Thicknesses of surface recombination velocity and back-surface-
of Si02 and MgF2 are 10 and 100 nm. field. As shown in Fig.5. higher conversion ef-
ficiency is calculated if surface recombination
velocity at the front. SF. can be reduc~d less than
1.000 Cll/S. A back-surface-field(BSF) with a com-
46r-------------------------~
plementary error function doping profile is the
- - - V-grooved
best as compared with structures without a BSF or
with a uniform BSF.
A limit conversion efficiency was predicted
as a function of wafer thickness for a low

28,---------------,

40
\ SiO, 26 BSr(Erfc)

\" "\
/
---\--------------
>- no BSF -----,
g 24
, , \ ....u
u \
BSr( Un iforlll)
\
37L---------~--~--------~~\~'~~'~
1 10 100
SiO, thickness (nm)
22
Fig. 4 Comparison of short-circuit current
density as a function of 5i02 thickness
for flat and V-grooved surfaces with 20~ ____ ~ ____ ~ __ ~ ____ ~ __ ~

KgF2/AR/5i02 triple and 5i02 single films. 10 10 2 JOl 10'

Surface recolllbination velocity .Sr ( CIII/S )
about 10 nm. The Jsc values for the flat and V-
grooved surfaces tended to decrease at 5i02 thick- Fig.5 Variation of conversion efficiency with
nesses more than 10 and 20 nm. respectively. Table surface recombination velocity at the front
1 indicates the maximum J sc values calculated by for crystalline silicon solar cells with
optimizing AR structures for flat and V-grooved various back-surfaee-field structures.
surfaces. The optimized film structures are
KgF2/ZnS/Si02 for flat and MgF2/5iNx:H/5i02 for V- surface recombination velocity. A lower curve of
grooved structures. respectively. The latter J sc Fig.6 was calculated based on the experimentally
value corresponds to 99.6 % of the theoretical highest value at an SF of 800 em/s and at surface
value and is higher than the highest realized recombination velocity at the back. SB' of 10 CII/S.
value(8). Until now. such an approach to fabricate The former value is obtained from the data reported
the V-grooved cell with MgF2/ZnS/Si02 films has from the relation between surface recombination
been fabricated. However. the improvement of J sc velocity and surface concentration in doped
values resulted in the decrease of Voc values. layers(9). The curve showed that conversion ef-
ficiency can be improved slightly to 25% even if
3.3 Limit Conversion Efficiency wafer thickness is reduced to less than 100 /J- m.
However. if a new idea is developed to achieve SF

64
 352, 1990
(3) T. Uematsu, H. Ida, K. Hane, S. Kokunai and
28~----------------------~
T. Sai toh, ibid., PP. 344-347, 1990
(4) O. Kallataki, S. Iida and T. Saitoh, Proc. 21st
IEEE Photovoltaic Specialists Conference, 1990,
5.=5.-1 Oem/ s p.363
¥! 26 (5) O. Kamataki, S. I ida and T. Sai toh, Technical
Digests of 5th International Photovoltaic
1J
c
Science and Engineering Conference, Kyoto,
.....ucu S,-8DDcm/s
1990, p.497
......... SI-IDcm/s
(6) D. E. Aspnes, J. B. Theeten and R. P. H. Chang,
W 24 J. Vac. Sci. Technol., Vol.16, p.1374, 1979.
(7) P. Basore, D. T. Rover and A. W. Smith, Proc.
of 20th IEEE Photovoltaic Specialists
Conference, 1988, p.389
(8) J. Zhao, A. Wang and K. Green, Proc. of 21st
22~ ______~~~________~~
IEEE Photovoltaic Specialists Conference, 1990,
10 lQZ p.333
Wafer thickness ( ~ ) (9) R. R. King, P. E. Gruenbaum, R. A. Sinton and
Fig.6 Comparison of cell efficiency with wafer R. K. Swanson, ibid., 1990, p.227
thickness for substrates with resistivity
of 2.0 Q • cm.

of 10 cm/s, conversion efficiency can be improved

drastically wi th the decrease in wafer thickness. A
predicted maximu. efficiency is expected to be
27.8%. The extremely low value for highly doped
layers is expected to real ized through profound
fundamental investigations on carrier recolbination
and development of a new passivation process.

4. CQNCLUS IONS

Complex refractive indices of Ti02' SiNx:H and

ZnS antireflection films have been measured by
spectroscopic ellipsometry. The indices were found
to vary with wavelength in a shorter wavelength
region.
A maximum J sc value of 44.8 rnA/cm 2 was calcu-
lated by optimizing the HgF2/SiNx:H/Si02 films for
V-grooved surface structures. A predicted maximum
efficiency was 27.8% for a thin base cell wi th a
very low surface recombination velocity of about 10
Cll/S.

5. ACKNOWLEDGMENTS

We are obliged to Mr. Y. Kida, Hitachi, Ltd.

and Dr. H. Watanabe, Kyocera Corp. for preparing
Ti02 and SiNx:H films on Si substrates. This work
was supported by the New Energy and Industrial
Technology Development Organization, under the Sun-
shine Project of the Agency of Industrial Science
and Technology.

REFERENCES

(I) H. A. Green, A . W . Blakers, J. Zhao, A. K.

Hilne, A. Wang and X. Dai, IEEE Trans. on
Electron Devices, Vol.37, No.2, pp.331-336
, 1990
(2) R. A. Sinton and R. H. Swanson, ibid., pp.348-

65
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THE INFLUENCE OF SURFACE RECOMBINATION ON THE

LIMITING EFFICIENCY AND OPTIMUM THICKNESS OF
SILICON SOLAR CELLS

E. Demesmaeker, J. Symons, J. Nijs, R. Mertens

IMEC, Kapeldreef 75, 3001 Leuven, Belgium

Abstract
In this work the influence of surface recombination on the optimum thickness and the
limiting efficiency of silicon solar cells with optical confinement is studied. To this extent
a model is set up for the carrier generation rate as a function of depth in cells with a
Lambertian back side re1lector. This model is then incorporated into a one dimensional
solar cell simulator and current-voltage characteristics are calculated. Results of these
calculations are shown as a function of cell thickness, for different surface recombination
velocities and for different cell acceptance angles.

1. INTRODUCTION present at the back side. The surfaces are tex-

tured, so that the light is fully randomised inter-
In the past several studies have been pub- nally.
lished dealing with the limiting efficiency of solar
cells [1], and more in particular with the limiting
efficiency of silicon solar cells with optical confine-
ment [2, 3, 4]. In [2J and [3J optical confinement
is modelled by setting up an expression for the
Figure 1: Solar cell structure with textured surfaces, zero
absorption probability, using the assumption that front surface reflection, and unity reflection at the back side
the cell material is only weakly absorbing the in-
cident light (aW « 1, where a is the absorption We have shown [4] that, if a monochromatic
coefficient and W is the cell thickness). For sili- light beam with intensity Ia(A) falls on the solar
con however this holds only for thin cells and long cell structure of figure 1, the carrier generation
wavelengths. Therefore we have set up a more rate G(X,A) can be written as:
general model, in which the weak absorption con-
straint is omitted [4J. In section 2 this model is
briefly described and an expression is given for the
generation rate as a function of depth in the cell.
In order to investigate the influence of sur- where A is the wavelength of the incident light
face recombination on the optimum cell thickness beam, x is the depth in the cell, W the cell thick-
and the limiting cell efficiency, we incorporated ness, a the absorption coefficient and n the refrac-
our model into DIODAN, a one-dimensional diode tive index of silicon, and E2 and E3 are exponential
and solar cell simulator based on Gummel's al- integrals defined by:
gorithm [5]. Current-voltage characteristics were 00 1
calculated as a function of cell thickness with dif- En(t) = /, - exp( -yt)dy (2)
1 yn
ferent values for the surface recombination veloci-
ties and with different cell acceptance angles. The The absorption probability is then obtained
main results are discussed in section 3. by integration of the generation rate (1) over the
entire cell thickness W:
2. OPTICAL CONFINEMENT MODEL
(3)
The solar cell structure considered in this study
is shown in figure 1. It is assumed that there
is zero reflection at the front surface, and that
a Lambertian reflector with unity reflectivity is

66
 In the limit of weak absorption (0'W ~ 1) the face concentration equal to 2 10 19 cm- 3 were used.
absorption probability formula simplifies to [2,3]: The direct AM1.5 spectrum, normalized to 1000
W /m 2 was used, and the surface recombination
0'(.\) velocities sp at the front surface and Sn at the back
a(.\) = sin' 8 (4)
0'(.\) + 4n(~)'W surface were used as parameters. The following
cases were considered:
With both models (3) and (4) we repeated the
calculations of the limiting cell efficiency, as per- - Sn = sp = 0
formed in [3]. The results are shown in figure 2 for
- Sn and sp as calculated in [6] from the mea-
the direct AM1.5 spectrum. Mainly for large ac-
sured density of surface states and capture
ceptance angles our model gives a higher limiting
cross sections of electrons and holes: Sn =
cell efficiency, but the same optimum cell thick-
350 cm/s and sp = 5.5 cm/s
ness was found with both models.
33 - Sn = 1000 cm/s and sp = 10 cm/s
- Sn = 1000 cm/s and sp as calculated in [7]
from the measured emitter saturation cur-
32
dP
e: A5- rent of transparent P-diffused junctions: sp
>.
u = 1000 cm/s
c
....IV Except for the first case it was assumed that there
.......u 31
e = 30'
...
IV
is a high/low junction at the back surface of the
cell.
....
....
IV Calculations were performed for 3 different accep-
u 30 P ~ :'10'
tance angles: () = 90°, 30° and 15°.

Some results are summarized in figures 3, 4

29'1~-----------1-O----------l-O-O--------~l~000 and 5. From figure 3 it can be concluded that
cell thickness [f m) the short circuit current J. c is not much affected
Figure 2: Limiting cell efficiency as a function of cell thickness by the recombination velocities in the range con-
for different cell acceptance angles. Full lines: our model, sidered here. As expected J.c increases with de-
dotted lines: model from [2,3). creasing cell acceptance angle. The open circuit
voltage however depends strongly on the surface
3. OPTIMUM CELL THICKNESS AND LIM- recombination velocities, but not on the cell ac-
ITING EFFICIENCY: INFLUENCE OF SUR- ceptance angle (see figure 4). The corresponding
FACE RECOMBINATION cell efficiency curves show that, for given surface
recombination velocities, the maximum efficiency
In order to investigate the influence of sur- increases and the optimum cell is thinner when
face recombination on the optimum thickness and the cell acceptance angle is smaller. Comparing
limiting efficiency of silicon solar cells with opti- cells with the same cell acceptance angle, but with
cal confinement, the model described above was different surface recombination velocities (figure
incorporated into the one-dimensional diode and 5), it can be concluded that the maximum is less
solar cell simulator DIODAN. This allows to work pronounced (flatter curves) and the optimum cell
with any doping profile and to take surface recom- thickness increases with increasing surface recom-
bination and band gap narrowing into account. bination velocity.
DIODAN is based on the Gummel algorithm [5],
Table 1 gives an overview of the optimum cell
a self-consistent iterative scheme for the numerical
thicknesses and the corresponding efficiencies for
solution of the basic equations for semiconductor
the cell structure considered in this work.
structures.
With DIODAN, enriched with our optical con- ACKNOWLEDGEMENTS
finement model, the current-voltage characteris-
tics were calculated as a function of cell thickness.
E. Demesmaeker is supported by the I.W.O.N.L.
In this series of simulations a substrate doping of
(Instituut tot aanmoediging van het Wetenschap-
10 16 cm- 3 and a O.4/Lm deep emitter with a sur-
pelijk Onderzoek in Nijverheid en Landbouw). Part

67
of this work is sponsored by the EEC Joule pro-
gram under contract number JOUR·0040-C(JR)
(MonoCHESS).

REFERENCES
[1] W. Shockley and H.J. Queisser, "Detailed bal-
ance limit of efficiency of p-n junction solar
cells", J. Appl. Phys., Vol. 32, p 510, 1961
[2] T. Tiedje, E. Yablonovitch, G.D. Cody, B.G.
Brooks, "Limiting Efficiency of Silicon Solar
Cells", IEEE Trans on El.Dev., Vol. ED-31, No
Figure 3: Short circuit current density as a function of cell
5, P 711, May 1984
thickness for cells with optical confinement (legend: see be-
[3] P. Campbell, M.A. Green, "The Limiting Effi- low)
ciency of Silicon Solar Cells under Concentrated
740~--------------------~
Sunlight", IEEE Trans on El.Dev., Vol. ED-33,
No 2, p 234, February 1986 720
[]
[]
iii iii
...... iii ••
[4] E. Demesmaeker, J. Symons, J. Nijs, R. > 700 ••
Mertens, "The Influence of Surface Recombina- E
...... • • •t
tion on the Limiting Efficiency of Silicon Solar xx. Xx" t
(,,) 680
Cells", Technical Digest of the Int. PVSEC-5, 0 ·1 ••
":"
•
OliO

Kyoto, Japan, 1990, p 501

> II II
OOOQo
660 0

[5J H.K. Gummel, "A self-consistent iterative

640
scheme for one-dimensional steady state tran- 0 20 40 60 80 100
sistor calculations", IEEE Trans. El. Dev., Oct.
Cell thickness [~m]
1964, p 455
Figure 4: Open circuit voltage as a function of cell thickness
[6J W.D. Eades, R.M. Swanson, "Calculation of for cells with optical confinement (legend: see below)
surface generation and recombination velocities
at the Si-Si0 2 interface", J. Appl. Phys. 58 (11), L egen dfor fi~gures 3, 4 an d 5
1 December 1985, p4267 Sn [cm/s] Sp [cm/s] e= 90 0
e= 30 0
e= 15 0

0 0 • III [J
[7J R.R. King, R.A. Sinton, R.M. Swanson, "Stud-
ies of Diffused Phosphorous Emitters: Satu-
350
1000
5.5
10
.&

M
•+ A
)(
ration Current, Surface Recombination Veloc-
ity, and Quantum Efficiency", IEEE Trans. El.
1000 1000 • 0 0

Dev., VoL 37, No 2, Feb. 1990, p 365

68
 25 • • 26
•• • •
...... • • ...... • • •
~
......
0 0=90° ......
0~
0=30°
>-
(,)
24
.. .. ...... ... >-
25
..
(,)
c c

- .. ••• •
-
II)
.!!! •+
(,)
• (,)
:;::: • ++++
w • w
+ +
• • ••• •
23 24
20 40 60 80 100 0 20 40 60 80 100
Cell thickness [Il m] Cell thickness [Il m]

27
...... D
D D
D D

......
~
0
26 0=15°
>-
(,)
c . .. . .. . ..
II) x x x x x x

-
·u
:;:::
25 o 0 0 0 0 0 0 0 0
Figure 5: Limiting cell efficiency as a function of cell thick-
ness for cells with optical confinement, in function of surface
W recombination velocities, for 3 cell acceptance angles: (J =
90°,30° and 15° (legend on previous page)
24
0 20 40 60 80
Cell thickness [Il m]

Table 1: Optimum cell thickness d and corresponding efficiency 1'/ for cells with optical
. f unctIOn 0 f
con fi nement, m celaccepl I an d surf ace recom b'matlon veIOCI't'les
i t ance angle
Sn =0 Sn = 350 cm/s Sn = 1000 cm/s Sn = 1000 cm/s
sp =0 sp = 5.5 sp = 10 cm/s sp = 1000 cm/s
d [/lm] 17 [%] d [/lm] 17 [%] d [/lm] 17 [%] d fIlm] 17 [%]
() = 90° 55 24.9 75 24.0 > 100 - > 100 -
() = 30° 40 25.9 25 24.9 50 24.6 75 24.3
() = 15° 30 26.5 20 25.6 30 25.2 50 24.9

69
10TH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DEVELOPMENT OF COST-EFFECTIVE, HIGH-EFFICIENCY

CONCENTRATOR SOLAR CELLS

Cecile M. Leboeuf and John P. Benner

Solar Energy Research Institute
1617 Cole Blvd.
Golden, CO 80401 USA

ABSTRACT. As an alternative to flat-plate PV systems, concentrating PV systems

represent a viable approach for reducing the total system cost of advanced, high-efficiency
technologies. New conversion efficiency records have been set in several materials systems,
in both single and multi-junction configurations. Going from laboratory successes to large-
scale, cost-effective production of concentrator solar cells requires development of an
understanding of fundamental research issues to establish technology and process data needed
for the engineering of production equipment. As part of the U. S. Department of Energy
National PV Program, SERI has awarded several three-year projects to industry and
university research groups that focus on important issues for this laboratory/production
transition. These research programs were initiated in mid-I990, and are described, along with
their relevance to cost-effective PV concentrator development.

1.0 INTRODUCTION processing needed to complete cells (metallization, coating,

testing, and dicing) is not dramatically different from that
The near-term goal of the U.S. National PV Program is required for conventional silicon solar cells. Additionally,
to reduce the cost of photovoltaic ~enerated electricity to because much of the cost of epitaxial growth is related to
$0.12JkW-h. To achieve this goal WIth flat-plate systems, it cleaning and handling substrates, the cost of epitaxy will drop
has been calculated that in large, multi-megawatt installations, with increasing area.
modules of 20% efficiency could be nearly three times more The U.S. National PV program has historically directed
expensive than 10% efficient modules and produce energy for most of its high efficiency solar cell research efforts toward
the same cost (1). However, relative to current flat-plate improvements in crystal growth of III-V compounds. These
module prices, the industry still must both improve module efforts, managed by SERI, have led to significant
efficiencies and cut production costs in half. As an alternative improvements in the material quality, and hence efficiency, of
to flat-plate PV systems, concentrating PV systems represent III-V solar cells. New conversion efficiency records have
a viable approach for reducing the total system cost of been set in several materials systems, in both single and
advanced, high-efficiency technologies. The allowable price mUlti-junction configurations. Additionally, the SERI
for concentrator cells has been calculated (2) for a wide range research program, through its support of research and
of cell efficiencies and concentration ratios. At concentration development in crystal growth, will yield further reductions in
ratios from 400X to l000X, prices in the range of $4.00/cm2 epitaxial film deposition costs. In the long-term, substrate
to $9.00/cm2 are allowable. If efficiencies exceed 35%, the costs can be further reduced or eliminated by the elimination
allowable cell cost rises to $12.00/cm2. Although the target of the substrate itself. Research towards the use of lower-
range for costs represents a decrease of nearly five-fold cost substrates such as silicon (which involves issues of
relative to current prices, no radical changes in process lattice-mismatched heteroepitaxy), and thin film GaAs
technology are needed to develop large-scale production technologies (such as the Kopin Corporation process CLEFT)
facilities that can meet the goal. is advancing rapidly.
Table 2 summarizes the projections resulting from
progress discussed above, giving a range of projected costs
2.0 POTENTIAL FOR COST REDUCTION for high-efficiency III-V concentrator cells. The optimistic
column assumes that the present rapid pace of progress in III-
Key cost factors in III-V photovoltaics are: 1) cost of . V technology continues. The conservative projections require
epitaxial film deposition; 2) post-epitaxial processing; 3) only the development of 4-inch germanium or GaAs
substrate area; 4) substrate fragility; and 5) substrate cost. substrates, with the balance of III-V technology fixed at
Current production and pilot-production facilities for GaAs present levels. In the near-term category, 4-inch wafers also
solar cells use wafers of either 4.5 cm x 4.5 cm square or 2- are assumed, and it is expected that the efforts to improve the
inch and 3-inch diameter round dimensions. If we assume efficiency and performance of the epitaxial processes will
that the wafer processing after epitaxy needed to complete Ill- result in significant cost savings.
V cells (metallization, coating, testing, and dicing) is not
dramatically different than that required for conventional
silicon cells, these costs should not be greatly different except 3.0 RESEARCH PROGRAM
that substrate fragility may decrease wafer yield. Using a
silicon concentrator cell cost of about $2/cm2 (active area) and Going from laboratory successes to large-scale, cost-
the standard wafer production oftoday, we estimate $75 per effective production of concentrator solar cells requires
wafer for the price of these steps. As shown in Table 1, this continued development of an understanding of fundamental
represents a high cost per cell for the much smaller wafer research issues to establish technology and process data
areas used in III-V production. Table 1 provides an estimate needed for the engineering of production equipment. To
of current III-V solar cell costs. address this need, SERI has awarded several three-year
In the near term, the most significant cost reductions can projects to industry and university research groups that focus
be realized through the development and use of larger-area on important issues for this laboratory/production transition,
substrates. The standard wafer size in silicon is 4-inch as part of a program entitled, "Fundamental Research in
diameter, while the largest available GaAs wafer is 3-inch Materials and Structures for Ultra-High Efficiency Solar
diameter. Use of larger wafers, when they become available, Cells". Proposals for research projects were solicited in late
should reduce the post-epitaxial costs for GaAs solar cells to 1989 in five categories: 1) Growth of III-V compounds and
the same level as currently obtained in silicon, as wafer alloys; 2) Solar cell design, fabrication and evaluation; 3)

70
 Table 1. Cost Elements of I1I-V Concentrator Cells Today
Silicon GaAs

Wafer Cost $ 0.15 $ 1.50 - $ 5.00 1cm2

Epitaxy Cost $ 5.00 - $10.00 1cm2

Cell Processing $0.90 $ 1.50 - $ 4.00 1 cm2

Yield--wafer area to active cell area 60% 50%

Yield--electrical 90% 80%

TOTAL COSTS $ 2.00 $ 20. - $50. 1cm2

Evaluation of new growth tech·niques and operating regimes; achieve superior performance in single-junction and tandem
4) Evaluation of new materials; and 5) Evaluation of solar cells. Reaction chemistry of chlorinated metalorganic
characterization techniques. Awards have been made in three sources is under investigation to assess the capacity of these
of t~e five categories (2, 3, and 4), and one program is chemicals to permit lateral epitaxy of high quality films.
pending award in the first category. A brief description of
each awarded project follows: 3.4 Purdue University
Two related projects are underway at Purdue University.
3.1 Colorado State University The first is directed at examining new, thin-film approaches
Colorado State University in investigating in situ for achieving sizeable efficiency gains in high-quality, GaAs
generation of arsine and hydride radicals for lower- cells, and the second is focused on developing material-
temperature MOCVD growth. The plasma processing also tolerant cell designs to realize high cell efficiencies with
offers potential for lower-cost and safer production-scale modest material quality requirements. The emphasis is on
deposition processes. In a series of experiments at growth demonstrating new designs, building on the knowledge
temperatures ranging from 250°C to 400°C, they have gained from previous basic studies and on the in-house
reported the capability of epitaxial growth at 300°C using growth and fabrication capabilities that have been established.
plasma-enhanced MOCVD of GaAs. The aftergloW plume of
a hydrogen plasma in this process provides both vacuum UV 3.5 Spire Corporation
photons and hydrogen radicals to stimulate chemistry in the Spire Corporation is developing means to improve the
sources. efficiency of UI-V solar cells grown on silicon substrates by
reducing the defect density in the OaAs-on-Si material.
3.2 Research Triangle Institute Success in this area can result in the achievement of high-
Research Triangle Institute is assessing materials and efficiency OaAs single-junction cells, or GaAs-based
structures for ultra-high efficiency solar cells. The objectives multijunction tandems, on a low-cost substrate.
of this program are to develop the necessary technology to
grow high-efficiency GaInAsP layers which are lattice- 3.6 University of Southern California
matched to GaAs, to demonstrate high-efficiency GaInAsP The University of Southern California is investigating
single-junction solar cells, and to demonstrate GaInAsP/Ge large-area, atomic-layer-epitaxy crystal growth of III-V
cascade solar cells suitable for operation under concentrated materials, incorporating device design enhancements such as
(500X) sunlight. a Bragg reflector to pei-mit thinner cells than previously
possible without sacrificing efficiency. Thinner cells use less
3.2 Rensselaer Polytechnic Institute costly source materials, are less likely to crack from high
Rensselaer Polytechnic Institute is investigating: 1) new stresses during handling, and require shorter epitaxy times.
reactor development and analysis for MOCVD growth that
can minimize convection and increase source utilization
efficiency; 2) characterization of reaction kinetics of novel 4.0 CONCLUSIONS
precursors including chlorinated metal organic sources for
GaAs growth, and 3) doping of UI-V ternary GaAs growth. In addition to fast-paced research results in efficiency
records, the U.S. National PV program is developing the
3.3 Kopin Corporation necessary material and device data for industry to use in
Kopin Corporation is performing research on structures making the transition to large-scale production of high-
that will improve and exploit the capabilities of the CLEFf efficiency concentrator solar cells. Current markets for
technology to provide free.-standing, thin, single-crystal GaAs photovoltaic modules are not directed at large concentrator
films. Cell structures are being designed and fabricated to systems. However, when markets develop, the industry will

Table 2. Cost Elements of Future 111-V Concentrator Cells

Optimistic Near Term Conservative

Wafer Cost $ 0.10 $1.00 $ 2.001 cm 2

Epitaxy Cost $ 0.01 $ 0.50 $ 2.501 cm2

Cell Processing $0.90 $ 0.90 $ 0.901 cm2

Yield 50% 50% 50%

TOTAL COSTS $ 2.00 $ 5.00 $ 11.00 1cm2

71
be poised with the technology for production of ·high-
efficiency concentrator devices in part as a result of the
research program discussed herein.

5.0 REFERENCES
1. U. S. Department of Energy (1987). National
Photovoltaics Program Five Year Research Plan 1987-
1221, DOFJCHl0093-7.
2. Chamberlin, J. L., and D. L. King (1990). Conference
Record of the 21st IEEE Photovoltaic Specialists
Conference, p. 870.

72
!Urn EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON. PORTUGAL

ENHANCEMENT OF SUB-BANDGAP ABSORPTION IN X-SI SOLAR

CELLS BY ALLOYING WITH GERMANIUM

SA Healy, M. Taouk, B. Chan and M.A. Green

Centre for Photovoltaic Devices and Systems

University of New South Wales
Kensington, 2033, Australia

ABSTRACT A method of incorporating a Germanium alloyed layer within a thin film crystalline
silicon solar cell is described. There is an increasing awareness that thin film crystalline silicon
is ~ P'?tential contender for a future cost effective photovoltaic technology. This technology
~apltallse~ on the well proven performance advantages of crystalline silicon while at the same
time offerrng much reduced material costs. The weak absorption by silicon of the longer
wavelength components of sunlight necessitates effective light trapping· for thin film cells to
maintain high efficiency levels while the increased proximity of the collecting junction to the cell
surfaces require~ good surf~ce passivation. Effecti~e light trapping and surface passivation are
bot~ wel~ establls.hed techniques as demonstrated 10 the present generation of high efficiency
X-SI deVices. ThiS study examines a means of extending the potential of thin-film X-Si devices
by enhancing sub-bandgap absorption. This is achieved by incorporating a Ge-alloyed region
of reduced bandgap wit.hi.n the cell. In ~his reQion longer wavelength photons of sub-bandgap
energy are absorbed glvrng the potential for rncreased cell current. Low temperature liquid
ph~se epitaxy is used for the fabrication of these devices, allowing the growth of alloyed
regions of good crystalline quality and only a few microns in thickness.

1. INTRODUCTION accommodated by strain, there will be the formation of

misfit dislocations, introducing mid-gap defect levels
. ~hin film X-Si photovoltaic technology offers a and enhancing recombination. The intrinsic
promlsrng aye~u.e towards ~he cost goals photovoltaics concentration in the alloy will increase in proportion to
must meet If It IS to take ItS place as a major power e"EglkT, with ~Eg the bandgap reduction in the alloy,
generation technology. This technology is predicated increasing the saturation current in this region and so
on the utilisation of effective light trapping with which bringing down the open circuit voltage. For these
pathl~ng.th e~han~~ments of up to 5~ tl":les are possible
reasons only Ge-dilute alloys are considered to be a
In prrnclple In SIlicon. An Investigation of possible feasible proposition for this scheme. It is believed that
s~~emes to implement performance improvements for
th.e use of such alloys, in non-critical areas of the cell,
SIlicon cells (1) suggests sub-bandgap absorption Will boost the short circuit current without a significant
processes may have this potential. Such processes loss of voltage. A result which will be enhanced by
Include absorption via defect levels and absorptIon due effective light trapping.
to an effective bandQap reduction by heavy doping or
~.II~xing. Free. carner a~~orption will be a primary 3. EXPERIMENTAL METHOD
IImltrng factor In determining the viability of these
schemes, with the negative effect also of the increase in The technique used in this work, liquid phase
saturati?n c.urrent density appearing to rule out epitaxy, offers a number of advantages over other
generation via defect levels. The lower equilibrium methods; simplicity, flexibility, lack of expense and the
carrier densities possible by alloying with germanium ability to grow high quality material. Prior to the present
make this the most attractive option. work, LPE work with Si'.xGex alloys had been confined
In this study, a method of incorporating a Ge-aliOred to their growth onto X-Si substrates [4]. The
layer of reduced bandgap within a silicon cel is
described.. The technique used, low temper~ture liquid incorporation of an Si'.xGex layer within a silicon cell
phase epitaxy, enables the growth of deVice quality demands the overgrowth of such a layer by silicon. A
layers of thickness from one to several microns. PC-l D technique, briefly described below, was developed to
is used to .mod~1 cell:; incorpor~ting a Ge-alloyed region achieve this and full details of this are available in
and p-n Junctions rncorporatrng such a region are reference [4].
reported. The LfSE work utilised a standard graphite slider boat
technique, and was performed under vacuum. The boat
contained two solvent wells enabling the growth of the
2. SiGe ALLOYS - CONSIDERATIONS FOR THEIR two subsequent layers. The layers were grown onto
USE nominally <111> oriented commercial X-Si substrates.
The initial Si'.xGex layer was grown with a fairly low
In recent years, considerable effort has been cooling rate, during which the solution for the
extended on the study of Si'.xGex heterostructures subsequent growth of the Si layer is saturated with
grown primarily by MBE (2). This interest was fueled in silicon. Meltback of this alloy layer was avoided by
the mid-1980's by the discovery that the bandgap of an supersaturating the second growth solvent and
alloy layer grown to accommodate the difference in performing the growth with a fast cooling rate.
lattice spacing in adjacent layers via strain was much This technique utilised the low deposition
rec;luced relative to that of the unstrained alloy [3]. No temperatures «500·C) made possible by the use of a
eVidence has yet been found for the growth of strained gold-bismuth alloy (5) as the growth solvent. These low
layers by LPE, with the critical layer thickness for the temperatures minimise problems of contamination and
most dilute alloys (~10 at% Ge) of the order of a micron thermal mismatch as well as providing the necessary
which is approaching the minimum growth thickness growth conditions for this crystal growth technique.
attainable via LPE.
Where the difference in lattice spacing is not

73
4. RESULTS AND ANALYSIS [4) Healy SA, Young T.L., and Green M.A., "Low
Temperature Growth of Silicon on Sil-xGe x by Liquid
4.1 Crystal Growth Phase Epitaxy", J. Crystal Growth accepted for
Smooth uniform layers of Si on Sil-xGex were growth publication.
with no evidence of meltback of the alloy layer. The
layers were typically grown to a thickness of 2-5 [51 Lee S.H., Healy SA, Young T.L., and Green M.A.,
microns with the germanium content of the alloy 5-10 ·Very-Low Temperature Liquid-Phase Epitaxial Growth
at%. The layers were characterised by optical and of Silicon", Mater. Lett. 9 (2,3), 1990, pp 53- 56.
electron microscopy, with compositional analysis
performed by energy dispersive X-ray analysis (EDX).
For a fuller description see reference [4].
4.2 PC-1D Modelling.
PC-1 D has been used to model cells incorporating a
Ge-alloyed layer. Various configurations have been
considered but the most effective has been found to be
where the alloyed region occupies the bulk of the cell
with silicon forming only a thin region at the surfaces.
The results presented here were modelled with thin cells
(10 11m thickness) with the Vee limited by surface
passivation although they appear to apply to the
general case. Tentative conclusions have been made
regarding a threshold dEg beyond which Vee drops to
unacceptable levels and a surface field effect between
the alloyed and the silicon regions.
For a dEg > 3-4 kT Voc falls more than predicted by
the simple reduction in bandgap. This appears to be
due to band edge spikes impeding carrier flow once
they are more than a few times the thermal voltage.
There is some evidence to suggest a second
mechanism in which alloyed layers in heavily doped
regions lead to a large increase in recombination.
The interface between the alloyed region and silicon
regions demonstrates a surface field effect with effective
surface recombination velocities reduced proportional
to e-6Eg/kT. In combination with the enhanced red
response such cells have demonstrated a power
increase of 14% relative to cells without an alloyed
region. This increase is approximately double that
given by an analytical model without the surface field
effect.
4.3 Experimental Devices
Initial device work has succeeded in the growth of
functioning p-n junctions incorporating Sil-xGex layers
although the first cells fabricated from these have
demonstrated only low values of Vee with large shunt
and series resistances.

5. SUMMARY
Preliminary rnodelling and device work indicates that
performance improvements are possible in silicon cells
by incorporating a layer alloyed with germanium. In
addition to the enhanced red response reductions in
effective surface recombination velocities are possible
due to a surface field effect between the alloyed and the
silicon regions. An effective method has been
developed for incorporatinQ germanium alloyed regions
within silicon cells, utiliSing the technique of liquid
phase epitaxy.
REFERENCES
[1) Green M.A., "High Efficiency Silicon Solar Cells",
Trans Tech SA, Switzerland, 1987, pp 33-36 & pp 105-
109.
[2) Bean J.C., "Technological Prospects for Germanium
Silicide Epitaxy", Mat. Res. Soc. Proc. 126, 1988,
pp111-121 and references therein.
(3) Lang D.V., People R., Bean J.C., and Sergent A.M.,
"Measurement of the band gap of GexSil_x/Si strained
layer heterostructures", Appl. Phys. Lett. 47 (12), 1988,
pp1333-1335.

74
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

NEW CONCEPTS IN 2·DIMENSIONAL SOLAR CELL MODELING

J.C.Jimeno, S. Uriarte, J.J. Zamora and C. Icaran

Dpto. Automatica, Electr6nica y Telecomunicaci6n.

ETSIIIT de Bilbao, Universidad del Pais Vasco
Aida de Urquijo sin, 48013 Bilbao (Spain)

Abstract. 2-dimensional modeling is essential for the analysis and optimization of the electrical behaviour of high-efficiency
solar cells. The conventional resolution of 2-dimensional models implies high data storage costs and running time which cannot
be dealt with by a personal computer. A solution is proposed to resolve this problem. It consists of dividing the cell into quasi-
neutral and dipolar zones. As the emitters are found in low injection and the contribution of the dipolar zones to the total current
is small, all these zones can be modelled analytically without producing large marginal error. The base of the cell is resolved
numerically and the grid extends preferencially in this region.

1. INTRODUCTION f2(n,p,E) = v(~nE) + v(DnVn) + G - R = 0 (7)

Computer-aided modeling is essential for the analysis and f3(n,p,E) = VU!ppE) - v(Dpvp) + G - R = 0 (8)
optimization of the electrical behaviour of solar cells. In 1985,
the PC-I0 appeared, the first computer program prepared As this is a system of differential non-linear equations, their
specifically for modeling solar cells, which can run in personal solution is not possible analytically unless approximations are
computers 11/. This program is capable of modeling single- made. The exact resolution is only possible by means of
dimensional devices with up to 3 terminals. Nevertheless, high- numerical methods. To do this, finite element and finite
efficiency solar cells have multi-dimensional structures /2-4/. difference methods are used. The latter method is simpler and is
The more recent versions of PC-I0 include some 2-dimensional also the method which is used more often for resolving
effects 151, nevertheless these are not sufficient for modeling the semiconductor devices and this will be the method we shall
complex structures of high-efficiency cells. The need for making utilize in this paper.
new models which incorporate 2-dimensional effects has been The resolution of a problem by finite differences consists
demonstrated previously /6/. of several steps: in the first place, the space is discretized with a
A large variety of programs which model semiconductor grid. Secondly, differential operators are replaced by differences
devices in 2 or 3 dimensions have been developed during the in the values of the variable in the nodes of the grid. In this
70s and 80s n-9/. However, all of them require computers with way, vv in node xi,yj can be expressed as
large calculation capacities. The object of this paper is to study
the feasibility of implementing 2-dimensional models in personal Vi+1/2,j-Vi·l/2,j Vi,j+1/2-Vi,j·1/2
computers. VVi,j= + (9)
Section 2 summarizes the system used in classical 2- kj
dimensional simulation models of semiconductor devices. An
estimate of the storage needs and the number of mathematical with hi=Xi+1/2-Xi.l/2, kj=Yj+1/2-yj.1/2.
operations to be made demonstrates the non-applicability of The error which is committed when linearizing v in
these models in personal computers. The resolution of relatively interval hi, kj is proportional to the second derivative ofv and to
simple cases would lead to the total occupation of the memory the size of the grid (hi, kj). The variables linearized usually are
and excessively long execution times. Section 3 analyzes a new the electric field and the current densities for both carriers. As
reformulation of classical semiconductor equations which may the electrical field varies very rapidly inside the dipolar zones, it
be more efficient for simulating solar cells. The approximations is necessary to use a fmer grid for the accurate resolution of
which can be made in order to simplify the use of these these areas. Figure 1 shows a typical grid for the analysis of
equations are also studied. finite differences of a pn junction. It can be observed that the
number of nodes used in the resolution of dipolar zones is
greater than 30% of the total nodes, which is the usual case.
2. THE NUMERICAL PROBLEM The discretization of the system (6-8) and its application
on the N=Nx.Ny nodes creates a system of 3N equations (flij,
The behaviour of a semiconductor device is described as a f2ij and f3ij, for each node xi-yj) with 3N unknowns (the values
set of five equations: Poisson's equation, electron and hole of n, p and E in each node), This system of equations is not
current equations, and the continuity equations of both carriers.

fs V E = q(NDtp-NA-n) (1) Q

In = <wnnE+ qDnvn (2) Emill('r -l-

N' -
Jp = 'lIlppE - qOpvp (3)

-vJn/q -G + R + lin/lit = 0 (4)

vJp/q -G + R + lipllit = 0 (5) I

P~~ype
By introducing (2) in (4) and (3) in (5), we obtain two
new equations which do not depend on In and Jp. The resulting I
system of 3 equations with 3 unknowns (n, p and E ), can be
expressed in stationary conditions (Ii /lit=O) such as

fl(n,p,E) = fs VB - q(NDtp-NA-n) = 0 (6) Figure 1. Classical grid used to solve a diode structure.

75
linear and so its resolution requires an iterative method. The 2.1. Decoupled solution technique
method which is usually utilized is the Newton method, to Gummel 111/ has proposed a plan for the decoupled
which slight modifications are made to ensure convergence /10/. solution of the system of equations (6-8). To do this, it is
Its application, for example, to equation f'l on node ij and on assumed that quasi-Fermi potentials are known and the electrical
the iteration k, leads to: field is calculated using Poisson's equation. With the solution
obtained, the concentrations of electrons and holes are
k N~y ( k+1 lif'lij I,.. k+1 lif'lij It k+1 l)f'lij 1,..) calculated, resolving both continuity equations in succession.
0=f2ij+2. ~Im - r +t\plm - - I +AElm--r (10) This process is repeated until a convergency criteria is satisfied.
1m =1,1 linlml liplml IlElml With this method it is only necessary to resolve 3 systems of N
equations and N unknowns instead of the original of 3.N
where the lif2ij/l)vlm element is a measurement of the influence equations. This process is between 9 and 27 times less costly in
of variable v in node I,m, on the function f2 in node ij. In a CPU time, than the coupled solution of the equations /12/.
conventional problem only variables in adjacent nodes, plus Compared to the coupled resolution technique a case of.2Ox20
those of the node itself, affect the functions of the node. nodes implies the resolution of 3 systems of 400 equations and
Thereby, equation (10) will have, in addition to the independent 400 unknowns each. This means that about 16,000 elements
term, 9 non-null terms for the single-dimensional case and 15 must be stored and 320,000 operations must be made for the
for the two-dimensional case. LU decomposition of the matrix corresponding to each system
The application of (9) and (10) on (6)-(8) leads to a This is 10 times less costly than the coupled solution.
system of 3N linear equations: Nevertheless, when any of the zones of the device
approaches high injection, the (1)-(3) equations begin to couple
K·V=-F (11) and the convergence process slows down and this process is
even slower than the coupled solution of (6-8) /12,13/. As it is
where matrix V represents the increases in variables for the usual for the base of high-efficiency cells to be in high injection,
iteration k+l, matrix F the values of the functions in the N this will not be a valid procedure for its modeling.
nodes and the k iteration, and K is the characteristic Jacobean
matrix of the system. In a one-dimensional problem with N 3. SOLUTION PROPOSED
nodes, K has a 3N·3N dimension and shows 11 grouped
diagonals next to the main diagonal. In a 2-D problem, with The previous section shows that the coupled resolution of
Nx.Ny nodes, K has the 3NxNy·3NxNy dimension and shows the equations of a semiconductor is. not viable for
17 diagonals, 11 next to the main diagonal and the other 6, in implementation in small computers. However, the classical
two groups of three at a distance of 3Nx from the main plans of decoupled equations are even slower than coupled
diagonal, as is shown in Figure 2. equations when a zone of the device enters in high injection.
The first idea will be to find a system of equations which
The resolution of (11) is the process which takes up more dermes the semiconductor and which remains decoupled, both
time and memory in modeling. One of the more appropriate in low and high injection conditions.
methods to do this is LU decomposition. In this, K=L. U, Secondly, the cell emitters are found in low injection in all
where L and U are triangular matrices from which the solution practical cases, due to which they correspond to linear
can be found easily. Matrix L+U -I, where 1 is the identity differential equations, and can be simulated with analytical
matrix, possesses all the information on matrices L and U. models. On the other hand, dipolar zones are highly non-linear
Figure 2 shows matrix L+U -I corresponding to a 2-D problem regions, all the variables evolve very quickly internally and their
of Nx·Ny node dimensions. It can be observed that the elements modeling is complex. Nevertheless, their effect on the total
of this matrix are null outside the most external diagonals on behaviour of the cell is insignificant, and only have slight
matrix K. Nevertheless, there has been considerable filling of relevance at low voltages.
matrix L+U-I with respect to the original, K. Some special The solution which is proposed here consists of using a
matrix permutation techniques (the nodes are ordered following numerical model for the base region with decoupled equations in
a special sequence), minimize the filling of the matrix and the low and high injection and analytical models for dipolar and
number of operations to be made. However, their analysis is not emitter regions.
a trivial problem and as their advantages are identical over any
of the methods described here, they will not be taken into
account. Therefore, the number of elements to be stored is 3.1. Decoupled equation system
proponional to the size of the main diagonal, D, and to the The analytical resolution on one dimension of the base of a
matrix bandwidth, 2d, and the number of floating point solar cell under variable injection conditions, has been proposed
operations will be the product of the number of elements previously /14/ and is widely used /15/. An ambipolar continuity
multiplied by the average number of operations per element, equation has been obtained under the hypotheses of charge
which is about d. For a typical I-D case with 80 nodes, it will neutrality and uniform doping. Its generalization is simple based
be necessary to store about 2200 elements and about 10,000 on the system (1-5). With the use of (2) and (3), the electrical
floating point operations will be carried out to calculate L+U-I. field can be found, E-=En+ED (En is the portion of E proportional
In the 2-D case, with 20x20 nodes about 150,000 elements to the total current density), and using vJ=O, the equation (13) is
must be stored and about 10,000,000 floating point operations obtained. Poisson's equation completes the system of three
must be made. The latter amounts cannot be dealt with by a equations with three unknowns (n, p and E in bases type p):
personal computer.
D Vijtnn) V(DnVo) (EV(J.In(P+Na)+DnV(P+Na»)
D =3~ x Ny EjJ.E --+j.lm - - - - I1P =R-G
r.--:ra-=
_3 N-x
1 r·....
~ ~ ~
~
(12)

VEn + EO Va/a =0 (13)

p/Na=A,2VF/VT (14)

with

J.l.ef = J.lp J.LnICJ.Lpp + ~n) ; J.lE = J.l.ef (p - n)

j.lm =J.lef( p + n) lip =J.l.em

p =p - n- Na a = q(J.lpp + ~n)
Matrix K Matrix L+U -I
Figure 2Matrices in a conventinaI 2-DimensionaI problem. A,2 = rsVT/qNa

76
It can be seen that the ambipolar equation (12) remains only I~~------------------------------~
slightly coupled from the other two (13 and 14) both in low and
high injection. Thus, for the very low injection limit , it is
converted into the traditional minority carrier continuity
equation:
1000
Dnv2n =R- G (15)

and for high injection, the following is obtained:

2VTIlhiV2n = R - G + EIlhiV(P + Na) - VIl.lhiV2(P + Na) (16)

with 1Jhl= I1p·J.IlIIUlP+jUI), that in neutral regions depends very

weakly on the p and E variables. In any case, the dependence is
proportional to VlE, which in conventional models /101 is often
disregarded (VE is constant in each interval). Injection
The low coupling which exists between equations (12)-
(14) allows their resolution to be made successively. To do this,
in iteration k, nt, r/' and Ek being known, with the use of (12)
Figure 3. Mobilities in the ambipolar continuity equation vs.
nk+1 can be calculated. With ot+1 and r/', Eok+1 can be obtained
injection level.
and USing (13), Eat+I. Lastly, with (14) the new charge value is
calculated, r/'+I, it being possible to make the following iteration.
3.3. Emitter regions and dipolar zones
3.2. Resolution of the am bipolar continuity equation The resolution of the emitter and dipolar zones, unlike the
The resolution of the ambipolar continuity equation is the base region, can be made with analytical models although the
most delicate process of the entire numerical problem. Its aim is grid extends across these.
to find a function which relates minority concentrations in a For the emitter, as it is always found in low injection, the
specific node, nij, with the adjacent nodes. In this way for some minority current density will be linear with the injection level on
specific surrounding conditions, the minority profile is found in the edge between the emitter and the dipolar zone. For a type n
the entire base region. emitter we write:
In order to resolve the equation (12), it is necessary to
make some simplification. Traditional methods coincide in pij - po
considering insignificant the generation and recombination Jpij = Jpsatij - - - + JLij (19)
within each interval. We consider that the coefficients in the po
differential equation are constant within the interval, Xi+h;f2,
xj+kj/2. Figure 3 shows the variation of the different parameters with
with the injection level. It can be observed that variation is slow, 'lie -Vh
and therefore the previous condition may be satisfied in large pij = po exp - - -
intervals. Under these conditions, equation (12) can be re- Vr
written: 'lie and Vb being the electrons and holes quasi-Fermi potentials,
Jpsatij, the saturation current density in node ij, placed on the
2JWn + v2n - n/L 2 = - G'(X,y)/VT jI.m (17) edge of the dipolar zone, and JLij, the photogeneration current
density at this node.
with On the other hand, Dember's field,ED, is insignificant in
low injection due to which in emitters equation (13) can be
P = JLE F/2VT jU\I ; L 2 = VT jU\I n/R written as:

G' = G + jLp(V1V2(p + Na) - Ev(p + Na» Va

VE+E- =0 (20)
As the coefficients are only constant in the Xi±hi/2, yj±kjl2 a
interval, the solution can be obtained in accordance with the
concentration of minority carriers on the edges of the interval The resolution of the emitter will be limited to resolving (20) in
(ni+I/2,j, ni-1/2,j, ni,j+1/2 and ni,j.I/2) and not in the adjacent intermediate nodes and (19) in the nodes situated on the edge of
nodes as it is intended. The solution to (17) will be: the dipolar zone. The value of ljIe and IjIb will be obtained from
the electrostatic potential calcu1ated for (20)_
sh(axhil2) pxb;f2 - Ihh;f2 In dipolar zones a good approximation will be to consider
nij = (ni+I/2,j e + ni-I/2.j e )= that quasi-Fermi potentials remain constant through these in the
sh(axhi) direction that they are crossed by the currents_ The introduction
of a recombination current, Jr, of the type
sh(aykil2) Pyki/2 - Byki/2
= (nij+i/2 e + 11i.j-i/2 ~) (18) Ve -Vh
sh(ayki) JR = JRsat [exp - - - - I] (21)
mVT
with
1\ 1\ will allow very small errors to be obtained in the assessment of
P=lhi+Pyj III} = 1/Lx2 + l/Ly2 these zones.

ax2 = 1h2 + 1/Lx2 ay2 = py2 + I/Lr

4. DISCUSSION
Lastly, the continuity of minority current on the edges of the
intervals provides the relationship between nij, the concentration Figure 4 shows the grid necessary for the resolution of the
of minority carriers in the adjacent interval, ni+i,j for example, pn junction of Figure I. It can be observed the non abundance
and the minority concentrations on the edges between intervals, of grid in the dipolar area because its analytical solution. It
ni+i/2.j.

77
 Q ACKNOWLEDGMENT
I !hi~ w~rk has been done under the sponsorship at the
Em il!. r Umversldad del Pais Vasco" and their contract UPV 143345-
w b-- ~
I~

EI04/90. .

REFERENCES

B~ype
e /1/ D.T.Rover, P.A.Basore and G.M.Thorson. "Solar Cell
p- Modeling on Personal Computers", 18th IEEE Ph. Sp
I Conf., 1985.
/2/ ~.A.Sin!on, Y.Kwark, LY.Gan and R.M.Swanson.
27.5% SI Concentrator Solar cells ", IEEE Electron Devices
Letters, EDL-7, No. 10, pp.567-569, 1986.
/3/ A.W.Blakers, J.Zhao, A.Wang, A.M. Milne, X.Dai and
Figure 4. Grid and structure used in the proposed method to M.A.Green. "23% Efficient Silicon Solar Cell" 9th
solve the diode in Fig .1. European Comunities Photovoltaic Soiar Energy
Conference, pp. 328-329, 1989.
/4/ A.Cuevas, R.A.Sinton and R.M.Swanson. "Point and
leads to a considerable reduction of the total number of nodes. Planar Junction P-I -N- Silicon Solar Cells for Concentration
On !he. other hl;IDd,. it will only be necessary to resolve the Applications. Fabrication, Performance and Stability" 21st
conunulty equatton m the base region (176 nodes of a total of IEEE Ph. Spec. Conf., 1990. '
225 in the example shown in the figure). /5/ P.A.Basore. "Numerical Modeling of Textured Silicon Solar
. Eq?a?o~ (13) is valid .for the entire grid, although in Cells Using PC-ID", IEEE Trans. on Electron Devices,
eontters It IS litnlted to resolvmg a fixed network of resistances ED-37, pp. 337-343, 1990.
and in dipolar zones is converted into a series of diodes the /6/ R.M.Swanson. "Point-Contact Solar Cells: Modeling and
currents of w~ich. ~ the sums of those obtained in (19), (21) Experiment". Solar Cells, No.17, pp. 85-118, 1986.
and that of mmonUes on the edge of the base with the dipolar n/ E.M.Buturla,P.E.Cottrell, B.M.Grossman and
~ne calculated through the profIle obtained with (17). Lastly, it
K.A:Salsburg. "Finite-Element Analysis of Semiconductor
will onI ~ be necessary to obtain the electrical charge, p, in the DeVIces: The FIELDAY Program", mM J. Res. Develop.,
base region. Vo1.25, No.4, pp. 218-231, 1981.
In order to compare this with the traditional method, the /8/ W.Fichtner, D.J.Rose and R.E. Bank. "Semiconductor
resol~tion of the case shown in Figure 4 would require the
Device Simulation", IEEE Trans. on Electron Devices ED-
soluUon of the following equations: a) the continuity equation 30, No.9, pp. 1018-1030, 1031-1041, 1983. '
on 176 nodes of the base region which would require about /91 W.L.~ngl, H.K.Dirks and B.Meinerzhagen. "Device
80,000 floating point operations per iteration, b) the electrical Modeling". Proc. IEEE, Vol.71, No.1, 1983.
curre~t on 225 nodes, wh~ch would require about 100,000
/10/ S.Selberherr. "Analysis and Simulation of Semiconductor
?pe~u?ns and c) the resoluuon?f base electrical charge, which
Devices", Springer-Verlag Wien New York, 1984.
IS tnvlal from E onwards. Wah these calculations, if the /11/ H.K.Gummel. "A Self-Consistent Iterative Scheme for one-
convergence process should not be affected, the method would Dimensioal Steady-State Transistor Calculations". IEEE
be about 50 times faster than the conventional method. Trans. on Electron Devices, ED-II, pp. 455-465, 1964.
/12/E.M.Buturla and P.E.Cottrell. "Simulation of
Semiconductor Transport Using Coupled and Decoupled
S. CONCLUSIONS Solution Techniques", Solid State Elec. Vo1.23, pp. 331-
334,1980.
The study of conventional simulation models of semiconductor 113/0.Mank, H.H.Heimeier and W.L.Engl. "High Injection in
devices in 2 dimensions shows their non-applicability to small a two Dimensional Transistor". IEEE Trans. on Electron
computers. Due to this, it is only possible to use models with Devices, ED-21, pp. 403-409, 1974.
decoupled resolution of the equations both in low and high 114/ A.Luque and G.L.Araujo. "Variable Injection Model for
injection. This paper has presented a model for these Bifacial Cells", ISES'85 Intersol Conference 1985.
characteristics. ~e us~ of simplific.ations, using analytical 115/ A.Luque. "Solar Cells and Optics for 'Photovoltaic
models for the .slmulauon of the emuters and of the dipolar Concentration", Adam Hilger, 1989.
zones allows this method to be much faster than conventional
methods without adding considerable errors.

78
10TH EUROPEAN PHOTOYOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

HYDROGEN EVOIlJTION ON CATALYST-COATED PHOTOELECTRODES p-InP and p-si

J. Schefold1 and M. Specht 2

1) Institut fur Physikalische Elektronik, Universitat stuttgart
2) Solar and Hydrogen Energy Research center (ZSW)
0-7000 Stuttgart 80, FRG

Abstract
Properties of the p-InP-Pt- and p-Si-Pt-electrolyte junctions are compared by
means of capacitance- and impedance measurements under hydrogen evolution condi-
tions. Discontinuous metal coatings leed to high barriers and efficient charge
transfer. The model of a Schottky-type solar cell in series to an electrochemical
reaction step is confirmed by impedance data for both semiconducting materials
and irrespective of deposition methods or electrolyte pH. Barrier heights and
diode ideality factors are derived under photocurrent flow and current-voltage
curves are given for H2-evolution and for a photoelectrochemical solar cell.

1. Introduction

Semiconductor photoelectrodes can be recalculated to the real hydrogen potential

used for direct solar to chemical energy (RHE). A 150 W tungsten halogen la~p served
conversion. Hydrogen evolution efficiencies as a light source. Impedance data were ta-
exceeding 13 % are reported for Rh- coated ken under potentiostatic control using a
p-InP [1] and 10 % for Pt-coated p-Si [2], Solartron 1255 frequency response analyzer.
but long term stability has to be improved.
Under photo current flow the shape of iu-
curves generally differs from those of so- 3. Deposition of IReta I coatings
lid state solar cells, the difference being
strongly dependent on the presence and the Vacuum deposition of Pt-films was done
amount of metal on the surface. by sputtering or electron beam evaporation
Large barrier heights are well documen- at a base pressure of 10A-9 robar. The thik-
ted both for p-InP-metal- [3] and p-InP- kness of the films was controlled by a
electrolyte contacts [1], whereas the for- quartz crystal sensor and values were com-
mation of a high barrier at the p-si-pt- pared to transmission measurements assuming
electrolyte junction contrasts solid state the same coverage of glass and silicon.
p-si-pt data [4,5]. Another matter of dis- Electrochemical deposition (formation
cussion is how the influence of unsuffi- of platinum islands) was done by photodepo-
cient charge transfer and enhanced surface sition from aqueous H2PtC16. The morphology
recombination can be distinguished. A com- of the Pt-islands is strongly influenced by
parison of both electrodes seemed therefore preparation parameters [2]. Best results
to be of interest. current-voltage curves, were achieved for p-si with galvanostatic
Mott-Schottky- and impedance data for Pt- deposition at -1 mA/cm2, a light intensity
coated electrodes are given, and emphasis of 5 mW/cm2 and a Pt quantity deposited
is put on barrier properties (electrochemi- corresponding to a charge of z100 mC/cm2.
cal reactions are discussed elsewhere in The Pt-deposition on p-InP was done while
more detail [6]), as the characterisation scanning the electrode potential between
of semiconducting materials in electrolytes the open circuit potential and the photo-
is thought to be a promising method not ne- current saturation region at illumination
cessarily limited to solar cell applica- levels which correspond to i h z -1 mA/cm2
tions. (total cathodic charge 1 -200 Pmc/cm2 equi-
valent to about 1 - 200 monolayers of Pt
assuming a current efficiency of 1).
2. Experimenta1

Single crystal boron-doped <111> sili- ". Barrier heights

con and zinc-aoped InP <100> with dopant
concentrations of z2·10 A15/cm3 (p-Si) and Mott-Schottky measurements yield values
zl.10 A17/cm3 (p-InP) were used as electro- for the flatband potential ~Fb (which may
des. Fermi levels are assumed to be above be taken as equivalent to the diffusion po-
the valence band by 230 meV (p-si) and tential ~D for solid state contacts). Pro-
120 meV (p-InP). Before front coating, p-si vided that the electrolyte redox level Ered
was cleaned ultrasonically in hex an and me- is kept close to the H2/H+ potential, bar-
thanol, etched in 40 % HF for 30 sec, rin- rier heights ~B for the hydrogen evolution
sed with bidistilled water and dried with reaction can be derived following equ.(l).
nitrogen. p-InP samples were etched in 5% In spite of the inherent limitations of the
bromine/methanol for about 20 sec and rin- Mott-Schottky method, it is widely accepted
sed with bidistilled water. for the characterisation of photoelectrodes
All measurements were taken in a single as the extrapolation of forward currents is
compartment quartz cell with a flat window, usually hampered by kinetic overvoltages
a Pt-counter electrode, and SCE- or HgO-re- and/or corrosion processes.
ference electrodes. Electrolytes were 1 M
H2S04 or 1 M NaOH. All data are normalized ~B = ~Fb - ~red + (EF -EVb)/e + kT/e (1)
to a surface area of 1 cm2 and potentials

79
 At p-InP large barrier heights are ob- Table 1: Barrier heights for p-InP-Pt and
tained (cf. Fig.1 and Tab.1), almost inde- p-Si-Pt electrodes under hydrogen
pendent of deposition conditions and the evolution conditions derived from
amount of pt on the surface (1 - 200 mono- Mott-Schottky measurements.
layers). A more complicated behaviour is
observed for p-Si with extrapolated barrier barrier height p-InP-Pt p-Si-Pt
heights close to the bandgap for electro- [meV] [meV]
chemically deposited Pt at high pH (Figs. 1
and 3) and low values for metal coatings
evaporated in vacuum which exceed the mono- electrochem. dep.
layer range. In the latter case quasi-ohmic (30 - 200 monolayers)
contacts are formed with ~B < 500 meV where acid solution (pH 0) 9s0±90 SSO±100
capacitance measurements result in shifted alkaline sol. (pH 14) 10s0±100 1000±ls0
flatband values due to high dark currents.
Flatband potentials for p-Si were found to thin layers «10 MI):
be more sensitive to traces of oxygen in pH 0 9s0±90 /
the electrolyte and apparent barrier pH 14 / 850±100
heights exceeding the bandgap which were
measured in some cases might be explained
by a positive redox potential shift. A sum- vacuum deposition
mary is given in Table 1. >0.5 nm average thick-
ness:
4 acid solution (pH 0) 900±100 ohmic
t1jC2['0"C~ ] alkaline sol. (pH 14) / ohmic
.. a
1.4 70 <0.2 nm: pH 14 / SOO±200

5.2 Under illumination

For higher metal loadings (>50 mC/cm2,
electrochemical deposition) iu-curves re-
semble those of solid. state devices with a
.. p-lnP-Pt photovoltage build up versus the reversible
hydrogen potential. Photovoltages under so-
a p-Si-Pt lar illumination conditions reach 650 mV
for p-InP-Pt and 500 - 550 mV in alkaline
-1 o 1 U[v.,,,J or 300 mV in acid solution for p-si, re-
spectively (Fig.2). Photopotential measure-
ments yield ideality factors n = 1.1 - 2.
Fig.l: Mott-Schottky data of p-InP-pt (1M with reduced metal loadings s-shaped iu-
H2S04, lS kHz) and p-Si-Pt (1M NaOH, curves are seen at higher photocurrent den-
40 kHz) under H2 in the dark. Elec- sities. Current flow corresponds to hydro-
trochemical Pt-deposition (100 gen evolution. Dark currents are negligible
mC/cm2) • in the scale of Fig.2 except for vacuum de-
posited Pt > 0.1 nm, where a gradual change
to ohmic behaviour occurs.
Frequency dispersion has little influ- Photovoltage data confirm high barriers
ence on the shape of Mott-Schottky curves derived from capacitance measurements at pH
indicating good quality materials and homo- 14. At pH 0 photovoltages are generally
genuous barrier profiles. (For bare p-InP smaller, a difference which is only partly
electrodes frequency dispersion was found reflected by capacitance data.
to be as low as reported values for UHV-
prepared Schottky contacts [3], with <1%
capacity change/frequency decade from 5 Hz
to 50 kHz) ..
It is known that barrier heights at the p-Si-Pt
p-InP-pt solid state contact depend on the --- vacuum dep.
presence of hydrogen, an increase is repor- - eledrachem.dep.
ted for solid state contacts from =SSO meV -.5 .5
(air) to =1040 meV (H2) [7]. Data given in
Tab.1 range between these two values, and a
more detailed comparison is limited both by
data scattering and different measurement
frequencies « 50 kHz for the electrochemi- 14 .:
cal- and 10 MHz for the solid state case). "
"" ,I
.o;~~-;:' 1!./
p-Si I - -10
5. current-voltage curves I

5.1 In the dark ,/2A. l[mA/em']

Dark currents give a m~n~mum estimate
for the barrier height (thermionic emission
assumed as the reverse current source). At Fig.2: iu-curves for p-Si (0.sA,lA,2A: ave-
potentials 1 V negative of flatband dark rage thickness for vacuum deposited
currents = -10 nA/cm2 were measured for Pt pH 0; others: 30 mC/cm2 electro-
p-Inp-Pt and -0.3 - -10 #A/cm2 for electro- chemical dep.) and p-InP (pH 0, 100
chemically deposited Pt on p-Si. This cor- mC electrochemical dep.). Illumina-
responds to barrier heights >Sso meV for tion intensity adjusted for identi-
p-Inp and >720 meV for p-si, a rough agree- cal photocurrent densities i ph '
ment with capacitance data.

80
5. 3 StabH ity The kinetic overvoltage n 0 for i = 0;
Mechanical stress due to the formation thus photovoltages can be compared to
of gas bubbles causes a gradual removal of Mott-schottky data. Values taken from the
metal coatings on p-InP. Thus, the fill intercept with the potential axis after
factor is steadily reduced until cathodic photo current flow were found to be more re-
corrosion of bare p-Inp can occur. Whereas liable than open circuit measurements as
changes in metal loading do not affect pho- this makes the saturation of the metal with
topotentials of p-InP in a measurable way, hydrogen more likely (Eredox = EH2/H+).
an increase is seen at p-Si electrodes for
decreasing loadings during operation. This
leeds to an optimum of the energy conver- 7. I.pedance spectroscopic .easure.ents
sion efficiency until, finally, the decrea-
se in fill factor compensates the increased 7.1 Basics
photovoltage. It is obvious from these mea- Even for the lowest metal quantities
surements that the effective pt-coverage deposited, a sufficient amount of charge
must be kept low on p-Si for high barrier can be stored at the surface which acts as
formation (ohmic barriers made by vacuum the countercharge both for the semiconduc-
deposition can be "activated" by etching tor space charge (Schottky-barrier) and the
the surface [4]). outer Helmholtz plane (metal-electrolyte
double layer). Accordingly, two capacitive
elements Csc and Cbb are detected with im-
6. Mott-Schottky shifts pedance spectroscopy. The impedance respon-
se of the contact corresponds to a series
For low metal coatings Mott-schottky connection of Csc paralleled by the recom-
curves generally shift under photocurrent bination resistor Rsc and the Helmholtz ca-
flow. It has been shown for p-InP that such pacity paralleled by the charge transfer
shifts are a consequence of the electroche- resistor Rct (see insert Fig.4). Rsc and
mical charge transfer (kinetic overvoltage Rct are derived from equ. 2 at W ->l 0:
n, [6]). Photopotential and overvoltage add
up to U-Qred = Qsc + n, or for the assump- dU 1
tion of a semiconductor-metal barrier: Rsc + Rct n· (kT/e) . - - - - + Rct
di i-iph+j·
U-Qred = (kT/e).n.ln((iph-i+j.)/j.) + n equ. (3)

(j. is the diode saturation current) (2) (As shown elsewhere [8] Rct amounts to

This relation distinguishes three ca- (kT/e)

ses: i) a constant Fermi level in semicon- Rct equ. (4) ;
ductor bulk, metal and solution (EF = Ered) Q c' }i 2 + 41. 2 '
in the dark (provided there is no corro-
sion); ii) a photovoltage build up under i. is the exchange current density for the
illumination without current flow while the electrochemical reaction and Q c the catho-
Fermi level of the metal~emains unchanged; dic transfer coefficient).
iii) under photocurrent flow Fermi levels For ir = i - i h (ideal current gene-
both in the metal and in the semiconductor ration in EEe space gharge layer) the dif-
differ from the solution redox potential. ferential recombination resistor equals to
An example for a Mott-Schottky shift
under illumination for p-Si is given in Rsc = n· (kT/e)/(i rec + j.). egu. (5).
Fig.3. No shift occurs for i < i. (i. is
the exchange current density of the elec-
trochemical reaction, see below), and for I\.
I
higher currents the shift becomes a loga- I
rithmic function of the photo current [6].
4 C.=15 nF C.. =5 J.l.F
c;;

-
(

0
R.=2.8 kO R..=2.6 kO
~

E
.c. 2

-
Eo

- 1 - .5
U [V~ en
rc
.....
E

N
I 2 4 6
Z.real. (ohm] (x10"3) --)

Fig.4: Impedance data for p-Si-pt (25 mC

Fig.l: Mott-Schottky data and iu-curve for Pt, photodeposition) in 1 M NaOH un-
p-Si-Pt (30 mC electrochem. deposi- der N2 at i -6.5 pA/cm2 and
tion) at pH 0 under H2, 56 kHz (0: i h = -17 pA/cm2. Measured data (A)
dark; A: illuminated). Dark current aRa fitted curves, parameters are
lidl < 5 pA/cm2. given in the insert.

81
 Impedance measurements at potentials quency dispersion for the vacuum deposited
where significant recombination occurs ge- sample indicating a less homogeneous Pt-
nerally reveal two time constants (Figs. 4 distribution. In all cases semiconductor
7); the elements in the high frequency barrier- and charge transfer elements are
range represent Rsc and esc, respectively. separated with ideality factors n = 1.S for
(other circuits could be excluded, as they electrochemically deposited and n = 1.8 for
were found to be inconsistent with Mott- vacuum deposited Pt. Helmholtz capacities
Schottky-data [8] and results from intensi- amount to 5 - 10 ~F/cm2. In acid solution
tiy modulated photo current measurements (photodeposition) ideality factors close to
[9]) . 1 and Helmholtz capacities =3 ~F/cm2 were
measured, again with clearly separated ele-
7.2 Impedance data for p-Si-Pt ments for recombination and charge transfer.
The data for p-Si-Pt in Fig. 4 are ta-
ken at a small current density. From the
high-frequency semicircle the diode ideali-
ty factor is n =
1.2 (equ.5 and j. «
i ec)' As there is no contribution of the 10k
electrochemical charge transfer to the iu-
curve (i < i.), from the low frequency se-
micircle an exchange current density i. = /\
I
10 ~A/cm2 can be derived from equ.4. t
In the two Figs. to follow impedance lk
data are given for electrochemically- and
vacuum deposited pt at constant illumina- E
J::.
tion and varying current densities. Quali-
tatively the same shape of impedance curves .3
is obtained with a slightly increased fre- 100
N
+ (n - 2.2)
10k
A n 1.4
"." 10
10 100 lk 10k lOOk
"I
I FreQuency. [Hz) -->
lk
e.c
~ /\
I
I
.100
N
c..
en

10 40~ '"
10 100 lk 10k lOOk
Frequency. (Hz) --> to
20a
......
<l

80 " 0
I
I 10 100 lk 10k lOOk
FreQuency. [Hz) -->
60
c..
en
~
Fig.6: Impedance data for p-Si-Pt (=0.1 nm
40 ~ vacuum depos.) in 1 M NaOH under N2
at i ph = -170 ~A/cm2 and current
.0 densities (d) -0.8, (*) -31, (0)
r; -87, and (+) -129 ~A/cm2 . Potentials
20 ....«c. from 310 mV (d) to 20 mV RHE (+) .
I
0 7.3 Impedance data for p-InP-Pt
10 100 lk 10k lOOk Parameters derived for p-InP-Pt are n
Frequency. [Hz) --> = 1.1 - 1.2 and ehh = 3 - 15 ~F/cm2 (Fig.7,
for more detailed discussion see [8]). The
value of Chh was found to give an indica-
tion for the degree of surface pt-coverage
with Chh = 1S - 20 ~F/cm2 for a completely
Fig.s: Impedance data for p-Si-pt (electro- covered surface and smaller values for coa-
chemical deposition, <10 me Pt) in 1 tings in the monolayer range). For p-InP-Pt
M NaOH at i h = -570 ~A/cm2 and va- Csc follows the KS-relationship under re-
rying curren~ densities (d) -2, (0) combination conditions (which corresponds
-34, (*) -270, and (0) -430 ~A/cm2. to measurements under forward current at
Potentials range from 340 mV (d) to the solid state contact) whereas an illumi-
-240 mV RHE (.0). Ideality factors in nation dependent capacity increase is ob-
the insert. served for p-Si-Pt under such conditions.

82
 1k 1) For both semiconducting electrodes
and independent of metal deposition condi-
80 1\ tions or electrolyte pH the photovoltage
1\
I
I
build up and the electrochemical charge

~
I
I
transfer steps are easily separated by im-
100 pedance spectroscopy. Diode ideality fac-
-: tors range from n = 1.1 to n = 2.
E 2) Poor charge transfer and not enhan-
s= ced surface recombination is the origin of
.£
bad fill factors. The electrochemical char-
10 ge transfer process is similar to the
N transfer at metallic electrodes and, conse-
quently, a large overvoltage appears for
small metal quantities deposited.
3) The barriers formed between semicon-
ductor and surface or electrolyte are al-
10 100 1k 10k 100k ways large for p-InP-Pt (analogy to solid
Frequency, [Hz] --) state p-InP-contacts [3]). There is a good
agreement of Mott-Schottky- and impedance
data, dark currents and photovoltages for
the assumption of the thermionic emission
Fig.7: Impedance data for p-InP-Pt in 1 M model.
H2S04 under H2-purge at i ph =-1 4) For p-Si-Pt barrier heights close to
mA/cm2 and i = -45 ~A/cm2 (n = 1.2) the bandgap are formed at low effective
(geometrical) Pt-coverage. This is achieved
either by submonolayer vacuum- or by elec-
8. The p-InP-V2+/3+ photoelectrochemical trochemical deposition. Due to the island
solar cell structure of the metal layer, larger
amounts of metal can be deposited electro-
Large barrier heights result in high chemically for comparable absorption losses
energy conversion efficiencies when the hy- in the metal. Efficiencies are highest at
drogen evolution reaction is replaced by a pH 14 at the cost of enhanced corrosion ra-
fast redox reaction (electricity generating tes.
photoelectrochemical solar cell PEC; the 5) It cannot be decided unambiguously,
oxidation of the reduced species occurs in whether the high barrier for p-Si-Pt is a
a dark process at the counter electrode). result of surface defects formed close to
Thus, kinetic overvoltages are reduced with the conduction band or whether it is a con-
tha electrolyte acting as a transparent sequence of the formation of a "true" semi-
front contact. An example is the p-InP pho- conductor electrolyte contact [4]. The ana-
toelectrode in the vanadium 2+/3+ redox logy to p-InP may be seen in favour of the
couple in Fig.8; Mott-Schottky measurements first possibility. Photovoltages = 550 mV
yield barrier heights close to the bandgap and ideality values close to 1 do not con-
(1.2 eV), and high photovoltages and fill tradict the assumption of a Schottky-type
factors of such cells result in excellent contact on silicon at the given barrier
efficieny data (limited to = 10 % due to heights.
absorption losses in the electrolyte for
the cell in Fig.8). 11.5 % efficiency is
reported under solar irradiation [10]). Acknowledgement

This work was part of the German/Saudi-Ara-

o .5 bian HYSOLAR-project .
References:

[ 1] E. Aharom-Shalom and A. Heller, J.

Electrochem. Soc., 129, 2865 (1982).
[2] M. Specht, H.-M. KUhne, J. Schefold,
2 p-lnP in "Hydrogen Energy Progress VIII",
Vanadium 2+/J + T.N. Veziroglu, P.K. Takahashi (Eds.),
Perqamon, New York, Oxford 815,(1990).
[3] N. Newman, M. van Schilfgaarde, and
W.E. Spicer, Phys. Rev. B 35, 6298
4 [4]
(1987).
H. Tsubomura and Y. Nakato, Nouv. J.
de Chim., 11, 167 (1987).
[5] M. Szklarcyk, and J. O'M Bockris, J.
Phys. Chem., 88, 1808 (1984).
[6] H.-M. KUhne, J. Schefold, J.Electro-
chem.Soc., 137, 568 (1990).
[7] D.E. Aspnes and A. Heller, J. Phys.
Chem., 87, 4919 (1983).
[8] J. Schefold, H.-M. KUhne, J.Electroa-
nal.Chem., 300, 211 (1991).
Fig.8 iu-curve of p-InP in V2+/3+/ 1 M HCL [9] F. Philpps, J. Schefold, M. Specht, to
under illumination. be presented at the 42nd ISE-Meeting
Montreux, Switzerland.
[10] A. Heller, B. Miller, and F.A. Thiel,
9. Conclusion/Summary Appl. Phys. Lett., 38, 282 (1981).
Results presented above can be summarized
as follows:

83
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EXPERIMENTAL DETERMINATION OF FIELD PROFILES IN DIODES FORMED ON AMORPHOUS

SILICON GERMANIUM ALLOYS

C. LONGEAUD, J.P. KLEIDER, O. GLODT, D. MENCARAGLIA, R. VANDERHAGHEN*

Laboratoire de Genie Electrique de Paris (URA 00127 du CNRS), Ecele Superieure d'Electricite, Universites
PARIS VI et XI, Plateau de Meulon, 91192 Gif-sur-Yvette Cedex, France
*Laberateire de Physique des Interfaces et Ceuches Minces (UPRA 0258 du CNRS), Ecele Pelytechnique,
91128 Palaiseau Cedex, France

The experimental study of the field profile in a Schettky er PIN diede can be performed by means
ef twO' methods based on the standard time-of-flight experiment. The first one was presented by Street
/1/ (method A) and the other. one by Vanderhaghen and Longeaud /2/ (method B). Advantages and
drawbacks of each method are discussed. Method B has been improved in order to take intO' account
interface recembination and trapping.
We present the field prefiles ebtained i) on Schottky diO'des processed on a-SiGe films with varieus
Ge contents (0-24%) and ii) O'n a PIN diode, the I layer being of pure a-Si:H. The field prefiles obtained
on the Schottky diO'des in reverse DC bias exhibit a linear variatiO'n with x followed by an exponential
?ne. From these features, it is pessible to' deduce the density of states (DOS) at the Fermi level Ep and the
mtegral of the DOS between midgap and Ep. The evolution O'f these parameters with the Ge content
clearly shows the degradation of the material when Ge alloying, in accordance with the results of
capacitance measurements performed O'n the same samples. The internal field prefile on the PI .side ef a
5 ~m thick PIN diode extends in the bulk with a reverse DC bias, as in SchO'ttky diO'des, whereas it
shrinks with a forward DC bias.

1. IN1RODUCTION profiles of a PIN diode for different DC biases,

forward and reverse, applied to' the device.
The determination O'f the internal field
profile and its evelution under DC biases in
devices such as Schottky diodes or PIN diodes 2- DETERMINATION OF THE FIELD PROFILES
achieved on hydrogenated amorphous silicon (a-
Si:H) or· O'n its germanium alloys (a-SiGe:H) The determination cf internal field profiles
presents two main interests. From the study of F int(X) in amorphous alloys diodes can be
the internal field on the first type O'f device some achieved by means of twO' different methods
physical insight on the density of states (DOS) based cn the standard time cL flight (TOF)
around the dark Fermi level Ep can be obtained, technique.
whereas the solar cell conversion efficiency of The first one, hereafter called method A,
the second type of device depends on its internal was prcPO'sed and detailed by Street /1/, and is
field extent. based on the transient phctccurrent respcnse
In this paper, after a presentation of the analysis. This method is valid if the photccreated
two main different methods available to' derive charges can be accurately measured and remain
the internal field profile, advantages and ccnstant and gathered in a narrcw packet during
drawbacks O'f each method are briefly discussed. their drift in the internal field: thus, it implies
Then, the evolutiO'n with the reverse DC bias that the transport is non-dispersive and that the
of the internal field profile of a Schottky diO'de interface trapping is negligible. It suppcses alsO'
obtained on an a-SiGe:H film containing 16% of that the response time of the apparatus is small
Ge is presented as an example. The evO'lutiO'n of enough so that the drift of the carriers at short
the internal field profiles with the germanium times can be well reccrded. Nevertheless, when
content in different Schottky diodes biased this method can be applied, it is quite easy and
under the same reverse DC bias is also straightfcrward to' perfcrm.
presented. All these results are cO'mmented in Ancther method, based on charge ccllecticn
the light ef a simple model based on the and hereafter called method B, has been
ShO'ckley-Read statistics for the occupation O'f the propcsed by Vanderhaghen and Longeaud 12/.
gap states /3/. FrO'm the ebserved features, it is Chcosing the appropriate sign and amplitude for
possible to deduce the values of the DOS at Ep an external pulsed field Fl applied to the sample,
and ef the integral O'f the DOS between mid gap we can create a region in the sample, say around
and Ep. These results are compared with those x=x 1 from the generation interface, where the
obtained from capacitance measurements whole field is zerced (Fl + Fint(Xl) =0). Carriers,
performed on the same samples. generated by a laser pulse of strongly abscrbed
Finally, we present the internal field light, will drift towards this region after their
generaticn and stop at x=x 1. The integration of

84
 the current in the external circuit gives a charge: TIK)
220
QI = Qo (xIIL) (1) 180
220
where Qo is the total charge collected in a
standard TOF experiment and L the thickness of
the sample. Thus. since the field is zeroed at X=XI
for an amplitude of the external field equal to Fl.
we have determined a point of the internal field
profile: Fint(XI) = - Fl. The whole field profile is
obtained by varying the amplitude of FI: the
higher Fl. the closer to the interface is XI.
We present in Fig. 1 the field profiles
measured at two different biases and
temperatures on a Schottky diode achieved on o 2 3
an a-Si:H sample (method A: full lines. method B:
symbols). Clearly. at the temperature of 220 K Figure I: Comparison betwun both methods used for
the internal field profile measurements on a Schottky
where the transport is not dispersive. both diode: method A (full line), method B (symbols).
methods indicate the same trends, which makes
us confident in the field estimate. With no bias
applied to the sample the field is exponentially carriers while the integration is performed or by
decreasing. With a reverse DC bias equal to -2 V. excessive interface trapping and/or
the field is mainly linearly varying. recombination. The influence of the interface
Nevertheless. whereas the integral of the field states will not be the same whether the
obtained from method B leads to a value close to measured point (Fint(xd. xd is close to the
-2 V, the integral of the field obtained by interface or not. In the first case the total
method A leads to a smaller value (close to -1 interface field (FI +Fint(O» experienced by the
V). Furthermore. if the temperature is lowered carriers will be smaller than in the second one
to 180 K, where the transport turns to be and, consequently, the influence of the interface
dispersive. method B leads to the same result as trapping/recombination will be enhanced.
at T=220 K, whereas method A is no more valid, Therefore. the whole charge Q'o drifting in the
as mentioned above. As a conclusion. both sample may depend on the external pulsed field
methods lead to equivalent results and trends F I applied to the sample to probe the internal
when the conditions of validity for both apply, field and can be different from Qo. Consequently.
but, since the transport in some of the a-SiGe:H the value of XI cannot be obtained through eqn.
alloys is dispersive even at room temperature, (1). Prior to any field profile measurements the
we have used method B to measure the field influence of the interface states have been
profiles and their evolution with the applied bias checked from a delayed field experiment /5/. as
in these alloys. well as from the dependence of Qo with Vapp and
However, the accuracy of method B depends Log(ti). To minimize the influence of these
on the reliability of the measurements of QI and interface states we propose a procedure in which
Qo, since both measurements can be affected by the total charge Q'o can be measured
interface and bulk deep trapping and/or simultaneously with QI. Two consecutive steps of
recombination. field are applied to the sample: the first one to
Prior to any internal field measurement. the stop the drifting carriers generated by the laser
effect of deep trapping and recombination can be pulse around XI where the total field is zeroed as
analyzed from a measurement of Qo as a function in method B. and the second one of the opposite
?f the. ap~lied pulsed voltage Vapp and sign to collect the carriers which have been
mtegratton ttme ti. When deep trapping gives a stopped around Xl. From the first step we obtain
non negligible correction factor, then we Q I and from the second a charge Q2 such that
extrapolate QO(Log(ti» to take the release of Q2+Qt=Q'O: Q2 and QI can also be estimated from
deep traps into account: this extrapolation is the Q(Log(ti» procedure. As expected. for small
possible since the deep traps density was found x I 's, when carriers are stopped close to the
almost flat from post-transit measurements /4/. generation interface the value of Q'o is smaller
The same procedure can be applied for the QI than for large Xl'S, when carriers are stopped far
measurement. from the interface. This method, called C in the
In case of low deep trapping the internal following. requires two identical samples, one
field measurements were performed using the which is maintained in the dark, the other being
same ti for the obtention of QI and Qo. ti being illuminated by the laser pulse. the difference
approximately equal to ten times the transit between the two signals being amplified. With
time observed when we measure QO. this technique we can almost get rid of the
For some samples. mainly the PIN diodes, transient response to a step of field applied to
the traps located at the generation interface can the studied sample.
alter the measurement of QI and Qo by releasing As a complement to all these methods. the
internal field value at the interface can be

85
 obtained from the value of the reverse field Fl is related to the integral 1* of the DOS between
that zeroes the initial current as proposed by midgap and the dark Fermi level.
Datta and Silver /6/. We present in Fig. 3 the evolution of the
Finally, despite .the difficulties described field profile of four reverse biased Schottky
above, an easy test of the validity of the diodes formed on materials with various Ge
measured field profile can be achieved: the contents. We observe always the same trend: a
integral of the measured internal field must be linear variation of the field, followed by an
equal to the value of the applied voltage exponential tail, as predicted by the theory.
corrected for the built-in voltage. Though this is Moreover, one can see that the field profile
only a necessary condition it has been checked becomes steeper when the Ge content is
successfully on the following results. increased. This reflects the degradation of the
material and shows that N(EF) increases with Ge
incorporation.
3- RESULTS AND DISCUSSION

All the samples were deposited by glow

discharge of a mixture of SiH4 and H2 (and also
GeH4 for the SiGe alloys). The thickness of the
semiconductor layer was ranging from 2 to 5 11m.
The percentages of germanium in the samples
were equal to 0, 5, 16 and 24 %. The Schottky
contacts were achieved by the deposition of a
semi-transparent platinum layer.
From the delay field measurements, we
have observed that the Schottky contacts on the
a-SiGe:H samples did not exhibit an excessive
interface trapping and we have used method B
to determined the internal field profile in these Figurc 3: Evolution of the internal field profile of
samples. We present in Fig. 2 the evolution of Schottky diodes containing an increasing content of
the internal field profile with the reverse DC bias germanium. The experimental results are shown by the
symbols, the dashed lines come from a rheorerical fir.
applied to a Schottky diode achieved on a sample
containing 16 % of Ge. With no bias applied to
the sample the field is exponentially varying,
whereas in reverse DC bias the field varies first The values of N(EF) and of 1* are
linearly then exponentially with x, the extent of summarized in Table I. We have also reported
the linear part increasing with the reverse DC the values of these quantities obtained on the
bias. The dashed lines are the results of a simple same samples from capacitance measurements
model detailed elsewhere /7/, valid for thick /8/. Both experiments give N(EF) values which
samples and based on the Shockley-Read agree within a factor of 3. Despite this fair
statistics for the gap states occupancy /3/. In agreement one can see that, though we observe
this model, the slope of the exponential part of the same evolutions of the integral of the DOS
the curves is related to the DOS at the Fermi between EF and midgap in both experiments, the
level, N(EF), whereas the slope of the linear part values of the integrals found from field profile
measurements are always lower than those
found from capacitance measurements. This
discrepancy comes from the fact that both
experiments are performed in different
% Ge=16 Vdc thermodynamical conditions. Indeed, capacitance
o0 measurements are performed at T=390 K in the
dark whereas field profile measurements are
-1 A performed at room temperature or even at lower
- 2 ., temperature, to minimize the influence of the
-4 • diffusion, and using a repetitive laser pulse to
create the carriers. Therefore, if deep traps have
not been emptied between two consecutive
pulses we will measure a lower DOS integral than
the actual one since the steady state regime will
be slightly different from the obscurity one. For
instance, at T=293 K the release time of a trap
located at 0.8 eV below the conduction band
1 2 x{pm) edge is of the order of 50 s assuming that the
Figurc 2: Evolution of the internal field profile of a attempt to escape frequency is equal to 10 12 s-l.
Schottky diode with the reverse DC bias applied to it. We have also studied the internal field near
The experimental results are shown by the symbols, the the PI interface of a PIN structure. Since the
dashed lines come from a theoretical fit.
delayed field experiment showed an important

86
 Table I

Field measurements capacitance measurements

%Ge N(Ep) 1* N(Ep) 1*
cm- 3 .ey-l cm- 3 cm- 3 .ey-l cm- 3

0 4.5 x 1015 2.0 x 10 14 1.5 x 10 15 2.5 x 10 15

5 1.0 x 1016 5.7 x 10 14 7.5 x 10 15 1.3 x 10 15
16 3.7 x 1016 1.0 x 10 15 6.0 x 10 16 4.0 x 10 15
24 2.0 x 1017 3.3 x 10 15 3.5 x 10 17 1.2 x 10 16
Table I: Comparison between the results of field profile mesurements and
capacitance measurements. 1* is the integral of the DOS between EF and midgap.

influence of trapping states at the PI interface 4- CONUUSION

we have used the method C to measure the field
profiles in such devices. We have shown that the study of the
We present in Fig. 4 the field profiles obtained internal field profiles in devices such as Schottky
on the PI side of a PIN sample. The I layer, 5 or PIN diodes could be achieved even when the
Jlm thick, of this device was made of pure a-Si:H electronic transport is dispersive. In particular, it
deposited in the same reactor and in the same can be achieved when these devices are made of
conditions as the a-Si:H layer of the Schottky silicon germanium alloys. For the first type of
diode whose results are presented above. We device the DOS at the dark Fermi level'N(Ep) and
observe that the field profile in zero bias the DOS integral 1* between midgap and the
exhibits the same trend as for the Schottky Fermi level can be deduced from the field
structure, a single exponential variation (see Fig. measurements. The results obtained on a set of
1), at some distance from the PI interface. The a-SiGe:H Schottky diodes with Ge content ranging
DOS value at Ep deduced from this exponential from 0 to 24% are in fair agreement with those
variation with x is similar to the one obtained for deduced from capacitance measurements: both
the Schottky diode. In reverse DC bias the N(Ep) and 1* increase when Ge alloying. The field
internal field seems also to behave as in the profile of a PIN device and its evolution in
Schottky diode: a linear variation followed by an forward and reverse DC biases can be measured
exponential one. Finally, for a DC forward bias, with the proposed methods. Therefore the
we observe a strong field narrowing as expected quality of a photovoltaic cell can be tested.
in a photovoltalc cell under illumination.
Nevertheless, though the internal field at the PI
interface seems to behave as in a Schottky REfERENCES
contact, some differences appear close to the PI
interface where the internal field seems to vary (1) R.A. Street, Phys. Rev. B 27(8), 4924 (1983).
more slowly than exponentially in zero DC bias. (2) R. Yanderhaghen, C. Longeaud, Mat. Res. Soc.
This bc!haviour may be explained by changes in Symp. Proc 149, 357 (1989).
the space charge density which may arise either (3) W. Shockley, W.T. Read, Phys.Rev. 87, 835
from changes in the DOS shape near the PI (1952).
interface or from boron contamination, the I (4) G.F. Seynhaeve, R.P. Barclay, G.J.
layer being deposited after the P layer without Adriaenssens, J.M. Marshall, Phys. Rev. B
any buffer layer in the present case. However, 39(14), 10196 (1989).
this problem and its influence on the internal (5) H.Steemers, W.E. Spear, P.G. Lecomber, Phil.
field profile deserve more investigations. Mag. B 47, L83 (1983).
(6) T. Datta, M. Silver, Appl. Phys. Lett. 38(11),
903 (1983).
Vdc (V) (7) J.P. Kleider, C. Longeaud, D. Mencaraglia, O.
-0.5 0 Glodt, R. Yanderhaghen, submitted to Phil.
104 o 4 Mag. B.
Ii! 60- +0.5 (8) J.P. Kleider, D. Mencaraglia, Z. Djebbour,
E 4
0
0
D

Journal of non Cryst. Solids 114, 432 (1989).

~ D
D
a
4 0
~ 103
4
4
0
"tl 4 o
~ 4
o
G: D 4 o
102 This work was supported by AFMEIPIRSEM and CCE
0 2 under EN3S-0062-F and EN3S-0054-F contracts.

Pigure 4: Evolution of the internal field profile of a PIN

diode with the biases applied to the device.

87
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ffiGHLY TRANSPARENT AND CONDUCTIVE ZnO:AI

THIN FILMS FOR PHOTOVOLTAIC APPLICATIONS

R.H.Mauch and H.W.Schock

Institut fiir Physikalische Elektronik, Universitit Stuttgart
Pfaffenwaldring 47, 7000 Stuttgart 80, Germany ,

ABSTl!-ACT. ZnO:AI films have been prepared by using reactive DC-Magnetron

~puttermg m~thod fr?m a Zn-target. AI-doping has b~ provided by cosputter-
mg of Al which partially covers the Zn-target. A variation of doping from 1 to
5 atomic% has been achieved by changing the 02-partial pressure from 4.10- 2 Pa
to 1.10- 2. Pa. Th~.filrns show eithe~ hexagona! or .wurtzite structure depending on
~reparat.lOn con~htIons. Enhanced hght sc~tterIng m the films caused by large grain
SIzeS which are In the order of t~e.film thIckness has been observed using elevated
substrate temperatures. Transmissions of about 90% have been obtained in the visi-
ble region whereas free carrier absorption occurs at wavelengths exceeding 1000nm.
At a~reparation teI,Dperature of ~50·C conductivities of the films are higher than
4000 Ocm)-l at carner concentratIOn levels of about 7.1020 cm- 3 • Films prepared at
350~. show very good. stability as ~roved. in several ageing procedures. The appli-
cablhty of. these ~lms In photovoltaic deVIces has been demonstrated by using the
films as wmdows In Cu(Ga,ln)Se2 based solar cells and as TCO in a-SiGe:H-based
solar cells.

1. INTRODUCTION The target voltage of 360 V and the Ar partial pressure

Doped ZnO is one of the most interesting materials for of 1 Pa have been adjusted to achieve film growth rates
window applications in thin film solar cells. In the case of about l/'m/h at 250·C substrate temperature. These
of Cu(Ga,ln)Se2-based heterojunctions, ZnO:AI as a win- parameters result in a sputter power of 20 - 50 W depending
dow gains more current compared to standard (Cd,Zn)S on oxygen partial pressure.
due to its higher bandgap. For a-SiGe:H-based solar cells, The optimization of the films has been obtained mainly
the good chemical stability helps to overcome interfacial by variation of oxygen partial pressure Po. and substrate
problems between TCO and amorphous films. In order to temperature T,u6.
realize monolithic thin film tandem structures, ZnO:AI is Different mechanisms are influencing the film growth
the ideal material due to its high temperature stability. as displayed in the following table:
Introducing enhanced scattering ZnO:AI-films, reflection
losses can be reduced which is useful for all applications po. T.ut
mentioned above. sputter power X
Various techniques for the preparation of undoped ZnO target oxidation X
as RF-sp-uttering fl/, DC-sputtering /2/, reactive evapo-
ration /3/, spray pyrolysis /4/ and CVD /5/ have been sputter yields X
reported in literature. The high conductivity of these films Zn - 0 reaction X X
is based on intrinsic doping by oxygen vacancies. This sticking of Zn X
causes stability problems of conductivity at elevated tem- film structure X
peratures /6f. Extrinsic doping with AI, Ga or B improves
conductivity and stability at a still high transmission over Hence, a change in either po. or T.u6 or both results
a wide wavelength range. in varied plasma and growth conditions. This is shown
In this work, ZnO:AI films prepared by reactive DC- in Fig.l where the sputter power dependent growth rate
magnetron sputtering technique have been investigated. vs. the oxygen partial pressures is displayed at different
DC-magnetron sputtering from a metal target is in terms substrate temperatures. In general, growth rates are de-
of cheapness and large scale production prior to all other creasing by increasing T.ut due to a diminution of sticking
methods referred above. coefficient of the elemental Zn.
In the case of T,u6 = 250·C a decrease in growth rate
by increasing Po. can be observed. This can be explained
2. EXPERIMENTAL by an enhanced target oxidation with increasing Po. which
2.1 Preparation cannot be compensated by the contemporary increased sput
The preparation of the films has been carried out in a ter power. In addition, Po. below 2.5·1O- 2 Pa leads to a
DC-magnetron sputtering system with a vertical arrange- more or less unoxidized, nontransparent Zn-film, caused
ment of target and heatable substrate-holder. The Ar/02 by an ineffective oxidation of the elemental Zn due to low
gas mixture has been adjusted by mass flow controllers. substrate temperature and not enough oxygen as reaction
The target in use was a 99.999% Zn-metal disc with 6 inch partner for Zn.
in diameter. Doping was provided by cosputtering of Al In contrary, higher T .ut leads to an increase of growth
partially covering the Zn-target. The different sputtering rate by increasing po.. In these cases, film growth is more
yields of Zn and Al and the desired doping concentrations dominated by the substrate reaction of Zn and 0 and the
between 1 and 5 atomic% lead to a calculated cover area of oxygen offer for the reaction adjusted by Po., because of
about 10%. the lower sticking coefficient of Zn. Hence, sputter power

88
.......
....... 40 d6
)I: in terms of conductivity and transparency can be found by
•
..c: adjusting substrate temperature and oxygen partial pres-
....... sure without changing the Al cover area.
""-
E
c
....... 3 . 00
....--.
..c:
.oJ
30 (/)
p.,
~ ()
0
r.. N
ao

-
a 2 . 00
.oJ ......
C
Q)
'0
cQ) 20 .....(tj
(l)

0.
Q)
• 250·C s... 1.00
'0 0 b. 300'C .....
r.. ~
Q) o 350'C ~
~ 0
0 ()
0. 10 0.00
1130 1132
1 2 3 4 kin. energy (eV)
O2 partial pressure (10- 2 Pa)
Fig.9: COllnt rate of the AP' phot~/ectrono and corn.ponding Al
concentration. for diff.rent po. of ZnO:AI film. prepared at T. u • =
Fig.}: Sp.tter power norma/izttl growth rate v•. po. for diffe ... nt 951J"C
T,u6
and target oxidation are less influencing the film growth 2.2 Electrical and optical properties
by increasing T .ut. However, structural changes can be ob- Hall measurements carried out at the differently pre-
served at different substrate temperatures shown in Fig.2, pared ZnO:AI-films yield carrier concentrations exceeding
102°cm -3. This indicates that all the films are degener-
ated semiconductors. Highest carrier concentrations have
been obtained for films prepared at T,ut = 350·C. In ad-
dition, maximum Hall mobilities are at 45cm2 IVs for films
.. 350·C prepared at this substrate temperature.

,....., • . 0 0 , . . . . . . - - - - - - - - - - - - - - - - - - - - - ,

....--.
S
K
/I \
() 3 . 00 :leO .c
S
1'//

...
COl _ ·c
a
' I ......
; I ....... T < !501/

' I \ >.
.....
250·C
I

" \ .>
:;;
I I () ,
.
~
't:l
.... ,; .f . ~
0
()
0 . 00
1.00 2 . 00 3 . 00 . .0
02 partial pressure (10 - 2Pa)
Fig.t: (00£) XRD pta'" of ZnO:AI film. p ...pa ...d al Tn! = £51J"C
and 951J"C Fig ../: Conductiviti.s vs. Po, of the film. p ... pared at diff....nt T,ul
where the XRD-patterns of the (002)-peak of a 250·C The conductivities of the films are displayed in Fig.4.
and a 350·C ZnO:AI-film are displayed. The angle of the The dashed area in this figure indicates the region where
first maximum in the 2500C case correlates with the (002)- nontransparent films occur caused by Zn-rich growth as
peak in a cubic modification, whereas the second maximum mentioned above. At substrate temperatures exceeding
correlates with the hexagonal structure. Hence, a struc- 3500C, no Zn-rich film can be obtained due to a complete
tural transition from cubic to hexagonal by increasing T.ut reevaporation of elemental Zn.
can be derived from this fifure. Highest conductivities obtained are in the order of 3 -
The cosputtering of A and its in<;orporation into the 4.1()3(ncm)-1 in the 3500C-case. In addition, higher sub-
film underlie the same considerations as made in the Zn- strate temperatures lead to a less critical dependence on
case, except that the sticking coefficient of Al is 1 inde- deviations of optimum oxygen partial pressure. The con-
pendent of the different substrate temperatures. The in- ductivity of such prepared ZnO:AI-films depends on both,
corporation of Al and therefore the doping concentration is intrinsic doping by oxygen vacancies and extrinsic doping
mainly influenced by the oxygen partial pressure as shown by AI. The influence of the two mechanisms on carrier con-
in Fig.3 where the XPS-spectra of the AP'-electrons for centration and conductivity is governed by the preparation
ZnO:AI films prepared at 350·C and different Po, are dis- temperature. Increased substrate temperature leads to a
played. The doping concentrations derived from Fig.3 are dominance of the extrinsic doping due to a more perfect
between 0.7 and 5.3 atomic%. The evaluation of the chemi- stoichiometric growth of the ZnO.
cal shift of the AI2'-peak gives a value of ~ 1.5 eV which cor-
responds to either AI20 a or the spinell-structure ZnAl,04
formed in the ZnO. Hence, the optimum AI-concentration

89
 I.~ r----------------------------------------'
at films prepared at different substrate temperatures indi-
cate a less pronounced free carrier absorption in the case
0 .. of 250'C-films in comparison to 350'C films. This is in
agreement with the Hall measurements where in the case
of 250'C the measured carrier concentrations are within a
factor of 4 lower than in the high temperature case due to
s:::
0." a lower AI-doping level.
o
'Vi
·senen 0 <0
2.3 Degradation
Several ageing procedures like ageing at 220'C in air
s::: (CulnSe.-treatment conditions), ageing cycles at various
CI
.!::; 0 . 20
temperatures and high humidity level, finally, treatment
in H.-plasma have been carried out on films produced at
various substrate temperatures and optimum po. derived
0 . 00~L-----------~----------~~------~--~.
from FigA. .
250 750 12M Fig.6 shows the degradation behaviour of sheet resis-
wavelength (nm) tance depending on substrate temperature. No degrada-
tion occurs in the 350·C-case whereas for films prepared at
T.ut = 250'C a loss in conductivity can be observed~ Age-
ing cycles under high humidity and various temperatures
lead to the same result, 350·C-films remain stable while
lower substrate temperature films are degrading .
•.OO.-------------------------------__~----_,

8 3 . 00

...:a
<II
.~
2 . 00

Fig.5: Optical transmission spectra (top) of (a) 180nm and (6) 1.9J.1m
'i
..d
thick films prepared at T. u • =
95(f' C and corresponding SEM pho-
<II

tographs (bottom) ~ 1. 00~::::::::::::.--------------------3-5-

0 -oc
Optical transmission measurements displayed in Fig.5
(top) of (a) 180nm and (b) 1.31'm thick ZnO:AI-films pre-
pared at the same conditions (T.u• = 350'C, Po. = O.02Pa), 0 . 00~----~----~----~--~~---4-----4~--~
indicate a pronounced free carrier absorption in the in- 0 .00 20 . 0 _0 0 60 .0
frared due to the high carrier concentrations which can be time (min)
achieved by AI-doping. An enhanced light scattering can
be obtained in case of the 1.3J.1m thick film caused by the Fig.6: Degradation of sheet resistance during ee(f'C air annealing oJ
large grains. The corresponding SEM-pictures are shown ZnO:AI films prepared at different substrate temperatures
in Fig.5 (bottom). Transmission measurements carried out

1.0
ZnO :AI. H2-piasrna :
c 0.8 Os: Ro= 260 0
0
. ,"""
30s : Ro= 250 0
Cf} 120s : Ro= 240 0
Cf} 0.6 \
SnOx. 30s : 5H10
.,"""

E
Cf}
c:: 0.4
l\S
s.....
~ 0.2 SnOx. 120s: 4000 0

0.0
500 1000 1500 2000 2500
wavelength (nm)
Fig.7: 1hmsmission and sheet resistance of ZnO:AI and SnO~ films
after a H2 -plasma treatment (PH. = ltP Pa, T.ut =
eO(f'C, ppl •• m•
= 100m W fcm 2 ) oJ different durations

90
 The degradation of conductivity in the 250°C-case is High conductive and transmissive films have been achie-
mainly determined by the occupation of oxygen vacancies ved by using substrate temperatures exceeding 350·C. Con-
partially responsible for doping. As mentioned above, ex- ductivities in the range of 4000(Ocm)-1 and electron mo-
trinsic AI-doping dominates conductivity only at higher bili ties exceeding 40cm2 /Vs have been obtained for layers
substrate temperatures. Hence, carrier concentrations and prepared at optimum conditions.
conductivities of these films are not affected by heat-treat- Several kinds of treatments and ageing procedures in-
ment in air and stay stable. However, in the 250·C-case a dicate an excellent stability of electrical and optical proper-
cubic-hexagonal phase-mixture (see Fig.2) occuring in the ties for films prepared at substrate temperatures exceeding
film may also affect stability. 350·C.
In order to check stability of the ZnO:AI against a hy- The applicability of the films has been demonstrated
drogen containing plasma used for fabrication of a-SiGe:H- in Cu(Ga,In)Se2-based heterojunctions as well as in a-Si-
based solar cells, films have been treated i~ H2-plasma (PH. Ge:H-based pin solar cells.
= 102Pa, T.U6 = 200·C, p pl •• m • = 100mW/cm'). Fig.7
shows the results in comparison to a commercially avail-
able SnO..-film for different duration of treatment. A small ACKNOWLEDGEMENTS
decrease in sheet resistance of the ZnO:AI-film can be seen, The authors wish to thank G.Bilger for performing the
caused by the reduction of the film resulting in the creation careful XPS-analysis, Dr.H.Dittrich for the XRD-measure-
of additional oxygen vacancies. In contrary, commercially ments, Dr.H.C.Weller for the H2-plasma treatments and
available SnO.. decomposes during 120s treatment com- PV Electric GmbH for carrying out ageing procedures un-
pletely, finished by a nontransparent Sn-metal layer with der high humidity level.
increased sheet resistance. From this ageing experiment, This work has been supported by the CEC under con-
no evidence for a degradation ofthe ZnO-films is given and tract No. JOUR-0045-C and partially by the Bundesmini-
it seems, that from the stability point of view, such pre- sterium fiir Forschung und Technologie under contract No.
pared ZnO:AI is prior to commercially available SnO.. of- 0328059D.
ten applied in a-Si:H cell production. However, additional
experiments on performance of pin-junctions prepared on
ZnO:AI-TCO have to be carried out. REFERENCES
/1/ T.Minami, H.Nanto and S.Takata, Appl.Phys.Let., 41
2.4 Applications (1982) 958.
ZnO:AI-films have been applied in CuGaSe.-based het- /2/ I.Petrov, V.Orlinov and A.Misiuk, Thin Solid Films,
erojunctions with a thin (Cd,Zn)S-buffer layer in between. 120 (1984) 55.
Efficiencies exceeding 6% have been obtained after intro-
ducing the new window. Detailed results are reported else- /3/ R.Menner, B.Dimmler, R.H.Mauch and H.W.Schock,
where /7/. J.Cryst.Growth, 86 (1988) 906. \
Applying these films as TCO in a-SiGe:H-based solar /4/ I.Aranovich, A.Ortiz andR.H.Bube, J.Vac.Sci.Technol.,
cells, a considerably improved Vo • in comparison to stan- 16 (1979) 994.
dard SnO.. can be obtained while j •• and FF is prior in /5/ A.P.Roth and D.F.Williams, J.Electrochem.Soc., 128
the SnO.. -case. This is mainly caused by using thickness (1981) 2684.
data of the a-SiGe:H-pin optimized on SnO.. also in the /6/ T.Minami, H.Nanto, S.Shooji and S.Takata, Thin Solid
ZnO:AI-case. Detailed results on the distinct interface re- Films, III (1984) 167.
actions and device properties will be reported elsewhere
/8/. /7/ R.Klenk, R.H.Mauch, R.Schaffier, H.W.Schock, this
conference, late news session.
/8/ H.C.Weller, R.H.Mauch, G.H.Bauer, 22 nd IEEE PVSC,
3. SUMMARY AND CONCLUSIONS 1991, submitted.

ZnO:AI-films prepared by means of reactive DC-mag-

netron sputtering from a partially AI-covered Zn-target
have been investigated. The films show either hexagonal or
a mixture of cubic and hexagonal structure depending sub-
strate temperature. Grain sizes of the films are in the order
of the thicknesses which lead to an enhanced scattering of
light using 350·C substrate temperature.

91
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THE HYDROGEN-CARBON COMPLEX: BUILD-UP, ELECTRONIC

PROPERTIES AND ANNEALING BEHAVIOR

A. L. Endros, J. Grabmaier
Research Labomtories of Siemens A G
Otto Hahn Ring 6, 8 Munchen 83, Germany

DLTS experiments are carried out on intentionally hydrogen (H) and carbon (e) doped n-type c-Si samples.
The measurements show that H in the presence of e forms an electrically active He complex after a reverse-
bias annealing treatment. The He complex is a deep donor with an activation energy located about 0.16 eV
below the edge of the conduction baud. The He complex is only stable in the positively charged state. It
anneals in the neutral charge state after the capture of free electro'ls from the conduction band with an
activation energy of 0.73 eV. The hydrogen species that is incorporated in the He complex comes from
hydrogen-passivated phosphorus (P). These HP complexes act as a "hydrogen reservoir" and dissociate
with an activation energy of 1.13 eV into P+ and a negatilJel" charged hydrogen species H-.

1. INTRODUCTION ments carried out on hydrogenated phosphorus-doped

c-Si:C are presented.
Among the main impurities present in silicon for solar
cells are - besides the shallow dopants - the element 2. SAMPLE PREPARATION
hydrogen H and the element carbon C. There is no
doubt about the importance of the role that hydroger. The c-Si samples used in the present investigation
plays in the technology of modern solar cell devices. are grown by means of the Czochralski (CZ) and the
Its small atomic size, its high diffusivity and its ex- float-zone (FZ) method. In order to achieve differ-
tremely high reactivity make it possible for hydrogen ent carbon contents in the crystals, different amounts
to bind to energetically deep as well as to energetically of carbon powder are added to the Si melt. Phos-
shallow defects in the semiconductor material. Usu- phorus is added in the form of highly doped n-Si
ally this leads to a passivation of both shallow and pieces to achieve a shallow dopant concentration of
deep levels. The resulting changes in the electronic 5 X 1015 cm -3. Special care is taken to avoid contami-
properties of the semiconductor influence the perfor- nation with metallic impurities by using only materi-
mance of solar cells severely. In line with this, hydro- als of the highest available purity. The concentration
gen is known to be vitally necessary for amorphous of metallic impurities should therefore be lower than
semiconductors like a-Si:H and a-Ge:H. Hydrogena- 1010 cm- 3 . All rods are found to be free of disloca-
tion of the material, however, is not only beneficial for tions.
amorphous semiconductors. Hydrogen is also used ex- The crystals are cut into (100)-oriented wafers
tensively to passivate recombination centers in poly- which are conventionally cleaned and polished. IR
and monocrystaJline silicon in order to increase the and DLTS measurements carried out on these wafers
minority carrier lifetime T and with this the minor- reveal that the added carbon is incorporated on
ity carrier diffusion length L and the conversion effi- substitutional sites while oxygen occupies interstitial
ciency 1]. Due to the fabrication process, solar-grade sites. The carbon contents range from ~ 10 16 cm- 3
silicon normally contains high amounts of carbon in up to the solubility limit of 4 X 1017 cm- 3 in both
the range from 10 16 cm- 3 up to the solubility limit of FZ and CZ samples, whereas the oxygen content is
4 X 1017 em -3 or even higher. typically ~ 10 16 cm -3 for FZ and 7 X 1017 cm -3 for
The fundamental properties of hydrogen in the Czochralski-grown silicon.
above materials and its interaction with carbon, how- For DLTS measurements mesa-structured p+n
ever, are not yet well understood. While there is now junctions as well as evaporated Au-Schottky contacts
quite a good understanding of the hydrogen passiva- are applied to the samples. The ohmic back contact
tion of shallow acceptor impurities [1,2,3J very little is consists of evaporated Ti/Pd/ Ag for the p+n junc-
known about H interaction with group IV impurities tions and of rubbed-on eutectic GaIn in the case of
in c-Ge [4] and almost nothing in c-Si [5,6J. In line Schottky diodes.
with this, it is still an unanswered question whether Hydrogenation of Schottky diodes is performed
H can form electrically active complexes, whereas C by wet chemical etching at T = 300 K for 30 sec in
is known to be involved in some electrically active de- an acid consisting of HN03:HF=5:1 [9,10J. Hydro-
fects in Si [7,8]. In the present work, the results of genation of the p+n junctions is achieved by exposing
"deep level transient spectroscopy" (DLTS) measure- them to monoatomic hydrogen in a plasma discharge

92
at 150°C for 30 min [3]. The Schottky diodes are de-
hydrogenated via a relatively mild annealing step at
250°C for 15 min in a nitrogen atmosphere [11]. The
4 - N,' 6 . 10"cnr' -
p+n junctions are initially free from hydrogen as a
consequence of the high temperature (870°C, 30 min) [C l content
during the boron diffusion process. The C -V profiles
'iL - 2.3 ' 10 " cm-'
of both types of dehydrogenated samples are proven ~ 3 - N", 4·'0"cnr· n !S; 10" cm-
to be flat. After hydrogenation, however, a 15% re- Ii
duction of free electrons can be seen within the first 0::
111 x 100
two microns. This is due to the build-up of electri- en
cally inactive hydrogen-phosphorus (HP) pairs which ~ 2
S liner RBA
lead to a reduction of electrically active phosphorus
b.for. RBA
(P) donors [3].

3. HC DISCOVERY

In order to detect potential carbon and hydrogen re- o

lated deep levels, DLTS measurements are carried out
on carbon and hydrogen rich samples. These DLTS o 100
spectra, however, reveal only negligible trap concen- Temperature (K)
trations in the upper half of the bandgap and show Figure 1: Typical DLTS BpectrlL of hydrogenlLted slLmpleB with
no peaks that can be attributed to hydrogen or car- IL high ILnd IL low clLrbon content before ILnd ILfter ILn RBA trelLt-
bon. For this reason the samples are subjected to ment.
an optional reverse-bias annealing (RBA) treatment
(320 K, -4 V) for several hours in order to look for
metastable or charge-state-dependent centers. After E(0.16)
this RBA treatment the samples are cooled to about 'iL Schottky Diode
200 K while still under reverse bias and DLTS mea- ~ 2
Ii
surements are carried out again. The spectra after the c
II>
RBA treatment now indeed reveal a dominant trap en
VI
level in samples containing high amounts of hydro- I-
.J
1 with H
gen and carbon. Figure 1 shows typical DLTS spec- C
without H
tra recorded in H rich samples having varying carbon
contents before as well as after the RBA treatment. 0f---~-~~~-----~

After RBA the DLTS signal at T = 90 K grows by 4 r--o--~-~-I--~-~_....:j

about four orders of magnitude in carbon rich samples E(0.16)
corresponding to a deep level located about 0.16 eV pf-n Diode
below the edge of the conduction band with a trap
Jr r - - -,
density of typically 5 X 10 14 cm -3. An evaluation of n aU.r RBA
- IV [e l'" 2 . 10 " cm-'
the Arrhenius plot reveals a capture cross section (7 ---l L - 4V

of 7 X 10- 16 cm 2 • The results of DLTS measurements ~ 2 r,; = 23 m •• c

II>
on hydrogenated samples having only a low carbon en
content are shown in Fig. 1 magnified by a factor VI
I-
100. The spectra demonstrate that the trap den- .J
C t
sity NT of samples with a high carbon concentration without H
reaches values about two orders of magnitude higher
(NT = 5 X 1014 cm- 3 ) in contrast to samples with 0r---~---Er-~-------~
a low carbon concentration (NT = 4 X 1012 cm -3).
Without RBA the trap density never exceeds values so 100 150
higher than 8 X 1010 cm -3. The measurements there- Temperetur (K)
fore clearly demonstrate that the trap concentration
after an RBA treatment depends sensitively on the C Figure 2: Typical DLTS spectra of carbon-rich samples with
content. va.rying hydrogen contents a.fter RBA. Hydrogenation is per-
formed vilL wet chemical etching (Schottky diodes) or vilL glow
In order to obtain data about the H dependence of dischlLrge (p+n diodes).
the trap, "high carbon" samples are measured after
the RBA treatment. Figure 2 shows typical DLTS
spectra of p+n and Schottky barrier samples both
with and without hydrogenation. The spectra show

93
 6E. = 1. 13 .v before RBA
"b::l 1.ISX lOll•• c· '

" (-3 V)

o 0 .5
OEPTH (IU")
Figure 3: Build-up (decay) kinetics of the He (HP) complex.
Inset: Arrhenius plot of the measured time constants n,. Figure 4: Typical C-V profiles (77 K, 1 MHz) during one mea-
surement cycle.

that the DLTS peak is present only if the samples con-

tain H, independently of the process of hydrogenation frequency shows that the process which controls the
(etching or plasma discharge). The trap level is the speed of HP complex dissociation is a single atomic
same defect as in Fig. 1 and typical trap densities are jump process.
in the order of 10 14 cm- 3 after hydrogenation. If hy- In order to get an insight into the spatial depen-
drogen is removed then the detectable trap density dence of He complex formation and the coupled HP
decreases to values of about 1011 cm- 3 . The observa- complex dissociation, C- V profiling measurements are
tions relating to Figs. 1 and 2 therefore provide strong carried out on samples that have been subjected to
evidence for the existence of a deep level related to the different treatments. These treatments ar~ a hy-
presence of both C and H. The oxygen content of the drogenation step to prod uce the HP complexes, an
samples is not relevant, since the results can be ob- RBA step to generate the HC traps and a room-
tained with CZ material as well as with FZ material. temperature annealing step to completely remove the
This HC complex is visible in the C-V-profile, i.e., it HC donors again. Figure 4 shows the typical C-V pro-
is a deep donor . In line with this, a Poole- Frenkel files of hydrogenated samples during the above men-
effect [12) is found to exist. tioned measurement cycle.
The C" V profile that is recorded before the for-
4. BUILD-UP KINETICS mation of HC complexes shows the typical passivation
of P which is due to the hydrogenation of the sample.
The existence of the HC complexes is directly cou- A subsequent RBA treatment then leads to both HP
pled to the availability of monoatomic hydrogen in dissociation and HC donor formation (dotted line in
the sample during the RBA treatment. DLTS exper- Fig. 4). The corresponding C-V profiles reveal that
iments show that the concentration of HC complexes the P passivation moves significantly during this RBA
saturate at typically 5 X 10 14 crrt- 3 after a 60 hour treatment. The centre of passivation moves exactly to
RBA (-4 V, 320 K) treatment although C is avail- the edge of the space-charge layer. The data in Fig. 4
able in excess of 1017 cm- 3 and although the abso- therefore provide strong evidence that there exists a
lute H concentration should clearly exceed 1016 cm- 3 negatively charged species of H in Si. The shape of
after H-plasma treatment. Corresponding C- V pro- the C -V profile that is recorded "after annealing" of
filing measurements, however, reveal clearly that the the HC donors shows on one hand that recombination
HC saturation value is directly coupled to the ini- of H- with P+ occurs predominantly at the edge of
tially available HP complex concentration before the the space-charge regime but on the other hand that
RBA treatment. The H species that is incorporated in recombination also occurs in the low-field regions of
the HC complexes comes from dissociating hydrogen- the space-charge layer. This means that a fraction
phosphorus (HP) complexes which act as a hydrogen of the drifting H- ions recombines with the ionized
reservoir for the experiment. The build-up kinetics of phosphorus donors even before the H- ions reach the
HC complexes are therefore investigated in order to field free volume. This fact suggests that H- moves
monitor the annealing kinetics of the HP complexes. via a "hopping" mechanism in an electric field.
Figure 3 shows the build-up of HC complexes - and
with this the dissociation behavior of HP complexes 5. ANNEALING KINETICS
- measured via isothermal annealing experiments at
two different temperatures. The dissociation of HP The annealing beh.avior of the HC complex is evalu-
follows first order kinetics (Fig. 3) with an attempt ated from D-DLTS profiles which are taken at 87 K
frequency Vb = 1.5 X 1013 sec- 1 and an activation en- after the traps have been subjected to several anneal-
ergy tlEb = 1.13eV. The magnitude of the attempt ing treatments. To gain insight into the decay behav-

94
 10 2. - - - . - - - , - - - - . - - .
AE.s O. 73 .v via a single atomic jump process (6.Eb = 1.13eV).
C ".= 8.6>c1o" •• c ·' There exists a nP{Jatively charged H species after the
!.." dissociation of the HP complex. This H species drifts
in an electric field and forms HC complexes if car-
bon is present. The HC complex is only stable in the
1I1(T ( lI eV) 40 positively charged state if T ~ 300 K and it dissoci-
ates after capture of electrons in the neutrally charged
state.

References
100 150 200
TIME (minI [1] K. Bergmann, M. Stavola, S. J. Pearton, J. Lopata, Phys.
Rev. B. 37, 5 (1988).
Figure 5: Annealing kinetics of the HC complex V8. time in
[2] M. Stavola, K. Bergman, S. J. Pearton and J. Lopata,
short-circuited samples. Inset: Arrhenius plot of the measured
Phys. Rev. Lett. 61 (1988) 2786.
time constants To.
[3] N. M. Johnson, C. Herring, and D. J. Chadi, Phys. Rev.
Lett. 56, 769 (1986).
ior, the activation energy and the attempt frequency [4] L. M. Falicov and E. E. Haller, Solid State Comm. 53
of trap annealing are determined. Figure 5 shows the (1985) 1121.
results of isothermal decay experiments with short- [5] A. L. Endros, Phys. Rev. Lett. 63, 70 (1989).
circuited contacts at different temperatures. The data [6] A. L. Endros, W. Kriihler and J. Grabmaier, Material Sci·
show that the decay follows first order kinetics with ence and Engineering B4 (1989) 35.
an attempt frequency Va = 8.6 X 108 sec- 1 and an ac- [7] A. L Endros, W. Krllhler, F. Koch, J. Appl. Phys. 61, 11
tivation energy 6.Ea. = 0.73eV. The magnitude of Va (1987).
indicates that the limiting process for annealing is the [8] A. G. Litvinko, L. F. Makarenko, L. I. Murin, and V. D.
capture of free carriers from the conduction band. Tka.chev, SOy. Phys. Semicond. 14, 455 (1980).
[9] A. J. Tavendale, A. A. Williams, and S. J. Pearton, Mat.
Res. Soc. Symp. Proc. 104,285 (1988).
6. MODEL
(10) C. H. Seager, R. A. Anderson, and J. K. G. Panitz, J.
Mater. Res. 2, 96 (1987).
The experimental data suggest that an anneal-
[11] C. H. Seager and R. A. Anderson, Appl. Phys. Lett. 53,
ing treatment leads to dissociation of hydrogen-
1181 (1988).
passivated phosphorus according to the reaction
(12) J. L. Hartke, J. Appl. Phys. 39, 4871 (1968).
(PH)O -+ P+ + H-. In the electric field of the space
charge regime the H ions start to drift but are likely
to recombine either with P+ or get trapped to C
atoms in carbon rich material. At those tempera-
tures where the HP complexes dissociate and H is set
free, however, the HC complex is only stable in the
positively charged state, i.e., a reverse bias must be
applied in order to be able to detect the HC com-
plex even in n-type samples according to H- + C -+
(HC)++2e"". The HC complex dissociates at temper-
atures T ~ 300 K in the neutral charge state after the
capture of an electron from the conduction band. The
H atoms which are emitted, however, are no longer
electrically active and are lost for subsequent mea-
surements. There might be recombination of H into
a molecular species H2 or into the postulated H~ due
to the overall reaction 2(HC)+ + 2e- -+ 2 C + H2.

7. SUMMARY

In this paper we show that hydrogen in the presence

of carbon forms a deep and charge-state-dependent
He complex. This defect is an electrically active deep
donor with a trap energy ~ O.16eV below the edge
of the conduction band. The hydrogen species that
is incorporated in the HC complex comes from disso-
ciating HP complexes. The HP complexes dissociate

95
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Defect Chemistry and Homogeneity Range of Pyrite

and their Influence on the Optoelectronic Behaviour

M. Birkholz, A. Hartmann, S. Fiechter and H. Tributsch

Bereich Photochemische Energieumwandlung
Abteilungen Solare Energetik und Materialforschung
Hahn-M eitner-Institut
Glienicker Str. 100, D - 1000 Berlin 39
Federal Republic of Germany

Abstract. Investigations of the sulfur deficit in iron pyrite, FeS2_x, by

chemical analysis, density measurements and X-ray diffraction are presented.
They revealed the non-stoichiometry to be in the atom% range. Precise
determinations of the cubic cell edge yielded only small variations of 0.001 A.
They were correllated with XRD-determinations of the sulfur positional
parameter to give the unit cell length as a function of the non-stoichiometry
(Vegard's rule for pyrite). The proportinality between both is two orders of
magnitude lower than for iron-monoxide. Investigations with transmission
electron microscopy gave no hints for a superstructure of the S-vacancies,
therefore the Schottky-defects are proposed to be homogenously distributed
throughout the lattice. Their concentration is rather high, ca. 1021 cm- 3,
corresponding to a low vacancy formation energy of 0.3 eV. Electronically
missing S-atoms should lead to the occurence of defect-states in the forbidden
zone of the stoichiometric compound.

done by inductively coupled plasma with atomic

1. The Problem emission spectroscopy (ICP-AESj. The investi-
gations revealed the crystals to be exclusively sulfur
Because of its high absorption coefficient for visable deficient by some atom%, corresponding to an
light (~lOS cm- i for hv > 1.3 eV) and its non-toxic average composition of FeSI .9l!2'
constituents pyrite is a promising candidate for thin-
film photovoltaic cells. The mineral is in the ideal Density measurements were performed using a
stoichiometric case of the composition FeS2' But helum gas and a conventional glass pycnometer with
density and chemical measurements previously done xylene as liquid. Both methods confirmed the sulfur
showed the material to be exclusively sulfur- deficit in· synthetic as in natural pyrite crystals to be
deficient /1/. in the atom% range.

This raised the question whether some of the

unexpected electronical properties could be
explained with such an intrinsic defect coupled to
the material. To give a first answer to this question
means one has to evaluate the structural
organisation of the sulfur deficiency in pyrite. This
was the aim of the work presented here. 3. XRD-investigations

The methods used were X-ray powder diffraction 3.1 Precise determination of the cubic cell edge
(XRD) and transmission electron microscopy (TEM)
both being adequate to detect non-stoichiometries in Table 1 shows the result of cubic cell edge
the percent range. measurements of synthetic pyrite powders prepared
by CVT and purchased from Cerac Inc. The values 0

vary in a range smaller than 0.001 A. The

temperature was controlled within ±O.S °C,
2. Chemical analysis therefore the measured effect can not be due to
and density measurements thermal expansion (the linear thermal expansion
coefficient of pyrite is 8*10- 6 I K /2IJ.
Chemical analysis of pyrite crystals grown by
chemical vapour transport (CVT) with bromine were

96
Table 1: Table 2: Results of structure factor
Measured unit cell edges a of different synthetic refinement of pyrite powders
pyrite powders produced by chemical vapour
transport (CVT) and purchased (Cer.). Only a- Sample sulfur positional sulfur population
deviations from 5.4100 A are displayed parameter uta) parameter Pop(S)

H3Cer 0.386§.(9) O.8I(2)

Sample (a-5.41 ) T (OC) HTCer1 0.385.5.(5) 0.92.(1)
Preparation
(10- 3 A) H2Cer 0.385.2.(5) 0.95.(1 )
CVT1002 0.387a(9) 1.0J.(3)
CVT990 8.05.(4) 27 ..2.(3) RuS2 0.3881(2) 1.002.(5)
Br2-transported

CVT775 8.25.(3) 27.!(4) Br2-transported

CVT975 8.3!(4) 26.I(2) Br 2-tran spo rted Now the unit cell edges can be plotted Vs' these
stoichiometry deviations to give a Vegard's rule for
H3Cer 8.3.6.(2) 27.Q.(2) Cerac powder pyrite, see figure 1. It results for FeS2_x
(8d heated at 120°C) a(x) = ao - kx
HTCer1 8.3a(3) 27.!(3) Cerac powder where ao = 5.418Z(6) A. and k = 0.0011(6) A..
(14h heated at 390°C)
H2Cer 8.5I(3) 27 ..0.(2) Cerac powder The proportionality constant k of Vegard's rule for
(not treated) pyrite is two orders of magnitude smaller than for
CVT1002 8.81(3) 27.!(3) Br2-transported iron-monoxide /3/.

a (A)
5.4189 .
3.2 Structure refinements .................. ·····..············i·....···..·.. ··_·r-··-
5 . 4188
XRDreflections are dependent on the occupancy or
r·. . _·. . · .
population of the crystal positions. Their intensities
are proportional to the complex square of the 5 . 4187
• _
-r..!_._-
.--.. -.--
structure factor F, which is for the case of pyrite
5.4185
F(hkl)=
5.4184 . . . _. ._. . : :~:~=. ~: :·t.~~:::=::~:t·==~:::~:f=:==I=_
iii
. _.
4{aw fp. + 2Pop(S)/scos21C(hu + h+k). ' - -......",.<1"..... "' ... -:-,........... +. .. :
4
i
k+l h-l 5.4183
COS21C(ku--)'cos21C(lu--)} 0 .84 0 .88 0.92 0 .96 1.04 1.08
4 4
Pop(S,u)
hkl Miller's indices of the reflection Figure 1:
fFe andfs Temperature dependent atomic Plot of unit cell edge a/A vs. population
scattering factors for iron and sulfur parameter Pop(S) for different pyrite powders.
= 1 for h,k,l odd or h,k,l even
= 0 for hkl mixed
Pop(S) Population parameter of the sulfur
position (=1 for stoichiometric case) 4. TEM-investigations
u Sulfur positional parameter (u=0.38505)
Figure 2 shows a shematic drawing of the diffraction
With this structure factor the reflections of pyrite x- picture of a pyrite splinter from a natural sample.
ray diffractogramms were fitted to obtain sulfur The diffraction pattern is interpreted as a hkO-cut
population parameters as presented in table 2. It through the reciprocal lattice. The high frequency
shows that the samples were all sulfur-deficient with which such patterns were obtained is correlated
except for one. But for this sample a population of with the fact that pyrite fracture faces are in most
100 % or less is still within the error range and the cases parallel to (100). The roundness of the
effect is proposed to indicate a statistical variation. observed diffraction spots indicated a good
A RuSrsample was measured for comparision, crystallinity of the sample. High densities of
because this compound crystallizes in pyrite stacking faults perpendicular to the plane of
structure, too, and we had no hints for a sulfur- diffraction would have caused elongation of the
deficiency in it. This is confirmed by the structure spots (reflection broadenig), but such effects were
refinement. not to observe. CVT crystals yielded very similar
results.

97
 /1 .,. Fe·S ~
.......... "'" le~ungsband \

• •• 9g -\
\
\
\
0 \ \
\ \
'2lI _ \ \
OkO \ \
\ Valenzband \
0 \ ; - spl
• • • • • • \
\ /
/

• observed
tit obs. , forbidden
\C., ..:...:Fe;..:.SO-----"V
/

Fe 2+
hOD 0 nol observed ok!

Figure 2: Figure 3:
Schematic drawing of observed and non- Basic features of the band structure model for
observed hkO-reflection in TEM-diffraction fully stoichiometry pyrite. Fe-3d-levels are
splitt due to their octahedral ligand field. t2g -
levels overlapp to form the valence band. while
A Jurther inspection of the diffraction pattern shows the conduction band consists of iron-e g - and
an interesting asymmetry, i. e. the occurence of OkO- su Ifur-3sp3-states.
reflection that have no hOO-equivilants (h. k odd).
The OkO reflections with k odd. which should be
forbidden by the space group of pyrite (Pa 3 ), were
always observed. They can be explained by a double Introducing homogenously distributed S-vacancies in
diffraction effect, where hk 0 and h2kO add to give the pyrite lattice results in the occurence of iron-
OkO. The non-observation of hOO (h odd) is explained atoms, that are not to six but to less sulfur-
neighbors bonded. Following ligand field theory this
by the fact that hkO (h odd) reflections are forbidden must lead to a new splitting of the iron 3d-levels as
in Pa 3 - in contrast to hkO (k odd). Therefore the shown in figure 4. Most important is the splitting of
adequate double diffraction hkO + 2hkO to give hOO the eg-states and the lowering of the energy of one of
(h odd) can not occur and the diffraction pattern those levels.
shows this interesting twofold rotational symmetry.
But the observation of forbidden reflections does not
indicate a superstructure of S-vacancies. The sulfur
deficit must therefore be homongenously distributed
throughout the crystal. This is in sharp contrast to
iron-monosulfide. Fel-xS, where iron-vacancies
induce tz lattice superstructure!4. 5/. Morevoer, the
oberservation of double diffraction is another proof
for a good long range ordering of the atoms and the
very good crystallinty of the samples under
(a) normal and disturbed coordination
examination.
octahedral tetragon ai-pyramidal

eg / __

/ a
5. Electronic significance / 6Dq
/
Their is a wide consensus among pyrite researchers 3d -(------~-----r~
that the basic items of the band structure are \ 4Dq 4 b2
represented by the model illustrated in figure 3 /6/.
Following these ideas the iron 3d orbitals are
t29\--J --- - e
....:_-

splitted by the octahedral surrounding of the atom

into filled t2g and emptyeg levels. The tzg-levels,
Figure 4:
filled with the six outer electrons of Fe 2 +. overlapp
Spatial coordination and energetic splitting of
with those of other iron-atoms in the lattice to form
iron-3d-levels due to ligand field theory in an
the valence band. The conduction band. however,
octahedral and a tetragonal-pyramidal surroun-
consists of iron-e g- and sulfur-sp3-states that
ding. These are the most abundant iron-
combine to form molecular orbitals.
corrdinations in sulfur-deficit pyrite.

98
If they combine with electronic sulfur levels their 6. References
energy will be in the forbidden zone of the
stoichiometric compound, see figure 5. By this I. Luck, J., A . Hartmann and S. Fiechter,
mechanism it can be explained how the occurence of Stoichiometry and impurity concentration in
sulfur-vacancies leads to the development of synthetically grown iron pyrite crystals and their
electronic defect states in the band gap of pyrite. constituents, Fresenius Z. Anal. Chem . 334: 441·
446,1989.

/ /
/1 (J' Fe ·S
oonduction band
~
,
2. Krishan, R. S., R. Srinivasan and S.
Devanarayanan. "Thermal Expansion of Crystals."
ag - ( , 1976 Pergamon-Press. Oxford.
\ __ - - - (J' defect slates \ ,
a -(\- \ , 3. Goodenough, J. B. and A. Hamnett. "Iron Oxide
\\ \\ (FeO and Fe 1 -x 0)." Landolt-Bornstein -
\'
12g
\\
'\
\
_ _ -\___ _
\'~ ~
Zahlenwerte aus Naturwissenchaften und Technik.
Madelung ed. 1984 Springer-Verlag . Berlin,
Heidelberg, New York.
\\ valence band '\

\\ :#- 4. Koto, K., N. Morimoto and A. Gyobo, the

\\
sp3
///
Superstructure of the Intermediate Pyrrhotite. I.
\ \ ,// / Paritally Disordered Distribution of Metal Vacancy
\ (J defect states // in the 6C Type, Fe lIS /2' Acta Cryst. B3J: 2759-

\1• oFe-S 1/_

2764,1975.

5. Nakano, A., M. Tokonami and N. Morimoto ,

Refinement of 3C Pyrrhotite, Fe 7 S 8' Acta Cryst.
B3S: 722-724, 1979.
Figure 5:
Band structure model of pyrite with defects 6. Bither, T. A., R. J. Bouchard, W. H. Cloud, P. C.
incorporated that are due to sulfur vacancies. Donohue and W. 1. Siemons, Transition Metal Pyrite
The splitting of some of the Fe.eg-states Dichalcogenides - High Pressure Synthesis and
causes the occurence of new energy levels in the Correlation of Properties, Inorg. Chem. 7(11): 2208-
forbidden zone of the stoichiometric compound. 2220,1968.

Until now it is not say at which energy these defect

levels excactely will be to find. Because of the high
concentration of S·Schottky defects in the pyrite
lattice we can not exclude that those states will
combine to form a defect band. At least we think that
they will account for many of the electronic
properties of pyrite, that are not to understand in
terms of the usual band models, as the low open
circuit voltage e. g. Further investigations where
electronic measurements are to combine with the
determination of the stoichiometry deviation are
underway.

99
lOTI! EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

COMPARISON OF DARK AND LIGHT I-V CURVES OF SOLAR CELLS

JuUa Beier, Klaus Biicher

Fraunhorer-Institut f"lir Solare Energiesysteme
Oltmannsstr.22, D-7800 Freiburg
Tel.: +49-761-4014-142
Fax: +49-761-4014100

Abstract
Measurements of I-V-curves of several solar cells at different temperatures and illuminations were made.
The results of transferring the fit parameters of the dark measurements to light I-V curves at different
illuminations are presented in detail.

1) Introduction Because of the different surface temperatures with/without

One of the most important properties for the characteri- irradiance the temperature of the cell is measured with a
zation of solar cells is their efficiency. Its value is normally surface temperature sensor (also PtlOO). The cell is contac-
presented for the internationally accepted Standard ted with Kelvin probes(4-wire probes). The voltage of the
Reporting Conditions (SRC: 25°C, AMI.5 spectrum and power supply is computer controlled and scans the I-V
1000W/m2). The SRC provide a common well-defined curve of the solar cell. All values are read by the computer.
point of reference for producers and appliers. But the By applying a negative voltage bias at the power supply the
efficiency varies over the year because of its dependence on
voltage drop in the circuit can be compensated. In this
irradiance and temperature; thus, the SRC only allow a
manner, any desired voltage can be supplied to the solar
rough estimate of the energy output which the solar cell will
cell.
deliver over a whole year.
For dark I-V curves the chuck with cell and Kelvin probes
With the concept of the Realistic Reporting Conditions
were covered by a light-tight lid. To measure light I-V
(RRC) /1,2/ one attempts to give a more exact measure for curves up to one sun, the cell is illuminated by a halogen
this average yearly efficiency. An experimental setup for
lamp which is fed by a highly stabilized Heinzinger power
measuring both dark and light I-V curves of solar cells with
supply (stability better than 10-5 ). The voltage of the cell is
one single apparatus was built for this purpose. In order to
measured with a Keithley DMM 197; the current is deter-
be able to extract the parameters of the 2-diode model from
mined with a Keithley picoAmperemeter (for I <2mA) or as
the curves, certain demands have to be made upon the
voltage drop over calibrated resistors.
measurement equipment; these will be discussed.
For the measurements with concentrated light we built a
For the RRC it is necessary to know the efficiency of a solar
frame with 9 halogen lamps with reflectors as a light source.
cell for arbitrary irradiances and temperatures. The aim of
our work is to determine these '? (E,T) with as few
The individual lamps are adjustable in 2 directions. Each
beam was directed towards the solar cell. By using several
measurements as possible.
lamps whose angIe of emission was not too small we
avoided the problem of non-uniform irradiance of the cells.
Naturally this is easier to realize for the small cells
2) Description or the measurement equipment
examined here (lcm2 and 4cm2) than for large cells,
In order to determine ,(E,T), light and dark I-V curves
especially in a test setup.
were taken w.ith the experimental setup shown in fig.I.
But the influence of the non-uniformity on the open circuit
voltage Voc and the fill factor FF does not seem to be as
large as we had thought before, at least for a small concen-
tration factor. We concluded this from a test experiment
where we concentrated the light with normal glass lenses.
We used different lenses [{S6 15crn, f=25cm} , [{S6 4cm,
f=30cm} ,[{st5cm, f=8cm}, [{ji'5cm, f=8cm+a transparent
v piece of plexiglass above one half of the lens in order to
alter the uniformity}] to achieve a concentration factor of
about 2.
The short circuit current Isc was adjusted to the same value
in each run by varying the distance of the different lenses to
the cell and by slight corrections of the lamp intensity. The
degree of non-uniformity in the measurement plane was not
measured but estimated by inspection and by considering
the different focal lengths of the lenses. In the first run the
fie.1: Experimental setup for measuring dark and cell was illuminated only over about 20% of its area; in all
light I-V curves the other runs the cell was completely illuminated. The
results are shown in tab. I.
The solar cell lies on a thermo-chuck. The stability of the
temperature is controlled by a Ptl00-sensor in the chuck.

100
run m_ illustrated in tab.2 .
1 nly 20% of cell illuminated
2
3
4
estimated order:
inhom.
V ~'I;r r01"ble Iwode/ A
10
1
10:5
10-7
5 hom. .01 10-9

1ill2.L.:.. Influence of non-uniform irradiance 1ab.l.; Resolvable diode currents for different Iph's
on VocandFF (5 digit DVM)

As one can see, the only deviation in Voc and FF occurs for The short circuit current was carefully determined by inter-
the case of extreme non-uniformity, namely the partial irra- polating the measured data in the 3rd and 4th quadrant
diance of the cell. around V=O.
The dark and light I-V curves were fitted to the 2-diode
model with the help of the simplex algorithm:
V-Rs*' V-R.*,
~
, = '0,(0 n1*Vt -1) + '02*(0 n2*Vt -1) + Rp <-'ph) (eq.1)
3) Measurements and evaluation
Dark as well as light I-V curves (approximately lOOW1m2
and 700W1m2) at 20"C and 50"C were taken with six diffe- (dark I-V curves : Iph = 0)
rent solar cells (2 poly, 3 mono, 1GaAs). In addition, dark
and light I-V curves at 8° and 70"C were taken with three of with Vt= k*T/e (k=Boltzmann constant, e=elementary
these cells (pol2'1:poly 2*2cm2, TFK #1:mono 2*2cm2, charge) and a constant n1 = 1. Fit parameters were Rs(series
B55:GaAs 1cm) at appro 100W1m2, 700W1m2 and resistance), ~(parallel. resistance), 101' 102 (saturation
400WIm 2( only 8°C). currents) and n2'
As an example for larger cells we have measured dark and
light I-V curves (250W/m2 and 1000W/m2) of a 4*6cm2
high efficiency Si cell at 50"C. 3.1) Relevance of the parameters
The GaAs cell B5S was measured at 50"C under concen- The relevance of the fit value of Rs was checked with
trated light (5X, 7X, lOX). several cells by putting an additional resistor Radd in series
The stability of the light source was most important for the with the cell. Then I-V measurements and parameter ex-
light measurements. For small voltages, the diode current tractions were done. If the parameter Rs is the assumed
increases very slowly with the voltage. This diode current is ohmic resistance, one expects the fit value of the cell com-
superposed on the photogenerated current Iph' If the fluc- bined with the additional resistor to be the sum of both,
tuations of Iph are larger than the diode current this current Rs cell +Radd' The last two columns of tab.3 list this arith-
cannot be resolved, leading to uncertainties in 102 and R p ' metic sum in contrast to the fit values.
which dominate the lower voltage part of the I-V curve. TIe For some of the measurements we just measured the vol-
stability. of Iph is illustrated in fig.2, where the contribution tage of the cell at the end of the cable leading to the con-
of the dIOde to the measured current I meas, IDiode =Imeas tacts of the cell, so that the cable served as an additional
+ Iph' is plotted in logarithmic scale for one of the measu- resistor. For the other measurements, a SOmn resistor was
rements. soldered in series with the cell. In both versions, the total
additional resistance- (inclutiing-eontact resistance) was
IigU 1/'1 FhG-1SE measured. The results for the different cells are shown in
d,~------------------------ tab.3. Considering the difficulty of measuring the small
10" additional resistances, this experiment confirms the ohmic
and lumped nature of Rs'

10" cell Imaterial R.cen Radd 11

~scc +Radd RsIit(Rscc11 +Ra
BZ1
I ? .58 .45 1.03 1.255
mono! Si .033 .093 .126 .147
mono1 Si .033 .24 .273 .286
~ _ 12."31 mA
".- 1.9'96"10-" A
B55 GaAs .068 .1 .168 .16
~,- 2.312 - 1(1" A B55 I GaAs .068 .24 .308 .28
10" R" = 0.066 n
R,.;;;; ....90).- ()' 0
10" a.,. 2.12'1 tab.3: Fit values for the series resistance of different solar
Boo = ~ooo
10"'4-.,----,.--.---,--,--..---.---.-----,-.,--,--1 cells with and without an additional resistor Radd
0.0 Q.I 02 0..) 0.4 Q.! 0.& 0.1 De 09 to u \l
(all values in Ohm)
U IV]

The temperature dependence of the fit parameter 101

fi~.2: lDiode=Imeas + ~h of one of the light I-V curves
showed the theoretical exponential variation, 101-TII*exp(-
in log. scale
Egap/kT), s.fig.3 . The dependence on the factor Til is by
Not only fluctuations of Iph have to be considered here but far outweighed by the exponential factor, so that it cannot
also the absolute magnitude of Iph : if the current meter be seen in the plot.
has e.g. 5 significant digits, the 5 h digit is uncertain. A When fitting the PV-relevant part of the I-V curve (V>O),
diode current up to the order of the 51h digit cannot be the fit values for Rp" 102 and n2 depend rather strongly
resolved (see eq.1). Then, the evaluation of the diode upon the start values for these parameters. So, we separated
current begins only at some 1O-5*lph.The smaller I h' the the fit procedure into two parts. As Rand 102 are most
better the resolution of Idiode .For a 5 digit DVMI'this is important for small voltages, we first fitPed the V < 0 part of

101
the curve, thus gaining RJ> and 102 values that varied only in authors of /3/ (always smaller for dark than for light
a small regime. The influence of Rs and 101 in the V < 0 measurements) did not occur here, either. The tendency of
part is negligible. These Rn and 102 values were taken as our results is rather in the opposite direction.
start values for the V>O fit.'The resulting n2 also varied less
strongly than in the fit procedure with the V>O part only. 5) Summary
Still, the physical role of this parameter n2 is not clear; after I-V curves of several solar cells were measured at different
all, we allowed n2 to vary because our aim was to reproduce temperatures and irradiances. The transfer of the dark fit
the curves as well as possible. For SRH recombination, n2 parameters to the different light I-V curves yields a very
equals 2. good agreement with the measured values of Voc and
For Iph we used the value of I given by the expansion of the P(MPP). Thus, it is possible to predict the efficiencies of
above equation for small voltages: Iph = Isc *(1 + Rs/Rp). solar cells at different temperatures only from dark I-V
When we examined the fit parameters of the light fits we curves at these temperatures.
found that 101 and Rs do not change much with irradiance
E, but the (102' Rp, n2) - sets are not unique; this could lead 6) Literature
one to assume a variation of these parameters with E. /1/ A. Raicu, H.R. Wilson, H. Fischer, KHeidler, 9th EC
In addition to fitting the light curves we inserted the fit PV Sol. En. Conf., Freiburg, 1989
parameters of the dark into the equation of the light I-V /2/ K Heidler, A. Raicu, H.R. Wilson, Proc. 21th IEEE
curves and calculated Voc and the maximum power point PV Spec.conf., Kissimmee, 1990
(MPP). The values of Voc, V(MPP) and P(MPP) from the /3/ D.S.H. Chan, J.C.H. Phang, H.W. Wong, Solid-State
measurement, the light fit and this transfer of the dark Electronics 30, 919-925 (1987)
parameters to the light curves are given in tab.5 . The plots
in fig.4 show the different measured data points together
with the curves from the dark- > light transfer. The good
results of this dark- > light transfer corroborate the doubts
that the parameters vary depending on the irradiance (at
Cell Tre '",/mA Rs/Ohm Ro'Ohm n2
least 102 and n2 which influence the MPP and 101 and 'OllpA '021 A
Rs(Voc». The variation of the parameters might result polyl 8 dunkel .0948 7.. 33*10 5 2.32 5.65 3.06
10.25 .106 9.80*104 2.39 2.86 2.81
from the numerical problem of fitting in a multidimensional 39.80 .105 6264 2.60 3.54 2.85
space, obviously with many minima in the curve describing a 61.11 .096 6329 2.58 11.3 3.36
good fit to the measured data. 20 dunkel .122 3.73*107 18.8 3.99 2.73
10.21 .098 2.27*10 7 19.0 11.5 3.09
61.96 .118 3.78*10 7 21.6 25.3 4.00
4) Results 3.80*107
I
50 dunkel .141 1556 22.7 2.76
The light and dark fits were quite good. The fit parameters 4.84*10 7

J:I
10.27 .138 1606 26.7 2.92
are given in tab.4 . 66.55 .108 1.72*10 7 1315 19.8 2.62
We compared the measured power output P(MPP) with the dunkel .133 7353 1.66*104 22.1 2.32
10.20 .122 3206 1.66*104 15.5 2.14
power output calculated from the dark -> light transfer (see 70.09 .109 1977 1.39*104 13.9 2.07
tab.5). The discrepancy between measurement and dark- -- --- -
> light transfer is small for this important quantity. For TFK "" i 8 I dunk. I .100 1.29*109 .588 .578 2.81
.11 1.77*10 7
I
.634 .677 2.87

I
almost all the measurements, it is in the order of a few 10.12
40.45 .108 1.91*10 7 .639 .992 2.97
tenths of a percent. The exceptions are two measurements
with the poly-Si cell at 100W/m2, where deviations of ! II 20 dunkel
60.46 .100
.117
10497
8.13*10 9
."(36
5.35
.908
.802 I
2.93
2.73
+ 2.6%(8°C) and -5%(20"C) occurred. The reason for the
large discrepancy of -5% may be an inaccuracy of measu-
II 10.01
. 62.02
I
.097
.104
3.01*10 9 5.74
2.35*108 : 5.98
!
1.73
1.93
3.08
3.05
50 t dunkel .093 6.13*10 7 514 2.79 2.60
rement, because for the 200C-measurement, the light fit is
I '"-" I-'~ I
1. 72*10 4 484 1.67 2.33
also the worst of the lot. The (opposite) deviations of the 1
i 65.85 .094 3.69*104 511 5.58 2.72
goC and 20°C measurements probably do not result from 1

I
70 I dunkel .092 2.73*104 6876 5.04 , 2.33
the low irradiance level, because at 50"C and 70°C the dark-
i
i 1.35*104 6511 2.29 : 2.00
10.21 \.087

I,
68.41 .087 8165 6873 4.36 I 2.14
> light transfer was quite good again.
For the GaAs cell B55, which is not especially designed for 'Ol,A '02"0' 1' A
concentration, the dark -> light transfer worked even up to 855 8 I dunkel 3.92*108 3.06*10- 21 2.99 2.17
a concentration factor of 10 (tab.5). 1..069
0757
2.24*10 5 3.14*10- 21

II
1.862 2.55 2.14
The open circuit voltage Voc was also reproduced excel- 7.245 .066 6.34*104 3.14*10- 21 2.39 2.13

I
12.14 4.90*104 3.00*10- 21 2.37 2.13
lently by the transfer.
The deviations for the 4*6cm2 high efficiency cell were
20 dunkel 1..0679
066
1.58*108 4.80*10- 20 8.55 2.14
1.883 .0678 1.99*10 5 4.59*10- 20 6.96 2.11
1.1 % (250W/m2) and 1.8% (I()()()W /m2) for P(MPP) and
2mVin Voc. ,, 50 I
I
11.77
dunkel
1..101
037 6.10*10 4
3.55*10 7
3.59*10- 20
1.91*10- 17
2.95
102
2.00
2.09
As mentioned at the beginning we want to be able to pre- I; 1.806
12.40
1.
I
077
.101
2.51*1051,.72*'0-'7
5.60*104' 1.73*10- 17
85.2 2.07
dict the power output of a solar cell for varying conditions l
92.7 2.08
6.46*103 12.03*10- 17
i
with few measurements. An obvious method would have I (5X) 101.71.075 94.9 2.06
i 1.01*104 2.12*10- 17
i
(7X) 151.2 .071 68.2 2.00
been an interpolation of the fit parameters gained from
light I-V curves at different irradiances. This was not possi- I (lOX) 211.7 .070
I
I
4.24*104 1.95*10- 17 71.1 1.74

ble because the parameter sets are not unique, as descri- 1..071 1.36*10 7 15.20*10- 16 536

I
70 dunkel 091 2.11
bed above. But the dark- > light transfer worked out so well 1.807 1.70*105 ! 4.45*10- 16 312
j .088 5.02*104 Is
.02*10 -16 440
1 2 . 02
12.93 2.07
that this method seems to be a good alternative to interpo-
lating between several light measurements. Tab.4: Fit parameters of the 2-diode model of the I· V
These results clearly deviate from those reported in /3 f. curves at different temperatures and illuminations
The variation of the series resistance Rs observed by the

102
 ~ ill
T"""I'C 8 20 50 70
Iph/mA 1. 862 7 . 245 12.14 1.883 11.77 1.806 12.40 101.7 151.2 211.7 1.807 12 . 93
~ meas. : 935. I 993.7 1013 908 . 1 986.8 833.6 925.2 997.4 1007.8 1015.8 785 . 6 883.8
mV fit : 935 .0 993.0 1013 907.5 986.5 833.5 925.0 997.0 1007 . 5 1016 . 0 785 . 0 883 . 5
dar.·>l ight : 938.5 997.5 1016 909.5 988 . 5 833.0 923 . 0 1001 1013 . 5 1023.0 787. 5 884.5

~ IX: 99.50 99.32 99. 41 99.70 99.5 100 100.1 101.0 101.3 101.0 100 99 . 58
P(d'>l igilt)
0.0 0.1 0.2
I
U [vI O.l o. OS
,
06

·~l
I

TetIlll'C : 8 20 50 70 -001
poly I 65mA
I phI"" : 10.25 39.80 61.11 10 . 21 61.96 10. 27 66.55 10.20 70.09
~ mess. : 517.3 559.6 573.4 485.6 549.1 418.3 486. 6 370.5 445.5
mY fit : 517.5 559.5 573.5 487. 0 548. 5 419.5 486.5 370.5 445.0
dar. ->l ight : 517. 0 563.0 576.0 491.0 549.5 417.0 483.5 373 . 0 443 . 0
as
fmL IX: 102 . 6 99.58 99.00 94 . 99 100.5 100.5 100 . 2 99 . 66 99.95
P(d->l,ght)

T"""rc : 8 20 50 70
I phI"" : 10 . 12 40.45 60.46 10.01 62.02 10.21 65.85 10.21 68 . 41
Uoc meas. : 563 . 2 600 . 5 607.6 534 . 0 582.1 462.9 518.0 413.3 473.0
mY fi t : 563 . 5 600 . 0 607.5 534 . 0 582 .0 463 .0 517.5 413.0 4n.5
dark ->l ight : 565.0 602.0 613.0 535.0 585 . 0 462 _5 518.0 413.5 474 .0
as 0.. O~ 06
f!!!L. IX: 99.82 99.24 98.60 99. 59 99.41 100 99. 05 100 . 2 99.05
P(d->l ight) j 1
Tab.S; Uoc from measurement, light fit and transfer of the dark I 1
fit parameters to the light I-V curves (explanation s.text) I 1 I
and the ratio of measured to calculated power P(MPP)
for different temperatures and illuminations.
I( +OS ~------------------------------------------,
7Q'C I / SO'C ; I

i
I
po ly1
I f
1 "
1

/ I 20"C/ /S"C
1[ +0'
• dark
+ 100W/m2
11

~0;;'
1[ +03
... 700W/m2
IO l/p
1( +0 2
~---------~~~~-.-- -
06
J
1[+01
I I
I J
" I J
I [+OO +-----~----~-----+----~------~----+_----~
2.9 3 J .I l .2 l.l J .' l .S l .6
~ ,
70'C / I
( 1/T)/1/Koe-3 , !so.C, I
I
fil:.3: 101 fit values of the cell poly1 at different
I
J
I
iiradiances and temperatures

, 2r1C!
0.0 0.2 0.. 0.6 0.8 to -0 os S'C

.,- / ~
0.00
,./
...
,
- 0.06 /' /
-O.os B55 cone. don< - > lig1t 50°C ------------~~--~~~
---.
-0 01
0.0 02 D. U [v) 06 0.8
,
10
0 .0
-0.10
5X 'lCr·
- 2!> .
~ B55 l 8mA
-o.1S
7X -~o

GoAs
- 1 .~

~
-0.20 OX ·10.0

-11.$
-0.2$

fil:.4: Light I -V measurements at different irradiances - IS.O

and temperatures (points) and curves calculated

- 11.$
from the Qru:k fit parameters
-20.0 ~

103
10TH EUROPEANPHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THE LDS SOLAR CELL- A NEW APPROACH TO HIGHER EFFICIENCY

K.W.J.Barnham, B.Braun, J.Nelson, C.Button, J.S.Roberts.

M.Paxman.
SERC III-V Facility,
Blackett Laboratory, Department of Electronic and Electrical
Imperial College of Science Technology and Engineering,
Medicine, University of Sheffield,
London SW7 2BZ, U.K. Sheffield 51 3JD, U.K.

ABSTRACT.
We have studied the forward bias behaviour of AIGaAs/GaAs
p-i-n Multi-Quantum-Well (MQW) photodiodes. In sampl~s with
low background impurity levels in the intrinsic reg~on the
high quantum efficiency observed in reverse bias is maintained
into forward bias even for carriers photo-excited in the
wells. We compare our MQW devices with structures which are
identical apart from having AIGaAs intrinsic regions without
quantum wells. The short-circuit currents in the MQW
structures are much higher than in the control samples though
the open-circuit voltages are somewhat smaller. In one case
the energy conversion efficiency of the MQW device in white
light is 110% higher than the control. We discuss the
implications of our results for the development of Low-
Dimensional Structure (LOS) solar cells.

quality produced by atmospheric pressure

Low Dimensional Structures (LOS) such Metalorganic Vapour Phase Epitaxy (MOVPE)
as Multi-Quantum Well (MQW) or the internal QE remains high as far as
Superlattice (SL) systems have been widely +0.6 V forward bias (5).
studied for applications in We report here on forward bias
optoelectronics. Until very recently they results from four p-i-n Al Ga1_xAs
had not been considered as the active photodiodes with x = 0.3. All sampfes were
absorbers in solar cells. However, it has grown on n+ GaAs buffer layers and
recently been suggested (1) that higher substrates and had 200A p-doped GaAs
energy conversion efficiencies are capping layers. They were grown by MOVPE
possible if such low dimensional at Sheffield University in a similar
structures are incorporated into the manner to the samples in Refs. 5 and 6.
active region of p-i-n solar cells. By In the MQW samples #la and #2a the
including thin layers of a lower bandgap intrinsic layer has a 416A undoped AIGaAs
semiconductor (quantum wells) within the spacer next to the top p-type region and
intrinsic region it is possible to IllO¥e 30 periods of 87A GaAs wells separated by
the cell's absorption edge to a lower 60A AIGaAs barriers, a total thickness of
energy. Provided that the carriers can 0.483}1fil. In the control samples #lb and
escape the wells the short-circuit current #2b tne intrinsic region is AIGaAs without
will be increased. The higher current wells with the same total undoped
leads to a net enhancement in energy thickness. The control samples were grown
conversion efficiency if it is maintained immediately after their respective MQW
far enough into forward bias. This samples and were identical in all other
approach is an alternative to a respects.
conventional tandem cell avoiding the These structures were deposited in
problems of Ohmic contacts between commercial horizontal reactors (MR
elements of the stack. Semi con) at 900 torr using adduct purified
To test this principle we have methyl-metalorganic sources and ASH3'
studied the behaviour of AIGaAs/GaAs p-i-n Further details are given in Ref. 6. In
MQW photodiodes in forward bias and particular the MQW i-regions were
compared with samples which are identical deposited at 650 ~ to reduce the common
in all respects except that they do not impurities of carbon and silicon.
have quantum wells. We observe a It is difficult to determine the
significant enhancement in short-circuit charged background impurity level Ni in
current and energy conversion efficiency the intrinsic region of a p-i-n system
in the MQW samples over the control when p and n region doping is high.
samples. However electrochemical etching of monitor
We are not aware of any systematic samples and C-V profiles of the completed
studies of p-i-n MQW photodiodes in devices suggest that the intrinsic region
forward bias. Their reverse bias behaviour of all samples is p-type w\th a residual
has been extensively studied (2,3,4) for Ni of around a few times 10 5 cm- 3 for #la
optoelectronic applications such as and #lb and less than 1015cm-3 for #2a and
photodetectors and optical modulators. #2b. The level Ni is critical to the
Good samples have very high internal performance of these devices in forward
quantum efficiency (QE) for collection of bias. A finite Ni causes the electric
carriers photoexcited in the wells at field to fall across the intrinsic region
their reverse bias operating points. In so that as the applied voltage is varied
many samples the QE falls near zero bias from reverse to forward bias the field may
although in one sample with good material reach zero at a point within the intrinsic

104
region. The lower Ni the greater is the samples have higher open-circuit voltages
bias at which the held cuts off. The than their MQW equivalents. However, for
loss of electric field impedes the carrier solar cell applications the important
drift and hence limits the QE. We feature is the product of the current and
anticipate that if the electric field is voltage at the maximum power point (broken
maintained across the i-region the high QE lines in Fig. 2) . For #la and #lb the
for collection of carriers photo-exci ted maximum power of the MQW sample is larger
in the wells will be preserved into than the control in the ratio 1.2 : 1.0.
forward bias. We therefore expect this QE For #2a and #2b the MQW sample maximum
to be maintained further into forward bias power exceeds the control by 2.1 : 1.0.
in #2a than in #la.
The samples were processed in the ~ 9r-----~--------------------_,
standard way (5) for MQW photodiodes as
mesa structures with top window-type gold 8
contacts either 400pJll x 400}llIn or 800,,-m x 7
800~m in size. Room temperature 6
photoconductivity measurements were made

!
using standard lock-in techniques with a ~ 5 #lb
100W quartz-halogen bulb and a O. 3m '4~~-r
: ----------~
monochromator. The external QE was ~ r································"[""· ...
determined by comparing the photocurrent
with the incident power measured with a
calibrated silicon diode power meter.
The photocurrent as a function of o1 _________ , ______________________.t __ --
I •

bias is shown in Fig. 1 for the MQW

~

samples #la and #2a at representative -1

I
:
wavelengths above and below the AIGaAs
bandgap. As anticipated, the photocurrent -2~~~~_'L_~~~~~~~~~L-L-L-~~~
resulting from photons absorbed in the -0.4 -0.2 o 0.2 0.4 0.6 0.8 1.2 1.4
well regions remains at its reverse bias Bias I V
value into forward bias in #2a whereas for
#la, with its higher Ni' the QE starts to b) 9 r-----~----------------------,
fall already in reverse bias. 8 #2a ~ ........................ -....... .
7

650nm '" 6
,~ 5
~30F=====~~~~1- II 4~~#-2-b~r..=···=.. ·=···=..=·-·=·..=---~~
>-
(.)
§
c:
Q) I

] 2or-7-5-o-n-m------~~~~~~~ 1
w o ·--------t --------.-----------_L __:_
§ -1
c -2~~~~~~~~~~~~~~~~
~ 10 -0.4 -0.2 0 0.2 0.4 0.6 0.8
o Bias I V

Fig. 2 : I·V characteristics under white light

-~~.0--~--~2.~O--~--1~.O--~---0~.O--~~1.0 illumination
(a) for #Ia with a MQW system and #Ib
Bias (V) having a bulk AlGaAs i-region.
(b) for #2a with a MQW system and #2b
Fig. 1 : External quantum efficiency as a having a bulk AlGaAs i-region.
function of bias at typical wavelengths above
quantum well barriers (650nm) and for non-
excitonic absorption in the wells (750run). The spectral response for #2a and #2b
at zero bias is shown in Fig. 3a and 3b
together with a theoretical calculation of
We have investigated I-V the quantum efficiency. The addition of.
characteristics in white light using the QWs in #2a has extended the cell's
light source described above (which absorption range from the AlGaAs band edge
approximates to a black-body source of (680nm) to close to that of GaAs (850nm).
about 3000K). The results are shown in In this region the sample shows the
Fig. 2a for #la and #lb and in Fig.2b for characteristic absorption features of a
#2a and #2b for samples with 800}Jffi x 800Jlm quantum well with a step-like spectral
windows. The results with 400pm x 400~m response and strong excitonic peaks at the
windows are very similar when allowance is absorption edge. It is this absorption in
made for the different effective device the wells and just above the wells which
areas. Again the poorer forward bias gives rise to the extra short circuit
performance of #la arising from its higher current.
N· is apparent. For both MQW samples the Fig. 3 further shows that for photon
sBort-circuit current is enhanced over energies above the AlGaAs band edge the
that of the respective control samples by QEs for #2a and #2b are similar. The
the collection of carriers produced by presence of 30 quantum wells in the i-
photons absorbed in the wells. This is a region does not appear to degrade the
much stronger effect in #2a where Ni is short wavelength performance of the MQW
smaller. In both cases the control device appreciably.

105
 In the high and low wavelength is no anti-reflection coating, no window
regions the theoretical fit is a layer, the p-region thickness has not been
reasonable description if we assume an optimised and there is a GaAs capping
internal QE in the i-region (for both layer. The last three factors are
wells and bulk AIGaAs material) of 80-90%. responsible for the rapid falloff in the
The underestimate of absorption around quantum efficiency at short wavelength
700nm arises from uncertainties in the through a high surface recombination
AIGaAs absorption coefficient just above velocity. We believe that identifiable
the band edge and the neglect of other improvements could be made which would
contributions to the absorption close to bring the QE closer to the best available
the top of the well. The theoretical AIGaAs solar cell performance without much
studies are continuing and will be adverse affect on the the MQW enhancement.
reported elsewhere. In fact it should be possible to increase
the number of wells and hence the MQW
contribution while still maintaining high
,.... 50 forward bias QE. This was the case in the
~ a) #2b sample in Ref. 5 which had 50 wells.
In conclusion we have demonstrated
"'"
W -OEtheory the principle that the short-circuit
0 ., p-reglon
•• I-region current and energy conversion efficiency
40 .,. n-reglon of a MQW p-i-n device can be enhanced over
- OEup. structures without quantum wells provided
that the intrinsic region background
impurity level is low enough. We
emphasise that this test has been made on
devices that are MQW photodiodes rather
than solar cells but we see no reason why
the enhancement should not be carried over
to a well designed MQW solar cell.
We would like to thank Gareth Parry,
Tony Rivers and Mark Whitehead of
University College London for advice,
support and for processing these samples
using the facilities of the
Interdisciplinary Research Centre in
Semiconductor Growth and Characterisation.
500 600 700 800 900 Geoff Duggan of Philips Research
Wavelength (nm) Laboratories, Redhill has provided very
helpful advice. We are grateful to the
Greenpeace Trust and the SERC for
~50r-------------------------~ financial support.
W b) #2a
o -•.• OE theory
p-reglon
40 •• i -region
... n-reglon
- OE expo

30
REFERENCES
(1) K.W.J.Barnham and G.Duggan,
J.Appl.Phys. ~, 3490 (1990).
20
(2) S.Schrnitt-Rink, D.S.Chemla and
D.A.B.Miller, Advances in Physics 1ji, 89
(1989) .
10
(3) A.Larsson, p.A.Andrekson, S.T.Eng,and
_._.-._.-.-.- A.Yariv, IEEE J . Quantum Electron. ~,
.. :'-'- 787 (1988).
500 600 700 800 (4) M.Whiteheaa, P.Stevens, A.Rivers,
Wavelength (nm) G.Parry, J.S.Roberts, P.Mistry, M.Pate and
G.Hill, Appl. Phys. Lett. ~, 956 (1988).
Fig. 3 : External quantum efficiency as a function of
wavelength at zero bias (5) M.Whitehead, Ph.D. Thesis. University
(a) for #2b without quantum wells. College London (1990).
(b) for #2a with quantum wells.
Full theoreticalllne is a representative fit with (6) J.S.Roberts, M.A. Pate, P.Mistry,
broken lines indicating contributions from p.i J.p.R.David, R.B.Franks, M.Whitehea6 and
and n regions. G.Parry, J.Crystal Growth 21, 877 (1988).
(7) H.F.MacMillan, H.C.Hamaker,
G.F.Virshup and J.G.Werthen. in
The external QEs of both cells are Proceedings of the 20th IEEE Photovoltaics
low in comparison with the best present Specialists Conference (IEEE, New York,
day AIGaAs cells (7). This is because we
chose to keep the samples as similar as 1988), p.48.
possible to known MQW photodiodes. There

106
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

SPECIRAL RESPONSE MEASUREMENTS OF MULTI-JUNCTION SOLAR CELLS WITH A

GRATING MONOCHROMATOR AND A FOURIER SPECTROMETER
K. Bucher, A. Schiinecker
Fraunhofer-Institut fiir Solare Energiesysteme
Oltmannsstr. 22, W-7800 Freiburg
Tel.: +49-761-40140
Fax: +49-761-4014100

Abstract:
During the past few years a great variety of multi-junction solar cells has been developed with the aim of a
further increase in efficiency beyond the limits of single junction devices. To characterise this development
adequately, new measurement techniques for spectral response and I-V characteristics must be investigated.
For spectral response measurements, new set-ups should be developed to deal with the problems arising
from the various kinds of multi-junction devices. In this paper, two set-ups for spectral response
measurement are described, which are based on standard calibration equipment with the addition of
coloured bias illumination and voltage biasing. The difficulties in spectral response measurements are
discussed in the case of two-junction (tandem) cells and some examples are given showing the errors arising
from the direct transfer of the measurement technique from single to multi-junction devices. In addition,
methods to measure the correct spectral response of two-terminal and four-terminal tandem cells are
explained.
1. INTRODUCTION This condition can be relaxed, if the spectral mismatch is
Much research effort in the last few years has been made mathematically corrected. The method to deal with this
on the improvement of multi-junction solar cells. The aim problem based on the spectral response is described in /5/.
has been to raise the efficiency beyond the limit of single Another application is the calculation of the efficiency
junction cells. This can lead to a reduction of costs for pho- under realistic outdoor conditions./6/ For two-terminal
tovoltaic systems, in spite of the increase due to the more tandem cells, the generated current depends very strongly
complicated production of multi-junction devices, as the sy- on the spectrum of the illumination and the spectral
stem cost will strongly depend on the area needed to response of the device, as the current is always limited by
achieve a given power. /1/ This development has led to a one junction. Therefore the spectral response of the device
great variety of multi-junction cells, which can be classified is necessary to calculate the power produced under realistic
according to the materials used, the kind of combination of outdoor conditions, if the methods described in /6/ are
the junctions, mechanically stacked or monolithic, and the used instead of long term outdoor measurements.
number of terminals used to contact the device. /2/ In Because of the importance of spectral response measure-
particular the number of terminals has a great influence on ments, we will describe two measurement set-ups developed
the behaviour of a multi-junction cell, as for example a four- from standard calibration equipment to fulfil the require-
terminal tandem cell can be measured in a way similar to ments of multi-junction devices. The problems arising for
two single junction cells, but the same tandem in a two- the measurement are discussed for various types of tandem
terminal configuration will lead to many problems, as it is cells and solutions are shown. Although the equipment to
not easy to separate the influence of each junction. Thus the measure three or even more junctions is available, it is
set-up used to measure the characteristics of a multi-junc- justifiable to deal with only two junctions (tandem cells), as
tion device can be completely different to the set-up for a the problems arising in this case are typical for multi-
single junction cell. For example, in a two-terminal junction devices, and a combination of more than two
configuration, the current of the device is limited by the junctions only leads to an increase in the amount of
junction producing the lower current under the prevailing measurement equipment without producing new effects.
illumination. If the spectrum of the incident light differs
from the optimum spectrum the cell is produced for, this 2. MEASUREMENT SET-UPS
can lead to a dramatic decrease in photocurrent and in The spectral response can be measured with two
efficiency as a consequence. different set-ups, a grating monochromator and a Fourier
In this paper special problems in the spectral response spectrometer. Both are developments from measurement
measurements of multi-junction solar cells are discussed. equipment used to measure single junction devices. /7//8/
These measurements are of great importance for several In both cases the measurement range was extended up to
reasons. As for single junction cells, the shape of the 2000 nm to meet the requirements of the extended spectral
spectral response of each junction in a multi-junction device response of tandem cells. In addition to the normally used
will give an indication of the internal losses of the white bias light, in both set-ups the tandem cell can be
photocurrent in the cell. /3/ Another aspect underlining the illuminated with coloured light produced by tungsten
importance of exact spectral response measurement is that halogen lamps with low pass, high pass and band pass filters,
this is the basis for I-V characteristic measurements with a which can easily be adapted to the different spectral ranges
solar simulator via the mismatch correction. /4/ This of various materials.
correction can be of even more importance for multi-junc-
tion cells than for single junction devices. For example, for 2.1 The grating monochromator
two-terminal cells it is very difficult to control the level of The grating monochromator set-up is based on a double
illumination for each junction under the condition that the monochromator with 250 mm focal length, which is equip-
spectral distribution is similar to the reference spectrum. ped with two gratings mounted on a turret to cover the

107
spectral range from 300 nm up to 2000 nm. The devices is the same as for the grating monochromator, i. e.
monochromator is illuminated by two sources, a 1000 W additional coloured bias light and bias voltage.
Xenon lamp, mainly used in the UV and visible part of the The schematic diagram of the Fourier spectrometer is
spectrum, and a 500 W tungsten halogen lamp, which is shown in Fig. 2.
used in the IR region. Switching between the two light
sources and the gratings is fully computer controlled.

lr
f!oIrc:h.l-s:on- lnt.,..rl'rONPt.,.

1000 \I

r
X.non- L.a. .... p
r IoP-Vectl"'o

XL O -~ B~
I ~ r·-f c
1~ bit AD-
on ..... ,. tvr
J
~17
!
Rod m) i [ rost-rourlerl
P'-'OCP$sot"

Blu~ m
~ .. - BL. ~S
HC
L.O .. -Pa.55-
Fli t.,..
W Mlr"ror - ]
r Cont,..oUt'r
~ .. - .. - ..- ..
AI"Ip Uf'le r

XL I X.non lQ,~p
Vol to.

"'M I
II'

Hovobl.
TC

",r"or BL , BIAS Llgnt

I CPU
I
HL I Tungsten-Hlllog.n LG"p H I Hr-rOl" r I ril 't~,..

C I Chopp.,.. C I CO"'IJi'nSQ to,. ~ I Monitor Ce-l1

r I f ll h'", BS I B~a."5"'lttC',. rc I T"st C"U
lIA I lock-ln- AMph'.,.. LD I Los.,.~tl'c:to"
TC ' Tost con
Me I Monl'tor Celt
FIG. 2: Schematic diagram of the Fourier spectrometer

FIG .1: Schematic diagram of the grating monochromator 3. SPEcrRAL RESPONSE OF TANDEM CELLS
The different problems and solutions in measuring multi-
The monochromatic signal is chopped after leaving the junction cells will be discussed for the case of two-junction
monochromator and the higher orders are blocked with devices called tandem cells.
different high pass filters. The data acquisition is done using Spectral response measurements for tandem cells can be
lock-in amplifiers for the test and the monitor cells to divided into two different categories depending on whether
separate the chopped monochromatic signal from the each pn-junction has its own contacts (e. g. four terminals)
constant bias light. In addition to this standard set-up, two or the complete tandem has only two contacts (two-
further components must be added to meet the needs of terminals).
multi-junction measurements. Additional coloured bias light
adapted to the specific type of solar cell must be used for 3.1 Spectral response of four-terminal tandem cells
two-terminal tandems. This additional light is produced by In the case of four-terminal tandem cells, each pn-
tungsten halogen lamps operated with a highly stabilised junction can be electrically separated from the other and
power supply and filtered according to the wavelength range therefore be measured independently.
of each pn-junction of the device under test. In addition to The top pn-junction can be measured in the same way as
the COloured bias light, a bias VOltage should be provided, a normal solar cell consisting of a single pn-junction, i. e.
which is able to maintain the voltage of the cell at different with a white light bias producing a short circuit current
bias voltage levels with high accuracy. In the case of two- analogous to an illumination of 1000 W1m2 and
terminal tandem cells, it is necessary to keep the pn- compensation of the voltage drop in the external resistor to
junction under test in the short circuit condition. keep the junction under short circuit conditions. The second
The schematic diagram of the grating monochromator is pn-junction acts as a passive substrate. The measurement of
shown in Fig.1. the bottom junction is also very similar to a single pn-junc-
tion, with the difference that the top junction acts like an
2.2 The Fourier spectrometer optical filter absorbing the high energy light according to its
Another approach to spectral response measurements is absorption coefficient.
with the Fourier spectrometer. It is based on a Michelson As for single junction cells there can be a non-linearity in
interferometer with an angle of 90° between the two arms. the dependence of spectral response on the illumination
The diameter of the optics is 90 mm, providing a very high level due to semiconductor effects such as the saturation of
light flux. The maximum optical path difference between energy states in the band gap. Therefore the measurement
the two arms is about 400 ~m leading to a resolution of conditions should correspond to standard conditions, which
61.7 cm-1• The main advantage of the Fourier spectrometer means that the pn-junction produces a short circuit current
is that for most single junction measurements there is no equivalent to an illumination level of 1000 W1m2•
need to use additional bias light for the solar cell, as the il- The correct level of bias illumination is not as easy to
lumination of the device consists of a high DC white light regulate as in the case of a single junction. The power of the
component with a slight modulation in dependence on the bias light cannot be determined using a calibrated solar cell
optical path difference in the two arms. The spectral infor- of the same type as the bottom cell and measurement of the
mation is calculated from the modulation of the measured short circuit current, as the calibrated cell is not filtered by
light signal with a Fourier transformation. A complete the top cell. Therefore the range of non-zero spectral re-
spectral response measurement consisting of the mean va- sponse is different for the bottom cell and the reference cell
lue of 50 single measurements lasts about 1 minute. As the and, as the spectrum of the bias light is different from the
complete spectral information is given in the interferogram, standard spectrum, the correct level of bias light for the ca-
which depends only on the path difference and the mirror librated cell is not the correct level for the bottom cell of
velocity, the time per measurement is independent of the the test device.
spectral range. There are two approaches to obtain the correct bias
The additional equipment necessary for multi-junction illumination on the bottom cell. The first one is to measure

108
the short circuit current produced under a solar simulator no change in the spectral response as the limitation in
having a spectrum close to the standard spectrum. As these photocurrent is due to the GaAs component. The slight in-
condititions are hard to meet, there will be an error due to crease in the spectral response of the 0.81 rnA bias curve is
the mismatch between the simulator and the standard caused by a second effect which will now be discussed.
spectra. The other possibility to determine the correct level
G"As/S, MECH. STACKED TANOEM---,
of illumination is to approach the short circuit current under 1.'
- GaAs Cot It!',..,..nal)
standard conditions by iteration. The first spectral response
can be measured without bias illumination in the dark. A "
.~
u 1.1
-
-
Sl ('4 te-P'"1II1nill>
BIAS 51 O. BlrA
first approximation of the short circuit current can be ~ -- BIAS 51 O. n~A
'" 1 .... BIAS SL o.Oi.,.A
calculated by convoluting the spectral response and the
I 0..
~

standard spectrum, and the level of bias illumination can be

adjusted to produce this current. With a second spectral 0-
Il<

response measurement a better appproximation to the ii 0.,

...
correct response can be achieved and a new short circuit u
current can be calculated. This method should be repeated Ii 0. '
'"
until there is no further change in the calculated short cir- .. ,
cuit current. The level of bias illumination corresponds now
to the standard conditions. :xxi .. DO &DD .ao 700 ICC SXI 1000 llOO 1200
LJoIv.l.nQ\h [n~)

3.2 Spectral response of two-terminal tandem cells

Two-terminal tandems present the more difficult case in FIG. 3: Mechanically stacked GaAs/Si tandem cell measured
spectral response measurements. As there is no direct as a four-tenninal and two-tenninal cell. In two-tenninal
connection to each pn-junction on its own, the current connection, the GaAs top cell is measured under different bias
produced by one pn-junction must also be transmitted by illumination levels of red light producing a photocurrent in the
the second to be measured with the data acquisition. The Si bottom celL The spectral response reaches the level of the
measured photocurrent is thus the minimum of the current four-tenninal measurement when the bias illumination is high
produced in the top and the bottom pn-junctions. In the enough to overcome the current limit of the Si cell.
extreme case of pure monochromatic illumination
(monochromator measurement with no bias light) this will The second problem is the short circuit condition of the
lead to a zero photocurrent in the spectral ranges where active pn-junction. In a measurement circuit for single junc-
only one component is illuminated, and a non-zero photo- tion devices, the solar cell is connected to a nearly zero
current can be measured only in a short region near the load. If the same technique is used for a 2 terminal tandem
band gap of the top cell. Therefore a direct transfer of cell with coloured bias light on the passive component, the
single junction test procedures to two-terminal tandem cells active component is not measured under short circuit
is not possible. conditions .
.To measure the correct spectral response of each pn-
junction in a two-terminal tandem, two additional Cur,..~nt
parameters have to be adjusted. The current limit of the
passive pn-junction not under test must be overcome and
the short circuit condition of the active test pn-junction has
to be guaranteed.
In the following, 'active' means the pn-junction actually
beeing tested (superscript a) while 'passive' denotes the pn-
junction which is under bias illumination (superscript p).
The solution to the problem of current limitation is the
use of coloured bias illumination adapted to the different Gct.v," pn- Junct10n
spectral ranges of the top and bottom cells. The range and
the intensity of the bias illumination must be chosen such
that in the passive pn-junction more photocurrent is Voltage
generated than by the monochromatic light in the active pn-
junction. Under these conditions the total current produced
by the monochromatic illumination in the active junction
passes through the passive pn-junction and can be Fig. 4: I-V characteristic of a two-tenninal tandem. The
measured. passive junction (superscript p) is illuminated with coloured
Fig. 3 shows the spectral response of a mechanically bias light. The dashed line characterises the current Ib
stacked GaAs on Si 4-terminal tandem, which is measured produced by the monochromatic light in the active junction.
in a 2-terminal series connection to show the influence of As a result the passive junction operates at the voltage vP =Vb
the level of bias illumination on the spectral response. The near V0:. This causes the active junction to work at V' =-Vb
bias light is produced by a tungsten halogen lamp filtered instead of under short curcuit conditions (V' =0, Ia = Is: ).
with a Schott RG830 filter, which transmits IR radiation The biased passive component will be current limited by
beyond 830 nm. The level of bias illumination is measured the active pn-junction as a consequence of the problem
by the current generated in an appropiately filtered silicon described above. Therefore the passive pn-junction is near
device. As the figure shows, the photocurrent of the GaAs the open circuit point and produces a voltage Vb depending
component approaches the true value with the increase of on the level of illumination near its open circuit voltage
the bias light intensity on the Si cell. The Si photocurrent
values in Fig. 3 are proprotional to this bias light intensity.
Vot (Fig. 4) As the complete tandem is under short circuit
conditions the active pn-junction is at a voltage level of -Vb
A further increase in the bias illumination produces almost and not at zero potential. This will normally lead to an in-

109
crease in photo current as the I-V characteristic slightly voltage produced by this bias light in the passive junction
increases for negative voltages. An example of this effect is must be compensated to keep the active pn-junction under
shown in Fig. 3 Bias Si 0.81 mAo short circuit conditions.
In Fig. 4 the I-V characteristic of a two-terminal tandem cell
is shown. The passive junction is illuminated with coloured 5. ACKNOWLEDGEMENTS
bias light. The dashed line shows the actual current produ- We gratefully acknowledge the help of B. Scheppat of
ced by the monochromatic light in the active junction. The Phototronics Solartechnik GmbH for providing the a:Si/a:Si
intersection of the dashed line with the I-V characteristic of test cell. We would like to thank S. Kunzelmann, K Heidler
the passive junction determines the voltage Vb produced by and B. Miiller-Bierl for various discussions and appreciable
it. As the complete tandem is measured under short circuit help with the measurements.
conditions, the operating point of the active junction is at - This work was partially supported by the Krupp von
Vb l!.nd not at V =0 as is necessary for correct measure- Bohlen und Halbach Stiftung and the German Ministry for
ments. Therefore the voltage drop Vb should be compensa- Research and Technology (BMFT) under contract number
ted by a power supply installed in the measurement circuit. 032-8650E.
As the voltage drop Vb cannot be determined without
additional information, V0/ of the passive junction under 6. LITERATURE
coloured bias illumination, which is only seen by the passive /1/ B.Voss, J.Knobloch, Solarzellen mit hohem Wir-
junction, would be a good approximation to it. kungsgrad und ihr Einflul3 auf die Kostenreduktion
von Solaranalagen, DGS: 6. Internationales Sonnenfo-
o . " - . - - - - a: SI/a: SI STACKED TA DE rum, Berlin, (1988), p. 468
BLUE: 0 ~U
/2/ M. B. Spitzer, J. C. C. Fan, Multi-junction Cells for
D.' -- QED: 0 U
Space Applications, Solar Cells, 29 (1990), p. 183
: /3/ K. Biicher, J. Bruns, S. A. Jahnke, H. G. Wagemann,
§. 0. 15 BLUE: 800 ~U
Determination of Recombination Parameters of Solar
~ - - QE~ : 800 ~U

a
~
0. >
Cells by Spectral Response Measurements, 8'h
,
'i ECPVSEC, Florence (1988), p. 1317
~ 0.1& /4/ K Heidler, B. Miiller-Bierl, Measurement of Multi-
~ junction Solar Cells, published this conference
.; 0.1 /5/ G. F. Virshup, Measurement Techniques for Mul-
tijunction Solar Cells, 21 51 IEEE PVSC, Kissimmee
0.05
(1990), p. 1249
/6/ KHeidler, A. Raicu, H. R. Wilson, A New Approach
100 for the Performance Evaluation of Solar Cells under
U.v.1 enq1 h [1'111'11
Realistic Reporting Conditions, 21" IEEE PVSC,
Kissimmee (1990), p. 1017
FIG. 5: a:Si/a:Si two-terminal tandem test cell (not current
/7/ J. Beier, A. Schtinecker, A. Zastrow, Spectral Response
matched) measured under blue and red bias light with and
of Large Area Solar Cells with a Filter
without bias voltage. The voltage was about half of the open Monochromator, 9'h ECPVSEC, Freiburg (1989), p.
circuit voltage of the tandem. 489
/8/ K Biicher, K Heidler, Photovoltaic Measurements and
Fig. 5 shows the influence of this voltage compensation
Calibration, Int'l PVSEC 5, Kyoto (1990), p. 1017
on the spectral response of an a:Si/a:Si tandem cell. In this
case, as both pn-junctions have the same band gap and both
are illuminated by a high level of bias light compared to the
monochromatic light, they are both operating at nearly the
same voltage near their open circuit voltage. Therefore the
necessary voltage compensation is very near to half of the
open circuit voltage of the tandem.

4. SUMMARY
We have shown the necessary extensions to modify a
spectral response set-up for single junction cells to meet the
requirements of various multi-junction devices.
The measurement method differs depending on the
actual kind of multi-junction cell. The problems in
measuring the spectral response of multi-junction devices
are discussed for the case of two junction tandems but the
solutions can be easily extended to three or more junctions.
The major difference in measurement procedures is caused
by the electrical contacts, depending on whether each pn-
junction can be contacted individually or only two-terminals
exist for the complete device.
The first case leads to a measurement procedure similar
to single junction devices while in the second case two
problems must be solved. The current limitation of the pn-
junction which is not Il1onochromatically illuminated must
be overcome by the use of coloured bias light and the

110
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

MEASUREMENT OF MULTI-JUNCTION SOLAR CELLS

K.Heidler and B.Miiller-Bierl

Fraunhofer Institut fUr Solare Energiesysteme ISE
OltmannsstraBe 22, W-7800 Freiburg / Germany
Tel. 0761 40140 Telefax 0761 4014 100

ABSTRACT: In this paper, first the different measurement techniques for the I-V parameters of multi-
junction solar cells are analyzed and classified. Some of the methods which have been derived from single
junction measurements become iterative which causes new problems. We then discuss the uncertainties in
the short circuit current. Finally we present our approach for multi-junction measurement, an improved
reference cell method which takes into account the coupling of the spectral mismatch factors of the sub-cells,
and we describe our current equipment.

2.1. Solar Simulator Method

Measurement is most easily done if the simulator spectrum
1. INTRODUCTION is so close to the reference spectrum that. no further
correction is necessary. Any photovoltaic device can then
Multi-junction solar cells have been developed to be measured without adjustment of the simulator. Using
increase the conversion efficiency of photovoltaic devices. properly filtered light sources which are combined in the
To date, cells with up to 3 junctions ("sub-cells") and 2 to 4 measurement plane, the simulator can be designed to
terminals have been successfully prepared, the highest minimize the difference between its spectrum and the SRC
reported conversion efficiency being 32.2% /1/. The spectrum as shown in /6,7/. For calibration, the absolute
measurement is much more complicated than that of single spectral irradiance as measured with a spectroradiometer
junction cells, especially when the sub-cells are not must be known. It can be transferred to an irradiance
accessible individually. The fill factor FF and the open detector (reference cell, thermal detector) for easy setting
circuit voltage Voc become dependent on the spectral of the simulator.
distribution in a two-terminal device, for one of the sub- This approach demands considerable technical effort,
cells will limit the current /2/. A confusing variety of especially if one requires high measurement accuracy.
measurement methods has been developed, but no However, it is the only way to minimize uncertainties when
"standard technique" has become established. For the sake the fill factor FF and open circuit voltage Voc and therefore
of clarity and in order to identify the principal problems the conversion efficiency T] may be spectrally dependent.
specific to multi -junction cells as opposed to single junction
devices, in the following sections the different methods are 2.2. Absolute Spectral Response Methods
analyzed and selected measurement uncertainties are In order to tune the simulator, some methods use the
discussed. A very promising method with the potential for absolute spectral response of the sub-cells which can be
high accuracy uses reference cells for each sub-cell. We measured even for two-terminal devices /8/. The aim of
therefore present an improved version of this method
these methods is to adjust the simulator, that the sub-cells
together with a description of our currently availiable
or, at least, the whole device is set to the Isc generated
equipment.
under SRC. To match the Isc of every sub-cell several
suitable light sources are needed. As the spectra of the light
2. MEASUREMENT TECHNIQUES sources overlap in general there are always contributions to
the Isc of one sub-cell from several light sources. Adjusting
Establishing the spectral distribution and intensity to one sub-cell to its proper Isc gives a change of Is< in the
correspond to the Standard Reporting Conditions (SRC: other sub-cells. The procedure therefore becomes iterative.
25°C cell temperature, AM1.5 spectral irradiance, 1000
W/m2 /3/) is the most important condition to be satisfied. 2.2.1. Absolute spectral response of current-limiting sub-
In most practical cases, the spectral requirement is reduced cell
to the condition that the current generated in each sub-cell For single source simulators with a spectral distribution
must equal the respective current under the SRC spectrum. close to SRC, an easy applied method was proposed by
If this condition is met, the Isc of the device can be Emery /4/. The aim of the method is to set the simulator
measured correctly. Open circuit voltage Voc and fill factor such that the Isc of the current-limiting junction is the same
FF, however, may be spectrally dependent as is the case for as under SRC. The current-limiting junction can be
example for amorphous Si, leading to an incorrect I-V calculated from the absolute spectral responses of the sub-
characteristic. Most methods use multiple light sources, cells under SRC. The intensity J1. of the source with the
where the number of light sources corresponds to the spectral distribution E is then adjusted that
number of sub-cells /4,5/. Table I reviews the published
multi-junction measurement techniques. They are analyzed (1)
in detail in the following sections.

III
holds. E sRC is the SRC spectrum and S denotes the absolute The index i =l..n denotes the sub-cells, j =l..n denotes the
spectral response where the index TCnmil means the corresponding light sources, index k denotes the k'th
current-limiting sub-cell (test cell). The dependent iteration step. This is an elegant method for measuring the
parameter A is omitted from E(A) and S(A) to improve I-V parameters of multi-terminal multi-junction solar cells.
readability. The advantage of the method is that only one The sub-cells not being measured should be held at the
source is needed. Care has to be taken that the spectrum of
maximum power point MPP to make sure that the I-V
the simulator is close enough to SRC, so that the current-
characteristic at least near the MPP, is not degraded by the
limiting sub-cell is the same for the simulator and SRC. In
addition, only the current-limiting sub-cell will be set to unmeasured sub-cells.
SRC, while the other sub-cells may be far from these
conditions and thereby produce errors in FF and VDC' 2.2.3. Absolute spectral response and absolute spectral
irradiance
2.2.2. Absolute spectral response of all sub-cells Without direct access to each sub-cell, not only the absolute
If direct access to the sub-cells is possible, then the spectral response, but also the absolute irradiances of the
simulator can be set for all sub-cells /5/ by adjusting the sources PiE. must be known. One can either adjust the Pi(k,'S
intensity coefficients Pi' so that in eq.(2) iteratively until eq.(3) holds or calculate the
proper intensity coefficients Pi by solving n dependent
(2) linear equations in Pi:
i=l .. n

holds until JESRC STC; cIA. == (4)

~i(k) = Ili(k- 1) i=l .. n (3)

This method is highly prone to uncertainty problems due to
the difficult measurement of absolute spectral irradiances.

Table I: Overview of multi-junction measurement techniques

method simulator setting Irradiance direct access aim of the remark reference
calibration to the subcells method
necessary ?

solar matching the absolute spectral no establishing the establishes SRC Sopori et. aI. /6/
simulator
method
reference spectrum
as close as possible
irradianoc of the
source
SRC spectrum
and intensity
Sf
not only for I but
also for FF an Voc
Bennell et. aI. /7/

absolute short circuit current absolute spectral nol ~.ofthewhole only onc source Emery et. aI. /9/
spectral of the device equal to responses of the eVlce same as nccessary with spectral
response the short circuit current subeells underSRC distribution close to SRC
method of the limiting subeell

short circuit current of absolute spectral yes Islis0f the sub- Virshup /5/
each subeell equal to SRC responses of the ce same as
subeells underSRC

absolute spectral no IS:ij of the sub- uncertainties in the

responses of the ce s same as absolute spectral
subeells and absolute underSRC irradiance calibrations
spectral irradiance may dominate
of the sources measurement
uncertainties

reference cell mismatch-method relative spectral nol ~c .of the whole only one source Emery et. aI. /9/
method applied to the current response of eVicc same as necessary with spectral
limiting subeell reference and sub- underSRC distribution close to
cells, relative SRC
spectral irradiance
of the sources,
calibration number
of the reference cell

short circuit currents calibration numbers no IS:ij of each sub- suitable reference Shimokawa et. aI.
of suitable reference of reference cells ee sameas cells necessary /15/
cells equal to SRC underSRC

mismatch-method applied relative spectral no IS:ij of each sub- potential for very high Glatfelter et. al. /16/
to each junction responses of ce sameas accuracy this paper
simultaneously reference and sub-- under SRC
cells, relative
spectral irradiance
of the sources,
calibration
numbers of the
reference cells

1 Current-limiting sub-cell must be the same under both the SRC and the simulator spectrum_

112
 Although easy to use as a measurement procedure
2.3. Reference Cell Methods (if iteratively applied, no calculations are necessary) the
Because absolute calibration of high accuracy is difficult, it aquisition of appropriate reference cells and calibration
can be advantageous to use calibrated reference cells numbers makes this method complicated if different multi-
instead of measuring the absolute spectral responses of the junction technologies are to be measured.
sub-cells. Reference cell methods make use of the
calibration number CN, which is the I of the reference cell
(RC) under SRC (some authors:frefe~ to take Isc divided by 2.3.3. Mismatch method for each junction
the total irradiance 1000 W/m for CN). It can either be The correct mismatch equations for n sources and n
calculated from the absolute spectral response of the reference cell /sub-cell combinations are /16/:
reference cell and the SRC spectrum or measured under
suitable conditions /9,10,11,12,13/. For single devices, a
mismatch method has been developed, taking into account eN;;: Mi f "in Jlj Ej EfC;
j= ,
d).. i=1..n (7)
differences in the spectral responses of test cells (i.e. the
sub-cells in a multi-junction device) and reference cells
/14/. Only the relative spectral responses of the reference f ESRC EfC; d).. f "in Jlj Ej
j= ,
STC; d)..

cell and the test cell and the calibration number of the with
reference cell must be known. This avoids uncertainties due f i="in, Jl'1 EI EfCi d).. J ESRC STC; d)..
to the absolute calibration of the spectral irradiance
measured in the plane of measurement. These coupled non-linear equations in the intensity
2.3.1. Mismatch method applied to the current-limiting coefficients J.lj can be reduced to the non-coupled equations
sub-cell
Using a single source simulator one can also apply the eN;;: Mi' Ili f Ei EfC; d).. i=' ..n (8)
mismatch method for single junctions to the current-
limiting sub-cell instead of using its absolute spectral in the limiting cases that the spectral overlap of the light
response (2.2.1.). The adjustment of the light sources is sources is neglegible, hence
made according to
fEi EfC; d).. = f "in Ej EfiCi d).. i=1..n (9)
eN ;: M limit· f "l Jlj Ej EfC d)..
j= ,
(5) j= ,

and that the spectral response of the reference and sub-

cells are similar, hence
where M1imit is the mismatch-factor calculated for the
limiting sub-cell (see eq.(7) below for the definition of M) i=1 .. n (10)
and S is now and in the following the relative spectral
response. Without knowing the absolute spectral responses hold. Then replacing Mi by Mi~k.l) and J.li by J.li(k)' an iterative
and the absolute irradiances of the sources, it is not adjustment which satisfies eq.l8) will produce the solution
possible to say whether the current-limiting cell of a two- of eq.(7). The iteration should be terminated, when
terminal device is the same under both the simulator and Mi(k)=Mi(k.l)' This iteration is normally done experimentally
by changing the individual light sources successively. Our
the reference spectra. The deviations in the mismatch
approach is to solve eq.(7) numerically thus avoiding
factors Mi for all sub-cells give an estimate for the experimental iteration steps. This approach saves time for
corresponding uncertainty /9/, for only if the deviations are setting the simulator and has the potential for higher
zero are all sub-cells (not only the current-limiting one) set accuracy.
to their proper I"".

2.3.2. Suitable reference cells 3. SPECIAL PROBLEMS OF MULTI-JUNCI10N

Using reference cells with the same spectral response as the MEASUREMENT
sub-cells to be measured is another method for adjusting a
multiple-source simulator /15/. Knowing the calibration A detailed uncertainty discussion is given in /4,17/ for the
numbers, the absolute spectral responses and the absolute measurement of the I-V characteristics of single junction
irradiances of the sources, one can calculate the intensities solar cells. Especially critical for tandem devices are:
J.lj by solving the n dependent linear equations:
Integral Errors:
- The temperature of the top cell will not be the same as for
eN;;: f i=f , Jlj Ej EfC; d).. (6) the bottom cell in general.
- Beam divergence may lead to different illuminated areas
for the top and bottom cells, especially for stacked devices.
with CNi being the calibration number of the i'th reference
- Non-normal incidence may cause wavelength-dependent
cell, RCi being the superscript to indicate the i'th reference
reflectance on the top cell, so the spectral distribution
cell corresponding to the i'th sub-cell. Without any spectral
detected by the bottom cell is distorted.
measurement one can also adjust the simulator iteratively,
- Non-uniformity becomes wavelength-dependent and
(replacing J.I. by /.I(k ) until eq. (3) holds.
therefore is more difficult to control.
This keth~a is in principle easy to use but will cause
- Degradation of the reference cells will be much greater
uncertainties if the spectral responses of the reference cells
when using filtered cells because of degradation of the
and sub-cells differ. When high accuracy is demanded, it
filters.
becomes difficult to get suitable reference cells.

113
- The working point of the sub-cells not being measured The calibration numbers of filtered reference cells are
influences the I-V parameters of the measured sub-cell. determined with a single-source simulator and the standard
mismatch method /19/. Efforts must still be made to
Spectral Errors: improve the uniformities and intensities of the light sources
- The spectral bias errors in multi-junction cells become to minimize uncertainties as described above.
complex because the light sources and spectral responses
influence each other during the tuning of the simulator.
- Non-uniformity of the light sources may lead to the 5. SUMMARY
spectroradiometer detecting a different spectral
distribution in the measurement plane to the device to be The different multi-junction measurement techniques have
been analyzed and classified. We presented · an improved
measured.
version of the reference cell method with mismatch
- Measurement of spectral responses and irradiances is
corrections which allows easier setting of the simulator. We
more difficult.
discussed the special uncertainty problems in multi-junction
.
j
measurement and presented in an outlook the multi-
junction calibration set up of the calibration laboratory at
- m-5. undpr SRC
ISE.
-- m-S. undpr X-25
6. ACKNOWLEDGEMENTS

The authors wish to thank Jutta Beier, Klaus Biicher,

Anatol Jung, Siegfried Kunzelmann, Alexandra Raicu and
Axel Schonecker for valuable discussions and technical
assistance. This work was supported .by the German
300 500 700 900 1100
uavelenqlh (nm) Federal Ministry for Research and Development (BMFT)
under contract number 032-8650E.
Figure 1: Spectral short circuit currents of a m-Si solar cell for simulator
and SRC spectra respectively.
Sometimes f =1-M is taken as a measure for the 7. REFERENCES
uncertainty of the mismatch method. However, even for
low f large spectral deviations from SRC are possible,
increasing the uncertainty in FF and Voc which may be /1/ L.M.Fraas, J.E.Avery, P.E.Gruenbaum,
spectrally dependent /18/. Therefore the spectral current RJ.Ballantyne, W.E.Daniels, PVSEC-5, Kyoto (1990) 89-94
of the test cell under the simulator and for SRC should be /2/ J.Burdick, T.Glatfelter, Solar Cells,18 (1986) 301-314
compared (fig.1). /3/ Measurement Principles for Terrestrial Photovoltaic
(PV) Solar Devices with Reference Spectral Irradiance
XE HU Data, IEC 904-3 (1989)
@ @ /4/ K.Emery, c.Osterwald, Current Topics in PV 3 (1988)
SR 301-350

<)
/5/ G.Virshup, 21th IEEE PVSC, Kissimmee, (1990)
/6/ B.Sopori, C.Marshall, K.Emery, 21th IEEE PVSC,
Kissimmee, (1990) 17A-28 ff
/7/ M.Bennett, RPodlesny, 21th IEEE PVSC, Kissimmee,
(1990)
vc XEnon IQ,.,p /8/ K.Biicher, A.Schonecker, submitted to this conference
HQLogen lo~p
Sppctr""oRQdloMeter
/9/ K.Emery, C.ROsterwald, T.Glatfelter, J.Burdick,
VOCUUr"I Chuck G.Virshup, Solar Cells, 24 (1988) 371-380
/10/ Requirements for Reference Solar Cells, IEC 904-2
(1989)
Figure 2: Block diagram of the multi·junction measurement set·up at /11/ c.R.Osterwald, K.A.Emery, D.R.Myers, RE.Hart,
ISE.
21tb IEEE PVSC, Kissimmee, (1990) 17A-6 ff.
4. OUTLOOK: MULTI-JUNCTION MEASUREMENT /12/ J.Metzdorf, T.Wittcben, B.Nawo, W.Moller, PVSEC-
ATISE 5, Kyoto, (1990) 705-708
/13/ RMatson, K.Emery, RBird, Solar Cells 11 (1984)
Figure 2 shows our experimental set-up: the light from 3 105-145
sources (one xenon arc lamp and 2 halogen lamps) is /14/ C.H.Seaman, Solar Energy 29 No.4 (1982) 291-298
combined using large area dichroic mirrors (350'495 mm2) /15/ RShimokawa, F.Nagamine, M.Nakata, K.Fujisawa,
in the measurement area. Variation of the height of the test Y.Hamakawa, Japanese Journal of Applied Physics, 28
chuck allows us to compensate for the different thicknesses (1989) 845-848
of test and reference cells. The test chuck is /16/ T.Glatfelter, J.Burdick, 19tb IEEE PVSC, New
thermoelectrically temperature-controlled and has a Orleans, (1987) 1187-1193
vacuum holder. We use the iterative mismatch method /17/ K.Heidler, J.Beier, 8th European PV Sol.En.Conf.,
described in (2.3.3.). The relative irradiance can be Florence, (1988)
measured in situ with a spectroradiometer from 350 to 3000 /18/ K.Emery, c.Osterwald, Solar Cells, 21 (1987) 313-327
nm. The spectral responses of multi-junctions are measured
/19/ K.Biicher, K.Heidler, PVSEC-5, Kyoto, (1990)
with a grating monochromator from 300 to 2000 nm /8/.

114
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CONTACT-FREE LUMINESCENT MEASURING

OF (n-p)GaAs-pAlGaAs SOLAR CELL'S PARAMETERS.
V.D.Rumyantsev
A.F.Ioffe Physico-Technical Institute
Polytechnicheskaya 26, Leningrad 194021, USSR

ABSTRACT. The purpose of this work is to control the main photoe-

lectric parameters of AlGaAs heterostructure solar cell wafers-
internal collection efficiency of carriers (Q~), the shape of p-n
junction I-V curve and sheet resistance (R~~) -before contact fa-
brication stage. Uniform illumination with a shadowed stripe is
used. Under open cirouit (O.C.) oonditions summarized photo- and
electroluminescent signal (npL+ n~~) is measured within the illu-
minated areas and only eleotrolum~nescent one (n~L) - within the
shadowed stripe. The lower limit for Q~ is estimated as Q"~nE4/
/npl.. + nEt." The Vo.c. values of SC wafer are measured by comparing
the nel.. signal of wafer and the one of reference AlGaAs-SC with
contacts. A number of V~c. values (in Volts) and corresponding
number of photoourrent values (in relative units) form the working
part of p-n junction I-V curve. The R~~ value is estimated by com-
paring the nCL distributions within the shadowed stripes of exa-
mined wafer and referenoe one. Thus, a basis of simple oontaot-
free control of SC parameters is the following: the higher is lu-
minescenoe intensity within the shadowed stripe - the better are
SC parameters.

1. InTRODUCTION rameters of AlGaAs SC-internal collecti-

on efficiency of photogene rated oarri-
Heterostructure (n-p)GaAs-pAIGaAs ers (Q~ ), the shape of p-n junction
solar cells include the direot gap pho- I-V curve and sheet resistance (Rs~ ).
toactive p-GaAs region with high inter- These methods can be considered as a
nal quantum efficiency of luminesoence. basis for a simpl.e and effective control
Thus the luminescent methods oan be used of SC wafers before contact fabrioation
for the oontrol of SC parameters. Lumi- stage.
nescenoe may arise due to the bulk reco-
mbination of photogenerated carriers wi-
thin p-GaAs layer - this is photolumi- 2. THEORY
nescence (PL), or due to recombination
of injected from n- to p-GaAs electrons, The energy diagram of SC under open
when forward current diffusion compo- circuit conditions is given on Fig.1.a.
nent flows through p-n junction - this
is electroluminescence (EL). The measu-
ring of PL-signal at a photoexitation
with variable wavelength has been used
for the cQntact-free spectral responce
characterization (in relative units) of
AlGaAs SC[1]. The observations of tiEL
map" on the photosensitive surfaoe at
forward currents were involved to iden-
tify different kinds of defects and dis-
tributed components of internal resis-
tance in GaAs-based SC with contacts
[2-41.It has been shown in[51, that at
relatively high photoexitation levels
both PL and EL can arise in SC under
O.C. conditions, if O.C. voltage is high
enough. This fact was used for the in- n..-&aAs ?- Gc.k. ~-AP.&aAs
vestigations of local eleotric-active Pex~~.
defects of p-n junction in SC wafers by
analyzing summarized "PL+EL map" of such
wafers under photoexitation [5]. Finally,
the method of PL and EL signals separa_ nEL
tion using the local photoexitation of
wafers has been proposed for a contact-
free estimation of Vo.t values L6] • The
description of luminescent methods men-
tioned above is given in (11 • In this Fig.1: Energy diagram of SC under O.C.
paper the quantitative contact-free lu- conditions (a) and scheme of currents'
minescent methods are presented for the flows in the SC wafer without contacts
measuring of the main photoelectric pa- (b) •

115
We assume the following arrangement. us
The short-wave incident light penetra-
tes the photoactive p-GaAs layer through c= ~.
n. E.L •
'D = cinpL.
tit\.. E.L
(7)
wide-gap "window" p-Alll.Ga'i-lI.As (x~ 0.8).
This light is absorbed completely wi- at given illumination level. It is ne-
thin p-GaAs layer and thus generates mi- cessary to oonsider the double-exponen-
nority carrier electrons under low exi- tional model for dark ourrent-voltage
tation level conditions. The diffusion charaoteristio of p-n junction:
length of electrons in p-GaAs layer is
assumed to be large enough so that the i.-Ii. =- ~ (o)c:l ex"r 4
K
Yo.c..
T •' (9)
distribution of electrons is practical-
ly uniform across photoactive region.
In the following sections the equipoten- . _. q, Vo.c. ( )
tial conditions of front region in p-n v"t, - "lo)"t. e.x. ~ AK T 9
junction are to be expected. Differentiating of (8) and (9) on Vo.c..
and taking into acoount the relations
2.1 Internal collection efficiency mentioned above we oan obtain:
Some photogenerated carriers recom-
bine radiatively and nonradiatively (10)
both in the bulk and on the front inter-
face of p-GaAs layer. The total value Since two unknown quantities Q~ and A
of recombination losses is proportional do not depend both on illumination level
to losses in bulk, which are accompa- and photocurrent density, it is possible
nied by photoluminescence. to calculate their values having only
We can consider the total losses luminesoent parameters C and D, but at
of photogenerated carriers introducing two different illumination levels - at
the value of "losses ourrent density" relatively low one (C 1 ;D1 ) and at rela-
itoss in opposite to the value of pho- tively high one (C1 ; D~):
tocurrent density ip~ for carriers
oolleoted by p-n junotion. In this oase A =. Cs - C 2. (11 )
the value of internal colleotion effi- D" -D,2.
ciency Q~ may be presented in the
form: Q.... (1+ Pi CC D2-C1 )-4, (12)
Q~ =. \..1'''' •
"to!>~ + Lpk.
(1 )
2.2
•
I-V curve
Di -D2.+ C,2.-C"

Here itov.. value is oonnected with PL Voe.. value is obtained from Eq.(4)
intensity np~ b~ expression: and Eq; (8), but Rand i{o)tlare unknown:
(2)
=..1U. tn.
Y\.I'L=R'LoBDI>!>'
were R is a ooefficient inoluding inter- V, t\.EL. (1)
nal quantum effioienoy of radiative re- a.c. '\'
R . \'(a)a.
combination and part of nonradiative We use GaAs SC with oontacts as a re-
reoombination in p-GaAs layer, optioal ference sample during SC wafers' tes-
and geometrioal factors oonnected with ting. It is necessary in this oase to
light bringing out and its photometry have similar heterostructure design
measuring. On the other hand i,~ value for referenoe and exeminated samples.
may be presented as a sum of diffusion Evidently the dependenses of 19 ne~
and recombination components of ourrent on Vo.~. form straight lines with equal
(i, and i~ ) flowing through p-n juno- slopes both for reference and examined
tion at O.C. conditions (see Fig.1,b): samples:
\',.1..= Ld,. ~ L't. •
The flowing of diffusion current oompo-
(3)
A(e~n.EL)=O.4?>4~ AVo.fI. ' (14)
nent is aocompanied by electrolum1nes- On the other hand the R and ilD)d. va-
oenoe owing to injeotion of electrons lues are relatively stable for~the sam-
and subsequent recombination in p-GaAs ples fabricated in different technolo-
layer under the same conditions as gical prooesses but having the same
photogenerated electrons (2): structures and heavily doped layers.
n.cL = R. • i.d. . (4) In contrast to this the conoentration
Taking into account the injection coe- of deep centres which are aotive in
ffioient Y= i-cl/ i.,~, we can rewrite space charge region of p-n junction and
the expression (1) in the form: reduce the photogenerated voltage may
Q.~ = (i T ~~~r(· r (5)
vary signifioantly, as well the diffe-
rent types of ourrent leakages may
arise in wafers before oontaot fabri-
Since '(6 1, the lower limit of Q~ is: cation stage. So the diffusion compo-
nent of current and EL intensity oould
Q.::.. n..Eb • (6) be used as a basis for oontact-free
~,.,. Y\.~L.+ Y\..EL testing of Vo.c • values (see Fig.2,a).
It is interesting that expression (6) The reference sample - one of the best
obtains the Q~ value in absolute units SC - and the examined wafer are illu-
minated under the same oonditions. The
substituting the DpL. and nt::L. values EL-signal, whioh is emited from shado-
measured in relative units. If in expe- wed stripe of reference SC, is made
riment npL~ n~L' that means in fact equal with one of examined wafer by
Qj.=1. variable shunting of SC. In this case
In general case the functional de- the VO~ value of examined wafer must
pendence between npL. and n~L. values be equal to voltage measured by volt-
at different illumination levels has to meter on reference SC. A number of Vo.c.
be used oalculating of Q~ value. Let values in Volts is possible to get at

116
 different illumination levels. The lo- the image of linear - like soreen on the
west value of Vo.c:.. is connected with the photosensitive surfaoe of wafer. The
sensitivit~ of EL-signal registration
s~stem.

a. Hq-2amp

5C wat~"t.. 1\\
came'lQ

~I... i !
~::.1
1\!on.~O
1

6 ....... ~t1=C
i~
Q~'-{

1ti=±=
:i 1.0 R!>",~O
~
.£ Q~::.- ~
--'
....
r-'
~
I
~~ Rs.h.:::"0
jH-;-
c...)
(/) ----~
0
.......
~
u
~ 11'11.2. I Q~.::.{

.....o
~ VI
-.j.o Iplll
I
b~!>t.at1c.e
R~'::"O

0
.J/.
,.......... • ..>.... lph.~
~ Fig.): Optical la~out of the in-
stallation forcontaot-free mea-
0 VO.C.tnCOl surements of sa parameters
Fig.2. Arrangement of samples (a) second lens projeots the "luminescent
and scheme of plotting of p-n junc- map" of sa wafer on the photocathode of
tion's I-V curve for se wafer wi- IR-camera. Here also the small-size pho-
thout contaots (b). todetector is situated. Visual oontrol
of "luminesoent map" and detector posi-
The ourrent soale of I-V ourv.e ma~ tion is oarried out via IR-oamera. The
be either in relative or in absolute inset of Fig.3 demonstrates the diffe-
(milliamperes, for example> units. In rent kinds of 1uminesoenoe intensit~
the first oase the fact of proportional distributions aoross the shadowed
dependence of photocurrent on illumina- stripe, whioh oan be obtained at diffe-
tion level both for examined wafer and rent values of Qv and R$~. It must
reference se is taken into account. p-n be olear that vertioal soaling of these
junction's I-V ourve of examined wafer graphs is conneoted with the Vo.c.. values.
is plotted in the wa~ demonstrated in The samples of AlGaAs SCs with oon-
Fig.2,b. In the seoond case the Q~ de- tacts were used for contact-free lumi-
termination procedure mentioned above nescent measurements and subsequent ve-
and measuring of illumination intensit~ rifications of obtained Qv and Voc
must be used in addition to the first. values b~ traditional methods. In this
case there was the additional possibi-
lit~ to measure the PL-signal regardless
2.3 Sheet resistance of EL one -under short circuit conditi-
As have been mentioned above the
measuring of Qj" and Vo.c . is carried ons.
out under equipotential conditions of
illuminated and shadowed areas of SC wa- 4. RESULTS
fer. In contrast to this for the esti-
mation of R.~ values nonequipotential A number of se's with different
conditions require, which oan be obtai- photoaotive p-GaAs l~er thicknesses
ned owing to either inoreasing in illu-
mination level or widening of the sha- were tested. The n"J,. and fiE\. values
dowed stripe. Under suoh oonditions the as well as C and D ~uminescent parame-
R.~ value of se wafer is estimated b~ ters were obtained at different photo-
comparing the EL intensit~ distributi- exitation levels. For three samples the
ons within shadowed stripes of exami- Qi, values oalculated using ECV. (12)
ned wafer and of referenoe one having were as follows: 0.99; 0.92; 0.80. On
known value of Rs~ • the other hand suoh values measured by
traditional "contact" method at .il =
633 nm, taking into acoount reflected
). EXPERIMENT part of radiation, were, respectivel~:
0.88; 0.85; 0.81. Thus, luminesoent me-
The PL- an EL-signals were measu- asurements allow us to seleot the SC sam-
red in the following wa~ (Fig.3). The ples and wafers having the best Q~ va-
~ellow-green light from do are Hg- lues.
lamp causes the infra-red luminescenoe Figure 4 shows the results of oon-
of sa wafer. The first lens projeots taot-free Vo.c measurements for one of
AIGaAs SC at reduced light intensities

117
 (circular points) . Here the I-V cha-
racteristic of p-n junction which was gues in Photovoltaic Solar Energy Con-
verters Laboratory.
r-:- to
:::s

-
REFERENCES

-
..:>
41
~ 1 • G.D.Pettit, J.M.Woodall, H.J.Hovel.
s:
01 Appl.Phys.Lett. a5 , 335 (1919).
~ OS 0.'1 ~ 2. Zh.I.Alferov, Kh~.Aripov, B.V.Ego-
~;:j :::s
u roY, V.R.Larionov, V.D.Rumyantsev,
u 0 O.M.Fedorova, L.Hernandez. Sov.
.....oo
~
0 Phys. Semicond. 14 , 403 (1980).
oS:
t:>- 3. G.W.Turner, J.C.~Fan, J.P.Salerno.
if. 0 O Solar Cells 1 , 261 (1979/80).
4. Kh.K.Aripov, ~.S.Koroleva, V.R.La-
0.'1 0.8 0.9 to rionov, T.A.Nuller, V.D.Rumyantsev.
C.C. voHa.Cle lVofh) Sov.Phys.Tech.Phys. ~t 202 (1983).
5. V.R.Larionov, T.B.Popova, V.D.Rumy-
Fig.4: The results of contact- antsev, L.I.Flax, Sh.Sh.Shamukhame-
free (circular points) and "tradi- dove Proc. 4th USSR Conf. "Physical
tional" (triangular points) mea- processes in semicond. heterostruc-
surements of p-n junction I-V tures" (in Russian), Minsk. Part 2,
curve. 220 (1986).
6. A.A.Alaev, A.B.Guchmazov, V.D.Ru-
myantsev. Proc. 1th USSR Coof.
"Photoelectric phenomena in semi-
constructed from measured with using of condo materials" (in Russian),
contacts If'" - Vo.(.. vaJ.ues is shown. Tashkent. P.414 (1989).
It is seen that the accuracy 5-10 mV is 1. V.M.Andreev, V.A.Grilikhes, V.D.Ru-
realized experimentally for contact- myantsev "Photoelectric conversion
free measuring of Vo.c. .• The lowest mea- of concentrated solar radiation"
sured Vo.C. values were approximately (in Russian), Nauka, Leningrad
0.8 V when EL signals from examined and (1989) •
reference SC s were compared by visual
or photoelectric methods.

5. CONCLUSION
Thus, the described contact-free me-
thod is quantitative and allows us to
measure the main SC parameters. But in
practice it is not necessary to do any
measurements for control of SC wafers.
It is enough to see (via IR-camera),
that EL intensity within the shadowed
stripe for SC wafer is higher than the
one for reference SC, in which a mini-
mal acceptable voltage is established
by shunting. If such procedure is car-
ried out at approximately 1/10 of wo~
king illumination level, it means that
maximum output power point voltage of
p-n junction in examined se wafer is su-
itable for us. In the course of this
procedure the prOXimity of Q~ value to
unit can be estimated - luminescence in-
tensity within the shadowed stripe must
be practically the same as the one in
illuminated areas. Finally, there are
no sheet resistance problems, if the
shadowed stripe on the examined wafer
looks "to be focused" at working illu-
mination level and at given stripe
width depending on contact grid spa-
cing of se. There are also other modi-
fications of contact-free luminescent
control. In any ease the general
rule is as follOWS: the higher is lu-
minescence intensity within the shado-
wed area of SC or wafer - the better
are SC parameters.

ACKNOWLEDGEMENT

The author presents his gratitude

for the help received from his col1ea-

118
 POSTER SESSION IB

Amorphous Silicon and Related Materials I

10TH EUROPEAN PHoroVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A COMPARISON BETWEEN FLUORINATED AID HYDROGENATED AMORPHOUS SILICON CARBIDE PREPARED BY

REACTIVE SPUTTERING.

F.De.ichelis, C.F. Pirri, E. Tresso

Dipartimento di Fisica Politecnico di Torino (Italy)
T. Stapinski
Dept. Physics and Electron Technology Acc. Mining and Metallurgy - Krakow (Poland)
L. Boarino
lEN Galileo Ferraris - Torino (Italy)
P. Rava
Elettrorava S.p.A. - Torino (Italy)

ABSTRACT
a-8iC:H and a-8iC:H:F films have been prepared by reactive sputtering in mixtures of argon,
hydrogen, methane or freon mixtures with different concentrations. The optical and electrical
properties, the chemical bonding configuration and the density of defects in these materials
have been studied in order to understand how the introduction of F influences the physical
properties of the a-SiC random network.

INTRODUCTION sputtering source. The target of monocrys-

talline electronic grade silicon (120 mm
Fluorine incorporation in a-Si:H and in diameter) was located 20 mm from the
a-SiC:H films grown by glow discharge substrates. The substrates of Corning 7059
decomposition of 8iH4 and CF4 have been glass, double side polished c-Si and quartz
investigated in previous papers (1-6). were placed on a rotating table. This
The aim of this paper is to present further configuration allowed to obtain homogeneous
detailed information on the influence of films. The content of the gas mixture of Ar
fluorination on the properties of amorphous (5N). Hz (5.5N). CH4 (5.5N). CF4 (4.5N) was
hydrogenated silicon carbide. 8iC:H and controlled by means 0; mass flow controllers.
SiC:H,F thin films have been prepared by The deposition temperature was kept constant
reactive sputtering in mixtures of argon, at 250 0 C and the rf power at 300W; the total
hydrogen, methane or freon gas mixtures with gas pressure varied in the range 0.40 to 0.67
different concentrations. Pa. The other parameters of the depositions
The measured strength of the 8i-F bond for the sets of a-SiC:H and a-8iC:H,F samples
compared to the Si -H bond and the larger (labelled SCH and SCF followed by a number
electronegativity of F with respect to H have indicating the flow of methane or freon in
suggested that the fluorination of amorphous sccm respectively) are shown in Table I.
silicon coUld increase photo and thermal Transmittance and reflectance measurements
stability, decrease- the Stabler-Wronski were performed with a Perkin Elmer UV-VIS-NIR
effect and increase solar cell stability. The Lambda 9 spectrophotometer. The extraction of
density of localized states in a-Si could optical constants was done following the
possibly be reduced by the introduction of F procedure described in Ref. 9. Information on
(7,6) . the spectral distribution of the sub-gap
The optical and electrical properties, states and their density was determined by
the chemical bonding configuration and the photothermal deflection spectroscopy (PDS)
density of defects in these materials have (10) which allows the detection of weak
been studied. absorption levels.
From the experimental results it was Electron spin resonance (ESR) measurements
deduced that due to the introduction of F into were performed by means of a Varian EPR-109
the a-SiC random network the spin density spectrometer. Both the g-values and the
decreases, the optical gap and the dark intensity of the spin resonance signal of the
conducti vi ty increases and a change in the samples were measured by comparison with the
structure occurs. value obtained from Varian pitch (g=2.0029)
and from DPPH (1.ldyphenil-2-pyrylhydrazil)
standard sample respectively.
EXPERIMENTAL Infrared (I.R.) spectroscopy measurements
were made in the absorption mode using a Bruker
Films -of a-SiC:H and a-SiC:H,F were IFS 65 Fourier Transform spectrophotometer
deposited in a high vacuum challber with a base between 400 and 4000 cm- 1 •
pressure of about lo-a Pa. The growth species
were generated by rf planar magnetron

121
 The dark conductivities at room temperature
were measured in vacuum (0.133 Pal by a
Rewlett-Packard high resistance meter 4329A.

TABLE I
Deposition conditions

H. flow Ar flow CH. flow CF. flow Dep. rate

Sample [sccm] [sccm] [sccm) [sccm] [A/s]
SHO 18 100 - - 1.40
SCR1 17 100 1.0 - 1.72
SCH2 16 100 2.0 - 1.53
SCH3 15 100 3.0 - 1.53
SCH4 14 100 4.0 - 1.38
SCH5 13 100 5.0 - 1.33
SCH6 12 100 6.0 - 1.42
SCH7 11 100 7.0 - 1.17
SCH8 10 100 8.0 - 1.24
SCH9 9 100 9.0 - 1.30
scn 17 100 - 1.5 1.39
SCF2 16 100 - 2.2 1.05
SCF4 14 100 - 4.0 1.39
SCF5 13 100 - 5.0 -
SCF7 11 100 - 7.0 1.02
SCF9 9 100 - 9.0 1.00
SCFll 7 100 - 11.0 0.61

SiF stretching modes (14) and at 1100 cm- 1

RESULTS AND DISCUSSION attributed to CFa mode (14). Fig. 1 shows the
typical I.R. spectra of two a-SiC:H,F samples
The chemical bonding structure of the having different fluorine content. An
a-SiC:H and a-SiC:H:F films was investigated absorption band at 3700 cm- I , present only in
by means of I.R. Spectroscopy. the fluorine rich samples can be assigned to
The absorption spectra of a-SiC:H exhibits the HF vibration mode (15). As the CF4 flow
different H bonds to Si and C. The 2100 and increases the intensities of the SiHz and SiH
2000 cm- 1 peaks were assigned to the SiH. and peaks decrease. At the same time a sharp
SiH stretching modes (11). The strongest increase of the SiF. (before) and of the CF.
features in-these spectra occur at 780 and (after) can be noted. Probably F radicals more
650 cm- 1 • The peaks were assigned to the Si -CHa than H radicals remove "weak" amorphous tissue
rocking or wagging and to the SiH rocking mode by etching it preferentially (16). The removal
respectively (12). The band at 1000 cm- 1 is of H or F from the growing surface can be
due to CRD rocking and/or wagging vibration justified by energetically very favourable
(13). It can be noted that with the increase reactions (17):
of the CH. flow the intensity of the peak of
the SiH. mode (2100 cm)-l remains practically Si-Si-F + F -) Si + SiFa
constant, whereas the peak of SiH decreases
until it disappears fot high carbon content AH - -364KJ Imol
and a new band with a doublet structure at
2900 cm- 1 appears. This band is assigned to
CHa stretching (13). On the other hand a sharp Si-H + F -) Si + HF
rise of the intensity of the peak due to Si-CHa
up to a saturation is shown when the CR. flow AH - -286KJ Imol
increases. This seems to indicate that H
prefers to bond to carbon in the form CHa or
SiCH3 or CHD, since the binding energy of The presence of CF. group is a direct
hrdrogen with carbon is higher than with consequence of CF4 dissociation. Only a weak
sllicon. CH bond can be observed; in fact CH bonds have
In the absorption spectra of a-SiC:H:F we a lower strength (81 kcal/mol) in comparison
observe, other than the characteristic peaks to CF bonds (128 kcal/mol), (18,19).
corresponding to the bond of H to Si and C,
absorption bands at 860 cm- I assigned to SiF.,

122
 a)
SiH

1000
SiH,
5iH I

E
u
/\ FIG. 1
Typical I.R. spectra of a-SiC:H,F
'--' b) films produced with different freon
d eF, flows:
1 a)SCFf b)SCFU

1000

o
400 2300
In Fig.J the values of the static refractive
index are reported. It can be seen that for
high CH. flow the refractive index saturates
In the region where the absorption at values n=2.05. The hydrogenated amorphous
coefficient is higher than 2 10. cm-' all carbon films have a refractive index that
samples can be described by Tauc's equation. covers the range from 1.8 to 2.3 (20). The
Tauc's gap Eg as a function of the CH. and present results indicate a tendency towards
CF. flows is reported in Fig. 2. The trend of a prevalence of the carbon atoms in the films.
Eg for a-SiC:H films indicates that the More fluorine than hydrogen causes a decrease
addition of carbon widens the pseudogap up to of the refractive index. The value of n=1.69
a saturation value. At high CII. flows the for high CF. flow indicates that probably the
pseudogap shows a decrease probably due to carbon atoms present in the films are
the appearance of graphitic bonds among the polymer-like.
carbon atoms. Further information comes from The density of spins for all the samples,
the measurements of absorption coefficient in as deduced from ESR measurements, is showed
the band-tail states transition range. The in Fig.4. It can be observed that fluorinated
addition of carbon results in a gradual films have a density of spins in the range
broadening of the Urbach edge. The char- 10'8-10'9 cm-', while for CH. deposited films
acteristic energy varies between 80 meV for the dependence on carbon content is more marked
pure a-Si:H to 280 meV tor the sample SCH9, and Ns reaches values of the order of 10'·
for fluorinated samples a less remarkable cm- 3 •
increase has been observed and the highest
value (230 meV) is reached for the sample 3.5
SCFll. The pseudogap tor a-SiC:H,F films shows 9 s- SiC:H
a trend dif~erent from that of a-SiC:H. For .0. a- SiC:F,H
low CF. flow'Ev remains practically constant. J
With. the increase of freon flow rate Ev sharply
increases up to values as high as 3.1 eV. 0 0
3.5
... a· S i C : F,H .0.
.0.
0 9
• a-SiC:H t:
.0.

0 0
0
.0.

0
>Q) 0
.0.

.0.
.0.
1.5
0 R 0.1
FIG. 3 Values of the index of refraction at
2000 nm as a function of methane or freon
relative flow rate (R).
1.5
o R 0.1
FIG. 2 Energy gap as a function of relative
methane or inoh flow rate (R).

123
 ~I~--------------------------------'----'
10 .a.SiC:H
A a-SiC:F,H

REFERENCES
"I (1) N. Saito. Y. Yamaguchi - Phys. status Solidi
E (a) 69, 133 (1982).
~
II (2) A.Morimoto. T.Miura. M.Kumeda. T.Shimuzu
Z - Jpn. J. Appl. Phys. 22. 908 (1983).
(3) T.Ueseqi. H.Ihara. H.Matsumura - Jpn. J.
Appl. Phys~. 24. 1263 (1985).
(4) N.Saito, T.Yamada. T.Yamaquchi, N.Tanaka
- Phil. Mag. B 52, 987 (1985).
101 ~ ______________________________ ~

(5) N.Saito, N.Tanaka - Phys. Status Solidi

o R 0.1 (a) 93, 341· (1986) •
FIG. 4 Spin densities as a function of
methane or freon relative flow rate (R). (6) G.Land Weer, M.E.M.de Wintel, J.Beremar
- Proc. 8th E.C.Photovoltaic Solar Energy
Conf., Kluwer Academic Publishers.
(7) A.Madan, S.R.Ovshinsky, E.Benn - Pllil.
Finally the dark conductivity as a function Mag. B 40. 259 (1979).
of CH4 and CF4 flows is plot ted in Fig. 5.
The dark conductivity of a-SiC:H,F is of one (8) A.Madan, "Hydrogenated Amorphous Sili-
order of magnitude higher than that of a-SiC:H con", Springer Verlag, (1984).
and for both sets there is no detectable
variation with flows. The experimental values (9) F.Demichelis, J.Kaniadakis,A.Tagliafero,
seem to confirm the prevalent influence of E.Tresso - Appl. Optics 26 (1987).1737.
carbon structure on all the properties of the
samples. (10) N.Amer, W.B.Jackson in "Semiconductors
and Semimetals", Acad. Press Inc., vol. 21
B,p.83 - ed.J.Pankove.
10'8~---------------------------'
A .·SiC:F,H
• a,SiC:H (11) M.H.Brodsky, M.Cardona , J.J.Cuomo -
Phys. Rev. B 16, 3556, (1977).
(12) H. Wieder, M. Cardona. C. R. Guarnieri
- Phys. Status Solidi B 92, 99 (1979).
(13) B.Dischler, R.E.Sah. P.Koidl, W.Fluhr,
A. Vokam - Proc. 7th Int. Symp. on Plasma
Chern. (Eindhoven 1985), vol.r p.45 •

• • • • • • • • A
(14) P.J.Zanzucchi in "Semiconductors and
Semimetals" , Acad. Press Inc. , vol.21B
p.113 - ed.J.Pankove.

18C~ ________________________ ~ (15) K.Nakamoto: "Infrared Spectra of· Inor-

ganic and Coordination compounds", J .Wiley
o R 0.1 Inc. (1963).
(16) C. C. Tsai in "Advances in Disordered
FIG. 5 Dark conductivity as a function of Semiconductors". vol.1, ed. H.Fritzsche -
methane or freon relative flow rate. World Scientific, Singapore (1988).
(17) Y.Okada. J.Chen, r.H.Campbell.
CONCLUSIONS P.M.Fauchet, S.Wagner Journ. of
Non-crystalline Solids 114 (1989).816.
The effects of fluorine incorporation
on the structuraL optical and electrical (18) K.A. Gingerich - J.Chem. Phys. 54, 2646
properties of a-SiC:H.F have been system- (1971) .
atically studied. Examining I.R. spectra it
was deduced that in samples with sufficient (19) J.G.Edwards, H.F.Franklin, P.W.Gilles -
fluorine SiFz and SiF absorption increases to J.Chem. Phys. 54, 545 (1971).
the det.riment of SiHz and SiH absorption. The
optical gap increases towards values as high (20) A.Bueberger - J. Appl. Phys. 54, 4590
as 3.7 eV. At higher CF4 flow the influence (1983) •
of carbon on the pseudogap. refractive index
and dark conductivity becomes predominant.

124
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

INFLUENCE OF SILICON ON PROPERTIES OF HYDROGENATED AMORPHOUS CARBON

F. Demichelis. G. Giachello. C.F. Pirri and A. Tagliaferro

Dip. Fisica - Politecnico Torino (Italy)

R. Spagnolo
lEN Galileo Ferraris - Torino (Italy)

ABSTRACT

Films of a-SiC:H having different silicon contents have been deposited at room temperature by
sputter assisted plasma CVO of a graphite target in a silane - argon atmosphere. The insertion of
small amounts of silicon in the' a-C:H network yield a material which has mechanical properties
similar to OLC and optical gap up to 2.2 eV. The room temperature electrical conductivities are two
orders of magnitude higher than typical PECVO deposited hydrogenated amorphous' silicon carbide.

INTRODUCTION conductivity is of about two order of magnitude

higher than that of glow-discharge amorphous
It is known that diamond-like films show silicon carbide rich in silicon.
unusual hardness, low friction coefficient. high
wear resistance and a remarkable resistance to
chemIcal attack. Because of ttJeir high trans-
parence, diamond-like films have been used as EXPERIMENTAL
antireflecting protective coatings for
semiconductor optics in the mid-infrared region The samples were de?osited by RF Sputtering
0.2). However. hardness as well as good of a 5N (purity) graphite target 20 cm in
adhesion of the films must be achieved for diameter in an argon + silane atmosphere. A
using the diamond-like films as coating for constant power level (300 W) but different
optical applications. silane flow rates from 1 to 8 sccm were used.
The pressure and the argon flow rate were
Hydrogenated amorphous carbon can have maintained at 27 mTorr and 99.8 seem
a wide range of properties falling between those respectively for all films. Electrode spacing was
of graphite and diamond, depending on the kept at 2.5 cm and deposition took place at
Sp2/Sp' ratio. In fact the optical and electricai room temperature (self heating increased the
properties appear to be dominated by the temperature up to 350 K). The used deposition
relative amounts of Sp3 and Sp2 bonding. So a method can be classified as a sputter assisted
very wide range of conductivities of carbon plasma CVD method, in which the sources of C
films have been observed. It has recently been and Si + H are independently controlled.
reported (3.4) that diamond-like films con-
taining low ,amounts of silicon, deposited by Absorption coefflcient In the high absorption
PECVD, show outstanding tribological region was extracted. following procedure
properties. The introduction of small amounts reported in Ref.5. from optical transmittance
of silicon in hydrogenated amorphous carbon and reflectance. measured by a Perkin Elmer
changes the structure of the samples and. UV - VIS-NIR Lambda 9 spectrophotometer. In the
therefore. their optical. electrical and mech- weak absorption region, Photothermal Deflection
anicai properties. Low content of sil1con appears Spectroscopy (PDS) was used. ESR measurements
to stabilize the formation of Sp3 C-C bonds were made at room temperature by a Varian
increasing transparency and band gap. On the EPR 109 spectrometer. Both the g-values and
other hand. because of the high carbon content, the spin densities were measured by comparison
quite an amount of the carbon atoms are Sp2 with DPPH (t.l dyphenil-2-pyrylhydrlzil)
bonded, so providing appreciable conductivity. standard samples and Varian pitch (g=2.0029).
In view of the application of such films as DC electrical conductivity was measured in
window in solar cells we have realized films of planar configuration in a vacuum chamber (to-.
semiconducting amorphous silicon carbide Torr) 'by means of a Hewlett Packard high
having. at the same time. optical properties resistance meter 4329A.
similar to those of diamond-like films. The films
were obtained by RF Sputtering of a graphite In the following. samples are labelled CSx,
target in Ar + SiH. atmospheres by varying the where x is the SiH. flow rate in sccm.
SiH. flow from 0 to 8 sccm. We have studied
their physical properties by optical absorption.
DC electrical conductivity and electron spin
resonance (ESR) measurements. RESULTS AND DISCUSSION
We will show that under some deposition
conditions the films become diamond';'like in The deposition rates (DR) are reported in
optical but not In electrical properties. Their Table I. We can observe an Increase of the

125
deposition rates with silane flow rate increase in a-C:II dark conductivity at room temperature
at low fluxes. while at higher silane flow rates is due to hopping between neighbouring
the deposition rate become a constant. This Is graphitic Islands embedded in a sp3 matrix if
due to the fact that. at low silane fluxes. the sp2 sites do not percolate and to hopping in
sputtering of carbon is still Important In the band tails when they percolate. The
determining the amount of deposited material. insertion of silicon atoms will then assume
while the glow discharge decomposition of silane Importance When the average distance between
becomes the rate limiting process at higher the islands is so Increased to make the "Islands
fluxes. contribution" to conductivity lower than the
Optical absorption coefficient spectra are activated conductivity value of the sp3
reported In Fig. 1 for films of different silicon network. This can readily be observed In Table
content. At first. the addition of small amounts I. if the conductivity for silane fluxes lower
of silicon to a-C:H (sample CSll causes the than 3 sccm are compared with those of samples
appearance of the Urbach tail. leaving with silane flow rates higher than 3 seem.
unchanged the absorption coefficient ' In the
high absorption region. Further increase of the
silane flow rate (sample CS5) completely change Typical ESR signals for a-C:H and silicon
the shape of the absorption coefficient curve. rich a-CSI :H (CS5) are shown in Fig. 2. whlIe
giving a shape typical of amorphous semi- ESR data for all samples are reported in Table
conductors. While no Urbach parameter can be l. An increase in silicon content causes a
defined In a-C:H sample. values of 280 meV decrease in spin density and a shift In the
and 180 meV are measured for samples CSI and g-value. For low silane flow rates . the inclusion
CS5 respectively. of silicon in the a-C:II networl( reduces the
number and size of the graphitic islands. and
in turn of the defects sitting on their border.
which are responsible for the /a-C:H ESR signal
(6). For higher silane fluxes. the g-value
106.~------------------------, increases. moving toward the a-SI:H value.
- a-C:H showing a predominance of dangling bond type
CS1 defects (8).
_ .- CS 5
In Fig. 3 a semllog plot of the spin density
as a function of energy gap Is reported. It can
be observed that an increase of energy gap. if
a not too large amount of silicon enter the
'/ ./ network. brings an approximately exponential
decrease of spin densities due to carbon related
'/ / defects (see g-values in Table I) . When a larger
amount of silicon is added and the amount oC
/ silicon related defects become Importan t. no
definite trend can be observed .
/
1
10~ __ ~ ____ ~ __ ~~ ____ ~~

o 4 a-C : rl
CS5
FIG . 1 Absorption coefficient for samples of
different slllcon content

The values of the Tauc gap Eu as a function ~

of Sl1l4 flow rate are reported In Table I. While
in amorphous semiconductor the Tauc gap Eg
is an Indicator of the energy spacing between
valence and conduction bands. it has been
shown (6) that In a-C and a-C :H it is an
Indicator of the size distribution of graphitic
like islands. The increase in Eg for low silane
fluxes « 3 sccm) Is then explained with a size
reduction oC the graphitic islands. At higher
silane fluxes Eg decreases as a consequence 0 f 331 H[mT ] 339
silicon prevalence In the network.
The data of room temperature DC conductivity FIG . 2 ESR signal for a silicon rich and a
are reported In Table I. The values oC typical silicon free film
a-SI:lI and a-SiC:H obtained by glow discharge
and of a-C:lI are also reported for comparison.
It has been shown in previous papers (7) that

126
 1'AIJLE I
Sample E. ". N. g-value
leV] (n"·em"J rC/Tl-~J

a-C:H* 1.4 1.0 10-- 4.0 10" 2.0029

GSI 1.6 1.4 10- 10 9.0 10'· 2.0028
GS2 1.9 2.1 10- 10 3.2 10'· 2.0029
GS3 ~.O 1./3 10- 1• 0 .7 10 2• 2.0035
CS4 2.1 1.1 10- 1• 2.0 10'· 2.0039
GS5 1.9 5.6 10- 1• 2.2 10 20 2.0040
CS8 1.8 2.8 102 • 2.0045
a-SiC:H8 2.1 10- 1:1 10 iO 2.0045
a-Si:H' 1.8 10- 1• 10 16 2.0055

Ref. 6 and 7.
@ Ref. 9 and 10
# Ref. 8 and 11

REFERENCES

21 r ______________________________
II A. Bubenzer. B.D. Dischler. G. Brandt and
~

10 P. Koidl - .1. Appl. Phys. 54 (1983) 4590

2) M.J. Mirtich. D. Nlr, D. Swec and B. Bani(s

- .r.Vac . Sci. and Techn. A4 (1986) 2680

3) K . Oguri and T . Arai - Thin Solid Films

165 (1988) 139

VI 4) J{. Oguri and T. Arai - J. of Mat. Res. 5

Z (I990) 2567

5) F. Demichelis. G. Kanladakis. A. Tagliaferro

and E. Tresso - App1. Opt. 26 (I987) 1737
19
10 6) D. Dasgupta. F. Demichelis. C.F. Pirr! and
A. Tagliaferro - Phys. Rev. B 43 (1991) 2131
1.2 2.4
7) D. Dasgupta. F. Demichelis and A. Tagliaferro
l"IG. 3 Spin densities and optical gaps - Phil. Mag. B On press)

8) M. Stutzmann - "Properties of amorphous

slllcon" in EMIS Data Reviews (INSPEC Publ.
UK) (1989) 63

CONCLUSION 9) S. Nitta - "Properties of amorphous silicon"

in EMIS Data Reviews (INSPEC PUbl. UK)
The effect of the addition of silicon to (I989) 80
a-C:H having diamond-like characters has been
studied. Increasing silicon content. sizes of the 10) S. Ray - ·Propertles of amorphous silicon"
graphitic-like Islands is reduced and a-SiG:H in EMIS Data Reviews (INSPEC Pub!. UK)
films having optical properties similar to PECVD (I989) 241
deposited a-SlC:H but higher conductivity can
be obtained. 11) V. Augelli - "Properties of amorphous
silicon" in EM IS Data Reviews (INSPEC Publ.
UK) (I989) 185

127
lOTI{ EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

A NEW APPROACH FOR CHARACTERIZING SURFACE INTERFACE AND BULK DEFECTS IN a-Si:H

G. Amato
Istituto Elettrotecnico Nazionale Galileo Ferraris
Strada delle Cacce, 91 - 10135 Turin (Italy)

ABSTRACT. A new analysis approach which combines photoconductivity and

photothermal deflection spectroscopies in a-Si:H is presented. An expo-
nenti a1 di stri but i on of defect states along the fil m is assumed, in
agreement with recent experimental and theoretical results. Fits of ex-
perimental data in 1-3 eV energy range allow to profile electronic de-
fects.

1. INTRODUCTION in a-Si:H films: this distortion consists in an

enhancement of the interference fringes amplitude
The determination of the photoelectronic at low absorptance values [8J. A quantitative de-
properties of Si-based hydrogenated amorphous termi nati on of the surface/i nterface -states has
films is generally carried out by means of very been carried out by assuming a step profile for
we 11 estab 1i shed techni ques 1i ke Photothermal the imaginary part of the refractive index along
Defl ect i on Spectroscopy (PDS) or Photoconducti ve the film depth [9], but thi s procedure, provides
Spectroscopy (PCS). The main reason of the success feasible results only if the defective region is
of these techniques is related to the ease in im- 1oca 1i zed withi n 60-70 nm and is at 1east 100
plementi ng them and to the possibil ity of i nfer- times more absorbing than the bulk [10].
ring the material quality without any huge sample I n some recent papers [11,12] the prob 1em of
preparations. the surface defects has been rel ated to the mOi"e
On the other hand, even if the theoretical genera 1 questi on concerni ng the i nhomogenei ty on
background for both techniques is simple and micrometric scale: it has been found that the as-
straightforward, a number of problems arise if PDS sumption of an exponential distribution of the
and PCS results are compared, or if samples of density of states:
different thickness are measured. In particular,
the film surface/interface status has gained popu- (1)
larity as the principal responsible for the mis-
match between PDS and PCS resu 1ts and for the where Nand Nb are the surface and bulk densities
strong thickness dependence of the films sub-band- of stat~s, x is the distance from the free surface
gap absorption [1,2]. and ~ is the distribution decay length, can effec-
Photoconductive measurements carried on films tively explain the dependence of PDS results with
having different thickness [3,4] also show a thickness D1] and the difference between PDS and
strong dependence of the photoelectronic proper- PCS density of states [12]. In these cases, values
ties on the film thickness, in the sense that thin of ~ of the order of 1 11m and more have been
films are poorer than thick ones. This means that assumed.
surface/interface recombination plays a non-negli- This exponential distribution of defects has
gible role in affecting the photoelectronic prop- been recently interpreted on the basis of a ther-
erties of the films, but this contribution is not mal-equilibration theory in which the, thermal an-
explained in a straightforward way: this is a key nealing of defects during growth can be responsi-
difference with respect to PDS measurements, in ble of such di'stribution [13].
which the surface/interface absorption simply adds The re I ati onshi p between the dens ity of de-
to the bulk one [lJ. fects and the absorption coeffi ci ent a at low
Surface Photovoltage Measurements (SPV) have energy has been pointed out to be [14]:
been carried out in the past on a-Si:H films [5,6J
yi e1di ng fi gures for wi dths of space charge re- N= K f 6 (a) dE (2)
gions, or Dember effect, or surface band bending. where ~ (a) is the difference between the absorp-
On the other hand, by means of Steady State tion coefficient values due to transitions involv-
Photocurrent Grating (SSPG) technique [7], the i ng defect states and the absorption coeffi ci ent
ambi po 1ar diffusi on 1ength can be accurately de- values due to transitions involving tail states, E
termined: it has been recently shown [4J that this is the ph~~on ~?ergy and K is a normalization con-
quantity strongly decays when the fi 1m thi ckness stant (cm eV ).
decreases. Eq. 2 allows to reasonably assume that the
The presence of a very 1oca 1i zed defect re- exponenti a1 decay of N wi th the depth x ref1 ects
gion on the surface/interface can be separately on a similar behavior for a:
detected by analyzing the distort1on occurring in
the absorptance spectra obtained by means of PDS a(x) = as exp [-xh] + a b (3)

128
where a and a are the absorpti on coeffi ci ent
_I
A' = as
b
exp [0. .t (e -Ll.t -1 )] +
s

I I
values It the f~m surface and in the bulk in the
defect-band transition regime.
The aim of this work is to study the capabil- - exp[o./(e -d/.t-1)] +a \ I
l-exp[o.s.t(e -Ll.t -1)]\+
ities of the model (3) in describing the PDS and
PCS behavi ors of films grown by different tech- (9)
niques.
+ as 0. t e -o.s.t [Ei(a .t) - Ei(o. t e-Ll.t )] +
S b s s
2. THEORY
-L/.t
The expression for the absorptance in a film +
B S
b o.b
.t e-as.t [Ei(o. t e ) - Ei(o. .t e-d/.t )J
s s '
with thickness d is [3]:

=i [I
where a is a phenomenological parameter accounting
for the fact that at low absorption 0. = 20. while
d at the energy gap a = 0. • Thi s deri ves freom the
A o.(x) exp o.(x') dx 'J dx (4)
fact that only one half of the defect-to-band
transitions give rise to photoconducting carriers
where the contri buti on of the ai r-fil m and fil m- (e 1ectrons) [1].
substrate ref1ectives is neglected for simplicity. Eqs. (5) and (9) are very complicated to
Substitution of (3) in (4) yields a compli- handle, but they have the advantage to hold in the
cate expression that can be solved analytically full spectral range in which PDS and pes measure-
under the simplifying hypothesis that exp(o.bx) - 1 ments are usually carried out. Thus, different ex-
for 02 x 2 d. This cond~~ion hOld~4in most prac- pressions can be extracted from (5) and (9) under
tical cases (o.b - 10° cm , d -10 cm). The so- different conditions: in particular, for tail-to-
lution of (4) 1S: tai 1 and band-to-band absorption, a(x) = 0. =0.:
this can be obtained by simply assuming .t »sd.
A:;: l-exp[a t(exp(-d/t)-1)]+ In the following table, the different expres-
s (5) si ons for A and A' are reported in the spectral

+0 .texp(-o. .t)
s s
I Ei(a .t)-Ei [0. .teXP(-d/.t)]!
s s
regions of interest.

Table I
where Ei(x) represents the exponential integral
[15J : I I

Ei (x)
'"
-f e
-T
IT dT (6)
I Surface I Tail I Defect
I abs~lption I abs~lption I absorp~~?r
I
I
-x Io.»d ,t»d,1 0.< d ,.t»d,la(x)=o.e +0. I
I S »S I S »S l.t < d,sS»S b l
Expression (5) represents the PDS absorption I bs I bs I bS I
for a fiim in which o.(x) is defined by (3). I I -o.L I I L/.t I
The PCS response can be cal cul ated in the IA'I,"\+Sbe I ~aSba(d-L) I"as [a.te- +1
same way, keeping into account the effective con- I I I I b s I
tribution of a photoexcitation to photocurrent: I I I I + 0. b(d-L)] I
I I 1_ _ _ _ 1 I
I I I I -Os.t I
IA I" 1 I... 1 -exp ( - 0. d) I ,,0. bde + I
I I I I I
I I I I +0. q l-e -d/.t) I
where 0. (x) indicates the absorption coefficient I I I I s I
for excrtati on of an electron to conducti on band
(in pes the transport of holes is commonly negli- Table Summary of the approximate solutions of
gible) and SIx) ;s a function introduced to scale (4) and (7) for different absorption
the contribution of d~fferent film regions to pho- regimes.
toconductivity.
For simplicity, we can define: As Table I shows, we are dealing with 6 dif-
ferent quantities whose knowledge allows us to
S x <L profil e the defect di stri buti on along the growth
SIx) s (8) axi s. The expressi ons for A' in the tai land in
Sb x > L the defect regimes are strictly correlated in the
sense that in both equations the term asbo.(d-L) is
This means that at a distance x = L from the present: this represents the bulk contr1bution to
surface, there is a transition from a surface con- photoconductivity.
trolled photoconduction (S ) to a bulk controlled The continuity of the function A'(~w) betw~rn
one (Sll)' Obviously, Sand Sb are proportional to the above spectral regions requires that 0. .te- L1 <
s
the effecti ve mobil i til ifet1me products, llT, on o.b(d-L). This can be accomplished by simply assum-
the surface and in the bulk. i ng .t .. Ll3. As a matter of fact, L" 3.t agrees
Solution of (7) with the conditions reported well with the viewpoint that considers the photo-
in (8) yields: conductivity as bulk controlled, where the density

129
of states is low: for x = L = 3~ the density of
states should be one order of magnitude less than lE+O
on the surface.
In the tail region a is determined by PDS. --:-
V,)

Comparison between PDS and PCS spectra allows to t:: lE-l

determine S (d-Ll. In the surface absorption re- ::s
gion, a is ~nown by means of optical measurement.
A fit of the PCS data yields the values for Ss' Sb ~
"-'
1 E-2

and L. 11.) LPCVD

The quantity a in the defect regime can be V,) lE-3
t::
di rect ly evaluated from pcs data cons i deri ng a b ~ 0
2 a . Thus, a can be ca lcul ated in order to ob- 0..
V,)
lE·4
tai~ a value ~f A equal to the experimental PDS 11.)
M
value.
:.E
p.. lE-5
3. EXPERIMENTAL RESULTS U
Freshly deposited samples, obtained by means lE-6
0 0,5 l,S 2 2,5 3 3,5
of Plasma Enhanced Chemical Vapor Deposition
(PECVD) and by means of Low Pressure Chemical Va-
por Deposition (LPCVD) were characterized in both Energy (eV)
PDS and Constant Photocurrent Method (CPM) [16]
apparatuses. PECVD films were grown by decomposi-
tion of silane at 250°C, while LPCVD samples were
obtained by thermally decomposing disilane at Fi g. 2 CPM spectra for 1 ~m thi ck PECVD and LPCVD
400°C. samples for front and rear (dashes) illumi-
CPM measurements were carried out both for nations. The graphs are scaled for clarity.
front and rear illuminations, in order to evaluate
the status of film surface (S , a , t ) and of the Table II
film-substrate interface (S . ,sa . ,s£ . )~
Fig. 1 shows the absdrpdon loefficient as
obtai ned by means of both PDS and transmi ss ion PECVD LPCVD
measurements on a 1 ~m thick PECVD sample and 1 ~m
LPCVD film.
In Fig. 2 the CPM spectra for front and rear d (~m)
illumination are presented for both samples.

--
-;'
E
lE+6~----------------------1

lE+S
.
~

~ .
s
,
(~m)

(~m)
5.5 10- 2

7 10- 2
2 10- 1

1.8 10- 2
u
"-' S/Sb 1.6 10- 3 1.5 10- 2
.....
t:: lE+4 2 10- 2 1.5 10- 2
11.) S/Sb
' (3
S lE+3 a b (cm -1 ) 1.35 4
11.)
0
u a (cm -1 ) 60 40
s
.-fr
t:: lE+2
0
.......
1 E+l Table II Summary of the principal results. The
0 ,, values of a b and a are calculated at
s
V,)
.0
1.0 eV •
< lE+O
0 0,5 l,S 2 2,S 3 3,S The concl usions that can be drawn are the
fo 11 owi ngs:
Energy (eV) - LPCVD films show a deeper surface di shomogene-
ity: this effect can be interpreted in terms of
Fig. 1 PDS and trasmission results for a 1 ~m defect annealing during growth so that the first
thick PECVD (dashes) and LPCVD samples. deposited layers show a lower density of states
than the last deposited ones;
- on the other hand, interface dishomogeneity is
The calculations have been carried on as de- much less detectable in LPCVD material: it might
scribed in the previous section. The results for be that the absence of ion bombardment at the
the two samples under study are reported in the substrate surface allows for a better control of
following table, the first deposited layers;

130
- the quantity S IS b (S./S b ) represents the ratio [3] P.D. Persans, D. Arnzen, and G. Possin, in
between surfac~ (1 ntetface) IlT and bul k IlT: it "Amorphous Silicon Technology", P.C. Taylor,
is shown that surface (interface) IlT is approxi- M.J. Thompson, P.G. LeComber, Y. Hamakawa, A.
mately of 2 or 3 orders of magnitude lower than Madan editors, Vol. 192, Materials Research
bulk IlT; Society (Pittsburgh) (1990).
- the surface density of states, as calculated by [4] L. Yang, I. Balberg, A. Catalano, and M. Ben-
means of thf7 (l Yhl ues_leported above is of the nett, in" Amorphous Sil i con Technology", P. C.
order of 10 -10 cm 16 wh_~l e the bul k densi ty Taylor, M.J. Thompson, P.G. LeComber, Y.
of states is about 10 cm for both ki nds of Hamakawa, A. Madan editors, Vol. 192, Materi-
samples. als Research Society (Pittsburgh) (1990).
[5] A.R. Moore, Appl. Phys. Lett. 40, 403 (1982).
4. CONCLUSIONS [6J A.R. Moore, J. Appl. Phys. 56, 2796 (1984).
[7] D. Ritter, K. Weiser, and E. Zeldov, J. Appl.
It has been shown that the exponential dis- Phys. 62 (1987).
tribution of density of states in a-Si:H film can [8] G. Grillo and L. De Angeli s, J. Non-Cryst.
effectively account for the PDS and PCS results. A Sol., 114 (1989).
detailed analysis of these spectra in the full en- [9] G. Amato, G. Benedetto, L. Boarino, and R.
ergy range yields information about the depth and Spagnolo, Appl. Phys. A 50, 503 (1990).
the degree of di shomogenei ty and about the IlT de- [lOJ G. Amato, L. Boari no, G. Benedetto, and R.
cay at surfaces and interfaces. Deep dishomogene- Spagnolo, in "Amorphous Silicon Technology",
ity are detectable in LPCVD specimens: this can be P.C. Taylor, M.J. Thompson, P.G. LeComber, Y.
interpreted as a resu 1t of thermal anneal i ng of Hamakawa, A. Madan editors, Vol. 192, Materi-
defects duri ng growth [13]. As a matter of fact, als Research Society (Pittsburgh) (1990).
it could be assumed that the high deposition tem- [11] H. Curtins and M. Favre, in "Amorphous Sili-
perature (400°C) enhances the thermal annealing con and Related Materi a 1s" , edi ted by H.
process on the layers previously deposited. Fritzsche, Vol. 1, World Scientific, Singa-
pore (1988).
5. ACKNOWLEDGEMENTS [12J G. Amato, G. Benedetto, F. Fizzotti, C. Man-
fredotti, and R. Spagnolo, Phys. Stat. Sol.
The author is indebted with C. Manfredott i ( a), 119 (1990).
and F. Fizzotti of Turin University for providing [13] N. Hata, E. Larson, J.Z. Liu, Y. Okada, H.R.
the samples, and with L. Boarino and R. Spagnolo Park, and S. Wagner, in "Amorphous Sil icon
of I ENGF, Turi n, for the help with PDS and CPM Technology", P.C. Taylor, M.J. Thompson, P.G.
measurements. LeComber, Y. Hamakawa, A. Madan editors, Vol.
192, Materi a 1s Research Soci ety (Pittsburgh)
REFERENCES (1990) .
[14] W.B. Jackson and N. Amer, Phys. Rev. B 25,
[1] Z.E. Smith, V. Chu, K. Shepard, S. Aljishi, 5559 (1982).
D. Slobodin, J. Kolodzey, T.L. Chu, and S. [15] M. Abramowitz, LA. Stegun (editors), "Hand-
Wagner, Appl. Phys. Lett. 50, 1521 (1987). book of mathemati ca 1 functi ons", Dover (New
[2J A. Maruyama, J.Z. Liu, J.P. Conde, and S. York) (1965).
Wagner, Mat. Res. Soc. Symp. Proc., Vol. 149, [l6J ~1. Vanecek, J. Kocka, J. Stuchlik, Z. Kozi-
Materials Research Society (1989). sek, O. Stika, and A. Triska, Solar Energy
Mater. 8, 411 (1983).

131
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

MODELLING AND CHARACTERIZATION OF PHOTOVOLTAIC

TANDEM CELLS WITH a-Si:H, a-SiC:H AND a-SiGe:H
INTRINSIC LAYERS

C.Gramaccioni, G.Grillo, F.Galluzzi

ENIRICERCHE S.p.A., 00015 Monterotondo (Roma)
G.Conte
ENEA-CRIF, 80055 Portici (Napoli)

ABSTRACT. Two-junction two-terminal p-i-n cells

based on amorphous silicon and related amorphous
alloys have been extensively studied in order to
obtain stable thin film devices. Cell design was
based on a realistic model taking into account
both optical and transport properties of the in-
trinsic layers. Experimental results are in fair
agreement with calculated data and confirm the
possibility of obtaining medium-high efficiency
with high photostability.

1. INTRODUCTION

Photodegradation of amorphous silicon 0, (nm)

represents a major drawback for an
extended use of a-Si:H solar cells.
has been shown that photostability can be
It 80 ~
greatly enhanced in p-i-n structures with 60
very thin intrinsic layers and thin
tandem cells have been proposed as an
optimal solution for obtaining high 40
efficiency and high stability.
We focused our work on two-junction
two-terminal p-i-n cells based on
amorphous silicon and its alloys with 20
carbon and germanium. The general
structure of the devices was
TCO/p-SiC/i-X/n-Si/p-SiC/i-V/n-Si/metal, 400 500
where X represents either a-Si or a-SiC 0
(Eg<2.0 eV) and V represents either a-Si
or a-SiGe (Eg>1.5 eV).
02 ( om)
0, ( nm)
2. MODELLING
90

When similar gaps are used in top and

bottom cells of a tandem structure, the
choice of intrinsic layer thicknesses is
critical. for obtaining relevant
60 ==
performance.
As an aid for correct device design,
we developed a realistic tandem cell 30
model, taking into account the optical
properties of each layer (also including
the effects of textured glass/tin oxide
substrates) and the transport properties
of intrinsic absorbers. o 80 160 240 320 400 480 560
More precisely, for the simulation of
the optical properties we used two 02 ( om )
different methods, as already reported
(1): a) an exact Rouard recursive method Figure 1
(2) based on the optical parameters of Calculated iso-efficiency contour maps as
all the layers, which gives the total a function of intrinsic layer thicknesses
absorption and reflection for each in top cell (01) and bottom cell (02).
wavelength and allows to evaluate the Data refer to single-gap a-Si:H/a-Si:H
individual absorption of each layer as tandem structures ~sing either flat TCO
well as light absorption per unit length with aluminium back-contacts (upper plot)
at different cell depths; b) a method or textured TCO with silver back-contacts
similar to that used by Schade (3) for (lower plot). Numbers on contours
in~luding the effect of light diffusion, represent efficiency values. Note the
wh1ch allows to calculate, as a funcion efficiency increase of about three points
of cell depth, the local absorption for when light confinement is present.
direct and diffused light travelling in
the two directions.

132
 The carrier generation functions for 3. EXPERIMENTAL RESULTS
the different absorbers were calculated
assuming unit generation efficiency andno Theoretical results have been compared
geminate recombination. The standard AM with experimental characteristics of
1.5 SERI data (4) were used for solar tandem cells containing a-Si:H, a-SiC:H
illumination. Optical constants n,k were and a-SiGe:H intrinsic layers of various
taken from experimental data on our thicknesses.
materials and on commercial glass/TCO As an example we show in fig.3 the
substrates. Literature data (5) for the experimen~al behaviour of short circuit
rear metal contact were used. currents versus the thicknesses of
For the collection efficiency we used a-Si:H, a-SiC:H top cells having a-Si:H
the variable minority carrier bottom absorbers. Data, referring to
approximation (6) for each intrinsic slightly textured substrates, are in fair
layer. The critical lenght Xc was agreement with calculated results.
determinated by the AM 1.5 light bias and
the local collection probability, as a
function of the distance, according to Jsc (mA/cm2 )
ref.(6). The mobility-lifetime values
10
used were in the range 10-7 - 10-9 cm 2 /V
depending on the absorber.
Eg2.1.72 .v
The reverse saturation current of the
diode shunting each elementary cell were 8
assumed to depend on the energy gap Eg
(i), while the diode quality factor was
held constant at a value of n=1.5. 6
The numerical results obtained for
each tandem cell, changing the
thicknesses of the two absorber, were 4
used to calculate isoefficiency plots by
a bicubic Spline interpolation. Typical
results are shown in figs. 1 and 2. As
2 •Egl=l.72 • Egl =1.85 •Egl= 1.95
a-Si:H/a-Si:H structures are concerned, a 02 = 270
maximum efficiency value of 11.5% in 02 = 300 02= 270
textured structures with Ag back contact
is calculated, in good agreement with the 0 50 100 150 200
results by Tawada and coworkers (7).
Dl( nm)

Dl(nm)
Figure 3
21)0 Experimental values of short circuit
Egl Eg2 currents in tandem devices as a function
of the intrinsic layer thickness in the
1.85/1.59 top cell (01). The tandem structures have
150 similar bottom absorbers (a-Si:H layers
with thickness of about 270-300 nm) and
different top absorbers: a-Si:H, a-SiC:H
1.85 I 1.72 with 1.S5 eVoptical gap, a-SiC:H with
100 1.95 eV optical gap. Maximum short
circuit currents are obtained for 01
1.72/1.59 values of about 70 nm in a-Si:H top cells
and 150 nm in a-SiC:H (Eg=1.S5) top
50 1.72 /1.72 cells.

In order to optimize photovoltaic

performance much work has been devoted to
o 100 200 300 400 500 600 the design of n-Si:H/p-SiC:H tunnel
D2(nm) junctions, connecting top and bottom
cells. Here doping levels and thicknesses
of recombining layers appear to play a
particularly critical role and best
results were obtained using heavely doped
Figure 2 layers with total thickness smaller than
Calculated iso-photocurrent curves as a 10 nm.
function of intrinsic layer thicknesses Typical current-voltage curves under
in top cell (01) and bottom cell (02). AM 1.5 light of thin amorphous tandem
Data refer to devices using flat TCO with devices, with matched photocurrents, are
aluminium back-contacts and containing presented in fig. 4 for a-SiC:H/a-Si:H
different intrinsic materials. As and a-Si:H/a-SiGe:H configuration
indicated by optical gap values, the respectively. The highest values of
different curves represent the following short-circuit currents (S-9 mA/cm2
configurations: a-Si:H/a-Si:H (lower), without optical confinement effects) are
a-Si:H/a-SiGe:H, a-SiC:H/a-Si:H, a-SiC:H/ observed in a-Si:H/a-SiGe:H structures,
a-SiGe:H. Numbers in parentheses refer to while the highest open circuit voltage
maximum efficiency values achievable by (1.6-1.7 Volt) are shown by
the various configurations. a-SiC:H/a-Si:H devices.

133
 1,2 "/"0
~sc =7.2 mAfcm 2
1.2

"'----.----- ..... -- --------- -- ..

.8
Voc= 1.6
FF = .62
=7.1
Volt

%
.4 1.0 ::--.-.-.--- ..... ,
...

.4
.8
a-SiC:H! a-Si:H
.6 • Single

~ Tandem
.4~i~/----~L--L ______-L__~I______~__
1.2 o 5 10 50 100 500
Jsc =8.2 mAfcm 2

Voc =1.4 Volt

.8 Time(hr)

FF = .61 .4 Figure 5
~ = 7.1
Relative efficiency versus continuous AM
% 1.5 light exposure time for a
single-junction a-Si:H cell and a
.4 two-junction a-Si:H/a-Si:H cell. After
500 hr the degradation of the latter
a-SI :H/ a-SiGe:H device is lower than 5%.

REFERENCES
(1) G.Conte, D.Della Sala, F.Galluzzi,
Figure 4 G.Grillo, C.Reita, Proceed. 8th E.C.
Current-voltage characteristics of tandem Photovoltaic Solar Energy Conf., Kluwer
cells under AM 1.5 illumination. Data Academic Publ., 1988, p.781.
refer to a-SiC:H (Eg=1.85 eV) /a-Si:H and (2) O.S.Heavens, 'Optical properties of
a-Si:H/a-SiGe:H (Eg=1.59 eV) con- thin solid films', Academic Press, 1955.
figurations, using slightly textured TCO (3) H.Schade, Z.E.Smith, J.Appl.Phys. 57,
substrates and aluminium back-contacts. 568 (1985).
Note the similar efficiency values. (4) R.L.Hulstrom, SERI Report
n.SP-215-2784 (1985).
(5) "Handbook of optical constants of
solids', E.D.Palik ed., Academic Press,
Degradation of photovoltaic 1985, p.369.
performances under continuous AM 1.5 (6) S.Nonomura et al.,Jap.J.Appl.Phys.
illumination has been tested for 1000 21,279 (1982).
hours on both tandem and single-junction (7) Y.Tawada et al., J.Non-Cryst.Solids,
devices. The results are shown in fig. 5, 115, 24 (1989).
where the behaviour of fill factor and
conversion efficiency, versus exposure
time, is reported. While single junction
amorphous cells suffer a relative
efficiency decrease of about 25-30 %,
tandem structures exhibit a particularly
good stability, with relative losses
lower than 5 %.

4. CONCLUSIONS
A realistic model for two-junction
solar cells has been developed, allowing
the design of optimized devices on both
flat and textured substrates.
Confirming the theoretical analysis,
the experimental results show that thin
tandem devices, based on intrinsic
materials with optical gaps in the range
1.6 1.9 eV, are able to give
medium-high initial efficiencies with
relative photodegradation lower than 5%
after prolonged AM 1.5 light exposure.

134
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A photoemission study of gdlLc-Si

G. Willeke\ K. Prasad 2 , R. Fischer\ A. Sha.h2 and E. Bucherl

1 University of Konstanz, Faculty of Physics, P.O. Box 5560

D-7750 Konstanz, Tel. (07531) 88-2079, Fax (07531) 88-3895

2 Institute of Microtechnology, University of Neuchatel

CH-2000 Neuchatel, Tel (0041) 38205459, Fax (0041) 38254276

Abstract. Total yield photoemission spectroscopy between 4 and 6.5eY has been used to
study p- and n-doped gdp.c-Si thin films prepared at 70MHz. The results show that the yield
increases from the most conductive n-tyre (u=100(Ocm)-1) to the most conductive p-type
specimen under investigation (u=l(Ocm)- ), consistent with a surface depletion layer in both
types of films. The experimental results are also discussed in terms of a two-phase model.

1. Introduction 2. Experimental details

In recent years there has been a renewed interest in The gdp.c-Si samples were prepared in a capacitively
glow discharge prepared microcrystalline silicon thin films coupled YHF glow discharge system at 70MHz [8J.
(gdp.c-Si). This materi¥, possessing crystallite sizes in the Depositions were carried out at T.=200-300· C and a
range between 20-200A and crystalline volume fractions dilution of silane in hydrogen of betwe~n 2-4%, thereby
between 20-90%, provides interesting physical insight into obtaining growth rates of about 0.6Asec- l . The film
the effect of stmctural ordering on the electronic properties thickness was typically 0.4-0.6p.m. The specimens
just beyond the amorphous phase. On the applied side, the investigated had room temperature conductivities of
incorporation of a highly conductive p.c-Si contact layer for between 6-100(Ocm)-1 and l(Ocm)-l for the n- and p-type
example in a-Si solar cells has proved very successful. samples respectively. Other parameters can be taken from
Recently the use of gdp.c-Si as a stable photoactive [9,10].
material has been proposed [1]. The first series of experiments has been carried out on
So far very limited information on the density-of-state specimens which have been lying in air for at least a few
distribution (DOS) in gdp.c-Si has been obtained. For months, such that the surface was covered by a thin native
example field effect studies on compensated material [2] oxide. In order to evaluate the possible influence of surface
suggest an essentially flat DOS between 0.4 and 0.geY states on the yield spectra, the samples received an HF -dip
below E e , with geE) in the range 2-4·10 l8cm- 3 ey-l. In this (2%, 20sec) and were then loaded into the UHY apparatus
study it was concluded that most of these states arise from within 5mins.
the boundaries between crystallites. In a seHconsistent The total yield photoemission experiments were carried
analysis of transport data of n-doped gdp.c-Si specimens [3] out in a standard set-up which has been described in detail
it was shown that the conduction band tail states become elsewhere [9,11].
delocalised in the p.c-phase due to the introduced long
range order. 3. Results and discussion
In the present work, total yield photoemission
spectroscopy (PES) has been used to gain some insight into In a previous publication [9] we have shown that the
the electronic changes likely to occur at the valence band photoemission yield of gdp.c-Si specimens strongly depends
edge of gdp.c-Si. PES has proved extremely useful in the on doping and is highest for the most strongly B-doped
study of surface DOS in cr-Si [4] as well as in the sample and lowest for the most heavily P-doped specimen.
determination of the complete DOS in a-Si:H and related Fig.1 reproduces some of the results. A very similar
materials [5-7}. doping dependence has been observed in work on single

Y[a.u)
10- 6 Fig.1: Total yield
photoemission spectra
for a series of boron
(0: u=l(Ocm)-l) and
phosphorus-doped gd-
p.c-Si specimens (.:
u=59(Ocm)-1; ~: U=
100(Ocm)-1).

-
4·0 5·0 hv leV) 6·0 1·0

135
 - - ---
ETp
s ETn
V
~'-- ----------- ,-
Fig.2: Surface
-- ;~ depletion layer for-

V
i - ----------- mation on n- and p-
\..... type semiconductors
'-- due to trapping of
~~ mobile charge carriers
at surface states.
Ev
f--"-I I-- ,,- Symbols are
plained in the text.
ex-

)..

crystalline Si cleaved in UHV [4,12]' and this effect has (100) FZ, 0.20cm). The experimental results for the cr-Si
been explained by surface band bending. specimen show that the yield can be expressed .by
The situation is illustrated in fig.2, which shows a
surface depletion layer in both n- and p-type semi- (1)
conductors due to trapping of mobile charge carriers at
surface states. In the former case this leads to an increase over a relatively large energy range (approximately
in the effective photoemission threshold energy ETn as 1.6eV) using E T =4.7eV and n=3 (solid line in fig.3). This
compared to its surface value E T •. In p-type specimens a experimental finding is consistent with previous work on cr-
smaller value ETJ> is expected. In fig.2 Au represents the Si cleaved in UHV [12J. In that study, the cube law
width of the surface space cha.rge layer which is of the dependence was found to be in agreement, within
order of the Debye-length Ln. In our heavily doped samples experimental error, with the 5/2 power law predicted for
investigated, the ca.rrier concentration for both types of an indirect excitation process. In a more recent study on
:films is expected to be in the 1019_102ucm-3 range with similarly prepared cr-Si specimens [4J a 3/2 power law was
corresponding values for Ln::;lOA. Since the photoelectron found to provide the best fit to the high energy part of the
escape depth A can be taken to be about 30A [12-14J, it yield curves.
can be concluded at this stage that emissions from strongly Turning to the strongly p-doped gdl'c-Si specimen, fig.3
p-doped samples should represent rather bulk-like shows that unlike the cr-Si sample most of the spectrum
properties, possibly including surface states. cannot be fitted to a single power law. Using eqn.1 for the
In the n-type case the surface barrier cannot be high energy part (i.e. 5.5-6.5eV) results in ~=4.geV and
neglected and will have a strong influence on the low n=3. In this single phase model the excess yield below
energy part of the yield spectra. The low energy limit of about 5eV could be ascribed to surface states or valence
the yield curves in fig.1 suggests a common work function band tail states in the I'c-phase, the latter appears to be
if! of about 4.5eV for the samples studied. This indicates a less likely. In addition an alternative explanation is
surface Fermi level pinning independent of doping similar possible. Since gdl'c-Si in general is composed of
to the crystalline silicon case [4J. microcrystallites embedded in an amorphous matrix, a two-
In fig.3 the yield spectrum 'of the p-doped sample of fig.1 phase model should be considered. In this model (fig.4)
is compared with a single crystalline silicon sample (p-type contributions to the photoemission yield arise from an

10-4
Y[a.u.J
Fig.3: Yield spectra of
the p-type gdl'c-Si
specimen of fig.1 (0)
compared with a
single cr-Si sample (.)
(p-type (100) FZ,
0.20cm). The solid
and dashed curves
denote fits using
eqn.1.

4·0 5-0 6-0 1·0

hv [eVI

136
 EVQC

Fig.4: Schematic re-

presentation of a sim-
ple two-phase model
I----r-------~--------~~ of gdJLc-Si indicating
- --c~,~ - --- crysta.lline-like regions
(1.1eV) and amor-
1-1eV 1·15eV phous-like phase (1.75
eV). .d denotes a
________ .c~,Q __ boron related band-
- gap narrowing in the
~ - amorphous-like phase.
-

amorphous-like region of bandgap 1. 75eV and a crysta.lline- The situation is more clearly shown in fig.6 for the most
like region of bandgap 1. leV. In figA surface band bending strongly n-doped specimen (oo=100(flcm)-1). Remarkable is
effects have been neglected and a constant vacuum level the almost exponential decay of the yield between 5-6eV.
EVa< assumed for simplicity, which need not necessarily be This feature has been discussed in terms of a broadened
the case. Using this model, the yield curve of the p-type Urbach tail at the valence band edge of phosphorus doped
gdJLc-Si sample of fig.3 could be explained as follows. A low gdJLc-Si [9]. Although this is highly unlikely, some donor
energy fit of eqn.l yields E T =4.55eV and n=3 and these related deep defect creation mechanism at grain boundaries
emissions would be ascribed to the crysta.lline phase, or in the amorphous matrix· cannot be completely ruled
whereas the high energy cube power law fit with E T =4.geV out.
would be attributed to the amorphous-like phase. The A more likely explanation is that the depletion layer
implicitly assumed significant contribution from band bending (see fig.2) leads to a smearing out of the low
amorphous-like regions is in agreement with the results of energy part of the yield spectrum. An increase of the
Raman studies on these samples, which indicate a crystal- valence band edge within the electron escape depth A of
line volume fraction of about 60%, which is well above the O.5-0.7eV appears possible, if EF at the surface is pinned in
percolation threshold for electrical conduction. the middle of the bandgap. In this single phase model the
Fig.5 summarises yield data for a series of strongly n- high energy fit of eqn.l resulting in E T =5.1eV and n=3
doped gdJLc-Si specimens before (closed symbols) and after (see fig. 6) would then represent the bulk valence band
an oxide removing etching treatment (open symbols). The edge.
yield has been normalised somewhat arbitrarily at 6.2eV In the two-phase model (figA) the low energy fit
and shifted in the ordinate for the two sets of E T =4.5eV, n=3 can be regarded as contributions from the
measurements, in order to clearly bring out differences in crysta.lline phase, whereas the above high energy fit
the shape of the spectra. As expected the etching represents contributions from the amorphous matrix. The
procedure influences the shape of Y only in the low energy difference in threshold energies of O.6eV leads to the
part, thereby revealing the influence of surface states on conclusion that most of the band offset between the
the yield curves. The effect increases with doping, crysta.lline and amorphous phases appears at the valence
suggesting doping related surface states at the surface. band edge, assuming sinlilar electron affinities for both

Fig.5: Yield curves

for a series of n-doped
YIQ.u.] gdJLc-Si specimens (.:
oo=lOO(flcm)'l; ..t: 00=
10-2 59(flcm)'l; .: 00= 6.1
(flcm),l). Open sym-
bols denote the same
specimens after a 20
sec 2% HF etching
procedure. Results
are normalised at
6.2eV and have been
shifted in Y for cla-
rity.

4·0 50 hv leV]
6·0 1-0

137
 Y[au] Fig.6: Normalised
surface states yield spectra for a
10-2 heavily n-doped gd/Lc-
Si specimen before (.)
and after an HF-dip
( 0 ). The solid lines
denote cube power
law fits (see text).

4·0 s·o hv leV]

6·0 1·0

phases. For the heavily p-doped specimen (see fig.3) this [8] H. Curtins, Proc. MRS, vol. 95 (1987) 249
difference is reduced to 0.35eV, which could indicate a
doping related bandgap narrowing of Ll=0.25eV (see fig.4) [9] G. Willeke, R. Fischer, E. Bucher, K. "Prasad, H.
in the amorphous phase. This effect has been observed Keppner, F. Finger and A. Shah, Proc.
before in the study of a.-Si:H. POL YSE'90, Schwli.bisch Hall, 1990, to be publ.
4. Conclusions and future work [10] K. Prasad, F. Finger, H. Curtins, A. Shah and J.
Baumann, Proc. MRS, vol. 164 (1989) 27
The interpretation of total yield photoemission data for
gd/Lc-Si is much less straightforward than in the case of a- [11] H. Berner. G. Kragler. J. Meier, M. Simon, H.
Si or cr-Si, due to the inhomogeneous nature of the Curtins and E. Bucher, J. Non-Crys. Sol. 114
material under investigation. The experimental results have (1989) 678
been tentatively discussed in terms of a single phase model
including band bending effects and a two phase model. [12] G.W. Gobeli and F.G. Allen, Phys. Rev. 127
Both models have led so far to a rather qualitative inter- (1962) 141
pretation of the yield data. In order to evaluate the relative
importance of these two models and extract quantitative [13] S. Aljishi, S. Jin, M. Stutzmann and L. Ley, Proc.
data on the DOS-distribution at the valence band edge of MRS, vol. 164 (1989) 51
gd/Lc-Si, detailed model calculations, etching studies on p-
type material as well as yield measurements on completely [14] J.M. Ballantyne, Phys.Rev. B6 (1972) 1436
crystallised specimens will be carried out [15].
[15] R. Fischer, G. Willeke and E. Bucher, to be publ.
Acknowledgements
The financial support by the German Ministry of
Research and Technology (BMFT) (project nr. 0328453-B)
and the Swiss Federal Research Grant EF-REN(90)045 is
gratefully acknowledged.
References

[1] C. Wang and G. Lucovsky, Proc. 21st IEEE

PYSC, Kissimmee, May 1990, 1614

[2] P.G. LeComber, G. Willeke and W.E. Spear, J.

Non-Crys. Sol. 59/60 (1983) 795

[3] G. Willeke, W.E. Spear, D.1. Jones and P.G.

LeComber, Phil. Mag. B46 (1982) 177

[4] C. Sebenne, D. Bolmont, G. Guichar and M.

Balkanski, Phys. Rev. B12 (1975) 3280

[5] S. Griep and 1. Ley, J. Non-Crys. Sol. 59/60

(1983) 253

[6] K. Winer, I. Hirabayashi and L. Ley, Phys. Rev.

Lett. 60 (1988) 2697

[7] H. Berner, Ph.D. thesis, University of Konstanz

(1987), Hartung-Gorre, Konstanz

138
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

MODELLING OF AMORPHOUS TIDN FIlM SOLAR CElLS

M. Block, D. Bonnet, F. Zetzsche

Battelle Institut e.V., D-6000 Frankfurt 90

ABSTRACf_ The development of a new computer-modelling procedure for amorphous silicon solar cells based on iterative integration
of the basic physical equations by means of a fourth-order Runge-Kutta procedure has been continued. It is now possible to model in
near-real-time structures consisting of up to nine individual layers having individual constant or graded material-parameters. This paper
illustrates the power of the model by comparative modelling of a few new solar cell structures of especially interesting composition. It
is shown, that the model allows new insights not achievable by other means.

1 Introduction - the ModeUing-Concept 2 Solar Cell Engineering - Representative &amples

The severe problems encountered in developing solar
cells based on amorphous silicon and its alloys, require theoreti- 2.1 Description of the Cells Studied
cal modelling tools, which can be closely adapted to the process The program in its present state allows to model cells
development by accepting realistic input data and giving besides made of up to 9 individual layers, each having nearly arbitral)'
I-U curves detailed space dependencies of critical parameters. properties, such as energy gap (which can be graded), charge
A comprehensive computer simulation based on a new carrier transport data, defect structure (traps and recombination
numerical approach has successfully been initiated 11/ and has centers) in the dark and under arbitral)' illumination conditions.
furnished significant results. Two layers are the p + and n + contact regions, which are as-
The model is based on the set of the well-known conti- sumed to be 10 nm thick in all cases presented here. This leaves
nuity- and transport-equations as well as Poisson's equation, stacks of up to 7 layers for the i-region.
using standard terminology, taking into account gradients in In order to demonstrate the power of the model, we will
electron affinity and energy gap: in the following give results of cell structures, which have been
"engineered" for didactic purposes, starting with a "re~llar"
!!£
dt
G(x) • R(x) -
1 d'
q a? = 0
1.74 eV cell, which will be stepwise modified. All cases are given
with i-layer defect concentrations representative for "New" and
dn 1 ~ significantly llegraded cells: 2·10'5 cm-3 and 1·iCi17 cm-3,
dt
= G(x) - R(x) +
q dx =0 respectively. Table 1 gives an overview of the diff~rent cell types
with characteristic results .
qpJ.Lp(x)E
.Qp d
jp = kT J.Lp(x) dx - PTx [x(x) + Eg(x)]
2.1.1 p-i-n Cell
jn qnJ.Ln(x)E + kT J.Ln(x) ~~ - nJ.Ln(x) ~x x(x) This cell-type has been chosen as reference standard,
being a p-i-n cell having an energy gap representative for opti-
dE g mum solar energy conversion in a Single cell. Two types of defect
dx = E (p - n + ND - NA + Nt) density are chosen for the i-layer: 2'10'5 and 1017 cm-3 , where-
as the p and n-Iayers (effective doping: 10'7 cm-3) are assumed
jn(X) + jp(X) = const. to have the "standard" defect density of 10 17 cm-3 • This means,
The resulting set of two nonlinear coupled differential we compare "new" celis of good defect quality with significantly
equations of second order for nand p is solved by iterative degraded cells. We have chosen two cells with constant energy
numerical integration using a Runge-Kutta procedure of fourth gaps of 1.74 eV, ill view of the classical a:-Si-cell, and 1.5 eV, in
order incorporating several accelerating, optimization and learn- view of the emerging potential of a;-Si/Ge alloys.
ing steps. The modelling program fully takes into account the
effective forces due to graded material characteristics by intro- 2.12 Heteroface Cell
ducing the related additional terms into the transport equations. This cell has a narrow region gap-narrowing extending
The results are given as position-dependent values of only over a region of 0.1 /Lm, followed by the "standard" i-layer
all interesting quantities, e.g.: charge carrier densities, band of 1.5 eV in gap.
bending for valence and conduction band as well, electric field,
current densities for both kinds of carriers, charge carrier gen- 2.1.3 Graded Gap Cell
eration and recombination. All results can be computed for all Graded gap cells have been in discussion for mixed
points of the current voltage curves in the dark and under compound materials, such as CdSe/CdTe and GaAs/GaP for
arbitral)' illumination. over 20 years and recently also for amorphous celis. For didactic
The program has been developed and implemented on purposes it is assumed that materials with energy gaps between
a Sun-Sparc 1 + station with a user-friendly graphic interface. 2.0 an.d 1.5 eV can be produced with the same quality achievable
The machine's rise-processor allows to calculate all the required in amorphous silicon. For defect concentrations we also have
data for an individual diode-structure within 30 to 60 minutes. chosen the "new" and the degraded cases, i.e. 2.10 15 and 1017
cm· a. Two types of graded gap structures have been employed,
differing by the doping sequence. In the first case the high-gap

139
Table 1: Overview of Cells Studied with ReICYaDt Data

"
Eg Uoc Jsc FF DEGRADA-
CELL TYPE (eV) (V) (mA/cml) (%) TlON
BY

A: HOMOJUNCTION AN 1.74 0.93 20.2 0.59 11,5

pin-aSi AD 0.83 17.9 0.36 5.6 51%

B: HOMOJUNCTION BN 1.5 0.70 27.6 0.57 11.7

pin-aSi/Ge BD 0.62 28.2 0.43 8.0 32%
I
C: HEIEROFACE-CELL CN 2.0 -15 0.62 18.4 0.52 6.2
pin CD 0.57 23.6 0.35 4.9 22%
~

I
D: GRADED GAP-CELL DN 2.0 - 1.5 0.82 21.6 0.66 12.3
pin DD 1.01 14.4 0.54 8.3 32%

E: GRADED GAP-CELL l EN 2.0 - 1.5 0.68 17.3 0.63 7.9

inverted doping: nip ED 0.57 7.4 0.46 2.1 73%

end is p-doped, in the second case the high-gap end is doped n- cases studied. So help is given to the "wrong-party".
type.The illumination occurs in both cases through the wide-gap
side. 3.2 Surprising Results

3.2.1 Increase of Uoc due to Degradation (Cell D)

3 Discussion of Results The open-circuit-voltage of cells with homogeneous
composition is decisively governed by the material's band gap. In
3.1 Overview eqUilibrium the Fermi level is close to the valence band in the p-
Figures 1 to 5 give the 1-U curves of all 5 types of cells, layer and close to the conduction band in the n-layer. As the
the Cllrves for the new and degraded cells having been plotted Fermi-level is flat, we find a band-bending, that is close to, but
into the same diagram. Figures 6 to 10 give the corresponding always smaIler than the energy gap. Under illumination at open-
generation and recombination rates across the volume of the circuit the generated charge carriers tend to flatten the bands,
cells (or the new and degraded cell. the amount of back-bending being equal to the U oc '
The "standard" a-Si cell (type A Figs. 1 and 6) shows In cell D we frod that the total amount of equilibrium
the expected efficiency of around 11% under the AM 1.5 global band bending for the new and degraded case are similar. Howev-
spectrum employed. Due to the increased light absorption, the er the open-circuit band bendings differ strongly. For the "new"
1.5 eV-ceIl (type B, Figs. 2 and 8) shows a higher short-circuit cell the valence band tends to be rather flat, while the conduc-
current, but smaller open-cireuit-voItage which lead only to a tion band shows a homogeneous descent due to the inhomoge-
somewhat increased efficiency of 11.7% compared to 11.5% of neous material composition, as reasonably expected (Fig. 11). So
pure amorphous silicon. As a consequence of degradation, the the open-circuit-voltage is governed be the smaIl energy gap of
efficiency of a-Si decreases by 50%, whereas that of the 1.5 eV the n-side, because the conduction band edge in the n-layer
cell only decreases by 32%. Charging of defects concentrates the cannot fall below the valence band edge in the p-layer, if we go
field in the region just below the p-region, where the carrier to equilibrium. In the degraded ceIl, the additional charged
generation in .this material is higher than in the a-Si cell, pref- defects lead to an unexpected transfer of the band gradient
erentially helping the electrons, which are the "workhorses" of (Fig. 12): The conduction band tends to be flat, forcing the va-
the amorphous celIs, due to their significantly higher mobilities. lence band edge in the n-Iayer to be higher than in the p-Iayer!
In the heteroface cell (type C, Fig. 3 and 8) the region IT we now go to equilibrium, the n-layer conduction band edge
of high generation rate and that of high field do not coincide, starts falling from a higher position than before degradation,
so the current is low and the recombination rate is significant, what means that the (potential for) open circuit VOltage has
resulting in low efficiency. Upon degradation this efficiency only increased. Of course, the strange form of the bands leads to
decreases by 22%. This is the lowest decrease in efficiency of all poor currents under load. Interestingly the maximum power
cells studied, confirming the experience that cells of low effi- point is reached at the same VOltage as before degradation.
ciency are rather stable.
The graded gap cell having a gradient in gap from the 3.2.2 Increase of Jac: due to Degradation (Cell B)
illuminated p-side towards the n-side from 2.0 to 1.S eV gives In the case of the low gap homojunction (B) the short-
the highest efficiency of all cells studied together with a degra- circuit current increases upon degradation from 27.8 to 28.6
dation comparable to that of the 1.5 eV-cell. The inverted grad- mA/cmz. An inspection of the field distributions of the cell be-
ed gap cell, however, in which the gradient runs from the illumi- fore and after degradation under short-circuit conditions (Fig.
nated n-side to the p-side of the cell shows the second worst 13) shows that the defects deform the previously homogeneous
efficiency. This is understandable, as in the second case the field such that at the p-i- and ion-interfaces the field is high, and
modified field distribution increases the field felt by the holes in the inner part of the i-layer the field is lowered. This has two
on account of that felt by the electrons, essentially reducing the effects: The lower field l'CgiOn impedes carrier flow, reducing the
field for the more important charge carrier, i.e. the electron, short-circuit-current, but the high field at the poi-interface is
which carry the current over the largest part of the cell in all optimally located to separate the high density of charge carriers

140
 •. ~nr---------~----------,----------,----------~
A
A

....
-I' .."., I .~·n

.............

6
-20
e.. OCIO. i! I
".., ............
'

----~
....-."'-...- ..--
. --..;-;..;;;;;.,;;;;;;;.;.=~~.=.;;=..~~
... ...
• . ...,... rch:",
···_· ..
-JO 0 . 00 0.1.0 0.20 0 , 30 0 . 100
~o , a ' .2 ' .1 ..0 I.' U pOal tian (,,1c,.0 tur)

B B

./
-10

a
./

,i"
-20

-lO
-0.' ... ' .2 ... ' .0 ' .0 I.' 1. 2
0 . 000-00
0 .00
"
A ....~ --' - - -
0 . 10
.- .-
o.~
pouuon r.1crolleter)
o.:tC ....
c .. ...,... C

~
1.000+22

-10

..'

E3
\
-20 ............ ""

->0
-0 . ... ... ... ' .1 '.1 I.' I.'
o._v--<
0." ~IO ....
pioahion (.. lcroll.eter)
• ... O. cO

o 1. 2tiO+n o
..•../
......... J . Ooo.Z2
.....
-10

-
-I~~••~--~.~••~---.~.~.----~••.•~---.~.~.--~~.~. •~==~,.~.====:J,
ES .. 0.000.00
....
\
"

0 . 10 O.aoo
............
1\.. . . . . .:..."" '_. _. _._"-"- --"- '--'-"- '::" '- - -J1
0,30
:.~

0 . 40
go.! tion (1I1CrQIIU!lhr-J

E I........ E

~
-10
.... _....... _...

EJ
~ ,OOO+:z s .............. .

..... ./

... .. ..
...........
O,ooo·O:.~"~:::"'-----:'.":'::-'-~===:'~."::=====';.~lO===----,JO... o
->0
-0 •• ' .0 ' .1 I.' I.' poaHion IlItiCI"Oftlll!llhr)

Ftgures 1 to 5 Figures 6 to 10
I-U-cum:s for the S different cells bcfwe and after (dotted) Generation rates (dashed) and recombination rates for the new
degradation (d. table 1)_ Cum:nt in mA/cmz, Voltage in V_ (solid lines) and degraded (dotted lines) cclJs, corresponding to
Inserts: Position dependent enexgy gap. the cells of Figs. 1 to 5

141
 Current densities
' •• r-------~--------~----------------~ ' .00 BO
1III:1IIctl"on" •. nolu _ toni

ON
r--------- l' .....
• -5 . t/o
I.·····
i.O!) ~

-----~ .
.! -100 . 00

1.00 ..
~

~
- s5.00

"
~
~
-20. 00

u
'- -~_oo
'-
'.00 r---------------------------:j ~
U
-lO.OO

00 . 00 O. H> o.n 0 .30 -o.4C ' .00 O. sO 0.20 0 .30 0 .•0

pes! Ion h.lc:rclAlleter )

Figure 11: EoeIgy-band-profile for undegraded graded- Ftgure 15: Cuae.nt-distribution between electrons and
gap cell (open ci:J:cuit case) holes in degraded type-B cell

'.00 r-------~-------~--------"T"""-------,
DO generated just at this place. Consequently the significantly more
mobile electrons take-over the biggest part of the total (con-
'.00 r-- stant) current earlier than before degradation, overcompensating

------------------------- the negative influence of reduced field in the interior, where

recombination is reduced by lack of holes (cf. Figs. 14, 15).
1.00
3.2.3 Low Efficiency of the Inverted Graded Gap Cell
This eell was designed after we had learned to under-
'.0' _________------------------- stand the behavior of the graded gap cell. It was intended to give
the holes better conditions to move across the i-layer, because
0 .00 0. 10 ' .>0 O.:JO
they are already handicapped by their poor mobility. This aim is
met in the inverted graded gap cell. The gradient of the energy
gap is fully transferred to the valence band, while the conduction
Figure 12: EoeIgy-band-profile for degraded graded-gap band is flat. So we help the holes on account of the electrons.
cell (open ci:J:cuit case) This is not effective, though, as the holes anyway carry only the
smaller average part of the current, if the whole bulk of the i-
region is considered. It would be much better to "help" preferen-
tially the electrons. This is more effectively done in the non-
.. 6.'
B
inverted graded gap cell (type D) .
~ After degradation the field for the electrons is virtually
~ .-. zero for both, short- and open-circuit conditions, resulting in the
' .' .I
".
strongest degradation of all cells studied.
"u
y

'-
... .......
....
4 Conclusions
.......... ..........
~
y
.~.
!
u
The most important issue in defining new cell structures
and modifications, is to care that the electrons take-over as high
..,..-____--:.~.,..,..--~---:-
o.~.~ • .':::
..,----~--:.~.,..,-------1
0 •• 0 a proportion of the photogenerated current as early as possible
potH ion [JillcroJllet.,-l
by modifying absorption and the field acting on electrons. Both
should coincide at the front part of the cell, as is the case in a
Ftgure 13: Field-profile for undegraded and degraded pin cell with light incident from the p-side. A physical limit to
(dotted) type-B cell (short cin:uit case) this will be backdiffusion of electrons towards the p-surface and
recombination via defect states in the p-region or at the surface.
Graded gap structures help to achieve a (limited) uncou-
Current densities pling of ficlds acting on electrons and holes, allowing to "con-
electronl .. holu _ t otal
struct" new mOIe efficient systems, at least on paper, helping
0 , 00 BN .............. developers in their exhausting practical work. Our modelling
1" ......... , ............... .
~ - " . 00 program will allow to optimize such conditions.
1. ..............
.....
- 10.00

-U . O<I REFERENCE

"
..~

-~.OO
/1/ Block, M., Bonnet, D., Langbein, D.: Theoretical Modelling
of the Amorphous Silicon Thin Film Solar Cell. Proc. 9th Photo-
volt. Solar Energy Conf. 1989, p. 30 - 32
~ -25 .00

.1
-JO. oo
0.00 •. ~ . ... ' .>0 • .~
posH ~on (micromett r ]

Figure 14: Cum:nt-mstribution between electrons and NOTE: The German Federal Ministry for Research and
holes in undegraded type-B cell Technology intends to support this work.

142
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

COz LASER CVD O~ AMORPHOUS HYDROGENATED SILICON:

GAS PHASE PROCESSES AND PROPERTIES OF THIN FILMS

E. Golusda, K.-D. LUhmann, G. Mollekopf, M. Wacker, and H. Stafast+)

Battelle Europe, Am R~merhof 35, D-6000 Frankfurt/M 90, FRG

ABSTRACT. During deposition of amorphous hydrogenated silicon (a-Si:H) by

continuous wave (cw) COz laser CVD in situ gas analysis was performed by
mass spectrometry. The gas sample passed through an orifice in the
substrate surface (T. - 400 OC) and subsequently via a molecular beam
into the analyser. Applying SiH4 and SizHo as precursors, polysilanes up
to tetrasilane and pentasilane were observed, respectively. Time resolved
mass analysis revealed a start-up period of several minutes becoming
shorter with increasing laser power. In case of SizHo special starting
conditions were necessary to suppress powder formation. Using a variable
shutter in front of the substrate, the average film growth rate was found
to decrease with the duration of laser deposition whereas the photo
conductivity of the a-Si:H films remained constant. Possible implications
of the start-up processes to the quality of solar cells are briefly
discussed.

1. INTRODUCTION CVD-Reactor Differential Pumping Unit

The energy conversion efficiency of mono junc-

i
tion photovoltaic solar cells on the basis of I
amorphous hydrogenated silicon (a-Si:H) was raised SiR., I
within one decade to 12 % in 1986 /1/. Since then, SiZK6
I QMS I
however, no further progress with respect to maxi-
mum conversion efficiencies of laboratory cells has Substrate I I
been reported. Following model calculations /2/, Heater
Molecular
Beam L,_,.J
the largest losses are assumed to occur at the two
interfaces between the transparent conductive elec- COZ ~ r= ~
------Ly---
I I

trode (TCO) and the p-doped layer as well as the p- Laser ~L --6
doped and the intrinsic layers. As a possible route
to improve a-Si:H solar cells, a hybrid process is ~V8S'I'1U.TE

proposed by which the bulk part of i- and n-layers

are deposited by conventional plasma CVD whereas
the p-doped layer and the critical interfaces are
3 mbar 10- 5 mbor 10- 7 mbc"
produced by CO 2 laser CVD. COz laser CVD is suppo-
sed to provide high quality interfaces by a "soft"
deposition following the energetically lowest pos-
sible decomposition pathway of the silane precursor Figure 1: Experimental set-up for cw CO 2 laser CVD
molecules /3/. of a-Si:H with simultaneous QMS gas analysis
So far, the most advanced work on COz laser
CVD for solar cell materials was essentially per- laser beam (Spectra Physics, model 971 with addi-
formed by two groups finally leading to the produc- tional power stabilization, 10.6~) of 10 mm
tion of a Schottky barrier /4/ and mono junction diameter and typically 160 W ± 1 % power (Ophir
pin and nip solar cells /5/. Based on the published 1000 W), corresponding to 200 W/cm z , entered and
results of these two groups and other related left the deposition chamber by KCl windows flushed
publications in this field /6/, a CO 2 laser deposi- with inert gas (Ar). The beam axis was placed 8 mm
tion facility and an apparatus for in situ gas in front of the vertically positioned substrate
analysis have been built up. Current experimental (Corning glass 7059, 1" x 1" x 0.016"). The proces-
work is focusing on selected problems inherent in sing gas entered by a nozzle mounted about 5 cm in
the deposition of layers and especially interfaces front of the substrate opposite to the outlet to
of high quality. In the following, emphasis is the pumping system. During the CVD process the gas
placed on the initial periods of gas phase and film pressure was controlled by a capacitance manometer
deposition processes and their possible implica- (MKS Baratron, type 122 AAX/01000 DBS and 122
tions in solar cell performance. Controlling the AAX/00010DBS) stabilized to 5.0 mbar using a
start-up processes, which imply non-ideal deposi- throttle valve (MKS, 235 A) between the deposition
tion conditions, might help to improve film and chamber and the rotary pump (Leybold, Trivac 40
solar cell quality. BCS). The gas fluxes were stabilized by mass flow
controllers (MKS 1259B). The substrate holder was
kept at a temperature of 400 ± SOC.
2. EXPERIMENTAL PART The large depOSition chamber has an inner
Two reactor systems were applied, a small one diameter of 25 cm and an optical path length of
for CO 2 laser CVD with synchronous gas phase ana- 25 cm. It provides a variable shutter in front of
lysis (Fig. 1) and a large one with a load lock the substrate. The total gas flow amounted to 10
system for a-Si:H film deposition under optimized sccm, with 2 sccm silane and 8 sccm Ar for window
conditions. With both reactors the horizontal CO 2 flushing. The small deposition reactor (15 cm inner
diameter, 30 cm optical path length, 2 sccm silane,
+)Author to whom correspondence should be addressed 2 sccm Ar) is connected to an intermediate pressure

143
 \.8
stage and an analytical chamber with a quadrupole
mass spectrometer (QMS, Balzers QMG 420) as sket- o 1.5
(a)
.c
ched out in Fig. 1. The deposition conditions were
equivalent to those in the large reaction chamber .s
(same gas supply and pumping unit) except for the •c
minimum gas flow into the molecular beam source !?
"i
(0.1 rom diameter hole in a thin steel plate) 2i
positioned in the substrate surface. On both sides '0
of the molecular beam source substrates were clam-
ped onto the heated substrate block to confirm a-
Si:H film deposition conditions during gas 0,5
analysis.
The a-Si:H film thickness was measured with a
profilometer (Dektak, 3030ST) after KOH etching of ~o
a.
a step or after removal of a silver dot placed onto
the substrate prior to deposition. The electrical o
photoconductivities were determined using a highly o 12 20 28
sensitive electrometer (Keithley, model 602) and
approximate AMI illumination (100 mWlcm2 ) . 0,9

(b)
3. RESULTS
With SiH 4 under CO 2 laser irradiation the
concentration of Si 2 H6 was found to evolve on a
minutes' timescale and to build up a stationary
concentration after Band 20 minutes with 190 and
·
c
!?
iii 0,5

·~
90 W of incident laser power, respectively (Fig. '0
2). With high laser power, the Si 2 Hd concentration

.t·
is exceeding its stationary value for a short time
after its initial increase. With low laser power
the final value is approached nearly asymptotical-
ly. The concentration of Si 2 Hd remains practically
constant after the initial starting period. The
~
~ 0,\
stationary SizHd concentrations, to which the cur-
ves in Fig. 2 are normalized, vary between about o
10- 3 mbar with 90 Wand 0.047 mbar (190 W) as given o 12 20 28
in Table 1. For the derivation of the absolute
pressure values perfect mixing of SiH 4 and Ar in
front of the substrate surface was assumed. 0:
0 ~
c o

r
!? (a) 1,5
"i .~.:R' 0 .. *~o.o I
0
" ."-' -_I a u ".! '0. o. w
'0
J"I ~• 0
. •c
c
!? • I
: .'
'/
CO

.
!?
I
j
~
i ~o:0
;::
....

...
C
0
'0
•
u
c

.J
c 0.5 <;
0 / g. 0.5

.. /.
'0 0
"• 1/ • c 90 W + l JO W I~O W
iii
c
<; , . a'
!?
E ./
~
z 0. 1 "' r;
~9 .. liO 'N X 190 W o
o 12 20 28
0
0. 1
(b)

·u
POWDER (d)

·
i 5"• E
0.5
Precur,or - Oi,ilane

·
Substrate: Temperature - ~OO C
~ o.os
0
o
Q:
Lo" r Power - t60 W
0:
.t:
i'" ic
3 (;
0,01 0.1
0
o 10 20 30 40 o 0 12 20 28
Duration of las.,. frrodiation (m in) Duration ot lO!Jer Irradiation (min)
Figure 2: (a) Temporal evolution of SizH6 during
CO 2 laser chemical decomposition of SiH4/Ar (5 mbar
total pressure) under irradiation at various power Figure 3: Partial pressures of Si 2 Hd (part a) and
levels; each curve normalized to its stationary of SiH4 (part b), and ion signal (elm = 85) of
value SisHB (part c) as a function of time during CO 2
(b) Average growth rate of a-Si:H during periods of laser irradiation (160 W) into 4 mbar of Si 2 H5 /Ar
15 min with CO 2 laser CVD from SiH4/Ar (5 mbar compared to the average a-Si:H film growth rate
total pressure) upon 160 W of incident laser power (part d) for three different ways (I III) of
(± 0.01 nm/s estimated uncertainty) processing (cf. text)

144
Table 1: Stationary pressures of SiH. and Si"H" 4. DISCUSSION
obtained with CO 2 laser CVD of a-Si:H from SiH./Ar In the above described experiments CO" laser
(5 mbar total pressure) as a function of laser CVD and gas phase analysis are closely related to
power (cf. text for details) each other: The gas sample is taken directly from
the CVD reaction gas mixture in a well defined
P1& •• r 90 110 130 150 170 190 (W) probe volume in the substrate surface. The gas
sample thus covers those species which usually hit
p(SiH.) 2.5 2.4 2.3 2.2 2.1 2.1 (mbar) the substrate surface and make up the film. By this
p(Si"H,,) 0.1 1.6 3.0 3.7 4.2 4.7 (10-" mbar) way, gas probing occurs at the latest possible time
before film formation. This fact together with the
(practically collision free) transfer of the gas
Film formation from SiH. under the above expe- sample via a molecular beam into the QMS analyser
rimental conditions was observed at a power level provides for minimum (negligible) chemical reaction
of 140 W or above. As evident from part b of Fig. 2 between sampling and analysing the gas.
the film growth rate averaged over deposition The substrate was heated to 400 DC and the
periods of 15 min decreased with increasing time of processing gas in front of the substrate surface
CO" laser irradiation. had a temperature very close to this value /7/. The
Similar investigations were performed with decomposition temperatures of polysilanes, on the
Si"H.. as prescursor for CO" laser CVD of a-Si:H other hand, are known to be relatively low, i.e.
(Fig. 3). Parts a through c of Fig. 3 are showing considerably below 400°C /8/. From this point of
the temporal evolutions of Si"H .. , SiH., and SisHa view the finding of Si.H~D and Si~H~" with SiH. and
under three different processing conditions (I, II, Si"H .. as prescursors, respectively, appears remar-
III) while part d displays average film growth kable. The presence of even higher polysilanes
rates as a function of the laser irradiation time. during CO 2 laser CVD of a-Si:H has been established
When a laser beam of full power is directed into recently /9/, but in those experiments gas sampling
the cold reaction gas mixture with its original was performed with a capillary far from the heated
composition, powder formation is observed (process substrate and the laser irradiated volume.
I). It is pointed out that the hot substrate It appears instructive to compare the concen-
surface induced partial Si"H .. decomposition during trations of Si"H o and Si5Ha probed in situ in the
the time necessary to establish stationary flow substrate surface with the time dependent a-Si:H
conditions and pressure stabilization (processes I film growth rate (Figs. 2 and 3): A correlation
and III). As a result, the pressure values of Si"H .. between the silane concentrations and the film
and SiH. at the beginning of laser irradiation growth is evidently missing. If Si"H .. , Si 5Ha , and
amounted t~ 1.8 mbar (part a of Fig. 3) and 0.2 higher silanes, on the other hand, are regarded as
mbar (part b of Fig. 3), respectively. Powder film forming species /9, 10/ our present finding is
formation (process I) can be avoided by either showing that unde, the present reaction conditions
slowly increasing the CO" laser power within three the generation of higher silanes in the gas phase
minutes (process II) or at full laser power by cannot be the rate limiting step of a-Si:H deposi-
slowly adding Si 2 H.. into pure flushing gas (process tion.
III). Again maxima in the concentration values are Looking at the maxima in the time dependent
observed. In all three cases of processing, on the Si"H" and Si~Ha concentrations it appears reasonab-
other hand, the same stationary concentrations of le to think of intermediate deposition conditions
SiH. and Si 2 H" are obtained. Furthermore, similar which may not lead to high quality films as expec-
to the CVD behaviour with SiH. as precursor·, the ted for the bulk film formation under stationary
average deposition rate is found to decrease with conditions. This assumption is supported by the
the duration of CO 2 laser irradiation into the gas observed powder formation in case of full power
mixture originating from Si 2 H". laser irradiation into the initial, cold Si 2 H6
Looking at the electrical photoconductivity reaction gas mixture. Following this idea, the a~h
(aph) of a-Si:H films obtained at different times values of a-Si:H films deposited at different times
of laser irradiation applying the variable shutter from the beginning of laser irradiation were compa-
in front of the substrate, no difference is obser- red (Fig. 4). Missing differences in the aph values
ved in the aph values (Fig. 4). are, however, due to a shortage of the measurement
method. It is nearly insensitive to low quality

10 nm H-sat. 10- 8 n-~cm-~ Rh

0 .0001
100 nm bulk
0.00001
t 10 nm start 10- 8 n-1cm-~ R.
E
u
10E-06
Ie:
x x )( x x x x x x x a(parallel) <---> l'
1 OE-07
)(
x x I
.t:
Q.
0 I
V
1 OE-08

1.0E-09
Shutter 15 30 15Min.
Deposition 15 15 15 30Min .

Figure 4: Photoconductivity values aph of a-Si:H

films obtained at different times of laser irradia- Figure 5: Scheme illustrating film resistance
tion into the processing gas (during "shutter time" measurements parallel and vertically to the film
film deposition on the substrate is prevented by surface (s start, b bulk, h hydrogen
the shutter) saturated; also cf. text)

145
sublayers within the film. This fact is illustrated
in Fig. 5. For an order of magnitude estimate, a
film consisting of three sublayers, namely a 10 nm
thick, poor quality film (a - 10-a 0-1cm- 1 ) deposi-
ted during the starting period (index s), the main
100 nm bulk (b) with a ~ 10- 6 0-1cm-1, and a 10 nm
hydrogen saturated cover layer (h) with a = 10- 8 0-
1cm- 1 are assumed. Measuring the film resistance R,
as usual, parallel to the substrate surface using
two contacts on the film surface can be regarded as
measuring R., Rh , and Rh in a parallel way with l!R
1!R.+1!Rb +1!Rh . As a result, an apparent
(effective) conductivity of a(parallel) = 8.4x10-7
0-1cm- 1 is obtained. Measuring, however, along the
carrier flux direction in a solar cell, i.e. verti-
cally across the film, corresponds to a serial
resistance R = R.+Rh+Rh and a(vertica1) = 5.7x10-a
0-1cm- 1 . Whereas the effective a(paralle1) value
differs only about 16 % from the bulk value,
a(vertical) is lower by more than one order of
magnitude.
Regarding the above result it appears neces-
sary to control the initial deposition period dili-
gently in order to avoid low quality interlayers.
Such inter layers Can evidently lead to poor solar
cell performance in spite of high quality bulk
material.

ACKNOWLEDGEMENT
Funding of this work by the German Ministry of
research and technology (BMFT) under contract num-
ber 0328965 A is gratefully acknowledged.

REFERENCES
/1/ C.f. e.g. R.H. Bube, Annu.Rev.Mater.Sci. 20
(1990) 19
/2/ C.f. e.g. H. Tasahi, W.Y. Kim, M. Hallerdt, M.
Konagai, and K. Takahashi, J.Appl.Phys. 63
(1988) 550
/3/ H. Stafast, Appl.Phys. A 45 (1988) 93
/4/ F. Curcio, I. Gianinoni, and M. Musci, in
Proc. E-MRS Conf., Strasbourg, 1986, pp. 117-
123
/5/ H.M. Branz, L.K. Liem, C.J. Harris, S. Fan,
J.H. Flint, D. Adler, and J.S. Haggerty, Solar
Cells 21 (1987) 177
/6/ Cf. e.g. D. Metzger, K. Hesch, and P. Hess,
Appl.Phys. A 45 (1988) 345
/7/ M. Meunier, J.H. Flint, J.S. Haggerty, and D.
Adler, J.Appl.Phys. 62 (1987) 2812
/8/ Cf. e.g. J.G. Martin, H.E. O'Neal, and M.A.
Ring, Int.J.Chem.Kinet. 22 (1990) 613 and
references quoted therein
/9/ K. Hesch, Ph.D. thesis, University of
Heidelberg, 1990
/10/ s. Veprek, Thin Solid Films 175 (1989) 129

146
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Simulation and Analysis of the Bias Dependent Spectral Response of a-Si:H pin Solar Cells

J.Bruns, S.Gall, H.G,Wagemann

Technische Universitat Berlin, Institut fUr Werkstoffe der Elektrotechnik
JebensstraBe 1, 1000 Berlin 12, Germany
Tel.: (int. 49) 30/314-22442(24482); Fax: (int 49) 30/314-23029

Abstract
Starting with our numerical model of a-Si:H pin solar cells, the bias voltage dependent spectral response has been simulated and
compared to experimentally obtained values (without bias light). The interchange between primary and secondary photocurrent
(i.e. between generator and load device operation) is explained by the interaction of field- and diffusion components within the
photocurrent. Experimentally obtained SPY values of a characteristic length provide support to this model.
l.Introduction exponential distributions of tail states and two Gaussian
ones of dangling bond states. Both types of traps obey the
Up to now the physics of amorphous silicon solar cells is not Taylor and Simmons occupancy statistics /1/ and are
understood completely. Experimental analysis combined considered independently. All localized states between the
with comprehensive numerical simulations are the tools trapped electron and hole quasi- Fermi levels act as
used to develop a physical model of the thin- film solar cell recombination centers. We also consider the effect of
made of hydrogenated amorphous silicon. To achieve dopants on the density of localized states. The model
unambiguous results it is useful to concentrate on assumes a 0.6.um pin structure, where the light enters either
pronounced peculiarities of the device, such as the bias through the p + - layer or through the n + - layer. The optical
voltage dependent spectral response changing its amount carrier generation rate is calculated for monochromatic
and sign when increasing the forward bias voltage of the cell. excitation and for AM1.5- spectrum. The value of the surface
In order to enhance the senSItivity of these investigations the recombination velocity at the contacts on the p + - and on the
experiments were confined to spectral response n + - layer may be chosen arbitrarily. A survey of the model
measurements without any bias light application. Thus the parameters is given in table 1.
transition wavelength from primary to secondary
photocurrent for different bias voltage was carefully Mobility gap EG = 1.72 eV
measured and analyzed, improving the insight into the Effective density of states Nc = Nv = 1019cm-3
device physics. Finally, experimental values of the Band mobility for electrons /-tn = 10 cm2 y-1 s-l
characteristic length obtained by SPY evaluation are Band mobility for holes .uP = 1 cm2 y-1 s-l
presented completing the model of interaction between field p + doping NA = 1019 cm- 3
and diffusion currents. n + doping ND = 1019 cm- 3
p + layer thickness 10 nm
2.Numerical Model n + layer thickness 10 nm
Intrinsic layer thickness 5BO nm
A complete simulation of a-Si:H pin solar cells is based on Tail state density at mobility edge NtsGap = 1021 cm- 3
the simultaneous solution of Poisson's equation and the Characteristic energy of conduction band tail EA = 60 me V
continuity equations for electrons and holes, Characteristic energy of valence band tail ED = 50 me V
Dangling bond density of intrinsic layer Ndb= 1016cm-3 eV- 1
a
dx
2 = -
c.L
e [p-n+Ptrap-ntrap+N})-NA- ] (1),
Dangling bond db + -level (above Ev)
Dangling bond db - -level (below Ec)
0.7 eV
0.4 eV
Standard derivation of distribution
't
d'
= q (R-G) (2),
d'
~ = -q (R-G) (3), of db- states
Neutral trap cross section
ad =150meV
an = 10-15 cm- 2
Ratio between charged and neutral
where n(p) is the free electron (hole) density, tp is the trap cross sections C =100
electrostatic potential, NA- (ND +) is the concentration of
ionized acceptors (donors), ntrap (ptrap) is the density of TABLE 1: List of parameters used in numerical modelling.
electrons (holes) trapped in tail states and dangling bond
states. R (G) is the recombination (generation) rate.
The electron and hole currents are given by 3.Experimental

. 5!!J!.. kT dn We have measured the bias voltage dependent spectral

jll=-q[.un n --.un-] (4) response of a-Si:H pin solar cells. The cells were made in a rf
• dx q dx
glow discharge deposition system at the Siemens Research
. 5!!J!.. kT 5!E. (5)
jp= -q[.upP dx +-q/-tP dx] Laboratories. They have an approximate thickness of 0.6 /-tm.
The measurement set-up consists of a tungsten lamp, edge
filters, a grating monochromator and mirror optics. Specially
where.un (t-tp) is the electron (hole) band mobility.
designed electronic equipment is used to apply the bias
The first step in solving the system of equations (1) to (5) is
voltage and to measure the photocurrent caused by a
to discretize them using finite differences. Next, the set of
chopped spectral irradiation in the wavelength range
coupled difference equations is linearized and solved by
between 350nm and BOOnm. There is no additional bias light.
Newton's method.
The density of localized states and their distribution through The bias voltage is varied between -lV and 1V. The
the band gap strongly influences the device behavior. The structure of the solar cells is glas/TCO/p(C)/i/n/Me, where
p(C) stands for p- doped a-SiC:H. Each cell is measured in
acceptor and donor-like states are represented by two

147
the annealed state (state A) and in the degraded state (state the wavelength of the incoming light and on the applied bias
B) after 64h of AMI illumination. This paper deals with voltage.
selected parts of our experimental investigation and Without bias voltage, a short wavelength causes a steep
modelling. generation profile near the surface. The excess electron
concentration gradient leads to a primary photo current
4.The bias voltage dependence of the spectral response within the i- layer (fig.3a), while the gradient of the
photogenerated holes causes a secondary photocurrent. The
To understand the effect of the bias voltage on the spectral diffusion part of photocurrent is dominated by the electron
characteristics of a-Si:H pin solar cells two aspects have to diffusion component due to higher electron mobility. An
be considered: irradiation with light of long wavelength (fig.3b) gives rise to
- why does S(,1.) change with bias voltage, and a more homogeneous generation rate through the device.
- how can the photocurrent change its direction? Both the electron and hole concentration gradients support
A set of measured spectral response characteristics of a pin a secondary photocurrent.
solar cell is shown in fig.la. The spectral response first
decreases and finally changes its direction with increasing S.o ---------------------
forward bias voltage. The primary photo current changes into S,bJAW·1 Sample AD2
secondary photocurrent. This transition is dependent on the '.0
voltage and is characterized by the transition wavelength
,1.T = f(V) where the spectral response disappears. Because of '.0
this behavior the solar cell works as a generator at a certain
bias voltage in the short wavelength region and as a load in •• 0

the long wavelength region. By increasing the bias voltage

the transition wavelength ,1.T(V) is moved to smaller 1.0

wavelengths, until S(,1.) finally becomes completely negative.

A comparison between simulated and measured
characteristics in fig.la/b shows a good correspondence. We -1. 0
use our numerical simulation model to examine the voltage
·dependency of the spectral response. For that purpose we
3.5 4.0 4.5 5.0 5.5 6.0 6.S 7,0 1.5 8.0 *102
investigate the simulated characteristics in fig.lb at the two ).fnm
wavelengths ,1.1 = 460nm (short wavelength) and,1.2 = 620nm
Fig. 1a: Measured spectral response of the sample AD2 for different bias
(long wavelength). The photocurrent of the a-Si:H pin solar
cell comprises a field and a diffusion component. Both parts ,'0- ' voltages.
are defined by the difference between the current 5.0 -----------------~

components with monochromatic excitation and without it . SabJAW'\ Simulation

The bias voltage influences the field and the diffusion part •. a
differently.
•. a
4.1. The field component of the photocurrent
2. a
The profile of the electric field strength within the device
depends on the bias voltage and to a smaller amount on the 1.0

wavelength of incident light. The profile of the electric field

is shown in fig.2 for the wavelength ,1. = 620nm at different o. 0 -t-----;-;;;(V::o:iiV7'~~....:.....7:f;::::::==:::::::f_
: ~r(Y ~ O.8Y)
bias voltages. A reverse voltage increases the electric field t
-1. 0
strength (not shown), while a forward voltage decreases the
field strength until it even changes its direction. In order to 'T',...~j'1'T'l~T'1'~'l'fTtr'111 I '1"1'1
_10 2
examine the behavior of the field part of the photocurrent, 3.5 4.0 4.5,1,\ 5.0 5.5. 6.0,\1 6.:5 7.0 7.5 8,0

let us analyse for example the difference between the F'Ig. Ib: S'Imu Iate d spectral response of a O.61J-m pin device Alnm
for different
electron field current components for monochromatic bias voltages.
p" n'
excitation and without any excitation, which essentially is the 10 6 !. ----.-----.-- - - - - _ . _ - - - - - - 4 -
field part of the photocurrent, ,t =620nm
ENcm'l

jn ph field = q IJ-II (n;, EA - ndark Edark) . (6) 10'

The interchange between the primary and the secondary 10'

field part of the photocurrent depends not only on the
electric field strength E but also on the signs and the
10'
magnitudes of the products (nOE) resp. (pOE).

4.2. The diffusion component of the photo current 10'

3 t
,
t
t t
The diffusion part of the photocurrent is determined by the
10 ' ~-~~~~i~~~~-~~~~--i
gradient of the photogenerated excess carriers (fig.3a/b). We
0.0 1.0 2.0 3.:::1 -4.0 5.0 xlnm B.O ';J0 2
examine the excess carrier concentrations An and Ap, which
Fig. 2: Electric field pror.Je at wavelength A2 = 620nm. Without bias
are given by the difference between the free carrier
voltage V = av, moderate bias voltage V = O.6V, high bias
concentrations with and without monochromatic excitation.
voltage V = a.8V. The inverted parts of the electric field are
The gradients of the excess carrier concentrations depend on indicated.

148
A forward bias voltage causes two variations in the device: compensated by a secondary diffusion current. Irradiation
- The concentration of photogenerated excess carriers generates excess carriers which are collected by the electric
depends on the electric field profile through the device. A field. At short wavelengths the diffusion of excess carriers
decrease in the electric field changes the slopes of the causes a primary photocurrent. Field and diffusion parts
concentration profiles. have the same direction. At long wavelengths field and
- The carrier injection into the i-layer of the device moves diffusion parts have the opposite direction. Hereby the
the quasi- Fermi levels to the respective band edges. primary field part dominates.
Recombination increases because carrier recombination For moderate forward bias voltages the electric field through
occurs via additional states. the i-layer is reversed (fig.2, V = O.6V). Nevertheless, the
Both effects give rise to a primary diffusion part of field part of the photo current remains positive.
photocurrent for higher bias voltages. For long wavelengths, At higher bias voltage, the field part of the photocurtent has
an increasing bias voltage decreases the secondary diffusion changed from primary to secondary direction. At short
current, and a reversal of profile slope is observed for wavelengths the primary diffusion part of photocurrent
voltages between V = 0.6 ... O.SV changing the secondary increases due to a rise in the gradients of the excess carriers.
diffusion current into a primary one. For short wavelengths, At long wavelengths the gradients of the photogenerated
the initial primary diffusion part of photocurrent improves excess carriers change their direction and now form a
with increasing bias voltage. diffusion current in the primary direction.
In both cases the field parts of the photocurrent have the
secondary direction opposing the diffusion parts. This is
shown in fig.4, where the total photocurrent is divided into
field part and diffusion part for a bias voltage of V = O.SV at
both examined wavelengths. At the short wavelength A,1
(A,1 <A,T) the photocurrent is determined by the diffusion of
photogenerated carriers. The total photo current is negative
(primary direction), the spectral response is positive (fig.4a).
-,.

":~.]\
A-460nm .)

"]\~
:: i -t---------------::-
-1.0
n'

'f
10"~~--------------------------------~

1\
-2.0

L~"."".'r "'-" .", .

.6n1em.J
b)
10'"

....
10'2 v_ ov -'.Q

10"
1.0 2.0 10 .0
n" ''''''',}
'S.D 'om ' 102
10'·

' ..
10'

10' j,.I~::,,·n1 - - --::-0 V

).0 lr'Alld
10 7
2••
10'

.:': ~ --
10'
C.O 1.0 1.0 11: 40 1.0 e.c
Fig.3: Excess electron distribution through the device for three
investigated bias voltages:
a) Short wavelength excitation withAl = 460nm. The gradients
-z.O 1
,,,1,.1
.>0 -;

i ,
of excess electron concentrations cause a primary diffusion

..
photocurrent. -'.0

..
b) Long wavelength excitation with .1.2 = 620nm. The gradients
-S.Q ~II 11"1111111 i'm' 1111'1,11
of excess electron concentrations change their direction with
, Z.Q '.0
increasing bias voltage. The secondary direction of the
Fig.4: Profile of photocurrent density in the i· layer separated in field
diffusion part changes to primary direction. part and diffusion part for a bias voltage V = O.8V. The diffusion
part causes a primary photocurrent, the field part causes a
secondary photocurrent.
4.3. The interaction between field and diffusion currents a) Short wavelength excitation with,!! = 460nm. The diffusion
part of the photocurrent dominates. The total photocurrent
Th.e interaction between the field part and the diffusion part is negative (primary direction) and the spectral response is
of the photocurrent is responsible for the bias voltage positive (generator).
b) Long wavelength excitation with .1.2 = 620nm. The field part
dependence of the spectral response. Furthermore, both
of the photocurrent dominates, the photocurrent is positive
parts depend on the wavelength. In the equilibrium state the (secondary direction). The spectral response is negative, the
electric field causes a primary field current which is solar cell acts as a load.

149
 18000~---------------__;

For the longer wavelength ,h (A2>AT), the field part ).JS /nmWA" S"mpie AD2
dominates the device. The total photocurrent is positive
(secondary direction) artd the spectral response has changed 15000
its sign. The solar cell is a load device (fig.4b).
Now we are able to explain why the photocurrent disappears V ~O.61 V

at the transition wavelength AT: the secondary field part of '2000

the photocurrent compensates the primary diffusion part.
The dispersion of the transition wavelength AT with rising
gOOOr
bias voltage is caused by an increasing field part, which
dominates over the primary diffusion part and shifts AT to
smaller wavelengths.
6000

5. Experimental extraction of a characteristic length Leoll

JOOO
The SPY method of extracting the diffusion length from
measured plots of spectral response, which is well-known
and well-understood for the application to crystalline solar O,~~--~----,_--.-_r--.-_.--.-~
cells /2,3/, has been transferred to a-Si:H pin solar cells and -0 50 c,oo 050 1 00 I 50 200
the results are presented here. The basis for any evaluation Ilfl llirn
is a plot of the wavelength A divided by the spectral response Fig.6: Extraction of characteristic length Lcoll. Sample AD2, degraded
SeA) vs. the reciprocal absorption coefficient a-I, providing state B, plot for two different bias voltages. At this state, LcoH
depends strongly on bias voltage.
the diffusion length of minority carriers of crystalline cells as
the intersecting value within the negative part of the
abscissa. In the case of a-Si:H pin solar cells the incident
light is not absorbed entirely within the cell and the YBiasN Leoll/.um Leoll/.um
photocurrent essentially consists of electrons and holes. State A State B
Thus the physical meaning of the extracted characteristic
length LeaH has to be newly established for a-Si:H solar cells. 0 0.38 0.36
The evaluated values of the characteristic length LeaH show 0.31 0.38 0.32
distinct changes in behavior according to voltage value and 0.41 0.38 0.31
light soaking. With increasing voltage values LeaH decreases 0.61 0.36 0.17
and even changes its sign. Light soaking means diminution of 0.70 0.28
Leal! and essentially a higher dependence on voltage value. 0.73 -0.07
Fig.5 presents S vs. A. plots for two different values of bias
voltage. Fig.6 shows the SPY extractions; a set of values is Table 2: Extracted characteristic length of sample AD2 for
presented in table 2 for the solar cell in the annealed state different bias voltages in the annealed State A and degraded
(A) and in the degraded state (B), which obviously ~rovide a state B.
quality feature of any a-Si:H solar cell. The expenmentally
introduced characteristic length LeaH for a-Si:H solar cells
may provide some arguments for the above (Chapter 4) 6. Conclusion
developed model of interaction between field and diffusion
components within the photocurrent. In the vicinity of the Detailed analysis of .the bias dependent spectral response
short circuit current the characteristic length may be within the region of interchange between primary and
understood as the drift length. For increasing bias voltage secondary photocurrent provides insight into the physical
the field component cuts down and a counteracting influence behavior of a-Si:H solar cells. For monochromatic light
of diffusion becomes important, shortening the LeaH values traversing parts of or the entire cell, the interaction of field
and reversing their sign as the primary photocurrent changes and diffusion components within the photocurrent
to a secondary one. completely decribes the phenomenon of transition from
generator to load behavior. A readily available characteristic
length extracted experimentally serves as a quality feature of
a-Si:H solar cells under test.
'.0 -r-
s-.-m-PI-.-AO- Z - - - - - - - - - - - - - r l l
S",JAW"
7. Acknowledgement
'.0 L
The financial support of Fa. Siemens AG. is gratefully
VaOV
l-D acknowledged. The solar cells were supplied by Dr. W.
Kusian. The authors highly appreciate fruitful discussions
with Dr. W. Kusian and Dr. H. Pfleiderer.
' .0
...----- V -O.6IV
8. References
'.0
/1/ G.W. Taylor, J.G. Simmons, J. of Non-Crystalline Sol.
8-10, 940 (1972)
/2/ E.D. Stokes, T.L. Chu, Appl. Phys. Lett. 30, 425 (1977)
>. , •. 0 • 5 :5.C ~,5 6.0 o.~ 7.0 7. 5 I.D _,oii! /3/ K Bucher, J. Bruns, SA Jahnke, H.G. Wagemann,
Mom
Proc.of 8th European Photov. Solar Energy Conf.,
Fig.5: Measured spectral response of sample AD2 at degraded state B. Florence, 1317 (1988)

150
10TH EUROPEAN PHOmVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON, PORTUGAL

INFLUENCE OF DEPOSmON TEMPERATURE ON THE STABILITY OF a-Si:H SOLAR CELLS

J.M. Asensi, J. Bertomeu, J. Andreu, J. Puigdollers and J.L. Morenza

Universitatde Barcelona, Departament de F(sica Aplicada i Electronica
Av. Diagonal 647, E-08028 Barcelona (Spain)

ABSTRACT

In order to study the degradation of a-Si:H solar cells, three temperatures (250,
300 and 350°C) were used for the deposition of the intrinsic layer of p-i-n devices
with identical p-Iayers and n-Iayers. The results show that the cell with intrinsic
layer grown at lower temperature has an initial poor efficiency and a higher
degradation than cells grown at 300 and 350°C. The cells grown at 250, 300 and
350°C present a degradation of 35%, 14% and 9% after 530 h illumination. We
attribute this behaviour to the stability of intrinsic a-Si:H with a low content of
hydrogen evolving at low temperatures and a small SiHz/SiH ratio.

1. INTRODUCTION influence of i-layer deposition temperature on the

degradation of a-SiH p-i-n devices.
In addition to the improvement of the photo-
voltaic performance of a-Si:H solar cells, the stabil-
ity for long-term use is considered to be a key 2. EXPERIMENTAL
factor for the commercialization of a-Si:H solar
cells for large scale power applications. A capacitively coupled plasIna deposition
The reversible (by annealing over 150°C) system was used to prepare a-Si:H pin cells. The
light-induced defect generation (Staebler-Wronski reactor consists of a single chamber with a turnable
effect [1]) has been associated with the degradation substrate holder which allows to hide the samples
of a-Si:H devices. Luckily, this degradation ap- from the plasIna influence. This feature has been
pears to be self limiting, since the number of creat- used to avoid contamination from the walls. So to
ed defects saturates [2] and very useful techniques prevent contamination of i-layer in p-i-n devices,
for practically suppressing the light-induced effects before deposition of i-layer, intrinsic a-Si:H has
have been reported [3]. been grown for 15 min while the substrates have
The origin of Staebler-Wronski effect and its reInained hidden.
connection with solar cell degradation are not yet The p-i-n solar cells used in this study have
well understood. However there is a certain agree- the structure glass/Sn02/p-i-n/Cr and a surface of
ment about the role of hydrogen in this effect. In 0.25 cm2 • Three substrate temperatures for the
general, films with low hydrogen content tend to be i-layer deposition have been used (250, 300 and
more stable [4,5]. Also the way in which hydrogen 350°C). To ensure that doped layers were identical
atoms are linked to the Si network seems to affect in all devices, the p-Iayer was always deposited at
the light-induced instability. It has been shown that the highest temperature and the n-Iayer at the low-
films with a substantial concentration of hydrogen est one, so possible changes in previously deposited
evolving at low temperatures [4], and with a high layers due to an increase of substrate temperature
fraction of hydrogen incorporated as dihydride were avoided.
groups (SiH0 [5], show the largest increase in Before solar cells were deposited, we had
defect density after light-soaking. studied the properties (deposition rate, hydrogen
The aim of this work is to improve the stabil- content and optical gap) of the intrinsic Inaterial
ity of a-Si:H solar cells. As the substrate tempera- obtained in the same conditions. In Table I we
ture during the growing process in plasIna de- present the results for the three temperatures. Then
posited a-Si:H films influences significantly the the deposition times for the intrinsic zones have
incorporation of hydrogen, we have studied the been calculated to obtain the same thickness in all

151
devices. All the films were obtained by using a p-i interface, because of the excess of gap states,
pressure of 40 Pa, a total flow rate of 20 seem, and which enhances the separation of carriers created in
a power density of 20 mW/cm2• The dopant levels the vicinity of this interface and competes with
in gas phase (BzH6 and PH) in SiH4) were 0.1 % for carrier recombination [6].
p-type doping and ·0.5 % for n-type doping. The
thicknesses were 10,400 and 20 nm for the p-, i-
and n-Iayers respectively.
.......,1 .0
<
-S
~0.8
Vi
250 0.19 19.5 0.85 1.74 Z
w
300 0.21 15.8 0.75 1.73 ~0 . 6

350 0.23 10.5 0.17 1.69 I-

Z
)t! 0.4
Table I. Substrate temperature, deposition rate, a::
::>
hydrogen content, SiHz/SiH ratio and optical gap u
0.2
values corresponding to intrinsic layers deposited at
the same conditions that cells used in this study.
0.0 =---::-':--::-'.:--=-'::---7"":--::":--:-':'-:-'=---="'=--=-'
0.0 0.1 0.2 O.J 0.4 0.5 0.6 0.7 0.8 0.9
VOLTAGE (V)
The light source for the prolonged illumination
Figure 1. Current intensity vs. voltage for the cells
was an halogen lamp with an intensity of
with intrinsic layer deposited at 250 (circles), 300
100 mW/cmz. The total illuminating time was 530
(triangles) and 350°C (squares). Solid lines repre-
h, and the samples were left in the open circuit
sent the initial state and dashed lines the degraded
condition during the illumination. The photovoltaic
state after 530 h illumination.
parameters I.e, Vce' FF and f/ were measured peri-
odically. All the experiments were carried out at
room temperature and in the air.
It has been noted a non-ohmic behaviour (in-
flection of I-V curve for applied voltages above
3. RESULTS AND DISCUSSION Vor:> and a deterioration of the fill factor for the
devices with n-Iayer deposited at lower temperature
Figure 1 shows the light-induced change in I-V than i-layer (Figure 2).
characteristics for the devices after 530 h illumina-
tion. With respect to the initial characteristics, it 1.25
---
r--T'"-..----.--...,....---.--.--~__.____,

can be seen that cells deposited at higher tempera-

tures show a higher short current intensity I.e and 1.00

a lower open circuit voltage Voc' The behaviour of <-E

........ 0.75
the current intensity can be related to differences in
the optical gap of intrinsic layer (the gap is lower ~
iii
in layers deposited at higher temperatures). The Z 0.50
w
I-
dependence of Voc on the i-layer deposition tem- ~
perature cannot be so easily explained. We believe, I- 0.25
z
in agreement with other authors [6], that in our w
case the Vor: is predominantly limited by the recom- ~ 0.0 0 1 - - - - - - - - - - - - + - - - - i
::>
bination at the p-i interface, more than by the built- u
-0.25
in potential. In fact, we have found similar Voc in
cells deposited at a constant temperature for all the
-0.58 =--;:-':---;;"':-:-":--;:-':--::-':---=-'=:--::-'.:--=-'=--="'.
layers and the same three temperatures (Figure 2), .0 0 . 1 0.2 O.J 0.4 0.5 0 .6 0.9
so it seems that the modification of the doped VOLTAGE (V)
layers has not a significant influence on the Vor: and Figure 2. Current intensity vs. voltage for a cell
the limiting mechanism of this parameter is related deposited at a constant temperature of 350°<:;:
with a property of the i-layer or the interface. Then (dashed line) and a cell with p- and i-layers de-
the high Vor: in the cells with lower deposition posited at 350°C and the n-Iayer deposited at
temperature could be attributed to a compensating 250°C (solid line).
effect due to the local increase of the field at the

152
 The degradation behaviour of photovoltaic 4. CONCLUSION
characteristics for the three cells is shown in Figure
3. where V00' I tc • FF and ", are presented as a The degradation of p-i-n solar cells with the
function of the illumination time. It can be ob- same structure. identical p- and n-layers and intrin-
served that V00 increases whereas lac. FF and 1j sic layer thickness depends appreciably of the i-
decrease. It is also observed that the degradation is layer deposition temperature. Whereas the cell with
higher in the cell grown at 2S0°C (35 % after intrinsic layer grown at 250°C shows an initial
530 h) and lower when the deposition temperature poor efficiency and a high degradation of 35 % after
of the cells is increased (14% for the cell grown at 530 h illumination. the cells grown at 300°C and
300°C and 9 % for the cell grown at 350°C). 350°C show an almost identical conversion effi-
ciency and a degradation after 530 h illumination of
0 .80
r---e-
~6
e
6
e
...
~
....
14% and 9% respectively. This later result is attri-
buted to the low light induced creation of defects in
materials with low H content.
We also have checked the complex connection
~a II a
0.70
.."
between the changes in the material and the solar
cell degradation. Redistribution of the electric field
1.00 Gease a
..
II 9
,..... ~666
6 6 6
-()
inside the device due to the increase of trapped
:;)
charge can produce positive effects in non opti-
~0.75
~ e e e
mized solar cells (initial increase of V00 and FF).
..,¥ ~

::
0.50

0 .65
~~ ACKNOWLEDGEMENTS

~ This work has been supported by the DGICYT

of the Spanish Government (project PB89-0236).

-.~ : : -~
REFERENCES
1:"0.75

0 .50
~ II e
[1] D.L. Staebler. and C.R. Wronski. J. Appl.
~
Phys. 51. 3262 (1980).
o 100 200 300 400 500 600
ILLUMINATION TIME (h) [2] H.R. Park. I.Z. Liu. and S. Wagner. Appl.
Phys. Lett. 55. 2658 (1989).
Figure 3. Open circuit voltage (VoJ. normalized [3] Y. Uchida. M. Nishiura. H. Sakai. and H.
short circuit intensity (J,J. fill factor (FF) and Haruki. Solar Cells 2. 3 (1983).
normalized efficiency (",) for the three cells vs. [4] M. Oshawa. T. Hama. T. Ichimura, T.
illumination time. Symbols are the same that in Akasaka, H. Sakai, S. Ishida, ~d Y. Uchida,
Figure 1. J. Non-Cryst. Solids 77&78. 401 (1985).
[5] E. Bhattacharya. and A.H. Mahan, Appl.
The decrease in FF and I tc could be explained Phys. Lett. 52, 1587 (1987).
as a consequence of the increase in the density of [6] S.S. Hegedus. N. Salzman. and E. Fagen, J.
recombination centers in the bulk of the intrinsic Appl. Phys. 63, 5126 (1988).
layer. The surprising increase of V00 in all cells.
could have the same interpretation that we pointed
out above: the increase of the electric field in the
p-i interface due to the charged traps. It has also
been observed an initial increase of the fill factor
for the devices with n-Iayer deposited at lower tem-
perature than i-layer. which even produces a initial
increase of the efficiency with the light-soaking.
This behaviour could be attributed to the unusual
poor i-n interface of our devices which is also
responsible for the non-ohmic behaviour. A light-
induced increase of the electric field in this inter-
face could counteract the excess of recombination.

153
lOTI! EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THICKNESS MAPPING OF THIN FILMS BY

LOW-CON1RAST OPTICAL IN1ERFEROMElRY

I. Solomon, M. Bhatnagar and M. Rosso,

Ecole Polytechnique, Laboratoire P.M.C., 91128 Palaiseau, France.

. . .The e1abor~tion of a thin-film solar panel requires, among other things, the
OpUtmzatlOn of the thickness homogeneity of the fIlm We present a method for a rapid
thickness mapping of the deposited film: with a simple experimental arrangement and a
proper computer treatment of the image, a mapping with a resolution of a few
Angstroms can be obtained.

1. INTRODUCTION variation of the thickness of the film by a quantitative

measurement of the variation of the intensity of the transmitted
The measurement of the optical transmission (visible and light. In a very inhomogeneous sample, we observe a series of
near infrared) of a thin film of semiconductor provides a interferences fringes: a "half-fringe" (variation from a maximum
powerful tool for the study of the physical properties of the to a minimum of transmission) corresponds to a variation of
material( 1). Because of the very high precision of the thickness ~dO obtained from Eq. (1) and (2) :
transmission measurements with a double-beam spectrometer,
the optical parameters of the film, thickness d, refractive index (4)
nO.) and absorption coefficient a(A) (n and a are in general a
function of the wavelength AJ can be determined precisely(2). For a film of amorphous silicon (n == 3.6 -3.8) and for an
Not only the optical gap is deduced from a(A) and the thickness operating wavelength around 800 nm, which is within the
d is measured within a few tens of Angstrom, but a careful study transparent region of amorphous silicon ( "->750 nm), we have
of the optical dispersion n(A) can give important information on ~do == 500 A. The optical contrast of such a half-fringe, (Tm -
the band structure of the material(1,3). Tmin) I T min == 1.6 (see Fig. 1), is quite comfortable and canl!f,e
The method is very sensitive to the quality of the film; in seen with a Simple near-infrared viewer. However, for a good
particular the optical homogeneity of the film, generally limited film of amorphous silicon, deposited by glow discharge, the
by the thickness homogeneity, is very critical to obtain reliable overall thickness inhomogeneity ~d is in general smaller than
results(4). In the present study, we discuss the possibility of an ~dO and one has to measure much smaller contrasts, and some
optical mapping of the film in order to evaluate its homogeneity opucal engineering becomes necessary.
and possibly to select a stationary region for further
spectroscopic measurements. We consider here reasonably good
films of amorphous silicon deposited by glow discharge(5) 3. LOW CONlRAST IN1ERFEROMElRY
where there is, in general, less than one optical fringe across the
whole sample: the variation of transmission from one part to the The minimum contrast that can be resolved is limited by the
other of the film can be very weak and the problem is that of the signal-to-noise ratio p of the optical captor. For a typical CCD
treatment of an image with a very low contrast. video camera, this signal-to-noise ratio for a single frame image
is around p==20. For a given variation of the thickness ~d, we
observe a variation of the transmission ~T (Fig. 2) and the
2. OPTICAL IN1ERFEROMElRY sensitivity of detection of ~d is proportional to the slope ~T/~d
of the curve T(d). This slope is obtained by denvation of Eq. (1)
The transmittance spectrum across a dielectric thin film and a maximum is found for
shows a series of oscillations due to the multiple reflections on
the surfaces of the film (Fig. 1). In its simplest form, the trans-
mission T for a light of wavelength A across a non-absorbing cos <p= 21F (1-"';1+8F 2 ) (5)
fIlm of thickness d and refractive index n, is given by(6):
where F=(Tmax-Tmin)! (Tmax+Tmin)
~ = ~(_1_ +_1_) + ~(_1___1_) cos <p (l )
T 2 T min T max 2 T min T max In practice the precise value of this point of maximum sensitivity
would be difficult to reach exactly, and to simplify the
wher,e <p =4~ n d I A (2) discussion, we consider the point near cos<p=O : this point is half
way between a maximum and a minimum of transmission and
For a free standing film of refractive index n in air, the maxima corresponds to a sensitivity which is not too far from the
and minima of the transmission are given by: optimum value and is quite representative of what is obtained in
practice. For this point, we obtain to the first order in the
variation of d:
2n
=(-2)
2
T max =1 T min (3)
1 +n ~T=4 ~ n ~d T max - T min (6)
T A T max+ T min
In fact, the film is deposited on a transparent substrate (glass or
fused silica) and the actual values are more complicated(2)' But Taking T I ~T = P == 20, the optical signal-to-noise ratio; we
the refractive index of amorphous silicon (around 3.7) is much obtain (for A= 800 nm and n = 3.7) a resolution ~d == 20 A. If
larger than that of the substrate, and the correction of a few the average thickness of the sample does not satis(y
percent is irrelevant for the present discussion. approximately the condition cos <p =0, the sensitivity decreases
The variation of the transmission T with the wave-length A and becomes zero around cos <p =±1. To avoid this difficulty, it
(transmission spectrum) or with the thickness d of the film is is then necessary to have a series of optical filters centered at
mostly due to the variation of cos <p in Eq. (1) : we neglect here different wavelengths, so that if accidentally one of them
the small variation of the refractive index n with A, which is not corresponds to a value of cos <p too close to ±l, the next one
important in the present study. In this work, we measure the will give a value of cos <p close to zero. For a film of amorphous

154
 the low-contrast interference is then obtained by taking the ratio,
"pixel" by "pixel" ("pixel" is the short term for "picture
element"), of the film image by the teference Im~ge. It IS to be
1.0
remarked that in this operation, the randox,n nOise of .the tv.:0
0.9 images add up quadratically, so that the slgnal-to-nOise rat~o
0.8 decreases by a factor ..f2, resultit,lg in the same decre~se m
z 0.7 sensitivity. Besides, in order for thiS .double-beam correction to
0 be significant, it is clear that the expenmental set-up must be very
;;; 0.6
stable both optically and mechanically. . .
V>

ior. 0.5 2°) It is possible to reduce the nOise of the recorded picture
Z 0.4 by signal averaging. This can be perf~rmed, in the simple~t way,
<
;:.: 0.3 by recording a certain number, N, of1l1~ages and then ~akl.ng the
:- 0.2
average, again pixel by pixel, of the N Images: the nOise IS then
reduced by a factor --IN. However, since each image, of 640x480
0.1 pixels of 256 levels of gray, uses more than 300 Kbytes: we .end
up with a very large amount of data to handly, takmg mto
600 700 800 900 1000 1100 account that the signal treatment has to be applied to the reference
WA ELE 'CHI A (nm) images as well. . . . c

If a resolution of 640 by 480 IS not necessary, which IS tIre

case for a reasonably smooth film, we can reduce the random
Fig. 1. Transmission spectrum of a film noise of the camera by signal averaging within one image. ThiS
of hydrogenated amorphous silicon oj is realized by grouping p successiye lines, and similar~y q
thickness d=0.92 p and optical gap of 1.74 eV. successive columns. A pack of pxq pixels of one group of hnes
The film is transparent (ad or 2%) for
and one group of columns is now considered as a ~it,lgle ~ixel,
A>750 nm. The spectrum has been corrected with a value which is the average value of the pxq ongmal pIXels.
for the effect of the glass substrate, so that the The noise is then reduced by --Ipxq, at the expense of the
maximum transmission is Tmax=1. resolution of the picture which is now reduced to 480/p lines by
640/q columns.
We show in Fig. 4 and Fig. 5 a typical result of such an
image processing applied to a film of hy~ogenated amorphous
silicon of about 10xlO cm2 and of thickness d=0.81~. As
silicon of thickness of about d == l~, for example, Eq. (2) shows calculated above, the starting resolution was <id=20*--I2=28 A.
that two filters differing by <i A. == 20 nm would be adequate. In By taking the average of N=4 images, we gain a factor of 2 and
practice, a small number of filters (5 ?r 6) is sufficient for the by reducing the resolution to 160 by 120 (p=q=4) we gain an
study of a series of films of different thickness. .. other factor of 4 so that the final resolution is <id=3.5 A, which
The experimental arrangement, as shown schen:atlcally In is hardly more than the thickness of one atomic layer.
Fig. 3, is very simple in principl~, but reqU1re~ some
precautions. As will be shown below, With t!'tese pre~~utlons and
after some corrections and treatment of the Images, It IS possible
to reduce significantly the systematic errors and the noise, and
thus to push the resolution down to a few Angstrom, i.e. close to
one atomic layer.
1°) The illumination of the sample must be very
homogeneous. However, since ~e are d~aling ~ith contr~sts of
a fraction of percent, it is very difficult In practice ~~ ~chleve a
sufficient optical homogeneity. In any case, the sensitivity of the CCD
... CAMERA

10
CCD retina of the camera is not specified to be uniform to bener

~
than 1% over its whole surface. It is then necessary to perform a
correction which in fact amounts to a "dou~le-beam" experiment:
two pictures, one with the sample ("film" image) and. one
without the sample ("reference" image) are taken succ.esslvely II II
and recorded in the memory of the computer. The actual Image of HALOGEN
LAMP
~ YSAMPLE
OPTICAL
t
DIFFUSERS FILTER

Fig. 3. Experimental arrangement

(schematic). An image of the low-contrast
interference pattern in the film, as "seen" by the
CCD camera with a resolution of 640 by 480
pixels and 256 levels of gray, is stored in the
computer for further treatment. The purpose oj
the optical diffusers ("opaline" glass) is to
make the illumination of the sample as uniform
as possible.

4. CONCLUSION

The precise mapping of the thickness of a thin film of

amorphous silicon, as presented in this article, was found very
useful for optical studies of the material. It allows in particular a
rapid determination of the stationary point of the film for a
Fig. 2. Schematic variation of the
precise determination of the dispersion n(A.) of the materia~.
transmission, for a given wavelength A, as a
which gives a direct information on its band structure(1,3). But It
function of the thickness of the film (Eqs. (1)
provides also a direct diagnosis of the deposition process. In a
and (2». The sensitivity for the measurement capacitive glow-discharge reactor, for example, some
of the hamogeneity of a sample is proportional optimization of the inter-electrodes distance is required: if the
to the slope ,1TI,1d of the curve.
distance is too large, the thickness of the film is larger in the
centre; if the distance is too small, the thickness is larger on the

155
Sides of the sample. A picture like that shown in Fig. 4 or Fig. 5
provides a rapid and convenient "signature" on the functioning of
the reactor. The method can. be used for any device that requires
the deposition of a reproducible and homogeneous thin film on a
flat substrate: for example, the fabrication of photovoltaic panels o
based on thin-film technology (this was the main motivation for
the present study), but also the deposition of conducting
transparent electrodes on glass, or the control of flat screens
based on the electroluminescent effect

5. REFERENCES

(1) I.Solomon, M.P.Schmidt, C.Senemaud and M.Driss

Khodja, Phys. Rev. B .18., 13263 (1988)

(2) J.Perrin,Thesis, University of Paris, 1978 (unpublished)

R.Swanepoel, J. Phys. E~, 1214 (1983)

(3) I.Solomon and M.Fang, to be published.
o
(4) R.Swanepoel, J. Phys. E 11,901 (1984)
(5) J.Perrin, I.Solomon, B.Bourdon, J.Fontenille and
E.Ligeon, Thin Sol. Films, Qb 327 (1979)
(6) See for example: M.V.Klein, Optics , publisher John Fig. 5. The same film as in Fig. 4 is
Wiley and Sons (New York, 1979) p. 205 mapped as contours of constant thickness. The
numbers on the contour curves are in
Angstrom relative to an arbitrary"zero" level.
We remark that the noise of a contour curve i~
about 0.10-0.15 of the distance L!d=25 A
between two conto!'rs so that the thickness
resolution is 3 'to 4 A.

100 A
soA
o

Fig. 4. "Perspective" view of the variation

of the thickness of a film of hydrogenated
amorphous silicon of average thickness
d=0.8111. Note that the scale on the z-axis is
about 1()6 times larger than on the x and y-axis.
The thickness is larger in the centre and on the
sides of the sample, which corresponds to the
optimization of the inter-electrodes distance in
the glow-discharge reactor.

156
lOTI! EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S--12 APRIL 1991 LISBON, PORTUGAL

DOPING EFFICIENCY OF BORON TRIFLUORIDE IN THE PREPARATION OF P-TYPE AMORPHOUS SILICON CARBIDE THIN FILMS

J,J. Gandia, M.T. Gutierrez & J. Carabe

Institute de Energias Renovables - CIEMAT
Avda. Complutense, 22. E-28040 Madrid - Spain

ABSTRACT. The properties of two sets of samples, prepared using different RF-power
densities and SiH 4 /CH 4 proportions have been analysed as functions of BF~ percent~
ge with respect to SiH 4 . In both cases, the optical gap EG, after a slignt initial
decrease, remains at a value of approximately 2.1 eV withou~ quenching in the do-
ping ranges covered. At a high RF-power density of 88 mW/cm using a gas mixture -
containing 37% CH A, conductivity improves only_eor dop~~t concentrations not beyond
9%, where it reaches a maximum value o~ 1 x 10 (ocm) • On the other hand, when
a smaller RF-power density of 44 mW/cm is applied to a gas mixture with 26% CH ,
no cond~ctivitr saturation is observed up to 15% BF • The best value obtained i~ -
2 x 10- (Ocm-). IR spectra allow to associate the~e features with film structu--
ral quality.

INTRODUCTION ternary (SiH 4 +CH 4 +BF 3 ) gas mixtures has been inves
tigated and correlated with optical and structural
I t is well knO\ll1. (1) that efficiency optimisa-- properties.
tion of amorphous-silicon-based solar cells of
structure GLASS/TCO/p-i-n/METAL (where TCO stands EXPERIMENTAL PROCEDURE
for transparent conductive oxide) requires a wide-
bandgap p layer. A high optical gap is associated P-type a-Si:C:F:H thin films have been prepared
with a good transparency and therefore with the mi by RF-glow discharge of SiH 4 +CH +BF gas mixtures
nimisation of optical losses. In combination with- in a PECVD (plasma-enhanced che~icar vapour deposi
a low activation energy of the extended-state con- tion) flat-bed reactor with two 38-cm-diameter alu=
ductivity, it provides a high built-in potential, minium electrodes 1.5 cm apart. Pressure was kept
thus improving open-circuit voltage. On the other at 1000 mTorr and substrate temperature at 250°C.
hand, a relatively high conductivity is necessary Two sets of samples have been made with different
in order to maintain the cell series resistance at RF-power densities and CH gas proportions, syste-
low values as a consequence of a low contact resis matically varying dopant gas concentration. Table
tance. 1 summarises the preparation conditions used for -
With the aim of obtaining wide-bandgap window la either of these sample sets.
yers, many authors have prepared different alloys-
by glow-discharge of mixtures of silane (SiH ) and SET A SET B
different gases containing mainly carbon ang ni-
trogen (2). Good results have been obtained (3) P (mTorr) 1000 1000
with hydrogenated amorphous silicon carbide (a-Si:
C:H) films prepared using methane (CH ). As a draw T roc) 250 250
back, the experimental analyses reveaf a certain - RFP (mW/cm2) 88 44
degradation of the electronic quality of the films
associated with alloying mechanisms. An important ~(SiH4) (sccm) 57 28
consequence of this is a decrease in doping effi-
ciency (2) with respect to a-Si:H. This feature is C{CH 4 ) (%) 37 26
highly dependent on deposition conditions.
Otherwise, since Ovshinsky and Madan (4) repor- C{BF 3 ) (%) 0-70 0-15
ted their first results on fluorinated amorphous
silicon alloys with good properties such as low Table 1: Preparation conditions
density of gap states, high doping efficiency and
no photostructural changes, many authors (5,6) ha- Methane and boron trifluoride concentrations
ve investigated the properties of a-Si:F:H and con (C(CH 4 ) and C{BF 3 ) are defined according to the fo
firmed its suitability for photovoltaic applica--- llowing expressions
tions and in particular for making cells showing C(CH 4 ) 100 x \I>(CH 4 ) / [\I>(SiH 4 ) + \I>(CH 4 )) (1)
no light-induced degradation. According to Matsumu
ra et al. (7), the incorporation of a certain - C(BF 3 ) 100 x ~(BF3) / \I>{SiH 4 ) (2)
amount of fluorine in the lattice results, in cer-
tain conditions, in better electrical properties. where ~ stands for flow rate in sccm.
Furthermore, Asano and Sakai (8) have reported re-
suIts on a-Si:C:H samples doped with boron trifluo The UV-VIS-NIR spectral transmittances and re-
ride (BF 3 ), according to which this gas shows a b~ flectances of the films have been measured in the
tter doplng efficiency than B H • range 400 - 2500 nm with a Perkin Elmer Lambda 9 -
Taking into account all tJe~e considerations, - spectrophotometer. From these data, the optical --
BF3 has been used as an alternative dopant gas with thicknesses and refractive indexes have been dedu-
the aim of improving properties such as stability ced using a combination of the methods reported by
against light exposure and doping efficiency. In - Aranda et al. (9) and Goodman (10). The absorption
the framework of a research plan for the prepara-- coefficients have been obtained according to (II),
tion of optimum p-type widow-layers, the influence and their Tauc (12) plots (indirect transition) ha
of RF-power density on the doping efficiency of BF ve yielded the optical band gaps.
for the preparation of hydro-fluorinated amorphous 3
silicon carbide alloys (a-Si:C:F:H) from

157
 The samples have also been electrically charac- bonded boron. Such results are in accordance with
terised. Dark- and 1-sun-AM1 ambient temperature - the IR transmission spectra, shown in figure 4,
conduct ivies have been measured on films deposited which, besides the already mentioned a-Si:C:H tyEi
on Corning 7059 glass, using two aluminium evapora cal features, reveal the appearance of a 1010 cm
ted coplanar electrodes 1 mm apart. The extended-- band attributed to SiF 4 groups. The SiH wagging and
state-conductivity activation energies have been - strectching bands do not disappear, indicating --
determined from conductivity Arrhenius plots. that the basic structure of a-Si:C:H is not appre-
The results have been analysed taking into --- ciably distorted by the addition of BF3 to the gas
account structural information form IR transmission mixture.
spectra. Beside the stabilisation of EG at a high value
in all our films, it is also remarkable that our
RESULTS AND DISCUSSION best sample has an ~l7tical g;~P of .2.1 eV and a co~
When no doping gas is added to the mixture, the ductivity of 2 x 10 (Ocm) . These properties co
conditions used for the preparation of set A (ta- rrespond to a suitable window-layer for amorphous-
ble 1) yield undoped films with optical gaps as -- silicon solar cells and are comparable to or even
high as 2.3 eV. The effect of doping in the opti- better than those commonly reported for B2H6-doped
cal and electrical properties can be seen in figu- a-Si:C:H, especially concerning otpical gap.
re 1, where the optical gap EG and the dark- and
AMI-conductivities are represented as functions of CONCLUSION
the BF concentration. The most outstanding featu
re sho~ is that, after a slight initial decrease, The influence of RF-power density on BF doping
E keeps at a high value of approximately 2.1 eV efficiency in the preparation of a-Si:C:F:J3by RF-
atl along the doping range covered, up to 70% BF • glow discharge has been investigated. The applica-
This is an important result when compared with t5e tion of moderate RF-power densities has a positive
widely reported optical gap quenching due to do- influence on the doping efficiency and on the struc
ping. The evolution of the conductivities corres- ture and electrical properties of the films. No op-
pond to a sharp transition from a nearly intrinsic tical gap degradation is observed for wide ranges-
photosensitive material to a doped one for which of BF3 concentration. The incorporation of fluori-
dark- and AMI-conductivities coincide. Maximum con ne to the lattice when doping with this gas has -
ductivity is obtained for a dopant gas concentra-- been experimentally observed. The possibility to -
tion of about 10%, while- for higher concentrations, make good a-Si:C:F:H window-layers adding BF3 to -
the films are not only less conductive, but also SiH 4 +CH 4 mixtures has been demonstrated. Our best
more photosensitive, revealing that the further - films so far have ~70ptica!lgap 2.1 eV and a con-
addition of BF3 does not contribute to increase - ductivity of 2 x 10 (Ocm) • This B2H6 • Further
the concentrat~on of dopant impurities in the la- investigations about the influence of process pre-
ttice. Furthermore, these results evidence a de- ssure, substrate temperature and other parameters
crease in doping efficiency above 10% BF that we are necessary in order to optimise the material.
associate with film structural damage in~uced by Furthermore, the assessment of its stability aga-
the excess of dopant gas. This is confirmed by fi inst light exposure would be of great interest.
gure 2, where the evolution of the IR transmission
spectra with doping is shown. The spectrum of the ACKNOWLEDGEMENTS
undoped sample is that of a typical a-Si:C:H film,
with the SiH waggi~T and stretching bands shifted This research has been carried out in the frame
to 650 and 2080 cm respectively, due to the pre work of the Community Research JOULE Programme, -
sence of carbon in the lattice (13), ~~d absorp-- contract number JOUR-0077-C, through which approxi
tion bands centred at 770 and 1000 cm ,attribu- mately 15% of its funding has been provided. The --
ted to SiC and CH bonds (14). As BF3 is added to Spanish CICYT has also partly sponsored this work
the gas mixture, the SiH wagging band is quickly by means of contract No. MAT89-1141-CE.
replacled by other bands corresponding to the in-
corporation of fluorine and boron to the lattice REFERENCES
in_~ifferent ways. These ban~~ are centred at_t015
cm (SiF 3 , (SiF 2 ) ), ~~O cm (SiF 3 ), 828 cm - (1) S.Guha, J.Yang, P.Nath, M.Hack. Appl. Phys.
(SiF) and 1400 n cm (SiB) (this last one only Lett. 49 (1986) 218.
for the highest dopant concentration).
The analysis of these data allows to~onclude
(2) G.H.Bauer, G.Bilger, H.D.Mohring, C.E.Nebel,
that for an RF-power density of 88 mW/cm in the S.M.Paasche. Proc. 6th EC PV Solar Energy
conditions used, BF3 concentrations not beyond 10% Conf. London. Ed. W.Palz, F.C.Treble. Reidel
should be used, as the further addition of dopant (1985) 660.
gas determines the incorporation of fluorine and (3) D.Kuhman, S.Grammatika, F.Jansen. Thin Solid
boron in such a way that doping efficiency decrea- Films 177 (1989) 253.
ses and electrical properties degrade.
Set B was then prepared in similar conditions (4) S.R.Ovshinsky, A.Madam. Nature 276 (1978) 482
but applying an RF-pow2r density of one half of the (5) H.Matsumura, S.Furukawa. JARECT Amorphous Se-
previous one (44 mW/cm ) with a CH 4 concentration miconductor Technologies & Devices. Ed.
of 26%. The undoped sample has an optical band gap Y.Hamakawa. OHMSHA-North Holland (1982) 88
of 2.3 eV. As for the films prepared at high RF-
(6) M.Janai, R.Weil, B.Pratt. Phys. Rev. B 31
power density, the optical gap is maintained at a
(1985) 5311
constant value of approximately 2.1 eV after a
small initial decrease (see figure 3). On the (7) H.Matsumura, T.Uesugi, J.Ihara, J. Non-Cryst.
other hand, the conductivities also evolve in Sol. 77 & 78 (1985) 841.
accordance with the incorporation of dopant impu-
(8) A.Asano, H.Sakai, Appl. Phys. Lett. 54 (1989)
rities, i.e., increasing and converging. No rever
904.
sal of this behaviour is observed up to 15% BF .-
This reveals that the application of a lower ~F­ (9) J.Aranda, J.L.Morenza, J.Esteve, J.M.Codina.
power density improves the dopability of the films. Thin Solid Films 120 (1984) 23.
Please remark that the best_gonduct~yity at high
(10) A.M.Goodman, Appl.Opt. 17 (1978) 2779
RF-power (set A) was 1 x 10 (Ocm) ,while at :7
low R~lPower (set B) a conductivity up to 2 x 10 (11) K.L. Chopra, S.K.Bahl. Phys. Rev. B1 (1970)
(ocm) is obtained. This is also an evidence of a 2545
more efficient incorporation of tetrahedrally -

158
(12) J.Tauc, R.Grigorovici, A.Vancu, Phys. Stat.
Sol. 15 (1966) 27.
(13) S.Ray, D.Das, A.K.Barua. Sol. En. Mat. 15
(1987) 45
(14) P.J.Zanzucchi. Semiconductors & Semimetals
Vol 21B. Ed. J.I.Pankove. Academic Press
(1984) 113

EG (eV) log a (0 em)-l

2,4 -7

2,1 -9
•
1,8 -1 1

1,5 -13
o 20 40 60 80
C(BF3) (%)

Fig. 1.: Evolution of optical band gap and dark- and AM1-
conductivities of samples prepared applying 88
mw/cm 2 to ternary gas mixtures containing 37%
CH4 •

0%
9%
T (w) 18%
(a .u.)
35 %

70%

3000 2000 1500 1000 500

w (em -1)

Fig. 2.: IR transmission spectra of the samples of fig. 1.

BF3 concentration as indicated.

159
EG (eV) log 0 (0 cm)-l
2,5 -5
2,3 -7

2,1 -9
1,9 - 11

1,7 ~------~-------L------~-13

o 5 10 15
C(BF3) (%)

Fig. 3.: Evolution of optical band gap and dark- and AM1-
conductivities of samples prepared applying 44
mW/cm2 to ternary gas mixtures containing 26 %

0%

5,4 %
T (w)
6,8%
(a .u.)
9,3 %

14,7%

3000 2000 1500 1000 500

w (cm-1)

Fig. 4.: IR transmission spectra of the samples of fig. 3.

BF3 concentration as indicated.

160
10TH EUROPEAN PHoroVOLTAIC SOLAR ENERGY CONFERENCE S--12 APRIL 1991 LISBON, PORTUGAL

STATIONARY PRIMARY PHOTOCURRENTS FOR THE CHARACTERIZATION

OF a-Si:H PIN-DIODES

C.-D. ABEL, H.R. PAES", G.H. BAUER

Institut fuer Physikalische Elektronik, Universitaet Stuttgart

Pfaffenwaldring 47, D 7000 Stuttgart-SO, F.R.G.
"COPPE!PEMM Universidade Federal, Rio de Janeiro, Brasil

ABSTRACT. Carrier transport in amorphous silicon pin-diodes has been analyzed by charge collection in transient
and in steady state mode. Whereas transient experiments yield JLT-products governed by deep trapping life time,
from steady state collection of electrons excited by different photon energies locally resolved JLT-products in
pin-structures have been evaluated reflecting recombination life time. A considerable decrease of JLT in undoped
bulk material in the neighborhood of the p/i-interface has to be attributed to the shift of FERMI-level towards the
valence band, which means an intrinsic effect, not necessarily related to a higher defect density at the interface.

1. INTRODUCTION port is provided by deep trapping and not by recombination

since in ideal reverse biased barriers prohibiting injection, no
The product JLT of mobility and life time of photoexcited partner for recombination is available, and
carriers is a meaningful feature for selecting semiconductors
with respect to optoelectronic applications. In particular in ii) due to the fact that carriers are used to travel across the
amorphous silicon based pin-diodes JLT-products of electrons entire sample thickness, and thus in terms of current decay do
and holes as well, in the undoped layer which serves as ab- not only reflect an integral figure in depth but in addition in
sorption regime and moreover provides transport of both the presence of band tails like in a-Si:H are used to drop to-
types of carriers, represent key figures for a satisfactory wards rnidgap as time propagates, an average figure indicating
function of these devices. Generally JLT-products have been the interaction with tail states is superimposed. As the ensem-
derived from several analyses for instants from charge collec- ble of carriers in dispersive transport moreover is locally
tion measurements in reverse biased barriers either in statio- spread, transient analyses will hardly signalize clear local
nary or in time-resolved single carrier injection experiments features in the density of states like interfaces (see fig. 1,
(primary photocurrents), by evaluation of voltage dependent upper part).
ratios of collected carriers at the rear contact related to photo-
generated ones close to the front contact [1-3]. In steady state charge collection, however any deficit in
carrier collection against generation - obeying continuity re-
In a-Si:H pin-diodes also slight modifications of the con- quirements - has to be balanced by recombination. So in
ventional charge collection approach either in transient or in Hecht's equation, T gets a different meaning, namely recombi-
steady state mode have been performed in order to get deeper nation life time. On the other hand, in ideal reverse biased
insight in carrier transport behavior through the i-layer [2,3], diodes, due to negligible injection of carriers from the con-
carrier transport across the p/i-interface [4,7,9], or in electric tacts, consequently incomplete charge collection must not be
field distribution [6,9]. attributed to carrier life time in the bulky transport region, but
the more to recombination in the region of photogeneration
In time resolved modes like in modified time-of-flight where both types of carriers are present. Outside this region,
analyses the ratio Q/Qo of charges collected within a certain according to continuity requirements and linear sample geo-
time window of the experiment can be represented by the metry, div(i)=0 has to be satisfied (see figure 1, lower part).
HECHT-equation [1] as a function of reverse bias voltage Thus the analytical relation for charge collection in steady
V rev which is assumed to drop linearly in the sample: state mode, according to these considerations, has to be modi-
fied.
Q(V rev)/Qo =TdJLV rev L-2(1-exp(-L2/JLTdVrev » (1),
In the case of negligible space charge, say negligible
where Qo' L, and I' represent photogenerated charge, sample departure of the quasi FERMI-level from the equilibrium
thickness, and mobility. The life time T d reads in deep trap- position in the transport regime, carrier motion in steady state
ping life time, determined by the density of deep defect states charge collection over the entire sample thickness is per-
Ndb and their respective capture cross section udb' from formed with a fairly constant mobility 1'.
which thermal reemission within the experimental time win-
dow is unlikely: Td=(V1hUdbNdtJ-1. The mobility p. represents In steady state charge collection additionally a very sensi-
a number averaged over the mobilities of the energetic posi- tive spatial resolution can be achieved by variation of the ab-
tion of the transport path which during motion of the carriers sorption depths via different photon energies.
to the rear contact more and more drops into the band tail
(see fig. 1, upper part).
2. EXPERIMENTS
Two important features of transient charge collection ex-
periments are worthwhile to be mentioned: Charge collection experiments in transient as well as in
steady state mode have been performed on one type of barrier
i) the extraction of carriers from contributing to charge trans- consisting of a layer sequence such as glass/ITO/p + -a-SiC:H

161
(10 nm)/i-a-Si:H(3.9/km)/n + -a-Si:H (20nm)/Al (fig.2). The
metallic rear contact to some extent has been semi-transparent lra nsienl
in order to enable hole charge collection for comparison of hI!
space charge effects and to check for saturation of collection.
Reverse bias currents of these diodes have been very low '\J\J'+
« 5xlO- 11 A), so that recombination of photogenerated carri- llg h t pu l e
ers via leakage current carriers has been negligible. (al 0)

3. RESULTS AND DISCUSSION sLale

In figure 3 electron charge collection versus reverse bias

voltage is outlined, calculated from transient experiments. The
corresponding log(i)-Iog(t)-plots are shown in the inset. The
evaluation of /kTd via equ. (1) yields </kTd> =5xlO-7cm2/V. d lV o
From steady state charge collection versus reverse bias x
voltage in principal </kTr> can be deduced (fig.4), however, L
the influence of photon energy/absorption depths has to be
investigated carefully (see fig.5) hence:
Fig. I: Upper part: spatial-temporal propagation of carrier
on the one hand, an increase in photon energy decreases the pulse in transient charge collection (pin-diode). Lower part:
photogeneration regime and basically the main condition for spatial distribution of photogeneration and recombination of
single carrier injection gets better satisfied, but in diodes in carriers in steady state charge collection (pin-diode).
general and in particular in pin-barriers the regime which
photogenerated charges are extracted from, shifts closer to the
interface of doped/i-layer (for electron collection to the p/i-in-
terface), where the built-in-field contributes to charge separa-
~
tion and may mask the effect of the externally applied vol- Po.
tage. Moreover interface states at the p/i-interface will contri- o
o
bute to recombination and incomplete charge collection. '"
~
'0
On the other hand an increase in absorption depth by
hI! o'"
applying long wavelength photons decrease the effective layer
thickness for the type of carriers (electrons) probed to travel
through the sample, and simultaneously the other type of car- light puis
rier (in this case holes) is involved into charge transport,
directed to the front contact. In fig. 6 </kT> *-products ver-
sus reverse bias voltages at different photon energies/absorp-
tion depths are drawn, deduced from modified HECHT' s-equa-
tion:
(2) [5]

in a traditional manner, without correcting for photon energy

effects, mentioned above.

For a fair and more realistic evaluation of wavelength hI!

dependent steady state charge collection data, we propose a
subdivision of the sample thickness L into
i) a region 0 < x < Xc where generation rate g and recombina-
slationary
tion rate r don't vanish and both are assumed to be represented
by a constant average number <g> =go, <r> =n/Tn=p/Tp'
and
ii) into a region xc>x>L with g=O, and r=O, only serving
for transport. From one-dimensional continuity equation in
x<xc we get

(3),

where q, A,n, < Tn> represent elementary charge, contact Fig. 2: Pin-diode for transient (upper part) and steady state
area, mean density, and average life time of electrons. Since (lower part) charge collection.
in x < Xc electrons and holes are present and both contribute
to the current, we write

iphot=qA(n/kn + p/kp)E = qA(goxc-nxc/ < Tn» (4),

assuming constant electric field strength E. Combining equs.

(3) and (4) yields

162
 5
...--.. 4
en " (nm)
< 4
C\I
3 4V 400
I -4 .--.. 450
a...-< < 3
2 3: c: 500
550
a
v QV -5 .g, 2 600
1 .3 650
1 700
0 -7 -6
log t (sec) o
-5 0 5 10 15 20 25 30 35
0 2 4 6 8 10 Rev (V)
URev (V)
Fig. 6: Steady state electron charge collection vs. bias
Fig. 3: Transient electron charge collection vs. bias voltage voltage in an a-Si:Hpin-diode at different excitation
in an a-Si:H-pin-diode (300 K); inset shows transient currents wavelengths.
i=i(t).
-6

• ••••• • • • •
• 4~
,-.., 4
<
r:::
..c::
Cl.
3 •
.- 2 o K=O.9
on ... K=O.99
1~ o
-9 +--t---+--t---+--t--t---t--'
• o 1000 2000 3000
- 1 0 123 4 5
Xc (nm )
URev (V)
Fig. 7: Electron <Jl.7> ",,-products vs. absorption depths in
Fig. 4: Steady state electron charge collection vs. bias an a-Si:H-pin-diode (300 K)
voltage in an a-Si:H-pin-diode (300 K).

(5),
-5.5
with io = goxc indicating the saturation current, and
--
N
r'
- 60
A (nm)
650
I: 600
U 550
We express the fraction of photogenerated charges within
r.
- 6.5 O<x<xc by K,
500
l- 450
::!.. %.(1)
.:::::.. -7 0 400
Jgo exp(-ax) dx
00 lC = --:-0_ _ _ _ __ (6)
0
L
-7.5 Jgo exp( -ax) dx
0 10 20 30 o
URev (V) and we get Xc =-(1/ a)ln(1-K + Kexp( -aLl). (7)

Fig. 5: Averaged <p:r> '"-products of electrons in an Now we are able to plot the averaged Jl.7-products deduced
a-Si:H-pin-diode from steady state charge collection at from saturated electron collection named < Jl.7 > "" over the
different photon energies. generation depth xc' playing the parameter K in realistic ran-
ges of close to unity, say K=O.9-0.99 (see fig.7).

163
 As < p.r > 00 is related to saturated collection under suf- 4. CONCLUSION
ficient high electric field, a homogeneous potential drop
across the sample can be assumed; the decrease of <p.r> From charge collection experiments in amorphous silicon
with decreasing thickness of the generation regime Xc is not pin-diodes, in particular from steady state charge collection
entirely attributed to the influence of additional defects at the under different photon energies for carrier excitation, p.r-pro-
p/i-interface. In addition close to the interface also the shift of ducts with good local resolution in the bulk and in the neigh-
EF from slight n-type behavior in the bulk towards midgap at borhood of the p/i-interface can be deduced. The drop of
the p/i-interface decreases < p.r > e remarkably whereas by the p.r-products in the intrinsic layer close to the p/i-interface at
corresponding rise of <p.r>h the total drop of <p.r> as a least to a considerable degree results from the shift of FER-
sum of electron and hole contribution analogously to secon- MI-level towards midgap. Consequently this detoriation of the
dary photocurrent experiments <p.r> = <p.r> e + <p.r> his material has to be accepted as an intrinsic effect in a-Si:H pin
not completely compensated. (see fig.8 and [8]). structures, which by optimization of material certainly can not
be avoided. The only means for a reduction of its unfavorable
These arguments lead to a schematic picture for tile distri- consequences on transport of photoexcited charges might be
bution of p.r-products of electrons and holes as well as of the an optimized structure of buffer layers.
sum in an a-Si:H pin-diode as outlined in fig.9.

The rise of <p.r> '"-products for short wavelength illumi- Acknowledgement

nation and low reverse bias voltages, not providing saturated Financial support from Fed. German Ministry for Research &
collection obviously results from support to charge separation Technology by contract 032 8327-D is gratefully mentioned.
by the built in field; analogously for long wavelength photons
().. > 550nm) and low voltages, the field in the bulky region
which partially is generation regime for complete collection
surely is to weak. REFERENCES

[I] K.Hecht, Zeitschr. f. Physik 77, 235 (1932).

4
->
[2] R.S. Crandall, J. Appl. Phys. 53, 3350 (1980).
[3] H. Okamoto, H. Kida, S. Nonomura, Y. Hamakawa,
~

5
Solar Cells 8,317 (1983).
(\J
6 [4] Y. Yamanaka, M. Konagai, K. Takahashi, Conf.
E Rec. 20 IEEE PVSC, IEEE, New York, 1988, p.l60
0
7 [5] R.S. Crandall, K. Sadlon, J. Kalina, A.E. Delalloy,
MRS-Sympos. Proc. 149, 1989, p.423.
I- 8
::l.. [6] M. Hack, B. Shur, J. Appl. Phys. 58, 997 (1985).
v
-9 [7] B. Yan, L. Shi, Y. Liu, X. Geng, Techn. Dig. Int.
t>J)
Conf. PVSEC-5, Kyoto, 1990, p.243.
0 LO [8] J. Kocka, C.E. Nebel, C.-D. Abel, Phil. Mag.B 63,
221 (1990).
- 11
[9] R. Konenkamp, S. Muramatsu, H. !toh, S.Matsu-
0.4 0 .6 0.8 10 12 1.4 bara, T. Shimada, ref.[7], p. 239.
Ec - EFo ( V)

Fig. 8: Electron and hole p.r-products of a-Si:H vs. position

of FERMI-level from steady state photoconductivity
experiments (300 K).

-5

>-
~

N"- - 6
E
u -7
A
I-
::t - 8
v

QlJ -9
.3
- 10
x
0 L

Fig. 9: Schematic spatial distribution of electron, hole, and

total p.r-products in an a-Si:H-pin-diode according to fig. 8.

164
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CORRELATION BETIIEEN DI8CHARGB GBOMB'l'RY AND FILII PROPERTIES

IN PB-CVD OF a-8i:8

D. Mataras, D. Rapakoulias
Department of Chemical Engineering, University of Patras, Greece
and
P. Kounavis, E. Mytilineou
Department of Physics, University of Patras, Greece

ABSTRACT. In this paper we present some evidence of direct correlation between interelectrode
distance and the characteristics of a-Si:H films deposited in an RF silane glow discharge. The
variation of interelectrode distance influences both quantity and quality of a-Si:H in terms
of deposition rate and defect density. The explanation given to the observed phenomena, based
on these results and also the analysis of our previous work combined with theoretical
calculations, points out the correlation between plasma microscopic parameters, like radicals
denSity and distribution, and the film properties. In this manor ground and excited state SiH
spatial radical profiles and intensities are used to outline this relation. This is achieved
by using coupled Laser Induced Fluorescence and spatially resolved Optical Emission
Spectroscopies for the in-situ, non-intrusive detection of radicals.

1. INTRODUCTION Laser Beam

Silane glow discharge, as the primary method of
deposition of device quality a-8i:H, has been the
subject of intense research activity during the
last few years. This is due to the continuously
growing number of applications of the material and
its alloys. The understanding of the relation
d
between external deposition parameters and the
microscopic plasma characteristics, and consequent-
ly the properties of a-Si:H films, is of primary
importance in the effort to understand and control
the deposition process. In this work we present
RF~ G
some first results in this direction, concerning
the influence of interelectrode distance to the
microscopic characteristics of the discharge and
the film deposited under these conditions.
x
To obtain this information we have applied
Laser Induced Fluorescence coupled with spatially
resolved optical emission spectroscopy for the
detection of ground state and excited state SiH
radicals, and optical absorption together with CPM,
dark and photo conductivity measurements, for the
investigation of the film quality. We have recorded
entire axial profiles of the concentration of these
radicals as well as the variation of their concen-
tration at areas of specific interest.
As it will be shown the results exhibit an
interesting relation between the geometriC char- Figure 1 Schematic presentation of the
acteristics of the discharge and some parameters experimental setup.
that characterize the film.
The deposition electrode is heated to 250 C while
all the films examined here were deposited from
2. EXPERIMENT pure silane. The properties of the films, deposited
on Corning 7049 glass at 250·C, were measured by
The experimental apparatus and procedure is de- the Constant Photocurrent Method (CPM), and optical
scribed in detail elsewhere /1/. It consists mainly absorption by the reflectance-transmission method,
by a round stainless steel deposition chamber whereas in the low absorption region their response
equipped with quartz windows for the application of was estimated from CPM measurements. Additional
spatially resolved emission and LIF spectroscopy, tests included thickness measurements, dark and
and two round parallel electrodes. The deposition photo-conductivity.
electrode is based on a linear motion vacuum feed-
through permitting continuous adjustment of the
interelectrode distance. The chamber is based on a 3. RESULTS AND DISCUSSION
carriage slider in order to adjust the sampling
position in the interelectrode gap. Thus, by moving In figure 2 we present spatial concentration pro-
the chamber back and forth one can obtain entire files of the SiH radical between the two elec-
one-dimensional intenSity profiles of both emission trodes, for three different interelectrode dis-
and fluorescence (figure 1). tances (d).

165
The curves present an intensity maximum, at a
distance of 10 mm from cathode, ~hich is insensi-
tive to the variation of d. In fact the position of Vdc
this maximum was previously shown to be independent 70 -6
of all the other externally accordable parameters
(pressure, power, flow rate, gas composition) in
the range that we have examined 111. Based on these 60 -7
data and on a series of additional experimental and
theoretical evidence we have previously demonstrat-
ed that these profiles represent the spatial 50
OR A atinn nrn~ilg~ ~nr ~" r~n;~~'e I? ~J
-8
40
-9
55 30
I
LlFi -10
20
45
+1 d~25
I ! 10 -II
* i d~30
35 ! !
01 d~35 o +--.------,--..,.------,-----,-l- - I 2
i !
10 20 30 40 50 60
i
!
i
I d (mm)
25
i
I
15 Figure 3 LIF and emission intensity as a
I function of interelectrode distance, com-
pared with the same function of Vdc (40
mTorr, 101'1, 11 sccm)
5
0 5 10 15 20 25 30 35
J
d

X (mm) ILIF,OES; I(x) dx

o

The ratio of relative intensities ILIF/10ES

decreases as the distance of the two electrodes
Figure 2 LIF intensity (I in arbitrary decreases (figure 4).
units) as a function of the position (X)
in the interelectrode gap. For three
interelectrode distances (d). P;50 mTorr,
1'1;12 Watts, f= 10 sccm pure silane. l(LI F)/I(OES) 1.4
700
This last observation permits us to record the
variation of intensity at the position of maximum
in order to study the influence of d in a more
extended range (figure 3). 1.3
As it can be observed in this figure d has a 600
different influence on the concentration of ground
state radicals compared to their excited state
counterparts. This can be observed on the rate of
increase of each radical as well as on the satura-
1.2
tion tendency of the emission curve.
Also, we can observe that the inflection
point of these curves coincides with the minimum
500
autopolarization potential of the RF electrode (Vdc ) 1.1
that indicates the optimum electrical symmetry of
the two electrodes under these conditions.
The conclusion that we have reached by
recording a large number of similar curves, in 400+---~--~----~--~1
various conditions, is that : the dependence of the
total LIF intensity on d, as compared to the depen-
20 25 30 35 40
dence of the overall emission intensity, clearly d(mm)
shows that in conditions of constant total power,
the increase of power density does not influence in
the same way high-energy (dissociative excitation,
J(LlF) --+-- I(OES)

-
ionization) and low-energy (dissociation, rovi-
bronic excitation transfer) processes.
Namely there is a redistribution of the J(Ll F)/I(OES)
specific power consumption in favor of high-energy
processes, as the two electrodes come closer 13/.
This can be clearly observed if we calculate the Figure 4 Variation of the overall LIF and
total intensity increase as a function of inter- emission intensity as a function of the
electrode distance as : interelectrode distance.

166
 This phenomenon is affecting the gas-phase
composition and consequently, together with silane
depletion that is also increased, can play an
important role in the rate and mechanism of film
formation. 30~-------------------------'
This is immediately observed in figure 5,
where we have plotted the deposition rate of anodic
a-Si:H film as a function of d. The curve corre- d (mm)
sponding to 50 mTorr presents a maximum at around 25
d=27 mm, whereas in the case of the 100 mTorr curve
there is a tendency for saturation beginning from ~25 ~30 _35
the same d. It is interesting to observe that this
distance coincides with the inflection point of the
corresponding LIF intensity curve and the minimum
Vdc (fig. 3).
On the other hand, when the two electrodes
approach, high sticking coefficient radicals as SiH
or 5iH2 have enlarged probabilities to reach the
anode surface not only because it is closer but
also because their diffusivity increases due to
silane depletion. Thus by reaching the surface they
compete with 5iH3 radicals and they modify in a
more or less deCisive way the film formation
mechanism.
The amount of 5iH radicals reaching the
deposition surface can be roughly estimated by
simple calculations, based on the total probability
of 5iH radicals to reach the surface as :
x-d
CATHODE ANODE
~=f
x-o

where k the rate of reaction with silane, N the Figure 6 Total % percentage of 8iH radi-
radical's density, and 5(x) the source of radicals cals reaching the two electrodes for three
which is the normalized LIF spatial profile. different values of d. Same conditions as
Thus the results of these calculations in in figure 2.
figure 6 indicate that anodic deposition, compared
to cathodic deposition, should be much more influ-
enced by the decrease of d. Otherwise the influence
of this effect is always stronger for the RF
electrode deposition because the source of radicals
is significantly closer to it.
The role of these high sticking coefficient
radicals in the quality of the deposited films is
essential because, due to their high sticking
1. 1 Dr (Angstrom/s) probability, they are incorporated in the film very
near or at the same location where they are first
adsorbed on the deposition surface.
This can lead to an amplification of the
surface roughness and to columnar structure that
0.9 would increase the number of film defects.
Thus as it can be seen in figure 5 there is
a direct relation between the defects density and
d, that is more pronounced at lower pressures where
the diffusivity of these radicals and their proba-
0.7 bility to reach the surface is higher. The density
of these defects, deep in the gap of the film
deposited under these conditions, is responsible
for the optoelectronic properties of the material.
This denSity can be indirectly estimated from the
results of CPM, at low photon energies /4/.
0.5 * 50 mtorr In general the position of the Fermi level of
the films deposited under these conditions varies
+ 100 mtorr between 0.55 eV and 0.7 eV with respect to the
conduction band. In this case photoconductivity
increases to about 4 orders of magnitude as the
0.3 +----,--..----,--,---.----,r-""T"""---r.....J posi tion of the Fermi level moves closer to the
)9 21 23 25 27 29 31 33 35 conduction band and presents a maximum for the
value of 0.6 eV. However a direct relation between
d (mm) photoconductivity and d is not yet clear due to the
scattering of values.

Figure 5 Deposition rate (Dr) as a function ... CONCLUSIONS

of interelectrode distance, for two dif-
ferent total pressures. W=9 watts, f= 11 In conclusion the fiim characteristics are directly
sccm pure silane, T= 250 °e, Vrf =54 volts. related to the influence of at least some external
deposition parameters. This occurs through the
modification of microscopic plasma characteristics

167
 Nd
D
I.OE+19 D 50 mTorr
o 100 mTorr

Figure 7 Number of defects (Nd in cm- ) as

a functiQn of interelectrode distance.

as the spatially distributed electron energy

distribution function, and the generation and
transport of active radicals in rf silane plasma.
The influence of interelectrode distance on film
properties is more pronounced at lower pressures
when geometric effects are more pronounced and the
diffusivity of radicals is much higher.

5. REFERENCES

Mataras,D., Cavadias,S., Rapakoulias, D.E.,

J.Appl.Phys. 66,119 (1989).
2 Mataras,D., Cavadias,S., Rapakoulias, D.E.,
in Mat.Res.Soc. Symp.Proc. 169, (1990).
3 Mataras,D., Cavadias,S., Rapakoulias, D.E.,
submitted for publication to J.Appl.Phys.
4 Pierz,K., Fuhs,W., Mell,H. J.non-Cryst.
Solids 114,651 (1989).

168
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DIAGNOSTICS OF VERY HIGH FREQUENCY SILANE PLASMAS

A. A. Howling, Ch. Hollenstein, J.-L. Dorier, P. Paris, M. Favre*

CRPP, Ecole Poly technique Federale de Lausanne,
2l Av des Bains, CH-1007 Lausanne, Switzerland
*IMT, Universite de Neuchatel,
2 Rue Breguet, CH-2000 Neuchatel, Switzerland

A large increase in deposition rate is necessary to reduce the cost

of solar cells and other devices based on amorphous silicon (a-Si:H).
Film preparation by Very High Frequency (~70 MHz) Plasma Assisted
Deposition has been shown to produce device-quality material at
~20A/s (1). The purpose of this work is to develop specific plasma
diagnostics to complement the research into amorphous silicon films
by identifying those plasma properties which affect film quality and
deposition rate.

In this paper, we briefly address the

following three topics:
l.Powder Production. A simple white light
scattering ,experiment yields a global,
spatio-temporal visualisation of powder
dynamics during plasma processing.
2.Plasma Effective Power. A measurement of
the rf power dissipated in the plasma is
crucial for a valid comparison between
different plasma parameters.
3. Lecher Wire Microwave Interferometry.
Free electron density is measured using
microwaves guided across the plasma by
Lecher wires, giving good spatial
resolution and avoiding stray reflections
from electrode surfaces.

1.VISUALISATION 0F POWDER DYNAMICS IN PECVD

1.1 Description
The plasma is recognised as a source
of particulate contamination in plasma-
assisted deposition (2,3). Here, the plasma
is illuminated with white light. Global,
spatio-temporal powder dynamics and
particle trajectories are rendered visible Figure 1: Video image of an illuminated
whereas laser scattering only measures plasma with a 30sccm flow of silane at
transient particle passages across the O.3mbar, SW, 60MHz, electrode temperature
laser path and spatial information is 300K. The viewing angle obscures the right-
obtained by displacing the laser or the hand electrode edge.
reactor.
The parallel plate reactor has two steel
electrodes of diameter 13Smm. The On plasma extinction, the powder is ejected
capacitatively-coupled rf frequency is 10- laterally but some settles on the lower
lOOMHz. A halogen lamp served as the white electrode, hence the substrate is placed on
light source and the powder structure was the underside of the upper electrode.
recorded on video.
The usual pure silane discharge image is 1.2 Measurements
the blue emission of the SiH* molecular Particles collected from the lower
band around 414nm. Fig. 1 shows a plasma electrode, which were unoxidised, were
illuminated from the left with white light; o .3micron in diameter. Thus visible light
the stationary powder layers are is Mie-scattered and the intensity 1S a
indistinguishable without illumination. The convolution of the particle number density,
video film shows a uniform stream of powder size distribution, refractive index and
flowing radially from the layers past the wavelength. In-situ estimation of the
electrode edge, forming a yellowish band particle number density and size is
around the reactor wall. The electrodes are therefore difficult even with a laser
protected from particle contact by the source (3). Nevertheless, the scattered
sheaths. Powder formed at the lower intensity of white light is a guide to the
electrode edge by the fringing field quantity of powder in the plasma.
between the earth screen and the rf Digitisation of the video image can be used
electrode surface leaves the discharge to give, for example, a vertical profile of
volume directly. the scattered intensity as in Fig.2.

169
 20
E Grounded Electrode
E

~ 15
(IJ
I

10
~e~
,?O...::.:----
\\e(,\e~
5 ~e~
/
~
/
0
0 Rf Electrode 0.5 1.0 T im e[ 51

Figure 2: Right-hand side: Vertical intensity profile at the

substrate position of the light scattered from the powder layers.
Left-hand side: Schematic of the formation of powder layers and the
variation of reflected power with time.

1.3 Powder Formation and Trapping 8~-------------------------------

For hot electrodes (500K) there is a Scattered Intensity [a.u. ]
power threshold below which no visible
powder is formed and the matching condition
is independent of power. If the power is 6
increased up to the threshold, the matching
begins to change. At this instant, no
powder can be seen. Subsequently, the
evolution of the powder profile becomes 4
visible; Fig.2 shows a representation in
time. The reflected power measured before
the matching box increases almost linearly
with time until the layer reaches steady- 2
state (Fig.2). The initial phase of powder
formation produces macromolecules or
'clusters' which, although invisibly small,
introduce electron Coulomb scattering, 04-----~~~~~~~~~--~~~~-L-
attachment and recombination centres (4) 0.1 1 10
and alter the electron energy distribution Modulation Frequency [kHz]
function (5). These molecular-scale
reactions modify the plasma impedance and
matching condition. A change in reflected Figure 3: Effect of power sine-wave
power is thus a precursor to visible modulation frequency on the intensity of
particle formation which is useful because light scattering (peak power = 20W) .
even 20A particles could be deleterious to
the electronic properties of a growing
layer. 2. EFFECTIVE POWER MEASUREMENTS
These negatively-charged particles are
trapped in shallow, closed potential wells. 2.1 Method
The vertical retaining electric field is The discharge input power is usually
due to the sheath and the radial field is measured with a standing wave ratio (SWR)
probably the ambipolar field equalising meter in the 50 ohm line before the
electron and positive ion diffusion. Powder matching network. This means that the
streaming can thus be interpreted as a effective power in the plasma is generally
potential well overflowing due to mutual not known. The effective power was
repulsion when the charged particle estimated by a 'subtractive method' (8) in
concentration becomes too large for the which the input power without plasma (Pvac)
well to remain closed. Powder reduction by is subtracted from the total input power
modulation of the rf power (6) is (Ptot) to give the effective power (Ppl),
demonstrated in Fig. 3. At 1kHz, the only with the constraint that both Pvac and Ptot
remaining visible powder is due to the rf are measured for the same electrode
electrode fringing fields. voltage. This condition is achieved by
adjusting the amplifier output and slightly
1.4 Summary of Powder Measurements re-tuning the matching system; the latter
Illumination of the discharge in adjustment introduces errors of around 10%.
plasma deposition reactors with visible The electrode voltage is measured directly
light gives immediate insight into the behind the surface of the rf electrode
problem of particulate contamination (7). exposed to the plasma so that the reactor
Global, spatio-temporal motion of powder circuit losses can be assumed to be in
and particle trajectories can be recorded parallel with the plasma dissipation. Fig.4
on video. A change in rf matching gives a shows typical electrode peak-to-peak
practical method for detecting the onset of voltage (Vpp) and input power (Ptot)
cluster production. measurements for argon, hydrogen and silane

170
plasmas at 70MHz and the straight lines are
least-square fits to the data. Note that 1 ~-------------------------------.
(Vpp)2 in vacuum is proportional to power Efficiency
as expected for a linear circuit. The
inverse slope is given by Go/8, where Go is Argon
the vacuum conductance of the system. The 0.8
data with plasma is well-represented by
straight lines, demonstrating a plasma
conductance approximately independent of Hydrogen
power provided that no alteration in 0.6
discharge symmetry and/or onset of powder
production perturb the plasma impedance at
high powers. The plasma line is always
observed t6 have a positive intercept with 0 .4
the vacuum line, at a value which we
interpret to represent a 'minimum
sustaining voltage' for the discharge. The
inverse slope of this line is Gtot/8 0.2
(Go+Gpl)/8, where Gpl is the plasma
conductance. The power transfer efficiency
~ in Fig.5 is defined as follows:
~ = [(Ptot-Pvac) /Ptot] const vpp = Ppl /Ptot
The asymptotic efficiency is (l-Go/Gtot). o 10 20 30 40
At low powers close to the extinction limit Input power [W]
of the plasma, the efficiency tends to
zero because the fraction of power Figure 5: Power transfer efficiencies
dissipated in the vacuum circuit becomes deduced from Fig.4.
large compared to the vanishing plasma
power. This rapid variation in efficiency
with input power could lead to problems of the complete system with and without
with reproducibility for depositions made plasma is frequency-dependent and so it is
at low input power. important to know how the effective power
in the plasma depends on frequency,
2.2 Dependence of Power Transfer otherwise a scan in frequency at constant
Efficiency on Gas Type input power will result in a variation in
For the same external conditions, the true plasma power. The estimated power
fraction of input power is high for Argon transfer efficiency for a pure silane
(z90%), falling to z75% for Hydrogen and plasma is plotted in Fig.6 along with the
finally only z55% for Silane. It would electrode peak-to-peak voltage for a
appear that the different gas collision frequency scan at constant input power. The
cross-sections give different plasma SiH* emission intensity at 414nm confirms
conductances relative to the vacuum circuit the power transfer efficiency measurement
conductance resulting in power transfer to a first approximation. Powder formation
efficiencies dependent on gas type. in silane plasmas alters the plasma
impedance (see Section 1 above) and
2.3 Frequency-dependence of Power therefore the efficiency depends on the
Transfer Efficiency power used - further experiments are in
Several groups are studying the progress to clarify the relative importance
effect of excitation frequency on of powder, reactor and plasma impedances.
deposition conditions (1,10). The impedance
2.4 Frequency-dependence of Ion Impact
Energy
From Fig.6 we note that the electrode
2000~V~~2__~[V~2~]~____________________~ voltage decreases strongly with increasing
frequency. Lower frequencies therefore
o Vacuum result in much higher ion impact energies
on the substrate which could lead to
X Silane defects during deposition. On the other
hand, a sufficient amount of low energy ion
1500 • Hydrogen bombardment is considered to be beneficial
for growth by e.g. increasing the surface
¢ Argon mobility of reactive species this
mechanism would then be favoured in 70MHz
plasmas as compared to 13MHz. At 70MHz,
1000 then, the effective power and growth rate
could be increased whilst keeping the ion
impact energy within reasonable limits
concomitant with deposition of good quality
material (11).
500~'---~~~----------------~--1

2.5 Conclusions for Effective Power

Measurements
The estimated effective plasma power
is shown to depend on reactor impedance and
o 10 20 30 40 gas type. Powder production in silane can
also alter the discharge impedance (see
Input power [W]
Section 1 above) and change the plasma
Figure 4: Estimation of plasma power for effective power. The electrode voltage is
3 separate gases (0.3mbar, 30sccm, 70MHz). much reduced at high frequencies and this
probably influences deposition quality.

171
250 1.0 3.2 Preliminary Results
Fig.7 shows measurements for an
on/off-modulated argon plasma using a 30GHz
Gunn diode and 8mm waveguide tapered down
200 0.8 to two O.lmm diameter wires. A Langmuir
probe ion saturation current is shown for
comparison; the different response time is
150 0.6 probably due to the different evolution of
the chord-averaged and local densities
measured by the two methods. In silane
0.4 plasmas, Langmuir probes are rendered
}OO inoperable by the silicon deposited on
Rf-Voltage[V] their surface whereas the Lecher signal
only suffers a slow baseline drift
50 0.2 attributed to the change in dielectric due
to the silicon deposited on the wires.

O~~~~~~~~~~~~~~~O . O Acknowledgements
We thank P.Bowen, B.Duval and the IMT
10 20 30 40 50 60 70 80 90 Neuchatel group of Prof. Shah for helpful
Rf-frequency [MHz] discussions and technical assistance. The
work was supported by Swiss Federal
Figure 6: Frequency-dependence of eff- Research Grant EF-REN(89) 17.
iciency and electrode voltage for silane
plasma (0.3mbar, 30sccm) with constant REFERENCES
input power 15W (1) H. Curtins, N. Wyrsch and A. Shah,
Electron. Lett. ~ 228 (1987)
(2) G. S. Selwyn, J. Singh and R. Bennett,
3. LECHER WIRE MICROWAVE INTERFEROMETER J. Vac. Sci. Technol. Al, 2758 (1989)
(3) K. G. Spears, T. J. Robinson and R. M.
3.1 Method Roth, IEEE Trans. Pl. Sci..l:.S.=l.i, 179 (1986)
Lecher wire interferometry (12), a (4) A.Garscadden, 'Nonequilibrium Processes
new application in rf plasma processing, in Partially Ionised Gases', Edited by M.
measures the chord-averaged, free-electron Capitelli and J. N. Bardsley, Plenum Press,
density from the phase shift of microwaves NY, 541 (1990)
guided by two fine wires across the plasma. (5) M. J. McCaughey and M. J. Kushner,
The spatial resolution of the measurement Appl. Phys. Lett. ~, 951 (1989)
is determined by the wire spacing which can (6) Y. Watanabe, M. Shiratan, Y. Kubo, I.
be made much smaller than the inter- Ogawa and S. Ogi, App1. Phys. Lett. .5...l,
electrode distance. Stray reflections from 1263 (1988)
the electrode surfaces are thereby avoided (7) A. A. Howling, Ch. Hollenstein and P.-
and the electrically-isolated wires J. Paris, submitted to Appl. Phys. Lett.
minimise perturbations to the plasma. (8) V. A. Godyak and R. B. Piejak, J. Vac.
Sci. Technol. Aa(5) 3833 (1990)
(10) S. Oda, J. Noda and M. Matsumura, Jap.
J. Appl. Phys. za L1860 (1989)
(11) F. Finger et aI, submitted to J. Appl.
ImV I LECHER INTERFEROMETER I I Phys.
I (12) M. E. Bacon, D. T. Goodhead and G. H.
J. Wantenaar, Rev. Sci. Instrum.· il, 1201
6 I (1976)
I I
I I

\
I
I
I
o
I
I LA GMUIR PROSE : Ion soturoloon current I
I
tSO I
y

100

50

o
o 6 8 10 12
TIME tmsl

Figure 7: Signals from Lecher Wire

microwave interferometer and Langmuir probe
for an on/off-modulated argon plasma.

172
10TH EUROPEAN PHOTOVOLTAlC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Electric Field Profile and Charge Collection in a-Si:H Solar

Cells

R. Konenkamp*, S. Muramatsu, H. Itoh, S. Matsubara and T. Shimada

Hitachi Central Research Laboratories, Kokubunji, Tokyo 185, Japan

FAX: 0081-423-27 7748
*and Hahn-Meitner Institut Berlin, 1000 Berlin 39, Germany
FAX: 0049-30-8009 2434

We have studied the electric field profile in single-junction amorphous

silicon solar cells. Our results allow an estimate of the space charge
density at the p. i- junction and a simple quantitative evaluation of the
cic... trical properties of the junction. We find the field profile to be
approximately exponential, and to be largely determined by deep
defect states in the intrinsic a-Si:H absorber layer.

important cell parameters for a set of cells is

given in Table (1). All samples were prepared
IN1RODUCTION in a three-chamber-rf-glow discharge system
Over the last years single junction amorphous under optimized deposi tion conditions as
silicon solar cells have reached conversion described in more detail recently (l). In the
efficiencies around 12 % (I). These cells are following we will focus our discussion on
usually prepared in the sequence, sample 3 which showed the best
glasslTCOlplilnlmetal, with many intricate performance. This sample prepared with
variations in the preparation of the Po, i- and three ultrathin a-SiC:H layers with
n-doped regions. In this cell structure the pli consecutively lower bandgaps having total
junction receives the illumination, and hence thickness of 60A. The p-doped region was
the carrier generation profile is peaked near followed by a nominally undoped a-SiCH
this junction. Attempts to improve the buffer layer, the intrinsic absorber region
conversion effiencies have therefore strongly (6000A), and the backcontact, consisting
focussed on an optimization of the pli again of a buffer layer, the microcrystalline
interface. As pointed ,out by Tsuda (2), high n-region and an optically reflecting metal.
transparency for the p-doped la'yer(s), a large The determination of the electric field profile
built-in potential and a low density of used a time-of-f1ight experiment as described
in more detail elsewhere (3,4). Here we only
interface and bulk defect states are the
note that a small electron packet consisting of
crucial parameters for an optimized pli
-10 9 e I c m 2, is generated by a short laser
junction. We have attempted 'to find a novel
flash at the pli interface. These carriers
method to further characterize the electrical
traverse the solar cell in the internal (or an
properties of thi$- ,junction. This appears of
importance since for a detailed assessment of applied external) electric field, giving rise to a
photocurrent transient which is analyzed to
the junction performance it. is essential to
separate optical and electrical contributions to determine the spatial electric field profile
the device improvement. Here we report the across the cell. In our experiments Jhe spatial
determinatation of the electric field profile in resolution for the electric field is
approximately 500A, based on a time
the pliln configuration. An analysis of our
resolution for the photocurr~nt detection of
results indicates that in optimized cells the
<I ns. For a simple analysis as performed here,
electric field in the pli region is largely
deep trapping should not occur, and the
determined by charged deep defects in the
carrier transport should be non-dispersive.
intrinsic layer.
These conditions were usually met in good
EXPERIMENTAL devices with efficiencies <10%. The electric
field profile could then be measured in the
A diagram of the typical cell structure is range of small forward bias (-.2V) to a few
given in Fig. (l) and a list of the most volts reverse bias (3).

173
 AI
n: I' r. ~i

Fig. (1): Cell structure of pli/n solar cells

-
I
Ell 1I :f.~ 1.:F.f. 1I :r." . I.: F."
used in this study.

D
--
r p,
f.f. P, P, 1', I',

1 - l - l

TABLE I. Structure.and properties of the set of solar cells: Po:a-Si:H. EG = \.7S eV; PI:a-SiH:C. EG = 2.12 eV; pz:a-SiH:C; EG = 2.3' eV; Pl:a-SiH:C.
EG 2.4IeV.

"
Sample Yo< I .. Fill
No. Structure (V) (mA/cmz) factor (% )

PoIi/" 0.77 15 0.75 8.6

2 60 Apl/i/" 0.87 16.7 0.76 I \.0
20 Apz. 40 API/i/n 0.87 17.3 0.7S 11.3
4 10 APl' 20 APz. 30 Apt!i/" 0.87 17.8 0.74 11.4

RESULTS AND DISCUSSION The electric field profile can be used to

determine the space charge distribution, p(x),
and the band bending, Vex), across the
In Fig. 2 we show the electtic field profile of
junction: The approximately exponential
sample 4 of the table obtained under short
shape allows for a simple analytical
circuit conditions. i.e. zero applied voltage.
treatment of the problem. From eq.(1) we
With reasonable accuracy this profile can be
find for the space charge.
approximated by an electric field peaked at
the p/i interface and decaying exponentally p(x):;;;Po exp( -x/L), (2)
as,
Q:(x):;;;Q:oexp (-x/L). (1)
with Po:;;;1.6*I016 e cm- 3 . Due to the high
midgap density in the amorphous material
with Q:o",S2kV/cm and L",210oA. No field
the space charge IS nearly completely
peak occurs near the i/n interface. This may attributable to deep gap states. Hence Po can
indicate a lower sensitivity of the experiment
also be considered as an estimate for the gap
for the deeper bulk region of the cell due to a
state density in the intrinsic layer.
spreading of the carrier packet, but it may
Integrating eq. (I), it is easily seen that the
also indicate that the nli junction is
band bending, Vex), has the same exponential
somewhat weaker than the p/i junction. This
shape,
conjecture finds some support in the fact that
V(x):;;;V o exp (-x/L) (3)
iln junctions in a-Si:H form ohmic, rather
than rectifying contacts.
with Vo",I.OS eV.

174
 Thus an energy diagram This picture is in qualitative agreement with
as shown in Fig. (3) can be deduced for the the idea that the defect states in amorphous
junction. As illustrated in Fig. (3), at the p/i silicon exhibit a broad distribution in energy.
junction the Fermi-level, Ep, is displaced from It is noteworthy to point out that our simple
its bulk position, Epo , due to charge transfer analysis suggests that interface states, located
between the p-doped and intrinsic layer. This directly at the a-SiC:H/a-Si:H transition, have
displacement of the Fermi-energy is no major effect on the shape of the electric
associated with a deoccupation of midgap field. Instead it appears that mainly the
states, N(E). These are expected to be neutral midgap states in the intrinsic a-Si:H layer,
when Ep assumes its bulk value, hence in the here estimated to be -1.6*10 16 cm- 3. are
junction these states are positively charged. responsible for the shape of the junction field.
The space charge density can then be This conclusion is in agreement with the work
expressed as a function of the deoccupied of Tasaki et al. (5) showing that in optimized
mid gap states as, p/i junctions the interface defect density can
Epo(x) strongly be reduced by an undoped buffer
p(x)=e /N(E)dE. (4) layer. Indeed. other work by us on non-
Ep optimized cell structures shows that in
Substituting variables, and considering eqs. junctions without buffer layer the field shape
(I )-(3), then leads to the conclusion that the differs substantially from the one shown
energetic distribution of midgap states must here.
be constant in the interval where the Fermi-
energy is displaced,

N(E)=const for EPo<E<Ep+Vo.

60
p-aSiC:H
50
E
u
;; \

,,
~ 40
\ 1-aS1:H
.,
:!:!
30
u:: \
\ ---E.-
..
u

ti
20

CD
W 10

0
0 2000 4000 6000

Position (A)
Ev

Fig. (2): Electric field profile for sample 4

Fig. (3): Schematic diagram of the space
of table (1) (brokeh line) and
charge region near the p/i interface.
approximation based on eq. (1) (solid
line).

175
It is well known that capture coefficients for Development Organization under the Sunshine
charged centers exceed those of neutral states Project of the Ministry of. International Trade
and it can be <;Ai'cded that the deoccupied and Industry.
midgap states in the space charge region will
also affect the electronic transport and the REFERENCPS
recombination traffic in the device. It appears
useful therefore to investigate in more detail (I) T. Shimada, S. Matsubara, H. Itoh, and S.
trapping and recombination kinetics in the Muramatsu, Proc. 9th EC Photovoltaic Solar
space charge region of the p/i/n cell Energy Conf. Freiburg, Germany, 1989, p. 81
structure. (2) S. Tsuda, Proc. 5th International
Photovoltaic Science and Engineering Conf.,
In conclusion we have determined the Kyoto, Japan, 1990, p.21
electric field profile in single junction a-Si:H (3) R. Konenkamp, S. Muramatsu, H. Itoh, S.
solar cells with a typical thickness of -7000A. Matsubara, and T. Shimada, Appl. Phys. Lett.
In optimized devices with a conversion 57, 478, 1990
efficiency above 10%, the field profile is (4) R. Konenkamp. S. Muramatsu, H. Hoh, S.
approximately exponential and can largely be Matsubara, and T. Shimada, Proc. 5th
explained by a deoccupation of midgap states International Photovoltaic Science and
in the intrinsic layer with a density of Engineering Conf., Kyoto, Japan, 1990, p.
-1.6*10 16 cm- 3 . (4) H. Tasaki, W. Y.Kim, M. Hallerdt, M.
Konagai, and K Takahashi, J. Appl. Phys. 63,
550, 1988
ACKNOWlEOOMENTS

This work was supported in part by the New

Energy and Industrial Technology

176
lorn EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 USBON, PORTUGAL

RELA TION OF SMALL SIGNAL TRANSIENT RESPONSE TO DENSITY

OF STATES IN A-SI

.Toze Furlan, Elvis B.1s8neze, Fr,wc Smole

Faculty of Electric.11 <11J(/ Computer Engineering
University of Ljubljana
Ljubljana, Yugoslavia

ABSTRACT: Based on exact numerical computation of small

signal transient response of charge carriers in a-Si certain
simplifications were introduced in the transient analysis enabling an
analytical solution of differential equations. The resulting
analytical expressions for exponential time constants are functions of
steady-state charge carrier concentrations. The calculation procedure
is then reversed so that transient time constants and bias light
generations serve as initial data for determination of carrier
concentrations and of DOS distribution.

1. INTRODUCTION
Based on charge carrier dynamic
nonequilibrium in a-Si described by a set -t/T -t/T -t/T
of four nonlinear differential equations t.N(t)=N +N e 1+N e Z+N3 e 3 (1)
012
governing the rates of change of free and
trapped electron and hole concentrations. a where t.N(t) is one of the small signal
set of linearized differential equations charge carrier concentrations and No to N3
relating carrier densities to small signal are ampl i tude coefficients that belong to
perturbations around steady-state bias can particular charge carriers.
be derived using SRH generation-recom- Even in case of small signal excitations
bination approach. the resulting system of linear differential
The resulting system of differential equations can not be solved analytically
equations can not be solved analytically. due to the functional dependence of
Therefore the transient response due to variables on the energy in the bandgap. The
small signal pulse illuminations added to numerical solution of small signal
dc light bias was solved numerically. The transient response seemed therefore to be
inspection of computed results revealed the most suitable initial approach
that certain simplifications might be providing the feasibility of examining and
introduced in the system of differential introducing eventual simplifications
equations. Following this approach enabling analytical solutions afterwards.
analytical solutions were made possible An example of computed small signal
giving approximate expressions for turn-on transients of free and trapped
exponent ial time constants during charge carriers (t.n. t.p. An~. Ap~) is shown
transients.
Next, the calculation procedure was
M(t) lE12
reversed. The previously derived small
signal transient time constants were taken [cm- 3 ]
as initial data from which steady-state lEll
concentrations of carriers were calculated c
for various bias light generations, ~ lEl0
enabling finally the evaluation of E c
local ized states densi tie8 distribution in
the bandgap. ""c0 lE9 t.p(t)-2
"'- Ant(t)-3 IIpt(t)-4
lE8
2. SMALL SIGNAL TRANSIENT ANALYSIS ·Ec GL(O)=10 23 cm- 3 s- 1
Charge carrier dynamics in a-Si in u
lE7 IIG L =1019 cm- 3 s- 1
nonequilibrium conditions can be described "~
following the SRH approach /1/. The free 0
..:;
and trapped carrier densities u lE-8 lE-7 lE-6
(n. p. n t • Pt) can be represented by a set time [sec]
of four nonlinear differential equations.
In case of small signal pulse illumillation Fig.I. Small signal charge carrier
superimposed un dc light bias the turn-on transient.
differential equations are I inearized. In
the absence of current flow. the equal in Fig.1. As the basis for the computation
rates of negative and positive charge the model of gap states shown in Fig.2 was
maintain the electrical neutral i ty during adopted, consisting of donorlike and
transient, reducing the system of acceptorlike tail states and dangling bond
differential equations to thret· ..'quatlons defect states represented by peaks having
defining three exponential time constants Gaussian distribution of states. Other
governing the small signal charge carrier parameters for computer analysis are given
transient response ill the form in Table I.

177
 '0'
\,. '£20

'"E"
I I[ID
-
~
2(8

~ lE18

1[17
'"E I t = 10- 9 sec
W ~
a; 1£1115

"
.! 1(',

..
.2 1(1.4
;:;.
'0
;:;. 1[1.] 'nc
.,"
'iii
c 1[12 "U

"
"U
-1 .70 -1..36 -1 UI -1 .02 -C.M -0.6.8 -O~l -0.34 -0. 17 0.0 0
.!
energy in the gop E-E C (eV] .2
~ ~O~.J~4----------===~--~O.~'7~----~==~:=---O~OO
Fig.2. DOS model used for numerical "g" energy E- Ec
.
( eV]
analysis. <J
g (0)

Table 1
lEla
I
Parameter values used in computer model ,.,~ 9Ell

Bandgap EG I
E
BEll t=10- 7 sec
Density of states Nc ' Ny ~ 7£9

Capture cross-section: <fi BEll

- in charged states 0 nD =0 pA =0 =10- 16 cm 2 q5 5Ell

~ 'Ell
- in neutral states 0 =0 =0/R=lO-18 cm 2
pD nA -~ JEll

.,"
"U 2Ell
The computed exponential time constants,
___
~ 1(9
T1 , T 2 , T3 , in accordance with eq.(l) for a o
n o~---------------~~---- ~~~-L
wide range of bias light generations are -g -0.34 -0. 17 0.00
'0.
shown in Fig.3. From these graphs it can be :>
energy E - Ec [eV]
concluded that <J
<J
o (b)
'[-3 r~~---"---~--"-'-~---r-.........,
2El0
...[sec) - - numerical calculation
lE4 - - - - approximate equations
-'- ' - measurements [4) t = 10- 6 sec

A
0'
...:
5 lEIO
q
1(-7

---------: ---------.---------- _~!1_~_:~: :::: :!p:-:!i: ~;: ~~;~~: ~

'0;
c:
,E-8
2 "
"U

"'3 .,"
'[-~£-~17:--------:':-:'£I:-::'~-'E~'-'--'-'£20'----'(.2"'-.--,-'[2'--.-......:'a:.....-3--,.Ja • "0
;; ~0~.J4~------------_-0.~'7~------~B------_-10.00
bios light generations GL(O) [cm- 3 s- 1 J "U
"
'0.
:l
energy E-Ec [eV]
Fig.3. ExponentiaJ time constants during u
turn-on transient. g (c)
Fig.4 . Computed density of occupied ac-
- one time constant is very large compared ceptorlike s!~tes during t.!.pnsient
to the remaining pair of time constants, at times_l=10 sec (a), t=10 sec (b)
- both short time constants are practlcally and t=10 sec (c).
equal and do not change significantly
with increasing light bias . An gA- (An+RAp)f tA (O)gA
Wi th the aim of finding some sui table
simplifications in calculating small signal n(o)+n 1 +R(p(o)+P1)
transients, the density of filled
acceptorlike states inside of the gap R dAf tA
contributing to the trapped electron ov --crt gA
density, was numerically computed for n (0) + n
1
various moments during turn-on transient.
The computed plots, shown in Fig.4a, band
c, represent the terms contributing to the A (t, E) - B (t, E) (2)
time dependent occupied states density
given by From these graphs it is clearly visible
that the contributions of terms
participating in transient response is very
much different at short and long times
after small signal light turning-on.

178
Therefore the transient response analysis Both analytically derived approximate
can be simplified by introducing some short time constants. 1"2 and T 3 • presented
specific approximations during short and also in Fig.3. are again in fair agreement
long times following the light excitation wi th the exactly computed short time
step. constants. Analytical expressions (8) and
a) Long times: (9) show that these time constants depend
In accordance with plots in Fig.4 it primarily on the gap states densities
can be assumed that during longer times participating in the multiple trapping
following the light pulse the transient process /2/./3/. The main effect comes from
response is sufficiently well described by tail states neighboring the conduction band
neglecting the term B(t,E) in eq.(2). In in case of electron densi ty transient. and
this case the density of occupied states in states adjoining the valence band in case
the bandgap can be simply written as of free hole transient response.
E JE g dE It can be concluded that the small
J liftAgAdE
C C
signal transient response in a-Si can be
lin\.= = lin n(o)+n:+RCP(O)+Pl> reasonably well represented by a set of
E
v
E three time constants given by analytical
expressions (5), (8) and (9). The pair of
both short time constants is in the range
of nanoseconds, being nearly independent on
steady-state I ight bias. The longer time
constant decreases wi th the increasing dc
(3) light bias. The dependence of this long
time constant is in 11 relatively good
In a similar way a linear equation given by agreernl'nt wi th the measured time constant
given by Chen and Tai /4/. The calculated
t.Pt = b 1 lin + b 2 lip (4) curve could be well fitted to the measured
can be derived for trapped hole densi ty .. curve simply by increasing the parameter
The resulting pair of linear equations crv for a factor of 10.
reduces the system of differential
equations to a single differential 3. DETERMINATION OF DOS
equation. The deduced time constant can be After establishing the functional
written in the simplified form as dependence between steady-state
concentrations of carriers and time
T
1
= [crv (n I. (0) + p I. (O»)J-l (5) constants during transients. the
where n\.(o) and PL(O) represent trapped calculation procedure can be reversed.
Transient time constants and bias light
carrier concentrations depending on generations will be taken as initial data
steady-state light bias. Inserting the from which steady-state carrier
appropriate values of dc trapped carrier concentrations will be calculated. enabling
concentrations, the time constant finally the evaluation of localized states
calculated from eq. (5), follows reasonably density in the bandgap.
well the long time constant Tl which I t can be as sumed tha t there are the
resulted from exact numerical computation. two distinct quantities that can be derived
b) Short times: experimantally:
During short times after turning-on - the longer time constant during small
the small signal light pulse it is evident signal transient. given by expression (5).
from Fig.4, that the trapped carrier - the steady-state light generations. given
concentration has not yet changed and that in the form
the term A(t ,E) is nearly equal to the GL (0) = crv [nCo) P" (0) + p(o) n\. (0)] (10)
term B(t.E) resulting into an approximate
relation Since there are four unknowns (n(o). p(o).

d:~\. = T[linJECgAdE _ (t.n+Rt.Plnt(o)] (6)

nt(o) and p\.(o». the missing pair of
relations is obtained when assuming equal
concentrations of electrons and holes.
Combining eq.(6) with linearized continuity n t . (o)=p~ (0). and n(o)=p(o). Following these
equation of free and trapped negative assumptions. the steady state charge
charge carriers carrier densities can be written in the
simple form
dt.n
crt'" +
dlin
Tt = liGL -
[
crv n(o)aPt + Pt. (o)lin + n~ (0) = P, (0) (2 (TV '1)
-1
(11 )
+ p(o) lint + n t (0) liP] (7) nCo) = p(o) = GL (0) 'j (12)
and neglecting small signal trapped carrier In order to ver i fy the adopt ed
concentrations in eq. (7) results in a assumptions. the eqs.(1I) and (12) were
single short time constant of free used for calculation of carrier
electrons given by concentrations from numerically computed
time constant Tl versus bias light
Tn = T2 = [crv[P\. (0) generations GL (0). Figs.5 and 6 show the
calculated carrier densities together with
Using the same simplifying procedure a exactly computed carrier concentrations.

r
single time constant during the transient As the final step. a, numerical method
of free holes is obtained E enabling the evaluation of density of
states distribution. was developed.
T
P
T3 = [crv[n t (0) + ~J C gl> dE]
l
(9) Assuming again equal electron and hole
concentrations, n(o)=p(o). n~ (o)=p •. (0). and
designating free and trapped carrier

179
concentrations in the three successive of tai I states the calculated DOS agrees
po i n t s by n (i -I ), n ( i). n (i + 1 ) • and relatively well with the initial DOS from
nL(i-I). n\.(i). nl.(i+I). the density of which time constant T1 as function of light
IE'7 r-'--~~"""-"""-"""-"""-"""--,-~"",,,,~
bias GL (0) was numerically computed.
nCO) Assumption of equal electron and hole
p(O)
lEl0 densities results in a symmetrical
......... n(O)-p(O) distribution of states. This might be
tolerated in the regions of tai I states
while in the central region of bandgap the
assumption of symmetrical DOS as well as
the described itterative scheme for DOS
evaluation break down.
4. CONCLUSIONS
The exact numerical modelling of small
signal transients in a-Si can serve as a
IE1 ~E""S:-""E'-,e-''''[-'7-'~E,-a-,'''[,-.-'u o -'[2"-'-,w'---,wL.-~''''[2'--'..........J,w
....... basis for introducing simplifying
assumptions during transients. enabling an
bios light generotions GL(O) [cm- 3 s- 1]
approximate analytical solution of
transient response in a-Si. Using this
Fig.5. Exact and approximate free approach the approximate relations between
carrier concentrations exponential time constants. carrier
concentrations and DOS distribution can be
obtained.
REFERENCES:
1. J.Furlan.S.Amon. Proc.Eight E.C.Photo-
voltaic Solar Energy Conf.. p.836.
Florence 1988.
2. T.Tiedje.A.Rose. Solid State Commun .•
37. p.49. 1980.
3. J .Orenstein.M.A.Kastner. Sol id State
Commun .• 40. p.85. 1981.
4. Xing Chen.Chen-Yu Tai. Mat.Res. Soc.
1[13 Symp. Proc .• 149. p.333.1989.

'£I~[_L,-~-:1["',a~-'E"'17--,£L.-'.-''''£'--'t-,'''e-o-,'''u-,- 22 _'w
'(....... ......._'u.....-...J'U5
bios light generotions GL(O) [cm- 3s- 1]

Fig.6. Exact and approximate Lr'apped

carrier concentrations.
states at trap quasi-Fermi level can be
expressed by

g [Ea
A f. n
(i) ] =
gD
[Ed
I. P
(i) ] =
kr
1+R [ n (i -I)
n ( i) n (i - 1)
] ~

. [2 n I. (i) -n t (i + 1) -n t, (i -1) ] ( 13)

where

Using this itterative scheme. the

density of states distribution was
determined. shown in Fig. 7. In the region
,e,
I
>OJ '£20 o 0 0 colculoted DOS
r') 6
I 'EUII
E
~ ,EUI

W- 'E ,7
C
'"
'[IS

W- '['~
'<
'"
1[14

III ,[tl
0
0
U:I2
-1 .70 -D. ~

energy E-Ec reV]

Fig.7. Initial DOS and calculated
I)QS distribution.

180
HITH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

CHARACTERIZATION OF HIGHLY CONDUCTIVE PHOSPHORUS DOPED ~c-Si:H FILMS

E.A.T. DIRANI, A.M. de ANDRADE and I. PEREYRA

LABORATORIO DE MICROELETRONICA - ESCOLA POLITECNICA

UNIVERSIDADE DE SAO PAULO
Phone: (11) 8159322 ext.3256 - FAX (11) 2114308
Caixa Postal 8174 - 01051 - Sao Paulo - Brazil

ABSTRACT. In this work the electro-optical properties of phosphorus

doped microcrystalline hydrogenated silicon (~c-Si:H) films as a
function of their deposition parameters are investigated.
The films were deposited in a plasma enhanced CVD system. It is
observed that their properties are extremely sensitive to the r. f.
power density in the studied range of 250 mw/cm 2 to 500 mw/cm 2 .
very low ~esistivit~ films were obtained. For a r.f. power density
of 500 mW/cm , P=5Xl~- ohm.cm was obtained Whi~st, for a,r.f. power
density of 250 mW/cm a resistivity of P=3,7x10 ohm.cm is observed.
These results show the sensitivity of the process to small changes in
its variable deposition parameters. A low value of 28 meV for th~
activation energy was obtained with a r.f. power density of 500 mW/cm
which can be due to a large density of microcrystals embedded in the
amorphous tissue. The grain crystalline size as estimated from sm~ll
angle X-Ray scattering experiments is in the range of 70 A to 120 A.
'This promising material has immediate technological applications
and it is 'believed that the experiments on the way will provide more
understanding on the deposition phenomenology.

1. INTRODUCTION 2. EXPERIMENTAL PROCEDURES

Since the works for Matsuda et a1. (1) The samples were prepared onto corning
in 1980 and Veprek et a1. (2) in 1981 who glass 7059 substrates for the electrical
sistematically studied the deposition and the grain size measurements. Optically
conditions for the hydrogenated flat glass slides were used for optical
microcrystalline silicon (~c-Si:H), the characterization and the films thickness
interest for this material is continuously determination. The samples were produced
increasing. The possible applications in on a PECVD system already described (5).
large area electronic devices has To get an optimized deposition
encouraged many researchers to develop condition for the phosphorus doped films,
deposition techniques for this important a bias voltage of +75V was applied to the
material. Several utilizations of the grid electrode of the system which yields
~c-Si:H material have been reported in a maximum value for the electrical
solar cells, pressure senSors and other conductivity and a mininum value for the
devices (2-4). activation energy (6). To confine the
In this work the electro-optical plasma region to the substrate holder, a
properties of phosphorus doped ~c-Si:H magnetic field was applied (6). This
layers were investigated as a function of magnetic field reduces strongly the
their deposition parameters since their bombardment of the reactor walls thus
properties depend not only the purity of reducing contamination and changes in the
the utilized gases but also of the r. f. gas mixture due to uncontrolled degassing.
plasma power density , gas flow rates and The deposition conditions are shown in
the dilution of the gaseous mixtures in table I.
hydrogen (2). The characterization of the
deposited layers was done through the PARAMETERS CONDITION,S
activation energy, optical gap and
electrical conductivity.
The average grain size of the r.f.power density (mw/cm2) 250-500
crystallites was determined by Small Angle VG-VSUB (V) 75
X-ray Scattering (SAXS) technique. Magnetic field IB (A) 3,5
The obtained material shows high Substrate Temperature (oC) 260
electrical conductivity and very low
activation energy associated with certain PH3/SiH4 1/80
deposition conditions. SiH4/H2 1/200
As the production of hydrogenated Gas mixture pressure (torr) 1,5
microcrystalline silicon is only possible
when a quasi-equilibrium condition betwen TABLE I. Deposition Conditions
the deposition rate and the etching rate
(2) is reached and this condition is
critically related to the proce~s 3. RESULTS AND DISCUSSION
deposition parameters, small changes ~n
their values determine big variations in The simplified chemical reaction
the film properties. asc~ibed to the process of
micltocrystalline films deposition can be
wri~ten as:

181
SiHn(plasma) ~
+-
Si(solid)+nH(plasma) (1)
R 2
10
were D is the direct reaction rate and R
is the reverse reaction rate. The I
10
difference between D and R determines the
film deposition rate (7). .~ I
After Veprek (2), strong silane dilution
tn hydrogen (1:200) favours the reverse -I
reaction (erosion) moving the reaction t> 10
towards the equilibrium. A small
deposition rate value is an important -2
10
condition since there is sufficient time
to permit ·the formation of small -3
crystallites resulting in a structural 10 ~~--~~~~~~~~~
250 300 350 400 450 500 2
transition from a-Si:H to ~c-Si:H. r. f. power density (m W /cm )
Consequently, it is observed that the
deposition rates for ~c-Si:H are much
smaller than those seen in a-Si:H (8). Fig.2 Ambient temperature conductivity as
a function of the r. f. power
density.
Correspondingly to what is seen in
E figure 2, the activation energy, as a
·s...... 2.5 function of the r.f. power density, has a
similar behaviour. with the increase of
E the discharge power, the activation energy
..:. 2.0 decreases strongly as it can be seen in
., figure 3 where just by doubl ing the r. f •
~ 1.5 power, the activation energy is reduce an
~
order of magnitude.
•
0
0.
1.0

250 300 350 400 450 500 0 .3

r .f . power density (mW/cm2)

Fig.1. Deposition rate as a function of

the r.f. power density. > 0 .2
~
In figure lone can observe that the o
deposition rate increases with the w
increase of the r.f. power density in the
studied power range. This could be due to 0.1
a higher efficiency in the process of
fragmentation of the gaseous species when
high r. f. power values are del i vered to
the plasma. One apparent exception could
be the sample deposited with the ~mallest
value of power density 250 mW/cm which 250 300 350 400 450 5002
showed a deposition rate as l~ge as the r.f. power density (mW/cm 1
sample deposited with 500 mW/cm . But that
sample exhibits strong indications of
being an amorphous silicon film or being Fig.3. Activation energy as a function of
constituted of a transitional structure the r.f. power density.
between amorphous and microscrystalline.
This possibility is corroborated by the From figures 2 and 3 it is possible to
not so high (as compared to other samples infer that for low values of the r.f.
produced with all the same parameters, power, amorphous tissue predominantes and
except r: f. power ~ensitYL electrical for larger values of r . f. power, what is
conductiv1ty, (J =2.7x10 3 S.cm 1 and not so known as microcrystalline material is
small activation energy, Ea=0 . 28 eV. deposited. When there is not enough r. f.
Besides that, there is the fact that it power, even when all other conditions are
was not possible to determine its grain favourable to get microcrystalline phase,
size with the SAXS technique whose o lower only amorphous material is obtained. Films
limit of detection is roughly 50 A with deposited ,ldth the power density of
the utilized equipment. 250 mW/cm exhibits characteristics of
The electrical properties of the films amorphous material and films 2 produced with
are extremely sensitive to the deposition power densities of 375 mWjcm or more show
condi tions, specially the r. f. power. characteristics of crystalline phase.
Keeping constant all other parameters and Films depofited in the intermediate 250 to
varying only the r. f. power density from 375 mW/cm range of power density show
250 to 500 mw/cm 2 , the electrical characteristics of a material transitional
conductivity has shown variations of 5 between amorphous and crystalline.
orders of magnitude as can be observed in The average grain size of the
figure 2. crystallite was estimated by the SAXS
technique. Figure 4 shows the electrical

182
conductivity versus estimated average (6) I. PEREYRA, et.al., Proc. 7th Europ. PV
grain size of the crystallites. Higher Solar Energy Conf. p.528, 1986.
electrical conductivity values are (7) C.C.TSAI, et.al., Proc.MRS Symp.
associated with bigger grain sizes as p.175, 1990.
figure 4 shows. This effects can also be (8) M.S.SANEMATSU et.al. Solar Cells, 14-
thought as due to a bigger ratio of the p-281,1985.
crystallite total volume to the amorphous (9) C.WANG, et.al., Proc. MRS Symp. p.335,
tissue. Experiments involving annealing 1990.
steps are being done to verify if the
average grain size or if the ratio of the
crystalline phase to the amorphous tissue
is more significant to get higher
electrical conductivity values.

1
0e 10
o
~I

-I
10

70 80 90 100 110 120

grain size l; (ll)

Fig.4. Electrical conductivity versus

crystallite grain size.

In conclusion, the experiments show

the feasibility of obtaining
microcrystalline silicon films with high
electrical conductivity values provided
that not only the gas dilution rates, gas
flow rates and substrate temperature
conditions are satisfied but also a
minimum value of r. f. power density is
delivered to the plasma.
utilizing high enough value~ of r. f.
power density, l? r.f=500 mW/cm it is
possible to obtain microcrystalline
silicon films in the range of 600 A to
700 1 thick with very large electrical
conductivity values, a =20 s.cm- 1 , and very
low activation energy, Ea=0.03 eV.
ACKNOWLEDGEMENTS
The authours are indebted to Eng.
Elvira Fortunato and Dr. Braz Fernandez
from Universidade Nova de Lisboa, Lisbon
for the Small Angle X-ray Scattering
measurements of crystallite grain size.
This work was partially suported by
FINEP, CNPqjRHAE and OAS-Organization of
American States.
REFERENCES
(1) MATSUDA, et.al., Jpn. J. Appl. Phys.
19 p.L 305, 1980.
(2) S.VEPREK, et.al., Proc. 9th Internat.
Conf. Amorph. Liquid Semicond. p. C4-
251, 1981.
(3) K.NAKAMI, et.al., Journal of Non-
Crystalline Solids, p.287, 1983.
(4) W.GERMER, et.al., Sensors and Act. 4
p.183, 1983.
(5) R.MARTINS, et.al., Proc. 5th Europ. PV
Solar Energy Conf. p.778, 1983.

183
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE
&-12 APRIL 1991 LISBON, PORTUGAL

A MODEL FOR THE DIRECT BIAS C-V MEASUREMENTS ON

a-Si:H P-I-N SOLAR CELLS

D. Caputo, G. De Cesare, F. Irrera and F. Palma

Dipartimento di Elettronica Universita 'La Sapienza'
Via Eudossiana,18 00184 ROMA (Italy)

ABSTRACT. In this paper we present a new analytical model for the direct
bias capacitance in p-i-n amorphous silicon solar cells. The model includes the
contributions due to band tails and dangling bonds trap states in the
diffusion capacitance.
Experimental measurements of capacitance versus direct bias voltage are
presented, showing a very good agreement with theoretical predictions.
The presented model, requires a very short computer time, being based on
the analytical treatment of the p-i-n structure; this allows an easy
comparison with experimental measurements.

1. INTRODUCTION Experimental results are reported and

compared with theoretical prediction.
Capacitance measurement as a function of
the bias voltage (C-V) is commonly performed to
study the density of states (DOS) in
semiconductors. In recent years, many authors 2. 1HEORY
(1-6) applied the C- V and the Deep Level
Transient Spectroscopy technique (DL TS) to In a p-i-n diode the total capacitance is the
investigate the density of states (DOS) in the sum of diffusion capacitance Cd and the transition
forbidden gap of a-Si:H in Schottky barrier capacitance Ct. We assume that the transition
diodes, MIS tunnel diodes, p-i-n and n-i-p solar capacitance is simply expressed in the usual form
cell structures. On a-Si:H solar cells the presence valid in crystalline semiconductors Ct=u/d, where
and the eVl:>lution· of traps in the forbidden gap Er is assumed equal to value for the crystalline
dramatically influences the device performance, silicon. In order to calculate the diffusion
and the capability of monitoring and predicting capacitance we applied the definition C=dQ/dV,
their behavior is an extremely important step in where dQ is the total variation of the charge in
improving the cell performance. the structure produced by a variation dV of the
Aim of this work is to· in"estigate the p-i-n applied voltage.
structure under direct bias capacitance The charge Q is composed by the free and
measurements in order to obtain general trapped charge. The latter is constituted by
informations which could be directly related to charge trapped into band tails and dangling
the material used in the actual solar cell. There, bonds traps. The "double carrier model" /7/
very thin layers, textured substrate and allows to determine free electrons and holes
interfaces play a very important role' in the distributions in the whole p-i-n structure at a
overall behavior, and the used material must be determined bias voltage.
considered quite different from bulk material or Following Tidje /8/, we analytically relate
what is in a Schottky diode. Besides, the few the trapped charge to the free charge
measurements and models on p-i-n structures concentration. In the following, we refer to
reported in literature, were performed under acceptor-like traps, with no loss of generality,
reverse bias and/or with high testing frequency being the donor like expressions obtained with
/6/, in order to neglect the contribute of the similar procedures. From SHR theory we get the
diffusion charge. Such a condition allows eqUilibrium net generations Gp and Gn as given
analytical simplification in modeling and thus by:
easier interpretation of experimental data.
In this paper, we present a theoretical model
of C-f measurements under direct bias voltage in (1)
order to determine the slopes of the exponential
Gn=Cn[nlN--noN O]
band gap tails in the actual material of the p-i-n.
Furthermore, our model, with the implementation where N-(E t) is the density of trap states
of acceptor-like and donor-like dangling bonds, with energy Eh occupied by an electron, N°(Et) is
allows us to highlight the separate contribution to the density of empty trap states (neutral states),
the capacitance arising from deep trap states and N(E t ) is total amount of trap states in the bandgap
from band tails.

184
with energy Et (it holds: W(E) + W(E) = N(E», Cp
( Cn) is the capture cross section for holes (5)
(electrons), Po (no) is the free holes (electrons)
concentrations, PI (nl) is the density of thermally
emitted holes (electrons), as defined by:
After some algebra one obtains the following
expression for perturbation of trapped electrons
PI=Nv exp[-(Et-Ev)/kT] as a function of the stationary trap states
occupancy and of the signal pulsation 0):
(2) o
cnnN -cppN
nl=Nc exp[ -(Ec-Et)/kT] v (E) (6)

As usual, Nv (Nc) in (2) denote the density of

states at the edge of the valence (conduction) where the expressions for N-(E) and N°(E) are
band. Using detailed balance holding in thermal those found in stationary case. To find the total
equilibrium condition, emission and capture trapped charge we assume exponential band tails
processes of holes and electrons are equalized. as given by:
We obtain the following expressions for the free
and occupied states at energy E as functions of
the free holes and electrons concentrations:
(7)

(3) where Tc is the slope of the exponential tail

in the forbidden gap.
(c PI + c nO) To obtain the total variation nt of trapped
W(E)= P n Nc(E) electrons in the ,acceptor-like states at a fixed
[c n ( nO + n 1 ) + cp ( PO + PI)]
position in the p-i-n structure, we integrate v-(E)
To obtain the differential capacitance between Ev and Ec. In (6) pi can be neglected,
measurements, we superimposed a sinusoidal holes emission from acceptor-like states being
signal with pulsation 0). Thus, we obtained a free very small.
carrier perturbation and a trap states occupancy It is possible to define a demarcation level
variation as given by: E tn, above which the emission processes are
predominant, while recombination processes are
n= n exp (j 0) t) + no predominant below. With this assumption,
approximated integration can be performed as
the sum of two integrals as given by:
Ii = P exp (j 0) t ) + Po

N'(E)=v exp(jO)t) +N-(E)

(4) nt= fE.

Ev
v dE=

where n, p and v - (E) is the variations

respectively of electron and hole concentration
and of density of occupied acceptor-like trap The level Etn is defined as
states amount at the energy level E, while the
superscript _ denotes the total quantities. Etn= Ec + KT In«cnno + cppo} + 0)2)112/cnNv) (9)

Obviously, it still holds: N-(E)+No(E)=N(E). At Such an approximation was evaluated and

this point, we wish to point out that the free accepted after numerical solution of the original
carriers concentration is calculated using a quasi- integral.
static approximation, coming from the "double In order to take into account also the
carrier model". This assumption implies that free contribution to the capacitance due to charge
carrier distribution is reached in times much trapped by dangling bonds we assumed a ~
shorter then the signal periods, and that the distribution both for donor and acceptor-like
current due to trapping can be neglected with traps. We used the same previous relations to
respect to the large free carrier current under a calculate the trapped dangling bonds as a
direct bias. function of the free carrier.
Perturbed generation rates g, and g. combine Numerical integration of free and trapped
so that the continuity of trapped charge is charge over the thickness of the intrinsic layer
verified: finally gives the total diffusion charge.

185
3. RESULTS Table 1 reports contributions to the total
capacitance due to band tails (BT), DB/A, and
DB/D states at Vbias=O Volt and Vbias=0.6 Volt at
Our model allows to simulate capacitance the frequencies of 0.1 Hz, 10 Hz, 20 Hz. In the
behavior versus frequency under direct bias case of Vbias=O and f=O.1 Hz deep traps
voltage. contribution to the total capacitance is about 80 %
We separate three different regions: low and slightly decreases with growing frequency; at
frequency (0.1 + 10 Hz), medium frequency (0.1+1 20 Hz it is still about 25 %. On the contrary at
kHz) and high frequency (0.01 + 1 MHz) regime. Vbias=0.6 V, deep traps contribution is only 10 %
Following the "double carrier model" /7/, the at 0.1 Hz and vanishes above 10 Hz.
The different contributions due to DB/A and
material parameters used in our simulation were
DB/D again depend on different doping of n+ and
assigned as follows: p+ zone.
Na=2.3E11 cm- 3 , Nd=I.E13 cm- 3 ; ni=I.E7 In figA we report capacitance measurements
cm- ; Nc=Nv=1.E21.cm- 3 ; Sn=1.E8 cm- 1 V,
3 on a a-Si:H solar cell at Vbias=O Volt and
Sp=I.E8 cm- 1 V; Iln=l cm- 2 V-I s-l, IlP=0.25 Vbias=0.6 Volt, in the frequency range from 50
cm- 2 V-I s-1;'tn='tp=I.E-8 s-l; Hz up to 1 KHz. This device was realized at our
where Na and Nd are the effective doping on laboratory by a three-chamber RF Glow Discharge
the p-Iayer and n-Iayer respectively, ni is the from GSI. Comparison between experimental
intrinsic concentration of the i-layer, Sn and Sp results at Vbias=0.6 Volt and theoretical results
are the recombination velocities at the interfaces presented in fig.2 at Vbias=0.6, Tc=500 and
i-n and p-i respectively, Il nand Il p are the Tv=650 shows a very good agreement. The
experimental value at Vbias=O Volt (- 16 nF) is
electron and holes mobilities and finally 'tn and very close to the transition capacitance value, as
'tp the electron and holes lifetimes. A thickness of deduced by the formula valid for crystalline
6500 A was assigned to the intrinsic layer while silicon.
the n and the p-doped regions were assumed 200
A thick. With these values, transition capacitance 4. CONCLUSIONS
is about 15 nF.
The values of band tails slopes and capture In this paper we present a new analytical
probability were Tc=500 OK, Tv=650 OK, cn=5.e-14 model for the direct bias capacitance in p-i-n
cm 3 s-l, cp=I.13e-13 cm 3 s-l. solar cells. The model calculates the contributions
To put in evidence dangling bonds due to band tails and dangling bonds trap states.
contribution to the capacitance we introduced in Experimental measurements of capacitance
the forbidden gap acceptor-like trap level (DB/A) versus direct bias voltage were presented,
at 1 eV and donor-like trap level (DB/D) at 0.72 showing a very good agreement with theoretical
e V, with a density of I.E 16 cm-3 and capture predictions.
probability equal to I.E-8 cm3 s -1. The presented model require.s a very short
In fig.l we report capacitance at low computer time, thus allowing an easy comparison
frequency at Vbias=O Volt and Vbias=0.6 Volt. with experimental measurements; future
Some behaviors appear quite evident: 1) the application is to extract material parameters (tails
more the cell is directly biased the higher is the slopes and DB concentration) by a fitting
capacitance; in fact charge diffusion is enhanced procedure.
by band flattening. 2) Growing up frequency,
both the curves drop regularly. The 0.6 V curve ACKNOWLEDGEMENTS
steeply decreases, while the 0 V curve decreases
more slowly because the contribution of the Authors wish to thank Prof. F.P.Califano for
diffusion capacitance is smaller. helpful discussions and encouragements.
In fig.2 we present calculated capacitance at
medium frequency and Vbias=0.6 Volt for three REFERENCES
different values of band tails slopes . We can note
that the curve with Tc=650, Tv=500 is higher / 1 / P.Viktorovitch and D.Jousse, J. of Non-Crist.
than the other two, while the curve with Tc=500 Sol., 35 & 36, 569-574, (1980);
Tv=650 is superimposed to Tv=Tc=500 one. The
large influence of the conduction tail, with respect /2/ P.Viktorovitch and G.Moddel, J.of Appl.Phys.,
to that of valence band, can be explained by the 51 (9), 4847- 4854, (1980);
two different doping concentration of the p+ and
n+ layer. /3/ P.Viktorovitch, J.of Appl.Phys., 52 (3), 1392-
Fig.3 shows capacitance behavior at high 1404, (1981);
frequency, Vbias=O and Vbias=0.6 Volt. It is
evident that above 200 kHz the two curves tends /4/ T.Tiedje, C.R.Wronski and J.M.Cebulka,J. of
to overlap and the main contribution to total Non-Crist. Sol., 35 & 36, 743 -748,
capacitance is due to the transition capacitance. (1980);

186
 50
/5/ J.D .Cohen, 'Semiconductors
Semimetals',vo1.21, part C, Chapt.2,
and
40
... v-a.
.0- Volt
V.0.6 Voh
Academic Press ,Orlando, 1984;
~

/6/ R.S.Crandall, Phys. Rev. B, 36 (5), 2645-2665,

S
..
u
c 30
(1987);
.
.!!!
'u ~

1"-
Q.
/7/ F.lrrera, F.Palma, Solid-St. Electron., to be '"
<.J 20
published

/8/ T.Tidje, Appl. Phys. Lett., 40 (7), 627-629, 10

(1982). a 200 400 600 800 1 0 00
Frequency (kHz)

Figure 3: Theoretical capacitance versus

frequency in the high range at Vbias=O Volt and
Vbias=0.6 Volt.
1500

...
.0- v-a Volt
V_0.6 Volt 100

~
S
1200

900
80
...
.0- V. O. Volt
V. 0 .6 Voh

tl
c
~
S
U
..
.!!!
Q.
600 8c
.
60

<.J 1i 40
300 '"
Q.

'"
<.J
20
a
0 2 4 6 8 10
Frequency (Hz) o +-~~----~~~~ __~~~__~
a 200 400 600 800 1000 1200
Frequency (Hz)

Figure 1: Theoretical capacitance versus

Figure 4: Measured capacitance against frequency
frequency in the low range at Vbias=O Volt and
Vbias=0.6 Volt. on an a-Si:H solar cell at Vbias=O Volt and
Vbias=0.6 Volt.

300,------------------------------, Frequency ( Hz)

- Tc=Tv_500
.... Tc- 500. Tv- 6S0
-0- Tc- 6SO. Tv_soo
G:'" 200
S 0.1 10 20
OJ
u
c
.!!!
'"'"
.
Q u...
to
Q.
100 c
<.J Q)
u
C
.....tI3
U
a 200 400 600 800 1000 1200 C.
tI3
CDB/A
Frequency (Hz) tI3
U

Figure 2: Calculated capacitance versus frequency

in the medium range at Vbias=0.6 Volt for three
different values of band tails slopes
Table I: Contributions to the total capacitance due
to band tails, acceptor-like dangling bonds, a~d
donor-like dangling bonds states at Vbias=O and
Vbias=0.6 at the frequencies of 0.1 Hz, 10 Hz, 20
Hz.

187
IlJfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 1S-12 APKIL l~~l Ll~HUN. PUKWliAL

W.J. KOPETZKY·, R. SCHWARZ

Technische Universitat MUnchen, Physik Department E16

James-Franck Str. I, WD-8046 Garching, "6' (089) 3209 2332

Abstract
We examined with the help of numerical simulation the internal collection efficiency q(V)
of an a-SiC:H/a-Si:H pin solar cell with different buffers: a constant, a graded, an inversely
graded and a mixed buffer. The defect densities of the buffers are assumed to be the same.
Interface defects at the heterojunction a-SiC:H buffer and a-Si:H i-layer are neglected. To
obtain agreement between measurement and simulation an electrostatic field reversal at about
0.7 V is required. In comparison to the i-layer the surface regions dominate the recombination
and have a high number of defects. The q{V)-curves are partitioned into three ranges: The
surface range, the transition range and the diffusion range. A solar cell with a mixed buffer
of a thickness of 20 nm shows both the highest efficiency under AMI illumination and the
best internal collection efficiency under weak illumination.

1 Introduction 2 Numerical Simulation

Solar cells have been improved by introducing an undoped We solve selfconsistently the Poisson equation and the semi-
a-SiC:H buffer layer between the a-SiC:H p-Iayer and the conductor continuity equations. The Poisson equation is discre-
a-Si:H i-layer. The explanation of this improvement is avoid- tized with help of the Taylor series expansion. The continuity
ance of both boron contamination and additional defect states equations are discretized in two steps: The divergence operator
due to strained material of the first few tenths nanometers of is expanded with help of the Taylor series, the current at mid-
the i-layer /1/. The boron contamination would lead to addi- interval points is determined analytically /5/.
tional defect states and to a shift of the dark Fermi level to For the description of recombination and trapped carriers
lower energy. The latter reduces the electric field maximum in we use donor- and acceptorlike defect states /6/. The continu-
the p-region /2/. The a-SiC:H buffer itself can have a constant ous density of states consists of valence and conduction band
mobility ~ap (constant buffer) or a gap linearly changi.ng with tail states and two Gaussian like distributions of dangling bond
position (graded buffer). The abrupt buffer is expected to be states.
connected with a defect interface layer at the heterojunction The integral model is solved by integration over all defect
consisting of the buffer layer and the i-layer /3/. states. This model consumes a high quantity of computer time.
We want to compare different buffers on their effect on To bypass this problem we compare an analytical model /6/
the internal collection efficiency q(V). The effects on the solar and a l4-level model In the latter model the tails are re-
cell efficiency will be discussed below. Additionally we intro- presented by 4 levels each and the Gaussian like distributions
duce an inversly graded buffer and a mixed graded buffer. by 3 levels each.
Figure I shows the different buffers. The qualitative agreement of both the analytical model
and the 14-level model with the integral model is satisfying.
p-SiC:H I a-SiC:H The a4vantage of the 14-level model compared to the analyti-
cal model is mainly the guaranty for short calculation times
f i h.~:~!:H- and stable numerical solutions over a wide range of parameter
20 eY,
'72 Y
J. :1.v .~... CB: constant buffer variations.

Ec~ ..... J Solar Cell Structure and Model Parameter

The solar cell consists of an a-SiC:H p-Iayer and an atSiC:H
Ey GB: graded buffer buffer layer with thicknesses of 10 and 15 nm, resp. an i-layer,
-vt-..... a transition layer between i-layer and n-Iayer and an a-Si:H
n-Iayer with thicknesses of 5 and 15 nm. The total thickness
is eqllill to 0.6 11m.
1GB: inversely graded buffer The mobility gap of the a-SiC:H layers is chosen to be 20
eV, of the other layers it is equal to 1.72 eV. The valence band
tail slope Euv is coupled with the dangling bond (DB) density
MB: mixed buffer n/ and the conduction band tail slope EuC' The ratio of the
tail slopes is chosen to be 2/3. For instance, a valence band tail
Fig. 1: Different buffers slope of 65 meV is connected with a DB-density of about 1017
cm-3 and a conduction band tail slope of 44 meV. The value of
The valence band offset is assumed to be zero /4/. The as- Euv of the doped layers and the i/n transition layer is chosen
sumptions are, that all buffers have the same defect density- to be 65 meV, of the buffer layer 55 meV and of the i-layer
and there is no additional interface defect density in the case 45 meV. Fig. 2 shows the band structure of a solar cell with
of the constant, the inversely graded and the mixed buffer. an constant buffer.
Before we examine the influence of different buffers on
q(V), we discuss the q(V)-curves of a solar cell with the
constant buffer.

188
 ·6~.~~~,-=~--------------~~~~
: ~ • .siOH p-!.ler .-S-tH 1>.lalOft Darkn~
20000
:'7" •.siCH burror 1010r IrolUilioQ 101er~

.0
: -_ _ _ _ _ _"_.s_i±l_i._II_y_er_-.: Ec::..._ II S 10000
0
-.6 ..........
:>
'--'
J:c1CClrOa
Loquui-Fermi '0
.....II)
....
~
0

...;
al
J:holo
1;; -10000
"qll&Si·Fermi

-2.0
-20000
Ev
o 2 3 4 II II
2 3 • 5 II
Position [em] _10- 5
Position [em] _10- 5

Fig. 2lJaDd structU£e of a solar cell with a constant buffer. Fig. 4: The electrostatic field in the dark. U in steps of OJ V
U=0.6 V; Int.. 10 12/cm 251 ).=400 nm
The relative change of the electrostatic field under weak
The electron and hole mobilities are chosen to be 10 and 2 illumination (lQ12/cm"2s) are a few percent.
cm2 fV s. resp. The capture cross section of neutral trar ClN The crossings reflect the transition of voltage range I from
in the i-layer is 10-17 cm2 • of the other layers 10-16 cm . Tile 0.0 to about 0.6 V to voltage range II from about 0.7 V to 1.0 V.
ratio of Cl C (C: charged trap) to ON is equal to 100. In range I the collection of carriers generated by red light is
more effective than collection of carriers generated by blue
~ Internal Collection Efficiency light. Blue light is more sensitive to the p-region than red light.
There are two reasons: First. the surface has a higher number
The internal collection efficiency q(V) is measured under of defects compared to the i-layer. Second, in the p-region elec-
weak illumination with different wavelengths 181. The q(V)- trons are minority carriers and have shorter mean lifetimes
curves of different wavelengths cross in one point at about than electrons in the i-layer. For very long wavelengths the
0.6 V. The function q(V) is the normalized photocurre.nt. It is internal collection efficiency decreases compared with shorter
normalized to a value of -I at a voltage of -3 V. We Simulate wavelengths. because light with very long wavelengths be-
the measurement /81 of q(V) of a solar cell in annealed state. comes sensititve to the n-region. We name range I the surface
Fig. 3 shows the simulated function q(V). range.
.2~~~~---- __________________ ~

We presume, that in the voltage range II the electrostatic

Constant Buffer field reversal influences more strongly the collection of carri-
ers generated by red light than the collection of carriers gener-
ated by blue light. The field leads to the transition from prima-
.0 I~l: ~oii I ~1Il2.- - - - - - - - - - - - - - DiffuSiON 1 ry (q(V) is negative) to secondary (q(V) is positive) photocur-
rent. The field maximum in the p-region. which is mainly re-
t>.. RANGE / sponsible for the separation of photo electrons and holes gener-
o -.2 ated by blue light. will only be weakened by increasing voltage.
~
....o
II) Additionally. carrier injection may contribute to the strong de-
crease of the carrier collection with increasing voltage.
.............
I%.l
-.'
In the transition range, in which the field vanishes in the
i-layer. the transport of carriers is determined by diffusion. In
the voltage range II diffusion remains the main transport
..... mechanism. The field peaks near to the doped layers. work as
(; -.8 SURFACE RANGE TRANSITION suction pumps. In the i-layer the reversed field is weakly in-
u RANGE creasing with increasing voltage (Fig. 4). Range II is named

-
...; the diffusion range.
~ The q(Y)-curves are composed of photo drift and diffusion
-.6 currents of electrons and holes. We cut out a slice of the solar
cell and plot drift and diffusion currents as a function of volt-
age. Fig. 5 shows a slice of the middle of the solar cell. The
light wavelength is chosen to be 650 nm and the light intensi-
-1.0~~~
0.0 .2 .4
- ~-=---=~
.6 .8 1 .0
.. ty is equal to 1012/cm2s. The photocurrent Iph is the unnor-
malized q(V. ),=650nm)-curve from fig. 3. Photocurrents with
Voltage [V] a negative sign are primary. with a positive sign secondary.
Fig. l- The internal coJJeetion efficiency q(V) for a solar ceJJ The sign change of the photocurrent is determined by the
with a constant buffer. The wavelengths ). are equal to 400. photo electron drift current.
500, 580 and 650 nm. ~ Comparison of Solar Cells with Different Buffers
The requirements to obtain this q(V)-cuves are: In this chapter we want to discuss the internal collection
- field reversal at low voltage ('" 0.7 V) efficiency q(V) of solar cells with a constant (CB). a graded
-weak recombination in the i-layer (ON =10-17 cm 2) (GB), an inversely graded (1GB) and a mixed buffer (Ma). For
- weak illumination (int. = 1012 / cm 2 s) this purpose we show the electrostatic field and the photo re-
combination rate at zero voltage. Finally the solar cell
The first field reversal at any position in the cell occurs at efficiency is represented.
0.73 V in fig. 4. The q(V)-curves cross in a voltage range from Fig. 6 shows the internal collection efficiency q(V).
0.6 to 0.74 V in contrast to the experiment 18/. The crossings
occur mainly before the field reversal 19/.

189
 1.0
SUCE DIAGRAM IDl - 1012 lem'2s 500r--a.~~~CH~~----------~---------u--
.-on
--v---'

....1 l -6S0nm p-layer lul.IOI2 / c:mls

Pas. . 0.3 JIm ,........, l ·400 DID
....
0
1.0 El
o
* Idrlrt \
pb' D \ ...........
O r------7-----f--~~--------------__1
........, \ ~
..........
I :
N \
( .5 \
J'dirr ,-.. ~ EOGB) . E{CB)
\
,,
ph-p
El c:I
\
: E{MB) - E{CB)
r:r
0 U -500
........... ............\ : E{GB) . E{CB)
-< .0
...,
'----'
~~.~ .. -- ..... ---- ,.-..
~ c:I - 1000
IV l;b
::s
::s'0'""''
-.oo
- .5 ;
Ipb / rph.-po
drif • ; c:I
/ ;

...,
0 /
;

"
0-1.0
c:I -1500

--- "
.d ,dirr '--'
Il-. ph-..
....
/ ~

-1.5~--~-------------------- __--~--~~ E(-) : .leclrostalie fiel6

- 2000 ~--~~~--------~------~--------~
0.0 .2 .4 .8 .6 1.0 o 2 3 . 5
Voltage [V] Position [em] *10- 6
Fig. 7: The relative differences of the electrostatic fields of
Fig. 5: Slice diBJ,ram of a solar cell under illumination (1 = 650 cells with the GB. the 1GB and the MB compared to a cell
n~IIn.t. = 10 1 Vcm.<'s, I = 0.3 (1m). It shows photo drift and with the CB. The first section is the a-SiCH p-layer, the
diffusion current components of electrons and holes. The field second the a-5iCR buffer and the third the first 25 nano-
reverses its sign at this ~tion at a value of 0.77 V. meters of the a-SiR i-layer. Zero voltage is applied
The built-in effective force on electrons due to the bandgap
.2r---------------~--~~~--------~ variation is in the case of the OB and the MB equal to 1.87·l(P
CORStao~ Graded, lowsely Graded, Miled Buffer V Icm and 0.93.10 5 V lem. resp. The effective potential drop is
equal to 0.28 and 014 V over a buffer thickness of 15 nm. resp.
The 1GB leads to the inverse effective force than the GB
........, .0 buffer.
....
'----J
Inl • 10I2 lem'l"
The relative differences of the electrostatic fields in all
cases are small compared to the effective forces.
>. Fig. 8 shows directly the effect of the different buffers on
0 - .2
~ the photo recombination rate R..ph(x) of electrons and holes.
IV The function Rph(x) is the difrerence of the recombination
'0 rate under illumination and under darkness.
;::::
..... -.4 ., 5,------,,....---------------------,-:--------,

--
>"I '0 I t 1.8-1018 em-) $'1 U • 0.0 V
1.400 om .... 1 Jul. J0121cm2.
............ 1 .400 nm
.. 1GB
....... .... .................
• I
I
0 -.8 I
U 1

-
....; I lomselr Graded Burfer
1
I::l 1
- .6 1
1
1

t~~~~~~~~~~~~.~~~~~.:.:
~.. '=..~~~__--__---1
1
1
-1.0 I
0.0 .2 .4 .6 .8 1.0 1
Voltage [V] I
1
I
j--Graded Burrer
Fig. 6: The function q(V) of solar cells with the Cn. the Gn. \
I
the 1GB and the MB. The light wavelengths are 4(}() and 650 nm. o
q)
I
\
fi
/.
p::l i!
The red light (1 = 650 nm) q(V}-curves show almost the ; !
same dependences on the voltage. In the surface range under ...,o COostant Buffer
illumination with blue light q(V) is much smaller in the case o ---'-,.---- Mixed Buffer
of the cell with the 1GB than q(V) of cells with other buffers. .d
Il-. 0 I--~-~--~.....:..il;--_ ......._ _ _....--l
The solar cell with GB and MB have in this range the best o
internal collection efficiencies.
In the diffusion range the splitting of the q(V,l=400 nm.
CB}-curve and the q(V.>.=400nm.GBIMB)-curves is getting
stronger with increasing voltage. Fig. 8: The photo recombination rate at zero voltage. The used
Fig. 7 and fig. 8 points to this behaviour. Fig. 7 shows the buffers are the Cn. the GB. the 1GB and the MB
relative differences of the electrostatic field of cells with the
GB. the 1GB and the MB in comparison with the constant T~e Rph(x}-curves of the cells with the CB. the GB and the
buffer at zero voltage. Negative values mean. that the field at MB dlffe~ by about a ~actor 2 The photo recombination rate of
this position leads to a stronger photo carrier separation for a the cell with th,e I~B IS a factor of 1000 higher than the other
cell with the GB. the 1GB and the MB compared to the cell phot~ r~omblflahon rates. The peak of the photo re-
with the CB. Positive values mean the opposite effect. comblflahon rate of the cell with the OB at the transition of
th~ ~uffer and t~e i-layer (position = 25 nm) follows from the
missing conductIOn band offset compared to cells with the
other buffers.

190
 A slice diagram (fig. 9) illustrates the competition of the
photo drift and photo diffusion current of electrons of cells Q Conclusion
with the CB and the GB. resp. Photo diffusion works against The main requirement to obtain the measurement of the
photo drift. internal collection efficiency q(V) in the simulation is the field
The photo drift currents follow roughly the field reversal reversal at a voltage of about 0.7 V.
The cell with the CB shows a stronger splitting of the photo We partitioned the q(V)-curves into the following ranges:
drift and diffusion currents than the cell with the GB. The surface range from 0.0 to about 0.6 V. in which illumina-
tion with blue light is more sensitive to the surface than red
ConsWll G",ded Burr.r
light. In the transition range the q(V)-curves cross and the elec-
trostatic field reversal occurs. The third range is the diffusion
range, in which diffusion is the main transport mechanism.
For the examination of the influence of the buffer layer
on q(V) we introduced different buffers: a constant, a graded.
an inversely graded and a mixed buffer. The defect densities
of all buffers are chosen to be the same.
The solar cell with a 20 nm thick mixed buffer leads to the
highest efficiency and to the highest internal collection ef-
ficiency. The mixed buffer has two desired properties: The
built-in effective force on electrons leads directly to a separa-
tion of electrons and holes. which are generated in the buffer
layer. A back diffusion of electrons into the p-Iayer like in the
case of the graded buffer is excluded because of the conduction
,.
,.,.
band offset.

1 Acknowledgements
/
: 1,IIrr
f
I.l 10,2/=2,
~
0

0400 OlD
CB
• W.J. KOPETZKY acknowledges support from SIEMENS
AG through a doctoral fellowship. We thank H. PFLEIDERER.
W. KUSIAN and W. KRUHLER for numerous discussions.
~ References
ph- D Pos.o OJ pm
.2 .• .8 .8 1.0 II/ R.R. Arya. MS. Bennett, A. Catalano and K. Rajan.
Voltage [V] Pree. of PVSEC-3. Tokyo (1987) 457
12/ W. Kusian and H. Pfleiderer. Int. Meet. on Stability of
Fig. 9: Slice diagram of the electron photo drift and diffusion Amorphous Silicon Material and Solar Cells. Denver
currents of cells with the CB and the GB. The slice is cut out (1991). to be published
in the middle of the ceJJ 131 F. Smole. J. Furlan and S. Amon. Proc. of PVSEC-5.
Kyoto (1990) 339
Back to a general point of view. what are the effects of 141 R.-C. Fang and L. Ley. Phys. Rev. B 40 (1989) 3818
different buffers on the solar cell efficiency? 151 S. Selberherr. Analysis and Simulation of Semiconductor
Solar cells with a mixed buffer have .the highest efficiency Deyices, Springer Verlag, Wien (1984)
compared to cells with a constant, a graded or an inversely 161 M Hack and M Shur. 1. AppL Phys. 58 (1985) 997
graded buffer (Fig. 10). Cells with the MB and the GB show nl M Stutzmann. Phil Mag. B 60 (1989) 531
almost the same dependence of the efficiency on the buffer 181 W. Kusian. H. Pfleiderer and W. Juergens. Proc. of 9 th
thickness. The buffer thickness. which leads to the highest EC PV Solar Energy Conf. (1989) 52
efficiency. is 20 nm. The cell with the 1GB shows a strong 191 H. Pfleiderer. W. Kusian. E. GUnzel and J. Grabmaier.
decrease in efficiency with increasing buffer thickness. Proc.. of 20 th IEEE PVSC (1988) 180
7.0r------------------------------------,
AMI

/~~......
ff t .......~ ~~
• l.i.:..::.: •••• ,
hiab.., ~ ......... Mind Buffer
~
()
cmacncy ~ ....
-'-'=':~ •.
~ 8.0 Graded Buffer
.....
Q)

.....
C)

:::::
~
Conslant Buffer
5.5

I.,erselr Graded Buffer

6.0~-- ____~~--__--~--~--~--~--~~
o 20 40 eo 80 100
Buffer Thickness [nrn]

Fig. 10: Solar cell efficiency as a function of the buffer thick-

ness. The buffers used are the Cn. the Gn. the 1GB and the MB.

191
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE IH2 APRIL 1991 LISBON. PORTUGAL

OPEN-CIRCUIT VOLTAGE ENHANCEMENT IN MIS SOLAR CELLS BY THIN PLASMA-GROWN

TUNNELING OXIDES

J. Meier, G. Kragler, G. Willeke, E. Bucher

Fakult!i.t ffir Physik, Universit!i.t Konstanz, D-7750 Konstanz, Germany

H. Keppner, D. Fischer, V. Viret, A.V. Shah

Institut de Microtechnique, Universite de NeucMtel, CH-2000 NeucMtel, Switzerland

ABSTRACT. A microwave water vapour and oxygen plasma has been used to oxidize
a-Si:H at low temperatures. The grown oxides were analysed by Auger electron
spectroscopy (AES), ellipsometry and by determination of the wetting angle with a
water droplet. A typical thickness of 16 A was obtained from AES Si0 2 /Si-peak ratios
as compared with 11 A obtained by ellipsometric studies. As expected, the plasma
oxide reduces the MIS forward and reverse dark current and leads to an increase in
V oc and Jsc . The photocurrent enhancement is due to an increased blue response of
the cell. The most pronounced effect of the plasma grown oxide was observed on
devices with lightly phosporus doped i-layers. As demonstrated earlier Pt/SiOx/<i-n+>
a-Si:H MIS solar cells with open-circuit voltages up to 900 m V and AM1.5 efficiencies
of 5 % have been obtained. The V oc of cells stored for three years remained nearly
constant.

1. INTRODUCTION oxide free in terface.

It is known from crystalline silicon (cr-Si) that a thin

tunneling oxide between the semiconductor and the
Au soo A
semitransparent metal leads to an enhancement of the
built-in-potential and in turn to a larger V00 of the solar Ir or Pt 200 J.
cell /1-3/. This so-ca.iled MIS (Metal-Insulator- II' or Pt 30 A
Semiconductor) structure is an alternative to conventional
pn junction devices. Whereas for cr-Si extensive studies
have been carried out concerning the preparation and the
passivating properties of thin tunneling oxides, the
s ;
n+
SOOO A
500 1
2 a-Si:H
~
hydrogenated amorphous silicon (a-Si:H) system has only
been studied to a much lesser extent. Work by different stainless stell
groups on evaporated oxides /4-6/ led mostly to unstable
devices.
ell!drital l.. d
In the course of a-Si:H pin solar cell optimisation
studies, it could be of interest to replace the currently used Fig. 1: Structure of the a-Si:H MIS solar cell.
100-300 A thick a-Si:C:H buffer layer between the p- and i-
layer by a thinner (15-20 A) tunneling oxide, both for
single and tandem solar cell structures. The possible
advantage lies in the expected enhanced diffusion barrier as

-'
well as in the improved optical properties.
The oxidation of a-Si:H is not straightforward since it Inl .. t
requires a low temperature process because of hydrogen
,,
Cavity
Pump
effusion above 250-300' C. In previous work we have shown
z---+''
/7-10/ that a treatment of the a-Si:H surface in a H 2 0
microwave plasma led to a significant enhancement of Voo· ,,_ :
In the present work, the plasma oxide is compared Fig. 2: Open oxidation quartz-tube reactor with
with native oxides and the difference as measured by AES pump, gas supply and microwave power.
and ellipsometry reported. In addition, its macroscopic
properties and its influence on solar cell properties have
been investigated.
3. EXPERIMENTAL RESULTS AND DISCUSSION

2. PREPARATION METHOD 3.1 Auger electron spectroscopy (AES)

AES is a very sensitive standard method for a
Fig. 1 shows our MIS solar cell structure. After the chemical analysis of surfaces. Fig. 3 shows AES spectra of
deposition of the n+- and i- a-Si:H layers on polished different a-Si:H surfaces. clearly visible is the SiOl-peak at
stainless steel substrates in a VHF silane plasma reactor 76 eV for the three oxidized samples. This peak is absent
/11/, the surface was treated in a specia.ily built microwave for the etched specimen, as expected. Connected to the
apparatus (fig. 2). Subsequently a thin semitransparent oxide signal, the oxygen peak at 510 eV is observed in the
metal film (Pt, Ir) was evaporated. In order to evaluate the former cases. The Si-peak at 92 eV is largest for the etched
influence of the oxide, the MIS structures were compared sample or when a freshly prepared a-Si:H surface is studied
with pure Schottky barrier type cells, i.e. those with an and is reduced for oxidized specimens.

192
 100 200 300 400 S 00 60 80 100 120

~30 Ellipsometry: • 0 etched

p l o s r"r"lO ox i d . • 0 fresb

:i fIl
AES, •
Acetone -
.....Ii
fIl

~lli
~ 20

1.. .
.!I:
.~ 15
....
.c
Q) 10
'0
'>1
o 5

0L-~~~~~~10~~~~~1~0~0~-i~~10~0~0~
Time (days]
Fig. 4: Native oxide growth in air of an etched and a
fresh a-Si:H surface.

Plasma oxidation changes the ellipsometric angles L1

und rJt in contradiction to the model used for native oxides.
100 200 300 400 500 60 80 100 120 It was found that after etching a plasma oxide the a-Si:H
Elactron Enargy reV] surface appears to be more 'bulk'-like regarding the optical
constants, i.e. plasma oxidation appears to modify the
Fig. 3: AES spectra from different a-Si:H surfaces. It a-Si:H. It possibly leads to a reduction in surface
is clearly seen, that the Si0 2 -peak at 76eV emerges roughness . This behaviour is just opposite to that of native
after plasma treatment and for native or thermal
oxides: here etching reveals to be the inverse process of
oxides, while the Si-peak at 92 eV is reduced
correspondingly. A further interesting point is that no oxidation in the sense that the grown oxide is removed
carbon contamination (at 272 eV) is observed for the without change of the substrate /15/.
plasma treated surfaces, although these samples were In previous work we have shown that for a specimen
exposed to air shortly before bringing them into the position in the plasma reactor z > 20 cm, solar cells with
AES vacuum system. large Voc-values could be obtained. Fig. 5 shows d ox as
determined by ellipsometry as a function of z. d ox
An estimate of the oxide thickness d ox can be increases strongly for a decreasing z and is practically
obtained from an analysis of the Si- and Si0 2-signals. Such constant for z > 20 cm. A similar dependence has been
an investigation has been very thoroughly made for cr-Si found in AES measurements, whereby the determination of
/12, 13/, showing that d ox is related to the ratio (! of the relatively thick oxides for z < 10 cm is critical because of
oxide peak to Si-peakby the following equation: the resolution of this method. It appears that for z < 20 cm
the oxides are too thick for good quality MIS cells. This
d ox = A· In({!/E + 1); A = 6.5 A, E = 0.2 view is supported by lower fill factors for cells with small z-
values. Ellipsometry yields an oxide thickness of about 9-
Assuming that this relation applies to our material, a
11 A for standard plasma treatment conditions, which is
typical plasma oxide thickness d ox = 15-16 A is obtained
slightly less than the corresponding AES values.
for standard conditions (e.g. z = 23 cIrl, see fig. 2).

3.2 Ellipsometric studies

J,,
Ellipsometry is a very useful optical method for the
characterisation of thin layers in the sub-nm range. In the ~30 H2 0 - Plasma
present study a single wavelength ellipsometer was used. ~
lor = 20 min .
The results showed a fundamental difference between a -d' 25 POI = 10 mbar

plasma oxide and a native oxide grown on a-Si:H samples.

fIl
fIl
:\\,
Whereas the latter case can be properly modeled by a ~ 20
growing film on the unchanged a-Si:H surface, this is not .!I: "t ,
possible for the former. .~ 15
.c
..., : -.--
Fig. 4 shows the growth of native oxides as a function
of time. It appears that d ox continuously increases to Q)
'0
10 ---. ---. -. -J----. -- -- 1 -

values larger than 30 A within one year. A similar behavior 'x

o 5
has been observed by other research groups /13, 14/.
Comparative measurements with AES showed that the
growth of a native oxide should indeed proceed much °0~~-i-1~
0~~~2~0-L~~3~O-L~~4~0~-L~5~O

slower. The thus measured native oxide thickness was not Sample-cavity distance z (cm]
more than 14 A even for 3 year old specimens exposed to Fig. 5: Variation of the oxide thickness determined by
air. The one year old a-Si:H films with apparently 30 A ellipsometry as a. function of the position z in the
native oxide were therefore rinsed in acetone and the plasma. reactor.
ellipsometric measurement repeated. It was found that
after this procedure the largest fraction of the 'oxide' had In fig. 6 results of both methods are compared for a
disappeared. A comparison of the remaining oxide series of plasma oxidized specimens and those with a natiTe
thickness with AES results shows a good correlation, oxide. For the latter an excellent correlation is observed,
indicating that ellipsometry cannot distinguish between an which is not the case for the former. The observed
apparent surface contamination layer and a true native discrepancy of about 5 A suggests a possible modification of
oxide. We can therefore conclude that the growth of a the top a-Si:H layer due to the plasma treatment . This
native oxide on a-Si:H is a comparatively slow process. phenomenon is not properly understood at present and is

193
currently being studied with a spectroscopic ellipsometer.
5 :== ~rs !lolnnSlc. - .- 72mlr/ em'
-..:: -..:: ~rs! o - doped . • , (OOmlr /e m'
o
3:20
VJ
CzJ MIS
« 15 <:80
>. o
.0
(I)
(I) 10
Q)
C
~
o
..c:: 5
• • nalive oxides
o plasma - oxides
Eo- ~. 0
~-'--0
::-.'::'
2 -"---:-
0 ..L.4-,---,---::-':,-l-~-0..J..8--'--'--1...J.0
V
o0~-~-!5~~~1:':0:-'-"""""'-'-1"::5-'-~-'-2L.0-'-'-'----L.J25
Fig. 8: muminated 1- V characteristics of MIS (plasma
Thickness by Ellipsometry [Al oxide) and Ir Schottky barrier solar cells.
Fig. 6: Comparison of oxide thickness determined by
ellipsometry and AES measurements. presented in fig. 9, where the external quantum efficiency
(Q.E.) is plotted as a function of wavelength. The expected
3.3 Surface wetting of a-Si:H by water decrease in the long wavelength response due to a reduced
It is well known from cr-Si technology that oxidized space charge layer in the P-doped samples /17/ is indeed
and non-oxidized surfaces show significant differences in observed. Fig. 9 also shows, that the increase in J se of the
the wetting angle e of a water droplet, which is related to MIS structures is completely due to an enhanced blue
the oxide thickness /16/. A correlation is observed in the response, which can be related to an increased' electric field
range d ox = 0 A (e = 90 • , hydrophobic) and d ox = 40 A at the surface and thus to a larger built-in potential. This
(e = o· , hydrophilic). The etched a-Si:H surfaces dearly increase in barrier height is directly correlated to the larger
demonstrate the expected droplet form (fig. 7). The native Voe·
oxide tends to be hydrophilic. A contact angle of between
40· and 55· was .measured in this case, which relates,
according to the cr-Si work, to d ox = 10-14 A, in good

<, ~:\\\.
agreement with our ellipsometry and AES data. However, 0 .5 MIS MS
MIS I IDlrmsJc
the behaviour of a water droplet on a plasma oxidized ~Ss l o - dop d
surface is completely different. Although with a d ox of only
11-16 A (from ellipsometry and AES studies, see fig. 6), the
water droplet spreads over the whole surface. This greatly
enhanced hydrophilic nature indicates a large concentration
~
c
... 0 .3 . . ,---X,
of built-in charges or dipoles in the plasma oxide, leading to Q)
,/ -',
" ,
,
...>
an increased attraction of water molecules. " \
~ 0 .2 "
""
3.4 Properties of MIS solar cells ',\
0.1 '~
Although plasma oxides prepared by dry oxidation ... ~...
(0 2 ) have the same thickness range (14-16 A), it is water ",..- ..
.....,- .:..~.
vapour which leads to the best results /7,8/. In what O. 0 '--''--:"'::7''--'--'--~-'--'----,-:-'--:-'--L~-=::=,,",,----'
400 500 600 800
follows the influence of the wet tunneling oxide on solar cell Wavelenglh
properties will be presented. In order to exclude possible Fig. 9: Spectral response measurements of the cells
effects from the .a-Si:H deposition process, MS and MIS shown in fig. 8.
diodes prepared in the same deposition sequence were
systematically compared. Besides the Voe-enhancement, an The effect of the tunneling oxide is most pronounced
increase in J se is also observed for the oxidized samples. for the phosphorus doped a-Si:H. Fig. 10 shows the dark
Fig. 8 summarizes data for two different a-Si:H depositions. I-V characteristics of the lightly n-doped a-Si:H solar cells.
The observed increase in J se is further investigated Clearly a reduced J o and an increased nand Rs are
by spectral response measurements. The results are observed for the MIS structure, similar to effects found in

Fig. 7: Differences of the contact angle of water on a) an etched a-Si:H surface, b) a native oxide covered a-Si:H surface
(specimen more than one year old), c) a water vapour plasma treated a-Si:H surface.

194
corresponding cr-Si devices. Here the larger n-values arise ACKNOWLEDGEMENTS
from the bias dependence of the barrier height /18/, which This work was supported by the BMFT project No.
is affected by the introduction of a thin insulating layer. 0328453 and EC project No. EN 3S-0107-D(B), which is
gratefully acknowledged. In Switzerland, it was supported
1o-'r-----------------------------------. through Swiss Federal Research grant OFEN EF-
- idS: Jo '" 3 .1xlO- 1oA/em' REN(87)09 + (88035). We would like to thank Prof. S.
n '" 1.15 • Wagner, H. Berner and B. Sailer for useful comments.
10 -. R. '" 13.9 Oem
---- 10115: Jo - 3 ,4xIO- II A/cm"
n '" 1.59
Ro = 52 ,8 Oem" REFERENCES:
/1/ J. Shewchun, M.A. Green, F.D. King, Solid State
Electron., 17 (1974), 563.
/2/ D.L. Puifrey, IEEE Trans, Electr. Devices, Vol.
ED-25 (1978), 1308.
/3/ P. Munz, R. Schultheiss, Proc. 17th Photovoltaic
10 -t -- __ Specialist's Conference IEEE (New York, 1984).
/4/ A. Madan, J. McGill, W. Czubatyj, J. Yang, S.R.
reVerse forward Ovshinsky, Appl. Phys. Lett. 37 (1980), 826.
10 - II L_~1-L~~_~0~.5~~~~0~-L~~0~.5~~~~LJ /5/ D.E. Carlson, IEEE Trans. Electron Devices, Vol.
V [V) ED-24 No.4 (1977), 449.
/6/ D.E. Carison, C. W. Magee, 2nd E.C. Photovoltaic
Fig. 10: Dark 1- V characteristics of the MS and MIS Solar Energy Conf. (Berlin 1979), p. 312.
cells with a slightly n-type a-Si:H base material (see
also figs. 8 and 9). /7/ J. Meier, H. Keppner, E. Bucher, P. Munz, H.
Curtins, A.V. Shah, S. Paasche, W.H. Bloss, Appl.
The stability of a-Si:H solar cells to light exposure is Phys. A, 46 (1988), 31.
of great concern. Fig. 11 shows the results of the first /8/ J. Meier, H. Keppner, G. Kragler, E. Bucher, P.
degradation experiments on our MIS solar cells. The Munz, H. Curtins, A.V. Shah, S. Paasche, W.H.
degradation appears to saturate after about 20 hours of Bloss, Proc. 8th European Photovoltaic Solar Energy
AM1.5 illumination. A partial recovery of the cell Conf. (Florenz 1988), 924.
performance is observed at room temperature, without /9/ J. Meier, H. Keppner, G. Kragler et al., Proc.
intentional heating. Euroforum ~ew Energies Congress and Exhibition
Saarbriicken 1988, 127.
/10/ J. Meier, G. Kragler, H. Keppner, H. Berner, M.
1.0 Simon, Y. Ziegler, H. Curtins, A.V. Shah, Proc.

~ 0 ,9
.......... - Conf. III Surface Modification Technologies
(NeucMtel 1989), 841.
;:l /11/ H. Curtins, N. Wyrsch, A.V. Shah, Electron. Lett.,
"'iii 23 (1987), 228.
> 0 .8
~----
/12/ B. Lang, P. Scholler, B. Carriere, Surface Science,
"~ 0.7
Q) 99 (1980), 103.
/13/ S.R. Kelemen, Y. Goldstein, B. Abeles, Surface
Br eak of 5d
0:1
......... voc Science, 116 (1982), 488.
E 0 .6
I-
............ J.., /14/ J. P. Ponpon, B. Bourdon, Solid State Electron., 25
o -FF
~'1
Degradation al AM 1.5 (1982), 875. --
'" 0.5 4>. 100 mW/em' /15/ G. Kragler, H. Berner, J. Meier et al., Proc. 9th
European Photovoltaic Solar Energy Conf. (Freiburg
0 .4 L-0~-'--'-2::':0:-'--'----.J
40 L.......l.-'--'-6.1..0-'-"""-....8..J.O
--'-.............J 1989), 1013.
light exposure time [hours} /16/ R. Williams, A.M. Goodman, Appl. Phys. Lett.,
Vol. 25 No.10 (1974), 531.
Fig. 11: Change of normalized parameters Voe, J se, /17/ T.D. Moustakas, R. Friedman, B.R. Weinberger,
FF and 7] of a Pt-MIS solar cell as a function of
illumination time, The cell is held at constant Appl. Phys. Lett., 40(7) (1982), 587,
temperature and the degradation is performed under /18/ E.H. Rhoderick, Metal-Semiconductor Contacts,
open-circuit condition. Oxford University Press (1978), 123.

4. CONCLUSIONS

By using a plasma surface treatment we achieved a

substantial improvement in cell performance of Schottky-
type barriers. The analysis of the a-Si:H surface by AES
and ellipsometry showed the formation of a very thin oxide
layer due to the plasma treatment. The thickness of this
layer was found to be in the range of 9-11 A when
measured by ellipsometry and 15-16 A determined by AES.
We correlate the V oe-enhancement with the thin
interfacial plasma oxide between the semiconductor and
the barrier metal. When comparing with native oxides, the
plasma oxides seem to have different properties. For the
open-circuit voltage, the best improvement was found for
plasma oxidized layers,

195
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

MICROSTRUCTURE AND STABILITY OF VHF-GO AMORPHOUS SILICON UNDER ILLUMINATION

M. Vanecek
Institute of Physics, Czechoslovak Academy of Sciences, Na Siovance 2, Prague 8, Czechoslovakia

J. Holoubek
Institute of Macromolecular Chemistry, Czech. Academy of Sciences, 16206 Prague 6, Czechoslovakia

A. Shah, M. Brechet and A. Mettler

Institut de Microtechnique, Universite de NeuchiUel, rue Breguet 2, CH-2000 Neuchatel, Switzerland

ABSTRACT: Microstructure and Staebler-Wronski effect have been studied in VHF-GO amorphous silicon
deposited at substrate temperatures 50°C - 300°C. Elastic Light Scattering (ELS) has been used for a
study of microstructure. Voids in amorphous silicon, larger than 5 nm in diameter, can be easily
detected by ELS. A simple theory how to deduce void size from the spectral dependence of the optical
scattering coefficient has been presented and cross-checked with SAXS data. For VHF-GO device quality
material a characteristic void size distribution exists, the largest voids (or their agglomerates) having
an effective diameter of 20-30 nm. The light induced degradation decreases with increasing substrate
temperature but remains quite unchanged in the 200°C - 300°C region for our high deposition rate
samples.

1. INTRODUCTION and scanning electron microscopy [7], small angle

A crucial problem that has to be tackled be- scattering experiments [8] and from studies of
fore amorphous silicon solar cells can find hydrogen distribution by nuclear magnetic reso-
widespread use in energy applications is the nance [9]. Recent studies by small angle x-ray
question of stability. Investigation of amorphous scattering (SAXS) together with density mea-
silicon intrinsic layers with respect to the satu- surements give direct evidence that some 1019
rated density of deep defects created by illumi- cm- 3 microvoids with diameters of about 1 nm
nation, under well defined conditions [1], repre- are present in high quality conventional glow
sents a very useful approach towards the pre- discharge solar cell material [10].
diction of long term solar cell performance. Our optical studies on a-Si:H films indicated
Recent work conducted at the Solar Energy the importance of elastic light scattering ef-
Research Institute, Colorado, USA pointed to the fects [5,6]. Scattering losses may be caused by
importance of a low hydrogen content (1-4%) for any of several contributing factors. In films with
the reduction of light induced degradation negligible surface and interface scattering, the
(Staebler Wronski Effect SWE)' [2]. Furthermore, observed volume scattering is due to hetero-
this study also suggests an important role of the geneities in the refractive index. Local fluctua-
microstructure of amorphous silicon in SWE. The tions in the refractive index can be induced in
chemical equilibrium model [31. describing the a-Si:H films by stress, microvoids, voids or by
defect creation and annealing in amorphous sili- any regions of lower density.
con (a-Si:H), involves an equilibration between Here, we present a quantitative theoretical
weak Si-Si bonds and Si-H bonds, both having description of elastic light scattering on voids
specific distributions in energy. These distribu- (Rayleigh and Mie scattering) in amorphous sili-
tions would also be influenced by the con thin films; a more general theory will be
microstructu re of material. given elsewhere [11]. We will demonstrate that
This paper is just the first attempt to relate ELS can be complementary to SAXS (wavelength
the microstructure of Very High Frequency Glow A.= 0.1 nm) and small angle neutron scattering (A.=
Discharge (VHF-GO) amorphous silicon [4] with 1 nm), on the longer wavelength side, while re-
the number of deep defects in the light soaked, taining the basic advantage of simplicity.
saturated state. The main purpose is the presen- Using a simple scattering theory [12] we have
tation of the recently suggested Elastic Light calculated the influence of the size and concen-
Scattering (ELS) measurement method [5,6], as a tration of voids on the spectral scattering
quantitative tool for microstructure investiga- losses. Total extinction (attenuation) of radia-
tions; therefore, a simple theory of ELS will be tion is described [12] by the extinction coeffi-
given here first. cient CXe
2. THEORY CXe = CXa + CXs. (1 )
It is now well established that considerable where CXa is the optical absorption coefficient
microstructure exists in hydrogenated amorphous and CXs is the scattering coefficient. It is well
silicon (a-Si:H); this follows from an in- established that the scattering losses can be
vestigation of microstructure by transmission expressed in terms of the scattering coefficient

196
 (turbidity) as [13]: 10 2
(2)
where aR = A I 1.. 4 is the scattering loss con-
nected with Rayleigh scattering, aM is the Mie 10 1
scattering contribution that posesses a wave-
length dependence between 1.. 2 and 1..- 2 and aj is
~

E 10 0
wavelength independent scattering. ~
We will look for a possible range of as and its till)
spectral dependence, for defects of various
sizes, both in the Rayleigh and Mie scattering 10. 1
regions, and for different densities of voids. So,
we assume that scattering centers are spheres
(spherical voids with refractive index close to 1) 10. 2
randomly distributed in a medium posessing the 0 2 4 6 8 10
refractive index of bulk amorphous silicon in a
very low absorption region (n .. 3.5). We start a [nm]
with the explicit relation between the scattering
coefficient and other optical parameters in the Fig. 1 Dependence of the optical scattering
form [12] coefficient as on the void radius a at the wave-
length A= 1J.Lm. Concentration of the voids Ns is
as = Ns ( 1t a2 ) Qsca = Ns Csca , (3 ) the parameter.
where Ns is the number of scatterers per cm 3 , a
is the radius of the sphere and Q sca is the scat-
tering efficiency factor. Finally, C sca is the dent determination of the void fraction VI is
scattering cross-section. Q sca can be easily cal- necessary (this can be obtained from the film
culated for spheres that are small with respect density measurement or from calibrated SAXS
to the wavelength both outside and inside the data [10]). The procedure is demonstrated in Fig.
scatterer (this means in the Rayleigh region). The 2.
scattering behavior is similar for all scatterers
with sizes smaller than about 1../10 (that is
A.vac/35 in a-Si:H; so about 30 nm for A.vac = 1 /!m); a [cm- 1] =10 ...
in this case, we can extend the treatment to 5 3
scattering entities of non-spherical shape by the 10 (A.=1/!m) 1
introduction of the "effective radius" of scat-
0.3 •
terer. For non-absorbing spheres in the Rayleigh
region, the relation holds [12]
Q sca = 8/3 x4 (m2 - 1)2 I (m2 + 2)2 , (4) Vf =O.001 o
where x = (2 1t a m2)!A. vac is the size parameter, 0.003
m 2 is the real part of the refractive index of the 0.01
outside medium, A.vac is the wavelength of light in 0.03
vacuum, m is the relative refractive index (m = 0.1 D
11m2 in the case of a void).
For scatterers larger than about Ali 0, the 10 13 10 14 10 15 10 16 10 17 10 18 10 19 10 20
simple relation (4) must be replaced by the exact
Lorenz - Mie expression which holds for spherical N [cm-3]
s
scatterers of arbitrary size and refractive index
[12]. This case will be discussed elsewhere [11]. Fig. 2 Determination of the void radius a and
We have calculated as according to eqs. (3) and their concentration Ns from the value of the opti-
(4). The results are summarized in Fig. 1. To test cal scattering coefficient as at A=1 J.Lm and from
the. reliability of eq. (4) for a=10 nm (the the void fraction Vf. The set of curves A repre-
maximal radius of voids in Fig. 1), we calculated sents solution of eqs. (3,4), the set of curves B
the scattering effiCiency also from Lorenz - Mie represent solutions of eq. (5). The large circle
theory [14] for this particular size of voids. We marks the case of the sample (Ts=200°C), cali-
found a slight difference of about 5% for a=10 nm brated by SAXS. The large square marks the lower
and A. .. 1 /! m. As the other sizes fulfill the limit for void detection under present experi-
Rayleigh condition even better, we consider the mental conditions.
plot to be a reasonably good illustration of the
general tendency.
The experimentally observed range of optical The set of curves (A) gives us solutions of eqs.
losses can be obtained by various combinations (3, 4), as being the parameter. Another set of
of appropriate sizes and concentrations of curves (B) gives solutions of eq. (5), the void
spherical voids. To choose the correct combina- volume fraction VI being now the parameter:
tion of void size and concentration, an indepen- Ns nv = VI NSi , (5)

197
where ny is the number of silicon atoms missing by CPM. Temperature during the light soaking has
in the void (ny can be approximately determined been regulated by an air blower (monitored by a
from crystalline silicon data, comparing the flowmeter) and kept at 27°C, as determined by
volume of the void with that of the crystalline the Infrared Thermography method [17].
silicon elementary cell and making a correction Some additional samples grown at Ts = 200°C
for the surface atoms) and NSi is the number of and 300°C have been light soaked in a filtered
silicon atoms per cm 3 (approx. 5x10 22 cm- 3 ). white light system. Excessive heat has been re-
It should be noted that the scattering proper- duced by a water filter, approximately homoge-
ties are partly modified by deviations from the neous illumination has been ensured by optimal-
spherical shape. As concerns the effect of inter- ization of the red, long pass filters (Schott's
actions between dense aglomerates of voids, the type RG 620 - 680) according to the transmission
calculation indicates [15] that these interactions spectrum taken for each sample. Temperature of
are negligible for a separation of scatterers the samples has been kept at 50°C by regulated
greater than three diameters (a condition that is air flow. We have simultaneously monitored the
in average fulfilled for a void fraction VI < 0.015). photocurrent, as generated by the light used for
This is typically the case for good quality a-Si:H the degradation, until saturation was reached.
[16]. Thus, we do not need to include the effects The time was typically 50 hours for our typical
of multiple scattering (which would have to be carrier generation rate G above 1x10 22 cm- 3 in
solved for a specific model of agglomeration). 1.5-2 Ilm thick samples. An aluminium back re-
When the void size starts to increase, we will flector has been used to ensure a yet more homo-
get, already at a radius a = 30 nm, a distinct de- geneous degradation of each sample and as a di-
parture from the 1n. 4 dependence of as. The rect experimental check of the homogenity of
spectral dependencies of the scattering cross- degradation. (This can be directly verified by the
section in this case have been calculated by comparison of the photocurrent values with and
means of the BHMIE program [14] and the results without the back reflector).
will be published elsewhere [11].
4. RESULTS
3. EXPERIMENTAL The results of our optical and photoelectrical
A series of samples was deposited at sub- characterization of the samples are summarized
strate temperatures between 50°C and 300°C, in Table 1.
using the VHF glow discharge technique [4], under
conditions described elsewhere [4,5]. Deposition
parameters were selected so as to give the low- Ts[Oc] Eg[meV] Eo[meV] aa as Exponent
est deep defect density in the "as grown" state, (1.25eV) (llJ.m) n
for each substrate temperature Ts. Further opti- 50 1.73 68 13 .. 10 <2
malization with respect to the lowest deep de- 100 1.76 60 11 .. 5 <2
fect density in light saturated state is under 1.75 50 0.8 2 .. 3
150
progress.
200 1.74 44 0.1 3 4-4.5
Deep defect density has been determined from
subgap absorption measurement by means of the 250 1.72 50 0.3 3 3.5
Constant Photocurrent Method (CPM) and the 200 1.74 46 0.07 5 3.5
Photothermal Deflection Spectroscopy (PDS). 250 1.71 44 0.2
Measurement of Elastic Light Scattering (ELS) 300 1.68 45 0.2 1.5 4-4.5
has been described in our previous papers [5,6].
Basically, the CPM spectra are taken with two
different interelectrode spacings (the narrow one Table 1 Characterization of samples. Here, Ts is
being 0.2-0.4 mm and the wide one about 5 mm) the substrate temperature, Eg is the optical gap,
to vary the influence of scattered light on the Eo is the Urbach slope as determined by CPM, aa is
CPM spectrum. Great care has to be taken to the optical absorption coefficient at 1.25 e V, as
guarantee the spectral purity of light in a low is the scattering coefficient at 1 J.lm and the ex-
absorption region, otherwise any non-monochro- ponent n represents the spectral dependence of
matic light component (especially in the red part as",,11)'P.
of the spectrum, where the sample is very photo-
sensitive) can strongly influence the spectral
dependence of as. Also, the substrate and sample Optical gap Eg is given by the Tauc plot, Urbach
slope Eo is determined by CPM (PDS gives slightly
surfaces have to be perfectly mirror-like to re-
higher values: for Ts=200°C and 30QoC we mea-
duce their contributions to the scattered light.
sure Eo=47-48 meV). Deep defect density is
For a purpose of comparison and in an attempt
characterized by the value of the optical
to identify the scattering centers in VHF-GO
amorphous silicon, one thick Ts = 200°C sample absorption coefficient a at 1.25 eV (averaged
has been deposited on AI foil and measured by over interference fringes) and Elastic Light
SAXS [16]. Scattering (ELS) by the value of the scattering
Samples from the Ts series have been degraded coefficient as at 1 Ilm. The spectral dependence
up to saturation with a Kr laser (647 nm spectral of asp_) is deduced from measurements
line) under light intensity 1 W/cm 2 and measured throughout the photon energy interval 0.9-1.3 eV.

198
 Results of the light degradation study are this sample as .. 1/A 3 . 5 ), gives us a confirmation
given in Fig. 3, where the number of deep defects that the dominating scatterers have a radius
N dd in the light saturated state is displayed as a larger than 10 nm. So when we are interpreting
function of deposition temperature Ts . The exact our ELS data (Table 1) we should consider this
conditions of degradation were described in the data as a good description for that fraction of
experimental part. Subgap absorption has been voids which are the most efficient light scat-
converted to the number of deep defects terers (essentially the largest ones).
according to [18]. We should keep in mind that these large voids
can in fact be clusters (agglomerates) of small
voids and that the void radius is just some
effective radius of irregularly shaped voids or
o Kr laser light void clusters.
o The spectral dependence of as in a series of
+ filtered white light
o samples deposited at different temperatures
(Table 1) points to 2 different microstructures,
the sample deposited at Ts = 1S0°C being the bor-
derline. Samples deposited at a low temperature
(T s .. SO°C and 100°C) show a high as and very
weak spectral dependence of as(A). So, we can
o conclude that this material has very large voids
o o (or big conglomerates of interconnected smaller
+ voids).
Samples deposited at Ts - 200°C-300°C ex-
hibit approximately 1/A 4 spectral dependence of
o 50 100 150 200 250 300 350 as (Rayleigh scattering), the largest scatterers
T [0C] being identified as midsize voids with an ef-
s fective radius a = 10-20 nm.
Results of the investigation of Staebler-
Fig. 3 Density of deep defects Ndd in the light
Wronski effect(SWE) in our films, given in Fig.
soaked, saturated state as a function of the
3, exhibit again 2 different classes of behavior:
substrate temperature Ts during deposition.
samples deposited at low temperature (T s .. SO°C
and 100°C) have much higher density of deep de-
Preliminary SAXS results [16] give us for VHF- fects Ndd created by light soaking up to satura-
GD samples a void fraction Vf = 0 0 . 1-0.03 and a tion compared to the samples deposited at higher
broad distribution of the voids size (with temperatures. This value is the same (within
diameters ranging from 1 to about 20 nm and range of the 20%) for all VHF-GD samples de-
with a higher void fraction of smaller voids). In posited in 200°C-300°C region with deposition
particular, a thick sample deposited at Ts" 200°C rates S-20 A/s . The similarity of the samples
had the total void fraction Vf= 0.028 and void prepared above 150°C is discussed in more de-
diameters ranging from 1 nm (Vf=0.020) to 33 nm tails in [19], which points to some specific fea-
(Vf. 0.001). On this sample ELS measurement tures of the higher excitation frequency on
gives as(1Ilm) - S cm- 1 and the spectral depen- plasma and film growth kinetics.
dence of as(A) is proportional to 1/A3.5. Thus, we can see a correlation of the mi-
crostructure with the magnitude of SWE when we
S. DISCUSSION compare the low temperature and the high tem-
At first, we will discuss our data on the perature samples. It is what should be expected
microstructure of VHF-GD samples. The anal- [2,20]. But we should go to yet higher tempera-
ysis of SAXS data and ELS data, measured on tures to see if there will be a decrease in SWE
T s=200°C samples deposited in the same run, similar to the one observed in [2,20], and if this
gives us, with the help of our simple theory (Fig. decrease will be accompanied by changes in the
2), the radius a and the concentration Ns of the light scattering behavior of the material. The
voids which dominate in light scattering. These present study should be considered just as a pre-
should be the biggest ones seen by SAXS (that is liminary one; its main aim is to show the poten-
with diameter 33 nm, vf=0.001) because the mea- tial of the Elastic Light Scattering measurement
sured as (A) is a weighted integral of the void size for the study of microstructure.
distribution multiplied by the respective scat- What are the smallest voids that can be de-
tering cross-sections C sca . And as follows from tected by ELS? The detection limit for a void size
eqns. (3) and (4), Csca is proportional to a6 . depends on their concentration and also on the
For vf=0.001 and as(A=1Ilm) = Scm- 1 we will contribution of surface scattering to the ob-
get an estimate of the void radius a = 1S-20 nm served volume scattering . So far, we have ob-
and Ns = 1013 _10 14 cm- 3 (look at Fig. 2), in a served as=O.S cm- 1, weakly spectrally dependent,
good agreement with the radius 16.S nm, as de- in some 13.S6 MHz GD samples as a lower limit.
termined by SAXS. Also , the spectral dependence If we take this as a typical detection limit due to
of as(A), which slightly deviates from 11A.4 (in the surface contribution, we will obtain under

199
the assumption of a reasonable value for the void [14] C. F. Bohren and D. R. Huffman, "Absorption
fraction (vf<0.03), a lower limit a = 2.5 nm (look and Scattering of Light by Small Particles",
at Fig. 2). So, we cannot see those microvoids, J. Wiley, N. Y. 1983
that are present in usual device quality 13.56 [15] A. R. Jones, J. Phys. D: Appl. Phys. 12 (1979)
MHz glow discharge films [10], but we can 1661
observe larger voids typical for VHF-GD material [16] D. L. Williamson and A. H. Mahan, private
or for "hot wire" amorphous silicon (which is communication
particulary stable) [2,16] or, also for amorphous [17] H. Keppner, P. Chabloz, D. Fischer, R.
germanium [21]. Tscharner, A. Mettler, A. Shah and Ch.
Rossier, this conference
6. CONCLUSIONS [18] J. Kocka, M. Vanecek and A Triska, ibid [8], p.
We have demonstrated how Elastic Light 297
Scattering can be used for detection of voids (or [19] F. Finger, U. Kroll, V. Viret, A. Shah, W.
void agglomerates) in VHF-GD amorphous silicon. Beyer, X. M. Tang, J. Weber, A. Howling and
Voids larger than about 5 nm in diameter can be Ch. Hollenstein, submitted to J. Appl. Phys.
detected. [20] S. Nakano, S. Okamoto, T. Takahama, M.
For VHF-GD device quality material it is Nishikuni, K. Ninomiya, N. Nakamura, S.
characteristic a void size distribution with the Tsuda, M. Ohnishi, Y. Kishi and Y. Kuwano,
largest voids of about 20-30 nm in the effective Proc. 21 st IEEE Photovoltaic Specialists
diameter. For a deeper understanding of SWE a Conf. (1990), p.1656
more extensive study of microstructure is [21] R. Plattner, E. GOnzel. G. Scheinbacher and B.
needed. SchrOder. ibid [1]

ACKNOWLEDGEMENTS
The authors wish to thank Dr. D. L. Williamson
from Colorado School of Mines and Dr. A. H. Mahan
from SERI for preliminary SAXS measurements,
to Dr. F. Finger and U. Kroll, IMT Neuchatel for
samples preparation and to T. Plechacek, Prague,
for technical assistance. This work was partly
funded by Swiss Federal Research Grants EF-REN
(90)045 and CERS 1820.1

REFERENCE
[1] P. V. Santos, W. B. Jackson and R. A. Street,
in "International Meeting on Stability of
Amorphous Silicon Materials and Solar
Cells", Denver 1991 (to be published)
[2] A. H. Mahan and M. Vanecek, ibid [1]
[3] K. Winer, Phys. Rev. B41 (1990) 12150
[4] D. Fischer, R. Tscharner, Y. Ziegler, H.
Keppner and A. Shah, Proc. 9th E. C.
Photovoltaic Solar Energy Conference,
Freiburg 1989, p.995
[5] M. Favre, H. Curtins and M. Vanecek, J. Non-
Crystal. Solids 114 (1989) 405
[6] M. Vanecek, D. Cervinka, M. Favre, H. Curtins
and A. Shah, MRS Symp. Proc. vol. 192,
"Amorphous Silicon Technology-1990", eds.
P. C. Taylor, M. J. Thompson, P. G. LeComber,
Y. Hamakawa and A. Madan,p. 639
[7] R. C. Ross, A. G. Johncock and A. R. Chan, J.
Non-Crystal. Solids 66 (1984) 81
[8] R. Bellissent, in "Advances in Disordered
Semiconductors", vol. 1, ed. H. Fritzsche,
World Scientific, Singapore 1989, p. 93
[9] J. A. Reimer and M. A. Petrich, ibid [8], p. 3
[10] A. H. Mahan, D. L. Williamson, B. P. Nelson and
R. S. Crandall, Phys. Rev. B40 (1989) 12024
[11] M. Vanecek, J. Holoubek and A. Shah, to be
published
[12] H. C. Van de Hulst, "Light Scattering by
Small Particles", J. Wiley, N. Y. 1957
[13] L. E. Busse, G. H. Mc Cabe and I. D. Aggarwal,
Opt. Lett. 15 (1990) 423

200
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

VHF plasma deposition for low-cost a-Si:H solar cells

D. Fischer. H. Keppner. F. Finger. K. Prasad and A. Shah;

Institute 'of Microtechnology. University of N euchatel,
Breguet 2, CH-2000 Neuchatel, Switzerland
phone +41 38205 121, fax +41 38 254276

ABSTRACT: The use of the Very High Frequency (VHF) glow discharge for the rapid deposition
of amorphous silicon (a-Si:H), introduced by our group at N euchitel in 1987, offers interesting
possibilities for future low cost production of a-Si:H solar cells. We report on the optimization of
p-i-n solar cells entirely deposited at VHF comprising SiC window p-Iayer, 22 A/s deposition rate
intrinsic layer, and highly conductive microcrystalline n-Iayer reaching 8 % efficiency, as well as
on 32 cm 2 integrated modules achieving 7.5% active area efficiency. Furthermore, new results on
n- and p-~oped microcrystalline material deposited at VHF, with conductivities upto 100 (Qcm)-l,
and studIes of hydrogen bonding and microstructure of intrinsic VHF material are presented,
indicating both profound differences as compared to conventional GD material. These differences
are assumed to be due to a change in the detailed growth mechanism caused by the increase of
plasma excitation frequency.

1. INTRODUCTION excitation frequency of 70 MHz was used for all

depositions described. This frequency was found to
In order to allow for future low cost production of be optimal for the present reactor, leading to the
amorphous silicon solar panels, increase of intrinsic highest deposition rate for intrinsic material and to
layer deposition speed remains a major problem to be the best plasma confinement between the electrodes,
resolved [2]. Different approaches have been fol- as has been discussed before[I]. Electrode diameter
lowed to increase deposition rates. However, was 130 mm, allowing for deposition on 83 x
material quality often degrades with increased 83 mm 2 substrates.
deposition rates [3]. In certain cases, such as for the Corning 7059 glass was used for single layer
use of microwave glow discharge or for the use of deposition. The layers were characterized by dark-
higher sHanes [4], the possibility for large area and photoconductivity measured in gap cell con-
deposition and for low cost production remain figuration. The optical gap was deduced from optical
uncertain. transmission measurements. For microstructure and
By the use of the Very High Frequency (VHF) hydrogen bonding analysis, hydrogen effusion and
glow discharge a novel and promising method was IR-transmission data were employed.
introduced in 1987[1] to tackle the problem. Since Solar cells were deposited on glass substrates
then it has been established: With VHF plasma de- coated with textured TCO, providing 85% maximal
position it is possible to produce • at rates around optical transmission, 10% haze and 10 Q/square
20 A/s and using mono silane (SiH4) as a source gas, sheet resistance (ASAHI). The solar cells were
intrinsic a-Si:H layers with electro-optic properties characterized by IV measurement and voltage
that are just as good as those of conventional layers dependent quantum efficiency. Small signal absolute
deposited at 13.56 MHz and at lower deposition quantum efficiency was measured under white bias
rates. Furthermore, the combination of VHF and light and variable bias voltage. IV curves were mea-
Electron Cyclotron Resonance (ECR) offers sured under 100 mW/cm 2 AMI.5 equivalent ELH
interesting new deposition possibilities for large-area light; thereby, ELH light intensity was adjusted to
large-scale production [5], and recently, 10% match the current integrated from absolute quantum
efficient single junction solar cells having a VHF efficiency data multiplied by the AMI. 5 spectrum.
intrinsic layer deposited at 17 A/s have been
presented [6].
But still, there further research is needed to 3. SiC LAYERS
improve the understanding of material- and device-
related issues in connection with VHF deposition; In order to be able to produce state of the art p-i-n
there is, in fact, growing evidence for the presence solar cells, p-doped silicon-carbon alloys were deve-
of some interesting differences between VHF material loped to be used as top window layers. The layers
and conventional material: On the one hand, the film were deposited from a mixture of SiH4, CH4, and
growth and the resulting film structure. especially B2H6 diluted to 1000 ppm in H2 at a substrate tem-
regarding hydrogen bonding, which is an important perature of 220°C. Plasma excitation frequency was
stability-related property, show some interesting 70 MHz, and no frequency dependences were
features, and on the other hand, the application of studied. Investigations were carried out under low
VHF to doped microcrystalline layer deposition discharge power density (15 mW/cm 3) and high
reveals important benefits of the increased plasma ex- pressure (1 mbar) conditions. these being usually
citation frequency. favourable to high quality SiC layer deposition under
conventional glow discharge conditions.
At CH4/(SiH4+CH4) ratios of 0.6, amorphous SiC
2. EXPERIMENTAL films exhibiting optical gaps E04 of 2.0-2.2 eV were
found (to avoid interpretation difficulties. E04, i.e.
All films and solar cells reported on were deposi- the photon energy at which the optical absorption
ted in a single chamber reactor by capacitively reaches 104cm- 1 , was employed). By doping with
coupled glow discharge at VHF frequency. A plasma

201
B2H6 at 0.15 %, dark conductivity of 2·10-6 (ncm)-1 Solar cell performance was optimized by first opti-
was obtained at a gap E04 of 2.05 eV. This layer was mizing p-Iayer thickness, and then buffer layer thick-
applied for the preparation of p-i-n solar cells (see ness. During this procedure, thickness of intrinsic
section 4). layer and n-layer were kept constant at 5000 A and
10.5 ~--~---r---.---.r---.---' 500 A, respectively.
For the p-Iayer, thickness was increased until a
~. suitable for saturation of Voc was observed at around 120 A. For
<;-- solar call this thickness, a quantum efficiency of 55% is main-
p-layer
tained, confirming the good quality of the SiC win-
~ dow layer material (Figure 3).
',.0 900 100
-- · - dark
•• ..... ~,
90
o photo
.
o
s;-
800
6',.
~
• •~ 80

10.8 '----'-----'-----'------'----'-------' .§. 70

0
1.95 2 2.05 2.1 2.15 2.2 2.25 8 m
'u Q 60
> is)
.,..
700 '·0
E04 leV] ",0 50 0
• :> 0
::>
Figure 1: Dark- and photoconductivity of p-doped SiC layers 40 3
deposited at 70 MHz for various carbon alloying and doping. ~
The arrow marks the layer having G > l~(D.cm)"1 and E04 > 600 30
2 eV and considered suitable for a solar cell window layer. 60 80 100 120 140 160 180
p layer thickness [A)
Further improvement of SiC film properties could
be expected from the use of B2H6 diluted in SiH4 Figure 3 : Optimization of SiC p-Iayer thickness: Voc and
(rather than diluted in H2); this technique might QE(400 nm) vs. p-Iayer thickness.
enhance carbon incorporation.
In order to reduce interface recombination loss by
electron back diffusion to the p-Iayer, a constant
4. P-I-N SOLAR CELLS bandgap buffer layer was introduced between p- and
i-layer. The buffer layers were deposited from a
p-i-n solar cells were produced by the consecutive mixture of 20% SiH4, 20% CH4 and 60% H2 at
deposition of a p-doped SiC window layer, a 1 mbar and 15 mW/cm 3. The resulting solar cells
constant bandgap undoped SiC buffer layer, a high showed monotonic increase of Voc with increasing
deposition rate VHF intrinsic layer, and a highly buffer layer thickness , which can be understood if
conductive microcristalline n-layer (see section 5). one postulates that a growing fraction of light is ab-
Employing a single chamber deposition system, sorbed in the SiC alloyed region. Fill factor (FF)
solar cell deposition initially suffered from strong showed a distinct maximum at approximative)y 60 A;
phosphorus contamination of the intrinsic layer. This this value corresponds to maximal cell efficiency.
resulted in a generally low fill factor of the cells, of
less than 0 . 6. The doping polarity was identified by a 890 0.70
reduced red versus blue collection efficiency, as
measured under forward voltage bias [7] . As source
of this contamination phosphorus desorption from the
880
"
0- -' ' .
"Q
/
.. 0.65
reactor walls originating from the preceding p-i-n cell 0 "
-/ "
~

n-layer deposition was assumed. By the introduction 5" 870

'-
oS '.'.
/

-, "
of 'I 30 min SiH4 purging deposition on a dummy 0.60 ."
substrate and on the reactor walls between each p-i-n / " ."
>8 860 e-
cell deposition the phosphorus contamination could /

be suppressed, resulting in high fill factor and " region of maximal

0.55
cell efficiency b
uniform collection efficiency for red and blue light. 850

840
W/ft/~ 0.50

.2 .
5 10 15 20 25 30
0.9
buffer layer depoSition time [s)
i! ~ 0.8
Figure 4: Optimization of constant bandgap buffer layer: Voc
~~ andfillfactor FF vs. buffer layer deposition time.
~ur 0.7
..e>R Based on this optimization procedure, small area
a~ 0.6 (0.25 cm2) p-i-n cells could be fabricated with effi-
~0" 0.5
ciencies of 8%; they incorporate a VHF intrinsic
layer deposited at 22 Als (dj=6500 A, tdep=300 s).
0.4
At the present state of developement, efficiency is
300 400 500 600 700 800 basically limited by a relatively low short circuit
wavelength [nm)
current Isc of ~ 14 mA/cm2 only. As QE of the above
solar cells shows strong interference fringes,
Figure 2: Quantum efficiency ratio QE(0.6V)IQE(-lV)for p-i-n considerable current improvement could most
cells with and without a preceding SiH4 purging plasma. Conta- probably be expected from the use of a more textured
minated cells (no purging) show low QE ratio for red light and more transparent TCO.
indicating phosphorus incorporation, whereas uncontaminated Some small integrated solar modules were pro-
cells show homogeneous and high QE ratio for all wavelengths. duced to demonstrate the homogeneity of the VHF

202
 15~----------------------.
power densities as low as 25 mW/cm 3 with a com-
paratively high deposition rate of 0.65 A/s. For
boron-doped samples, the silane concentration needs
to be reduced further to less than 2 % in hydrogen
and the discharge power increased to 120 mW/cm3 •
Figure 7 illustrates the conductivities measured in a
coplanar configuration, for 0.5 Ilm thick films depo-
sited on glass, as a function of the gas phase doping
ratio. Conductivities as high as 100 and 40 ({km)·1
Ise : 13.7 mAlem2 are obtained for nand p type samples, respectively.
Voe : 870 mV The results illustrate that VHF-GO offers favourable
FF : 0.67 conditions for the growth of Ilc-Si:H. We suspect
elf : 8.0 % that the detailed growth mechanism is different at
VHF frequencies when compared to 13.56 MHz and
that this may explain why microcrystallinity is
0.0 0.2 0.4 0.6 0.8 1.0
favoured.
voltage (V)
Figure 5: N-characteristic of a small area p-i-n solar cell 10 2 I
deposited by VHF glow discharge at 70 MHz using monosilane n-type /:_ .
aS,a source gas. 10'
;:/. ~ \
process at laboratory reactor scale. The module 100
structure was formed by a simple screen-print- 10"
ing/wet-etching process. Figure 6 shows the IV- '5 p-type
characteristic of a 70 mm x 68 m.m module after ~
~
10.2
succesful reverse bias curing of previously observed
0
microscopic shunting defects. Subcell width is 10.3 un doped
10 mm, including 3 mm of interconnection spacing
loss, which explains the considerable' difference
between the active (32.3 cm2 ) and the total cell area
10-4 I
0

(47.6 cm2 ). 10.5 If 10.4 10'3 10.2 10.1

80~-------- ____________- .
(P H J or n1 "6 ) I iH.
Figure 7: Influence of gas phase doping ratio on conductivity of
microcrystallinejilms deposited at 160"(;
60
~
oS
c 40
~ 6. HYDROGEN AND MICROSTRUCTURE
:s
0
active area : 32.3 cm2
Isc :66.4 rnA The hydrogen content and the hydrogen bonding
20 Voc : 5.87 V str.ucture in amorphous intrinsic films prepared with
Prnax : 0.247 W the VHF-GO at 70 MHz has been extensively studied
FF : 0.63 using infrared spectroscopy and· hydrogen evolution.
eft : 7.5 % (active area) It was found that there is a surprisingly' small
0 increase of the total hydrogen content from 10 at% to
0 2 3 4 5 6 20 at% only, as one decreases the substrate tempera-
ture from 250°C to 50°C.
voltage (V]
Figure 6: IV-characteristic of a 70 mm x 68 mm integrated
module prepared by VHF-glow discharge, consisting of 7
series-connected subcells;

5. MICROCRYSTALLINE SILICON

Extensive work has been carried out on

~ 40
.§.
E
E 30
\ \.
-..
microcrystalline silicon (llc-Si:H) as potential n- and o •"" 13.56 MHz
u
p- type layers in a pin solar cell. Microcrystallinity is
g 20 ."--.....
induced by strongly diluting silane in hydrogen. The
e
.: I~'-
CIO
resulting plasma is understood to be enriched in
hydrogen radicals. This causes the plasma-film 'Q
>. .........
growth surface to approach the condition of partial :I: 10 70M}b .~
chemical equilibrium (PCE) [8], wherein a balance is
reached between deposition and etching rates,
resulting in deposition of Ilc-Si:H. Doping is per-
formed by adding phosphine or diborane diluted in
hydrogen (1000 ppm) during deposition. Substrate Temperature (0C)
Highly conductive phosphorus doped samples are
obtained at a subtrate temperature as low as 160°C,
and for a silane concentration of 3 %. Interestingly, Figure 8: Total hydrogen content in a-Si:H films prepared at
unlike the general observation using the 13.56 MHz - 70 MHz and at 13.56 MHz[9] as a function of deposition
GO process, the best films are obtained for discharge temperature.

203
Connected with this, only moderate indications of It is argued that the influence of the high
microstructure (voids, clustered hydrogen) are found excitation frequency on the process plasma plays a
in material prepared at low substrate temperatures. key role in this content. A detailed investigation of
This is illustrated in Fig.9 where the IR spectra of this problem will be published elsewhere [10].
films prepared between SO°C and 2S0oC are shown.
The important feature is that there is only little con-
tribution to the IR absorption at 2080, 84S, and 890 7. CONCLUSIONS
wavenumbers - the absorption bands usually con-
nected with voids and hydrogen clusters. By the optimization of 8 % efficient p-i-n single
junction solar cells it has been shown that the VHF
glow discharge technique is capable to produce state
of the art solar cell devices under single chamber
conditions, and at high deposition rates, which meet
the formulated requirements of economic mass pro-
duction for amorphous thin film solar panels .
Additional beneficial contribution of VHF techno-
logy might be possible in the field of microcrystalline
layer deposition, or, by the exploitation of its
different structural properties, as an intrinsic material
with improved stability.

ACKNOWLEDGEMENTS

The authors wish to thank C. Fortmann for helpful

discussions during his visit at Neuchlitel, and to
Asahi Glass Inc. for supplying TCO coated sub-
strates. This work has been supported by the Swiss
Federal Office of Energy (OFEN), under contract EF-
REN 90(04S), and by Swiss National Science
Foundation Grant FN 27 916.
2400 2000 1600 1200 800 400
Wavenumber (em· l ) REFERENCES
Figure 9: Infrared transmission spectra/or a series 0/ a-Si:H [1]:H.Curtins et aI., Plasma Chern. and Plasma Pro-
films prepared with the VHF-GD technique at/ast deposition cessing, 7(1987), pp. 267 .
rates at different substrate temperatures TS between 500C and [2]:H.Aulich, Proc. Sib Int. PVSEC(1990), pp.S23
250 "C. The spectra are shifted with respect to each other. [3]:W.Luft,Proc.19 th IEEE PV Spec.Conf.(1987),
pp.S64
We find, thus, that the correlation between micro- [4]:P . K . Bhat, 20 th IEEE PV Spec. Conf.(1988),
structure and total hydrogen co-ntent in VHF material pp.91
is similar to that observed in films prepared by other [S]:S.Oda et a1. ,MRS SYffiil.Proc. 118(1988),pp.117
methods. There is, however, a remarkable difference [6]:D.S.Shen, to be published at the SERI subcon-
for the dependence of the amount of microstructure tractor meeting (1990)
on the total hydrogen content as determined from [7] :W. Kusian, PhD thesis, MUnchen 1987
hydrogen evolution on one hand, and from infrared [8]:S.Veprek et aI., MRS Symp. Proc. 118 (1988)
absorption on the other hand; the reason could be that [9] :W. Beyer, 'Tetrahedrally-Bonded Amorphous Se-
VHF material has a large amount of medium-size miconductors', Plenum Press, NY,129(198S)
voids that are not interconnected. [lO]:F.Finger et aI., subm. to J.AppI.Phys.

204
 HYfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CONTINUOUS MONITORING OF VHF-GD DEPOSITED p-i-n SOLAR CELLS DURING UNIFORM

DEGRADATION TILL SATURATION.
H.Keppner, P.Chabloz, D.Fischer, A.Mettler, R.Tscharner and A.Shah
Institut de Microtechnique, Universite de Neuchlitel, Rue A.-L. Breguet 2, 2000 Neuchlitel, Switzerland
Phone: 41038/205451 Fax 41038/254276
Ch.Rossier
Entreprises Electriques Fribourgeois, Perroles 25, 1700 Fribourg, Switzerland.

ABSTRACT. This paper addresses the prioritary issue of the Staebler -Wronsky effect (SWE) in p-i-n
Solar cells. In previous publications it was shown, that VHF-GD yields individual layers and solar cells
with results comparable to 13.56 MHz GD. In this paper we present the first results of the degradation
behaviour of VHF-GD solar cells. To this end, we applied a fast degradation technique wherein the SWE
on the solar cell parameters (VOC, Isc , FF, TJ) was measured until saturation occured. In order to check the
uniformity we used LBIC, (Light Beam Induced Current) Thermography and CCD (Charge Coupeled
Device) beam shaping, thus making sure, that the solar cell parameters are isotropic values coming from the
degraded volume. By continuous monitoring it is possible to observe the saturation behaviour of the cell
parameters and the validity of the power-law. In situ monitoring permits to find a relation between the
saturation values and the intensity which is applied for light soaking.

a) '<ii'
INTRODUCTION. In literature different degradation techniques 'E:>
have been used, as for example AM 1.5 light degradation, i":-
f!!
concentrated white light degradation [1], e-beam degradation [2] ~
and laser beam degradation. The last three techniques are applied ~
to reduce the degradation time, especially when saturation values .~
~
are demanded. The AM 1.5 degradation, of course, gives results CD
that are easily transferable to actual solar cells. On the other hand .5
'.."
E
the complicated. non uniformly absorbed spectral distribution
permits no detailed understanding of the SWE. .c
For our light soaking experiments we decided to use high
intensity Kr laser light at a wavelength of 647 nm and a total -1.5 -1 -0.5 a 0.5 1.5
power of 650 mW. dlS1ance 1=)

EXPERIMENTAL
i) Samples used.

For the experiment we used annealed (1h at 1800 C) VHF-GD

deposited p-i-n solar cells (Glass I TCO I <p> a-SiC:H / buffer
layer / <i> a-Si:H / <n> Ilc-Si:H I AI) where the thickness of the
intrinsic layer was varied between 0.2 and 0.665 J.1m. yielding
thus to AM 1.5 efficiencies between 5.5 % and 7.9 %. The
thickness was measured with a surface profller. with an
accuracy bette,r than 5 %. The degradation was executed under
open circuit conditions with a 650 mW Kr ion laser at a
wavelength of 647 nm. The transmission of the intrinsic layer at
this wavelength is about 80 % for the thickest film, which leads
to a fairly uniform absorption. The generation rate was
determined from the reverse saturation current in illuminated
condition.The experimental set up is shown in Fig. 1:

p-i-n .".,~~
SUbsUlll.eSOlarcel~ ~ ............. ,
K/alaser @7 : :
"'C=;z:....-~fc~n::t~=====:tl I-V: Computer i
~"\.4. j conUl)1 i
~. t........... J
shuller-COl1lrol

Fig. 1,' experimental set up for uniform light soaking (647 nm). Fig. 2 a: Beam intensity profile measured by a CCD camera. 2b.
shutter. I-V characteristics (under halogen light 15 mWlcm2). local photo current profile for a light soaked solar cell, measured
temperature measurement (IR camera). beam profiling (CCD for the case of an ideal swface ratio as seen by LBlC, in 2c the
camera) and control system. ratio is too big for a given cell area.

205
ii) Experimental set up. RESULTS AND DISCUSSION.

The Kr laser beam is broadened in a suitable way by a spatial By our thermography measurements we made sure that during
filter and transformed to a broad parallel beam by a lens. During the high-intensity light soaking the sample temperature could be
the degradation the beam is pointed onto the solar cell. During kept below 30°C and that by a controlled air stream, the sample
the white light I-V measurement the halogen lamp is put on and a could be efficiently cooled in a reproducible way. The results are
shutter interrupts degradation. The intervals between shown in Fig. 4.
measurement and degradation are fully computer controlled.

iii) Control of lateral uniformity. (3'100

L
~ 90 . . . __- _,::.,vs
While performing fast laser beam degradation the proper beam ~OJ 80
shaping is a non trivial task. By careful beam broadening the Q.
E 70
ratio between active cell area and beam diameter must be suitably co
I-
optimized, leading to both high and laterally uniform 60
generation rates. To check for lateral uniformity, LBIC (Light 50
Beam Induced Current) is an ideal tool, as in the case of a high
degradation level (Tl < 0.5 Tl initial) short circuit current (Isc) 40
sensibly reflects the degradation level of the solar cell. The result 30
of this measurement is shown in Fig. 2 b for a given beam 20
intensity distribution monitored in Fig. 2 a. As LBIC source we 0.1 0.2 0.3 0.4 0.5 0.6 0.7
used a 680 nm laser diode spot focussed at about 20 11m 0. For 1·layer thickness [ itrn]
a big surface ratio (Laser spot 01 active cell 0), Fig. 2c shows
a lateral non uniform LBIC plot. Fig. 4: Solar cell temperature during light soaking as a function
of the <i> layer thickness and the air flUX.
iv) Temperature control.
Temoerature durin!! del!radation
As shown in Table I, the heating of the substrate by light 70 ° C 38° C 27° C
absorption has an influence on the saturation values of the solar saturated I Initial
cell. This is probably due to a self-annealing effect. Therefore
Voc /0.825 V 0.84 0.9 0.9
the lateral temperature distribution must be controlled.
Thermography is a useful method for doing this; it is also useful Isc /11.9 rnA/ern3 0.88 0.74 0.72
for the development of an efficient cooling system. ( For room FF /66% 0.72 0.70 0.70
temperature cooling an air stream of 32 11 min,S mm distance at T1 /6.5 % 0.51 0.44 0.42
an angle of 45° to the substrate streaming through a Imm 0 hole
was used). Fig. 3a shows a thermogram of the temperature Tab. 1: Example for self-annealing if the degradation
distribution on the rear Al contact of the solar cell, Fig. 3b temperature is varied in terms of saturated / initial values.
shows the distribution on the non metallized <n> I1c-Si:H cell (Generation rate: 8 x 10 22 lIcm3)
back-surface.
Under this well defined conditions the solar cells have been light
a) soaked; the i-layer thickness, the laser power, and the sample
cooling were varied. A typical degradation plot for the solar cell
parameters is drawn in Fig. Sa, the semi-log plot shows the
pronounced saturation behaviour of all solar cell parameters.
The ratios saturated I initial value of all experiments are
summarized in Fig. 5b. In Table 2 the absolute saturation
parameters as function of the <i> layer thickness are given.

i. -layer Voe (V) Ise (mA) FF (%) l] (%)

in urn Voc- Isc· FF- fl-
0.2 0.82 10.4 65 5.5
0.76 8.32 49.4 3.1
0.26 0.83 11.9 66 6.5
0.74 8.56 46.2 2.9
0.44 0.81 12.8 66 6.5
b) 0.73 7.55 44.9 2.4
0.58 0.85 14.2 63 7.9
0.73 5.96 37.8 1.7
0.67 0.85 14.9 57 7.2
0.72 5.21 34.2 1.3

Table 2 Ratio of saturation/initial solar cell parameter as a

function of the i-layer thickness.
* saturated value.

Fig. 3a: Thermogram of the cell rear contact; b: Thermogram of

the non-metallized cell.

206
 a) .. .) 0.22

-
~,
I I I
~ 0.9

i 0.8 ~v_
.................. ~
l;'
jO.215
1!
~
0.7
0.6
x, - "+-++-i+IHI+IIIIIf.
FF ..
~
"Xx ~ 0.21
x, Iso
0.5 x.x ~, "
1l
0.4
0.3
I-Ia)'orthiclcness: 0.44II1II
generation llIIe: 8x 1022
'''''x~~ 10.205
0
1\ c:
degr.tempe ure: 27°C
0.2 0.2
0.01 0.1 10 100 0.01 0.1 1 10 100

- - - --
degradation time [hI relaxaton time [hI

0.38
b) -e- ___ - b)

..
Voc 0- -0-- _ _ _

0.8 l,e
-0- __ c

~
0.375 ~ Ih laser degradation

-~
" ~::t:. ,. _: .... _ ... __
... --
:Ii! FF co I h sample relaxation
.5 3! 0.37
0.6 ... ~ ... :z:
co T c 3()6C
"-
-
1"1
'3 0.365
2 0.4
:I
'><.
--- • K- .
+.
+
..
1:
.1:1 0.36
0.2 - x §
- - oK 0 0.355
c
0 0.35
0.4 16 18 20 22 24 26 28 30
0.1 0.2 0.3 0.5 0.6 0.7
timelh)
<I> • layer thickness (lim)

Fig.7a: Tendency 0/ the solar cell to increase its efficiency after

Fig. Sa: typical saturation behaviour during light soaking; b: light soaking till saturation. b: non equilibrium effect shown by
Saturation values inJunction o/the <i> layer thickness. subsequent degradation -and relaxation steps.

If the intensity of the laser beam is increased after the saturation The slow process towards the new saturation values can, of
of the cell parameters, the cell tends to approach a new lower course, thermally be accelerated; in the extreme case until one
saturation value of all parameters. The two step (in situ) light reaches again the initial val ues.
soaking behaviour is plotted in Fig. 6.

1.2
CONCLUSION
~ 1.0
~<II We have degraded VHF-GO solar cells with a short- time high
intensity laser method; we could show by LBIC and
] 0.8
Thermography, that the observed cell volume has been degraded
iN uniformly. By this method we found that the saturation values of
0.6
i
(;
the solar cell parameters and its saturation efficiency depends on
the laser intensity and that the saturation values are stable as long
c 0.4
as the cell is illuminated. The state itself of the cell after the light
is switched off, again is in a non eqUilibrium il/ld tends, even at
0.2
room temperature, to stabilize at higher efficiencies.
We conclude, that the laser method gives a possibIiIity to
0
0.01 0.1 10 100 degrade solar cells faster than by the AM 1.5 intensity, but the
degrada~on ~m8 [hI method can only be used to observe the degradation kinetics.
The non equilibrium saturation values are again a function of
intensity and cannot be taken to provide the stable efficiency of a
Fig. 6: two step light soaking a/two VHF-GD solar cells (cellI
solar cell which is used under normal operating conditions.
streight line, cell 2 dashed line), where the generation rate was
increased qfter the saturation of the efficiency
ACKNOWLEDGEMENTS:
streight line until step: 2 x ]{j22 cm-J after step: 3 x ](j23 em-3
This work was supported by the Federal Office of Energy
dashed line until step: 15 x J(j23 cm-j" after step 3 x 10 23 em-3
(OFEN) Switzerland under contract EF-REN 90(045), and by
CERS 1820.1
The results show, that for VHF-GO p-i-n solar cells, the
saturation values are a function of the applied light intensity; the REFERENCES:
higher the light intensity, the lower the saturation values, when [1] A.Banerjee, S.Guha, A.PawIikiewicz, O.Wolf and
the temperature during the degradation is kept constant. I.Yang United Solar Systems Corp., Michigan,
For individual layers, the saturation of the density of states in International Meeting on Stability of Amorphous Silicon
function of the generation rate and temperature previously has Materials and Solar Cells. Feb. 1991
been published [3)
After the high intensity light soaking the solar cells are not in [2] U.Schneider and B.SchrOder Proc. of the 21 th
thermal equilibrium at room temperature: The parameters show Photovoltaic Specialists Conference 1990 Orlando p. 1521
the tendency to increase gradually to a new equilibrium and thus,
to new and higher saturation values. This effect on solar cells [3] P.V.Santos, W.B.Jackson, and R.A.Street. Xerox Palo
already has been reported after e-beam degradation [2]. For red Alto Research Center, International Meeting on Stability of
light degradation the effect is shown in Fig.7: Amorphous Silicon Materials and Solar Cells. Feb. 1991

207
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

REVERSAL OF FIELD AND PHOTOCURRENT IN a-Si PIN SOLAR CELLS

H.Pfleiderer
Siemens Research Laboratories
Otto-Hahn-Ring 6, DW 8000 MUnchen 83

ABSTRACT. Numerical simulations of dark and AMl characteristics are discussed. A field reversal supports the
photocurrent reversal. The principal source of reversed (secondary) photocurrents is surface recombination,
transmitting an electron current to a hole current and vice versa.

D ... The values of all the variables for arbitrary x are linearly interpolated
1. INTRODUCTION b~iween the neighbouring x.. 9
We put for cell A: S.. = vl·a.. = 10- cm3/s with some
.IJ 1J •
The photocurrent of an a-Si solar cell, i.e. the difference of the exceptions, VI thermal VelOClt}'. C.i1 = C j4 21 20 3 (ta.i!1 states),
light and dark currents, depends on the external voltage U, and vanishes c. 2 = c. 3 = 10 (bottom states). DiD = Ui4 = 10 (= eV) (valenc;,e
at the transition voltage U r The negative (positive) photocurrent for ~d 1 conduction band edges). Dk! = Dk3 = 5.101 •
U < U (U > U ) is termed primary (secondary). The reversal of the
photoafrrent at aT was observed for weak and also AMI illumination
D = 5.1015 (cm3eV)·1, k = 2 to 5. D .. for i = 0,1,6,7 (p- and n-layers)
co~iderably larger. Electron and IJ hole mobilities II-n = 7 and
fl}. II- = 0.7 cm2Ns . Built-in voltage U b = 1.0 V. .
The voltage U affects the photocurrent by its influence on the P The same parameters are used for cell B, but for the foIlowm~;
electric field through the pin structure of an a-Si cell. The field in the S =S =10·7cm3/s; D =D =5-1018, D =5'1017, D 52 =2·10 •
middle part of the i-layer vanishes at the flat-band voltage Up> and is rJ1 =6\=2-1018 (cm3e\7)-1. ifence at "2 anJ\ (near the p/i and i/n
negative (positive) for U < Up (U > Up). The field and its reversal at j~~io~ at ~ and x~ the r~combination factor S and the DGS. are
Up have to be inferred from theory. increased by a factor 100 (m most cases). These new assumptions
The flat-band and transition voltages are related to each other. stimulate a massive increase of surface recombination.
The simplest guess is Up = Ur Analytical modeling and numerical The above assumptions, defining cells A and B, are rather
simulation of spectral characteristics yielded Up < UT /2,3/. The arbitrary. The purpose of the present paper is by no means to suggest
simulation of AMl characteristics will be added here. In any case, the very realistic simulations. The comparison only of cells A and B should
conspicuous sign reversal of the photocurrent demands for an be consdidered seriously.
explanation.

3. CURRENTS, FIELDS, RECOMBINATION

2. WEAK AND STRONG SURFACE RECOMBINATION,
CELLSAANDB The simulated dark and light characteristics of cells A and B are
shown in Figs. 1 and 2. While both characteristics of cell A merge
Numerical simulation is applied in order to find the origin of beyond open circuit, the light characteristics of cell B cross over the dark
secondary photocurrents in a-Si pin solar cells. The basis of a simulation characterictic at voltages U T Fig. 3 gives ~ enlarged pictw:e ~f the
is laid by a certain set of parameters. Any set represents a particular cell. crossovers. The characteristics of cell B, Figs. 2 and 3, are similar to
We shall compare the characteristics of two cells that differ in only few observations fl}. Only the open-circuit and transition voltages are too
parameters. The frrst (second) cell termed A (B) is disposed to develop low, by about 0.2 V. The crossovers are concomitant to the appearance
a small (large) amount of recombination around the pli and i/n junctions. of secondary photocurrents, and indicate a good deal of surface
The recombination, located near these junctions, could be termed recombination SR.
"interface recombination" or "surface recombination". The latter Attached to a simulated characteristic are internal properties
designation will be used. Surface recombination, abbreviated SR, like the carrier concentration t1rrough the pin structure. The internal
provides boundary conditions in connection with simple modeling. The properties illustrate and explain calculated characteristics. Some
concept of 'SR" is to be understood in a broader sense here, however, examples will be presented. Figs. 4 and 5 show the potential derivative
embracing the enhanced recombination over a neighourhood around the (the negative electric field) versus coordinate x t1rrough cell B. Fig. 4
cited junctions, the extension of which is not exactly definable in gives an overall picture. It consists of 3 parts. The spikes on either side,
advance. Cell B shall be disposed for more SR than cell A. Hence cell A above the p/i and i/n junctions, define the boundary regions. The valley
(B) exemplifies the case of "weak" ("strong") SR. We shall see that only in between, with rather uniform fields, represents the bulk region. The
cell B exhibits secondary photocurrents. space charges are concentrated in the boundary regions. The applied
The physical assumptions used here will be outlined now. The voltage U mainly shifts the bulk fields. Fig. 5 gives an enlarged picture.
pin structure is divided in 7 regions, having boundaries l1' The With increasing U, the field in the bulk region reverses its direction. The
coordinates "0 = 0 and ~ = 0.6 ILIIl mark the outer boundaries. The lowest voltage that induces a vanishing field amidst the bulk region is
regions 1 and 7, extending between "o'~ and x6~ , represent the p- and termed flat-band voltage Up. Its value is Up = 0.49 V.
n-layers, and are 15 nm thick. Region 4, extending between ~,x4 ' Fig. 4 lets recognize well defined regions. In reality, the borders
represents the i-layer bulk. The other regions can guide gradual may be less distinctive. The field spikes in the boundary regions originate
parameter variations. The thicknesses of regions 2 and 3, and also of 6 from the tail states. Cell A shows a similar field distribution. But the
and 5, are 15 and 30 nm. The density of gap states (DGS) is composed of field spikes (the bulk fields) of cell A are lower (higher) than those of
4 branches, specific exponential functions D.(E), with index j from 1 to 4, cell B. The voltage Up = 0.675 V is also higher. Hence a low Up favors
and energy E through the gap of widfu EG = 1.7 eV. Branch 1 the appearance of secondary photocurrents. In Figs. 1 and 2, showing
represents the donor tail states, branch 2 the donor bottom states, the characteristics of cells A and B, the values of Up are indicated. For
branch 3 the acceptor bottom states, and branch 4 the acceptor tail cell A (B) the relation U > U o (Ur < UO> is valia, with U o ~ open-
states. The donor branches, bottom branches, acceptor branches cross at circuit voltage under AMf light. An mcrease of the bulk DGS m cell A

e
energies E = 0.3EG, E2 = 0.6E G , E = 0.8EG . The valence. an.d
conduction band edges are Eo = 0 and 4 = E G. The recombmatlon 1S
determined by the D. and the capture cross sections (for electrons and
(D = 2.5.1016 (cm3eV)·1, k = 2 to 5) leads to the relation U < U o
witfout admitting secondary photocurrents. Hence only strong SR and
low U F together bring forth secondary photocurrents.
holes) attached to th6m. The capture cross sections of neutral (charged) The recombination rate r of excess electrons and holes versus x
traps are a. (C.·a.). At coordinates x = x. the parameters ajJ' and CjJ. are is quite revealing. (In thermal equilibrium r = 0.) Figs. 6 to 9 show rex)
J • J J 1 • h
to be prescribed, and furthermore at energres E = E j t e parameters in the dark and under AMl light, Figs. 6 and 7 (8 and 9) for cell A (B).

208
i Current I
45
i 'f"
1.5

AM1 =1
-HlI
/I
rnA dark ~0 =0
crn 2
UF
dark

o
0=0 ~J
~
~ U=O 11
0=1 o
-0.15 0.8 V
-15
0.3 V 0.8 o 0.3 0.6
Voltage U --+ x ----+
Fig. 1: I-V characteristics in the dark and under AM1, Fig. 4:1'~otential derivative, x coordinate through pin, voltages V, cell B
flat-band voltage Vp' cell A

i Current I
45 r----------.--------------~~
i ;·~__----------U-F-=,0-.4-9-V--------~~

0=0
0 0.8

dark
0=1
-15 -1
0.3 V 0.8 o 0.3 /.!rn 0.6
Voltage U ---. x -+
Fig. 2: I-V characteristics between dark and AM1,
Fig. 5: As Fig. 4, other ordinate scale
flat-band voltage Vp' cell B

j Current
10
I
r-------------------------~

The function rex) increases with V and with an illumination, especially

for V = O. The two boundary regions and the bulk region are again
clearly distinguishable. The extra recombination in the boundary regions
represents SR. The boundary regions of cell B are thicker than those of
cell A. Insofar SR of cell A (B) is "weak" ("strong"). Bulk recombination
with V > 0 is strong for weak SR and vice versa. It can be observed
o from Figs. 7 and 9 that the amount of SR considerably increases as soon
as V surpasses Vp. This behavior is to be understood simply. The
reversed bulk field in case V > V p drives photoelectrons (photoholes)
towards the pli (iln) junction where abundant holes (electrons) are ready
to serve as recombination partners. SR obviously promotes (reduces)
secondary (primary) photocurrents. This connection was already
discussed on an argumentative level 14/.
-5 The behavior of some internal properties directly reflects SR.
0.5 V 0.65 Two examples will be given, concerning photocurrents and the electron
concentration. Fig. 10 and 11 show the photocurrents of electrons and
Voltage U-+ holes (light currents minus dark currents) of cell B. At short circuit
Fig. 3: As Fig. 2, other scales U = 0 the primary electron (hole) current is leading near the iln (p/i)

209
 r
1024 1~024
U = 0.8 ~ U = 0.8
A
~ .......
cm 3 s cm 3 s 0.6
~~
0.6
~ ~
~
0.4 OV

AMl light

1012L-________________ ~ ______________ ~

o 0.3 ).lm 0.6 0.3 0.6

x x -----.

Fig. 6: Recombination rate r(x) in the dark, x coordinate through pin, Fig. 7: As Fig.6, but under AMl
voltages U, cell A

i~024r---------------.---------------~ i ~024 r----------------r--------------~

U = 0.8 U = 0.8
0.6

__--------~--~0.~4----_ OV

AMl light

dark

10 '2 ~---------------L----------------U 1012 ~------------__..L-______________-.J

o 0.3 tJ-m 0.6 o 0.3 Ilm 0.6
x -----. x-
Fig. 8: Recombination rate r(x) in the dark, voltages U, cell B
Fig. 9: As Fig. B, but under AM1

junction. Beyond the flat-band voltage, for U > U a secondary

electron photocurrent builds up in the i-layer near fue p/i junction . Figs. 12 and 13 show the free electron concentration n(x) of cell
(electrons moving to the left), and is transmitted to a secondary hole A m the dark and under light. A forward voltage U > 0 injects carriers
photocurrent through the p-layer (boles moving to the right). In other and rises n(x) in the dark. But at the p-side n(x) decreases under light
words, the photoelectrons and photoholes move towards the p/i junction with increasing U. This inversion is a consequence of SR : The injected
from opposite sides. A less distinctive but similar situation exists around holes recombine with photoelectrons, reducing the concentration of
the i/n junction, the roles of electrons and holes being interchanged. The them. At the n-side, the respective hole inversion occurs. Cell B shows
collision of photoelectrons and photoholes at the p/i and i/n junctions an even more pronounced inversion.
gives rise to the SR hills in Fig. 9. More strictly, the photorecombination
rate, the difference of the plots in Figs. 9 and 8, would have to be
considered in the present context. Actually, the difference looks similar
to Fig. 9. The electron and hole currents of Figs. 10 and 11 add up to
constant total currents. A secondary photocurrent also builds up in cell
A near the p/i junction for U > U F. The total photocurrent stays
primary, however.

210
i Ip
20
Ip
20

electrons holes
rnA
0.8 V
crn 2

0.8 V

0
o 0.6
U= O
0.4

-15
-15
o 0.3 0.6 o 0.3 Jlm 0.6
x --+ x--+

F".g. 10: Photocurrent Ip(x) of electrons, voltages U, cell B Fig. 11: As Fig.lO, but holes

i~ 0 17 . . . . - - -- - - - - -, - - - - - - - - - - - , i ~0'7 r - - -- - - - - - r - -- - - - ---,

AM1 light

0.3 0.6
x --+ x ~

F".g. 12: Electron concentration n(x) in the dark, cell A Fig. 13: As Fig.l2, but under AM1

usually disregarded, however. Hence while the present s~u~a.tion has t~

4. DISCUSSION be improved, it nevertheless discovers important peculianties of a-S.
solar cells. The reduction of surface recombination by buffer layers is
Simulation leads to secondary photocurrents under two
already common practice. Renewed efforts t~ better understand and
previsions:
further improve these layers could be worth while.
1) Surface recombination is preponderant
2) Flat -band voltage is low
The first (second) item represents the main (an auxiliary) condition.
REFERENCES
Reversed electric fields activate surface recombination. This mechanism
/11 Y.M.Sun et al: 8th E.C. Photovoltaic Solar Energy Conference,
drives secondary photocurrents.
Florence (1988) 666
The qualitative concepts proposed here, strong surface
!21 H.PfIeiderer et al: 20th IEEE Photovoltaic Specialists Conference.
recombination and early field reversal, favor secondary photocurrents
Las Vegas (1988) 180
and are detrimental to primary photocurrents. Usual simulations avoid
13/ H.PfIeiderer: J.Non-Crystalline Solids 115 (1989) 57
these undesirable features. Indeed, the calculated characteristics of
/4/ W.Kusian et al: J Appl.Phys. 64 (1988) 5220
F".gs. 2 and 3 for cell B are quite bad. They are realistic only in showing
the turnover from primary to secondary photocurrents. Just this aspect is

211
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

OPTICAL MODELUNG OF .·SI:H AND .·Ge:H P·I·N SOLAR CELLS

.
R. Saeng-udom, W. Kusian, and B. Bullemer"

Siemens AG, Corporate Research and Development

Otto-Hahn-Ring 6, D-W-8000 Milnchen 83, Germany

• Federal Armed Forces Univ., Faculty of EI.Eng.,Inst.of Physics

Wemer-Heisenberg-Weg 39, D-W-8014 Neubiberg, Germany

ABSTRACf. We have developed an optical model for amorphous thin-film solar cells. The cells are represented
by plane-parallel multilayer thin-film systems. A single-junction a-Si:H solar cell model was achieved by means of
electromagnetic theory. The calculated i-layer absorption compares well to spectral response experiments.The
modeJ is extended to multijunction solar cells, i.e. the a-Si:Hla-Ge:H tandem and the a-Si:Hla-Si:H stacked solar
cells. Optimum i-layer thicknesses of the subcells are calculated in order to yield current matching. The ~c~ess
pairs for top and bottom cells are 90 / 500 nm for stacked cells and 500 / 81 nm for tandem cells, resulting In a
short-circuit current density of 7.0 and 12.5 mA/cmz respectively. Further increase in short-circuit current
demands greater i-layer thickness for the a-Si:H top cell and should be avoided as it makes the cell more subject to
light induced degradation.

1. INTRODUCTION For the optical model of our p-i-n solar cells,. we assumed
semi-infinite glass to be the incident medium, semi-infinite silver back
Since the first single-junction amorphous thin-film solar cell contact as the transmitting medium and plane-parallel boundaries
was reported /lJ, the development of such solar cells has been pushed everywhere. By applying electromagnetic wave theory and boundary
ahead in many laboratories around the world. However, for large- conditions which conserve the tangential components of electric and
scale power applications two main problems must be solved. Firstly, magnetic field vector amplitudes at every interface crossed by the
the efficiency of the solar cells must be increased and secondly the normally incident wave, we obtained the electric field vector
light-induced degradation must be diminished. Up to now, the a-Si:H amplitudes E of the positive- and negative-going waves in the
p-i-n solar cell does not operate at 15 % cell efficiency expected of a z-direction from the optical constants and the thicknesses of the layers
thin-film tandem cell and it still suffers from considerable Iight- involved. At a fixed wavelength, the intensity at any point z in the
induced degradation. Multijunction amorphous thin-film solar cells solar cell is calculated from the Poynting vector
have been reported that reach a theoretical efficiency of more than
15 % /21 and feature reduced light degradation (3,4/. In this paper we I(z) =.1.
2
Re [E(z) H" (z)]
will optically model both the single-junction and the multijunction
solar cells and provide design rules for optimum multijunction cell
structures~

With the aid of electromagnetic theory and in particular thin-

l
film optics;it is possible to model amorphous thin-film solar cells. We 1.0
will do this for a-Si:H p-i-n solar cells in the wavelength interval of 0.3 - - - Model /

\\
- - Experiment /-
to 1.5 IJ.lIl, a range which is of practical use for tandem solar cell ~ /

IV ""f\V\
c 0.8
design, and then compare our model with the optoelectrical
'0
:c
'"
characterization of real solar cells deposited on a smooth substrate.
c'"
The model will be described and subsequently applied to a-Ge:H p-i-n

V \\
.2 0.6
0
solar cells. The extension of this model to a-Si:Hla-Ge:H tandem and ~
"0
to a-Si:Hla-Si:H stacked solar cells will then be introduced and U

'i,
A

J
0.4
discussed.

2. OPTICAL MODEL OF SINGLE-JUNCTION P-I-N SOLAR

CELLS

To model the a-Si:H solar cell optically, we must know the

0.2

0,0
1/ "-
I~
0.3 0.4 0.5 0.6 0.7 jlm 0.8
optical constants of the constituent layers. Therefore, we deposited
Wavelength A - -
four types of amorphous films, namely p-SiC, i-S;, noS;' and i-Ge layers
separately on calcium fluoride substrates in rf glow discharge
reactors. We determined the optical constants of these films by Fig. 1: Measured collection efficiency (solid lines) of a pin diode
evaluating their reflectance and transmittance values in the under reverse bias and calculated absorption (dashed lines)
within the i-layer versus wavelength.
wavelength interval of 0.3 to 1.5 ~ /5/.

212
 The film absorption A at each wavelength was found by the
number of photons absorbed in the film normalized to the number of r
c
incident photons at the gIasslTCO boundary. Assuming no o
recombination losses within the i-layer, the maximum short-circuit 'eo-
current density, J sc' can be calculated from the absorption in the
D '"
'"Q;
i-layer, Ap0 for any wavelength interval '-t to ~ by the equation >.

-j'"
}.,

J sc = ~c (S(),) ~(>-) dA

>,.
where e is the elementary charge, h is Plank's constant, c the speed of
light and SeA) a given spectral irradiance. The current density losses
of the other layers are calculated from similar equations 151. 11 m
We tested our model by comparing it with a real a-Si:H p-i-n Wavele ngth ).

solar cell. The cell was produced on a smooth gIasslTCO substrate. A

silver contact was then sputtered on top of the n-Iayer to form the Fig. 3: Calculated i-layer absorption versus wavelength of an a-Ge:H
solar cell for different i-layer thicknesses.
back contact. Collection efficiency measurements with various reverse
bias voltages were performed in the wavelength interval of 0.3 to
0.8 ",m. Fig.l shows the calculated i-layer absorption and the
measured collection efficiency at a reverse bias voltage of -1 V for Ftg.2 shows the influence of varying the i-layer thickness on i-layer
which the no-recombination-loss assumption certainly holds 16/. Good absorption for an a-Si:H p-i-n solar cell with a g1ass/TCO(200 nm)/p-
qualitative agreement is found between the model and the SiC(20 nm)/i-Siin-Si(lO nm)/Ag structure. It can also be seen that the
experimental results, including the interference effects at long cell is particularly sensitive in the wavelength interval of 0.3 to 0.8 iLm.
wavelengths. Increasing the i-layer thickness improves absorption of red light and
The wavelength region over which solar cells are sensitive causes marked interference patterns in the long wavelength region. By
depends, among other factors, on the thickness of the i-layer and the means of AM 1.5 illumination, i-layer thicknesses of 100, 300, 500 and
optical bandgap. We calculated the effect of i-layer thickness on the 700 nm lead to Jsc values of 9.0, 12.9, 14.4 and 15.3 mA/cm2
a-Si:H (Eg = 1.75 eV) and the a-Ge:H (Eg = 1.1 eV) solar cells by respectively.
varying the i-layer thickness and keeping all the other layers constant. We also studied the effect of i-layer thickness variation on the
a-Ge:H p-i-n solar cell using a g1ass/TCO(2OO nm)/p-SiC(20 nm)/i-
Siln-Si(10 nm)/Ag structure. Fig.3 shows, paralleling Fig. 2, the i-layer
absorption of a-Ge:H p-i-n solar cell in the wavelength interval of 0.3
to 1.2 ",m. The i-layer thickness varies from 50 to 500 nm. The short-

1.0 circuit current density is calculated from an AM 1.5 spectrum in the

pina-Si:H wavelength interval of 0.3 to 1.2 j.I.M. For i-layer thicknesses of 50, 100,

c: 200,300 and 500 nm, J se was found to be 19.4, 26.8,29.2,31,2 and 32.6

e-o 0.8
.Q
mA/cm2 respectively. The bandgap of the a-Ge:H material, which is
II narrower than that of a-Si:H expands the i-layer absorption of the
'"Q; 0.6 a-Ge:H p-i-n solar cell towards the red part of the solar spectrum.
>-
~ Thus markedly enhanced Jse values are obtained.
/ ---
100nm ··
0.4 3. OPTICAL MODEL OF a-Si:H/a-Ge:H TANDEM SOLAR
CELLS
._--..-...
0.2 The a-Ge:H solar cells being developed that will form highly
efficient tandem cells with a-Si:H solar cells. The question now is how
0.0 to choose the thicknesses of the two active i-layers so that the
0.3 0.4 0.5 0.6 0.7 ",m 0.8 photocurrent in both subcells match. We therefore simply extended
Wavelength A our optical model of single-junction solar cells to tandem solar cells
based on the same assumptions as above. For the calculations, we
used the tandem structure g1ass/TCO(2OO nm)/p-SiC(20 nm)/i-Si In-
Fig. 2: Calculated i-layer absorption versus wavelength of an a-Si:H
solar cell for different i-layer thicknesses. Si(lO nm)/p-Si(2Onm)/i-Ge In-Ge(lO nm)/Ag. The first p-i-n junction
(top cell) is assumed to be made of amorphous silicon and is identical
with the structure described in section 2. The second p-i-n junction

213
 I
1.0 20
di,bottom therefore enhances Joe' As a corollary, the i-layer thickness of
mA
em' the top cell, ~,top' must then also increase in order to satisfy the
,..
Q;
18 current-matching condition, For ~,bottom values of more than 100 om,
~ Jso tends to saturate.
a.
g
0 A high J so value implies a high value for di,bottom and an even
16
.,~ greater one for di,top' This results in increased top-cell absorption that
c
><
0 is effective in the wavelength region in which both cells are active.
:c
I-
0.4 14 Therefore, Joe should remain below 15 mA/cm2 in an a-Si:Hla-Ge:H
tandem cell. Another limitation on the choice of suitable
di,top/di,bottom pair is imposed by light-induced degradation of the
0.2 12
a-Si:H top cell. Increased di,top would lead to stronger degradation of
this cell m and, with it, of the entire tandem structure. For this
~ __~____L-__~__~____~__~____~~ 10

40 100 120 nm140

reason, di,top should not exceed a value of 0.5 j.LID. The corresponding
60 80
Thickness of bottom i-layer di,bottom in this case is then about 80 om, leading to a short-circuit
current density of 12.5 mA/cm2, Fig.5 shows the calculated i-layer
Fig_ 4: Contour plot of optimum thickness pairs for i-layer of top and
absorption of the constituent solar cells for this thickness pair. Most
bottom cells in a-SiJa-Ge tandems along with the resultant
short-circuit current. of the blue photons (j,. = 600 nm) are absorbed by the top cell, while
the red light (~~ 700 nm) is absorbed by the bottom cell . .Thus, in
(bottom cell) is an a-Ge diode, in which the p-layer consists of silicon this optimum case, the intermediate region, where both cells are
carbide_ The optical constants of the n-Iayer are assumed to be sensitive, is quite narrow and extends from 600 to 700 om,
identical with those of the i-layer. Top cell and bottom cell are series-
connected. Thus, the greatest efficiency will be obtained when the 4. OPTICAL MODEL OF a-Si:/a-Si:H STACKED SOLAR CEllS
thickness of both i-layers is adjusted so that the photocurrents
generated by each cell are equal under AM 1.5 illumination. Starting Stacked cells "consisting of two a-Si:H cells on top of each

with the given i-layer thickness for the bottom cell, we calculated the other are under discussion as a means of enhancing the stability of

adjusted i-layer thickness of the top cell under the current-matching a-Si:H solar cells. To model this stacked solar cell, we used the

condition and thus obtained the optimum thickness pair, i.e. d.I,top and structure: glass rrCO(2OOnm)/p-SiC(20nm)/i-Si /n-Si(1Onm )/p-
SiC(20nm)/i-Si /n-Si(lOnm)/Ag. Again, short-circuit current matching
di,bottom'
FIg.4 shows the contour plot of these optimum thickness between both cells is our requirement for adjusting both i-layer

pairs along with the short-circuit current densities of an thicknesses. Fig.6 shows the contour plot corresponding to FIg,4, The

a-Si:Hla-Ge:H tandem cell for various thicknesses of the bottom matched short-circuit currents tend to reach saturation as the bottom

i-layer. A marked increase in Joe is seen for i-layer thicknesses i-layer is thicker than 700 nm. To reduce light-induced degradation, a

di,bottom of more than 70 nm. This thickness is necessary in order to p-i-n/p-i-n stacked solar cell structure should be adopted so that the
have a sufficiently long absorption path for the red light. Increasing total thickness of both i-layers is nearly the same as the i-layer of the

1. 0
110
a-Si:Hfa-Ge:H tandem solar cell

f\
100 8.0
c
~ 0 .8 Q; nm mA i!:"
>-
(;
Jlco
a'Si:Hj f'' k- a-Ge:H :!a. 80
em' iii
c:
7.0 "0
.,
~ 1\ ! g
0
c:

/
0.6 ~
co
~ V~J ,1 .,
VI
VI
60 6.0
;>
0

t\ ~ 5
.2 ~
.c U

/ I T\
0.4

'\
l- t::
40 5.0 0
.c
\", \ .,'"
0.2
! \ /".
"0
.c
u

) iii

~
20
\ I \ 4.0 :::;;

o
0.3 0.4
J
0.6 0.8
1 "
1.0 11m
\
1.2
0
a 0.4
a-SI:Hfa-SI:H stacked solar cells

0.6 0.8
3.0
1.0
11 m
Wavelength i..
ThIckness of bottom i·layer - -

Fig. 5: Calculated i-layer absorption of top and bottom cell in an Fig. 6: Contour plot of optimum thickness pairs for i-layer of top and
a-SiJa-Ge tandem cell versus wavelength; ~,top = 0.5 ,","In, bottom cells in a-Si/a-Si stacked cells along with the resultant
~,bottom = 80 nm. short-circuit current.

214
 tI 0.8
a-Si:H/a-Si:H stacked solar cell
5. SUMMARY

c: The comparison of theory and experiment demonstrates the

o
usefulness of our optical model for the a-Si:H p-i-n solar cell.
~ 0.6f-----t:---=:--:-:-t----t-hi---t----
Therefore it is also useful for optical modelling of multijunction solar
~
~ cells for which experimental data are comparatively scarce. We note
j 0.4~--_*---+--_A__I-----'hGl---_1
that the calculated short-circuit current densities under AM 1.5 solar
spectrum are rather low. This is largely due to the ideally plane-

\ parallel interfaces, which are assumed to be suppressing all stray-light

\ effects. However, that should not notably restrict the general
0.2~--I-_+----H'---__1H.----+lr__-_1
applicability of our findings: current matching appears to impose
\
\
'-
stringent requirements on i-layer thickness pairing. Stability
....
O~____~~~~____~____~~~~
considerations additionally limit the maximum thickness of the top
0.3 0.4 0.5 0.6 0.7 ~m 0~8 cell. As a consequence, the overall efficiency of two-terminal tandem
Wavelength A - cells will most likely not fulfiIl optimistic expectations /2/. However,
the current-matching restrictions can be circumvented by using four-
Fig. 7: Calculated i-layer absorption of top and bottom cell in an terminal tandem solar cells.
a-Si/a-Si stacked cell versus wavelength; ~,top = 90 nm,
di,bottom = 0.5 1Llll.
ACKNOWLEDGEMENT
We are thankful to Dr. E. Giinzel for supplying a-8i:H films
and solar cells. We would also like to express our appreciation to Dr.
W. Krtih1er and Dr. H. Pfleiderer for their helpful suggestions. This
work was supported under the auspices of technological program of

highly efficient single-junction p-i-n solar cells /8/. Fig.7 shows the top the Federal German Ministry of Research and Development.

and bottom i-layer absorption as calculated for the optimum thickness

pair ~,top = 90 nm and ~,bottom = 500 nm. The resulting short-circuit REFERENCES
current density is 7.0 mA/cm2• As expected again, the blue photons /11 D.E. Carlson and C.R. Wronski, Appl. Phys. Lett. 28 (1976) 671
are absorbed significantly by the top cell, and the long-wavelength /2/ J.C.C. Fan and BJ. Palm Solar cells,10(1983) 81
photons are absorbed more by the bottom cell. Note the interference (3/ K. Asaoka, M. Yamaguchi, H. Yamaguchi, WA. Nevin, H. Nishio,
pattern occuring in the i-layer absorption of the bottom cell. In T. Endoh, K. Tsuge, Y. Tawada, Jour. Non.-Cry. Solid.115 (1989)
contrast to Fig.5, the intermediate range of sensitivity is more 24
important. Results for such stacked structures were reported recently /4/ S. Fujikake, S. Saito, H. Sato, T. Yoshida, Y. Ichikawa and H.
/4,8/. The cells were rather small in area (about 1 cm2) and the i-layer sakai, Proc. of Int'l PVSEC-5 (1990) 623
thicknesses quoted by the authors were di,top = 80 nm and /5/ R.Saneg-udom, W. Kusian and B. Bullemer, Proc. of Int'l PVSEC-
di,bottom = 300 nm. It was found that these cells indeed showed a 5 (1990) 331
reduced photodegradation of about 14 % after 1000 h AM 1.5 /6/ W. Kusian, H. Pfleiderer, B. Bullemer, Mat. Res. Soc. Symp. Proc.
illumination. The initial efficiency was 11.4 % and the stabilized final Vol. 118 (1988) 183
value was 10 %. Our design of double-stacked a-Si:H solar cell my. Uchida, M.Nishiura and H. Sakai, Solar Cells 9 (1983) 3
justifies this choice. /8/ W. Krtih1er ( to be published)

215
10TH EUROPEAN PHOTOYOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

An a-Ge:H pin Solar Cell wHh High Red Efficiency

W. Kusian, E. GQnzeI, R.D. Pliittner

Siemens AG, Corporate Research and Development, Munich

ABSTRAcr. Solar cells with pin structure were manufactured from pure hydrogenated amorphous germanium in
order to obtain a red sensitive sub cell in a tandem stucture. Different deposition conditions were tested. A substrate
temperature of 150·C and bombardment with ions of energy greater than 200 eV proved to be the best conditions.
These conditions and a cell constructed with a p-SiClp + -Si front contact and an i-layer thickness of round 0.3 IIJll
yielded an overall cell efficiency under AMl illumination of 2.5 %. The external short-circuit collection efficiency for
an illumination wavelength of 900 nm in this cell was 2S % and increased to over 40 % under a reverse bias voltage of
-1 V.

1. INTRODucrroN
200 eV). High rf-power (150 W) and just 5 % GeH4 diluted in ~ as
Thin-film solar cells based on hydrogenated amorphous silicon process gas constituted the hard-deposition conditions. The substrate
(a-Si:H) have reache.d high efficiency values /1/. However, they still temperature T S ranges from 130 DC to 230 ·C. The hitherto best structure
undergo light-induced degradation during operation, so that further for a-Ge:H solar cells utilizes deposition on a glasstrCO substrate,
improvements in efficiency and, especially, stability are necessary. followed by a p-silicon window layer with a bandgap of roughly 1,7 e V, an
Tandem solar cells combining a blue-sensitive a-Si:H cell with a red- intrinsic a-Ge:H layer, an n-doped a-Ge:H layer and finally a palladium
sensitive solar cell made from hydrogenated amorphous germanium back contact. The i-layer thickness varies from 0.18 IIJll to 0.49 1IJll.
(a-Ge:H) appear to fulfill these demands. Amorphous germanium can
also be deposited via the plasma enhanced CVD technique. So it should
be possible to prepare monolithic a-Si:H/a-Ge:H tandem solar cells. We 3. RESULTS AND DISCUSSION
started to develop pure a-Ge:H material with the aim of using it as the
active i-layer in pin solar cells. The development of a high-performance amorphous germanium
Two sets of plasma conditions have proved favorable for solar cell that is especially sensitive to red light imposes several demands
preparing intrinsic a-Ge:H layers showing high norma1ized on the technology. First, it is necessary to establish the plasma conditions
photoconductivity values ('If.llr products) 12,3,4/. One ("hard" deposition) for the preparation of amorphous germanium material with high
comprises deposition conditions of high . rf power, strong ion photoconductivity, a low defect density and a sufficiently high minority
bombardment on the growing film and extensive dilution of the germane carrier lifetime. Then the solar cell structure must be designed so as to
process gas (GeHJ with hydrogen. The other ("soft" deposition) utilizes have good junction properties and a strong electric field across the i-layer
the opposite conditions, i.e. low rf power,low ion bombardment and no in order to limit recombination losses. Of the various possible deposition
hydrogen dilution, and is similar to the deposition of high quality a-Si:H. parameters that have to be optimized for depositing high-quality
Better films were obtained via the hard deposition mode and the first amorphous germanium material, we choosed the process pressure and
a-Ge:H solar cells were therefore made with "hard" deposited material the substrate temperature in our early work /5/. For this paper, we
/5/. Despite of this the efficiency of these first cells was rather low (2 %), refined these investigations and also tried to demonstrate the influence of
a fact which was due to the high defect density of the a-Ge:H material ion bombardment on cell efficiency during deposition.
and to the poor IJ.'" product of the minority carriers (holes). The preliminary studies showed that a high T s (2OO.C) leads to
We report here on further developmental work on a-Ge:H solar improved red response of the solar cells, but at the same time the dark
cells, in which we focused our interest on the influence of substrate current of the diode under reverse bias shows no saturation /5/. In
temperature, i-layer thickness and ion bombardment on the red response contrast, solar cells deposited at low Ts (100 ·C) showed less red
of the cells. sensitivity. Fig. 1 revealed the external collection efficiency Q versus
wavelength A of a-Ge:H solar cells deposited at different Ts (smaller
temperature steps then in ref. 5). The i-layer thickness in all cells was
2. SAMPLE PREPARATION 0.5 IJ.m. Again, there is a tendency for Q(>.= 1000 nm) to increase with
increase in Ts' The overall efficiency oeeured at a substrate temperature
The solar cells were deposited by means of a conventional rf of 150 ·C. At this Ts value, the diode behavior is also the best for this
glow-discharge reactor having an asymmetrical electrode configuration as sample series. Consequently, a plot of AM1-efficiency as a function of
described in /2/. The substrates were mounted on the smaller rf electrode substrate temperature shows a maximum as well at T s = 150·C (see
in order to grow the samples under strong ion bombardment (about Fig. 2).

216
 100 100
% %
0 0
.,c~
.y .,c~
.,
:t:: :::.,
{}

c: c:
0 Q
~ U
8
~

1200 1200
Wavelenglh ). Wavelength ).

Fig. 1: Spectral external collection efficiency of a-Ge:H solar cells Fig. 3: Spectral external collection efficiency of a-Ge:H solar cells
deposited at different substrate temperatures versus wavelength deposited at Ts = 150 ·C with different i-layer thicknesses versus
wavelength

The absorption of intrinsic a-Ge:H films is very high in the voltage-dependent spectral response measurements 151. In relation to the
wavelength region of importance to tandem cell operation. Therefore, built-in voltage of approximately 600 mV, which is a consequence of an
i-layer for preparing the solar cell can be thinner than 0.5 fIoID. Fig. 3 Eg value of 1.1 e V and the Fermi-level position in the p- and n-layers, this
shows the short-circuit external collection efficiency versus wavelength of estimated value for the electric field strength is about one order of
a-Ge:H solar cells of different i-layer thicknesses (deposition temperature magnitude too low. The losses in electric field strength are caused by the
150 0c). As can be clearly seen, the spectral response towards the red electric field spikes at the pi- and in-junctions and the high defect density
part of the spectrum increases if the i-layer thickness d j lies between in the hitherto achieved a-Ge:H material. Both effects can be reduced by
0.18 fIoID and 0.29 fIoID. Over this range, the collection efficiency is limited lowering the midgap density of states and the slope of the tail states at
by the small number of absorbed photons. We developed an optical both sides of the mobility gap, which we hope to obtain by increasing the
model for a-Ge:H solar cells and found that the theoretical absorption ion bombardmant energy towards 300 eV.
within the i-layer at a wavelength of 900 nm also tends to cause saturation The efficiency of solar cells under AMI light is shown in Fig. 4
for d j values of 0.3 ....m /6/. The electric field across the i-layer of solar as a function of i-layer thickness. Like the spectral response in Fig. 3, the
cells having thicker i-layers is not sufficient to collect all photogenerated efficiency peaks for an i-layer thickness of 0.3 fIoID. This value seems to be
carriers, with the result that Q again decreases if d j is too high. In fact, a good compromise between the two demands for adequate absorption
the electric field in the middle of the i-layer is already to low when and the strong electric field within the i-layer and so confirms the
d j =0.3 fIoID. An upper limit of 5*103 V/cm was estimated from theoretical calculations.

2.6

° O
J.0
,-......
"~
2.4

/
~

.,c::~ r-....

/\
G 2 .2

/
.:
w
1.8
r 2.0
'/ -
\
1.6 '-----

1.4

1.2
" ~
- :::::::::...
I 1.4
1.8

1.6
I
C~~ j
1\
j
120 140 160 180 200 220 C 240 100 200 300 400 500 nm 600
SubSlrate temperature Ts I·tayer thickness

Fig. 2: AMI efficiency of a-Ge:H solar cells versus substrate temperature Fig. 4: AMI efficiency of a-Ge:H solar cells versus i-layer thickness

217
 80

%
U - OV
0
100
.
0/.

0 80

(
>-
~
c:
60 .,c<> ' \ strong ion bombardment

,l/. . . . . "'
'0
~., with p+.Si.layer
::
., 60
c
g
0
40
~
c:
.2

40
\ 1\

\
~ "0
I V~eak
<.)
<.) Ion \
bombardment \
,,
y
20ri-n~---+-----4~~--~----4
I
20

O~~~__~~__~~__~~~~
o
400 600 900
"'"
" ... .... _- ~
1000 nm 1200
400 600 800 1000 nm 1200
Wavelength A Wavelength A

Fig. 5: Spectral external collection efficiency of a-Ge:H solar cells Fig. 7: Spectral external collection efficiency of a-Ge:H solar cells
deposited with a p + -Si and a p-SiC front contact versus deposited under weak and strong ion bombardment versus
wavelength wavelength

Another way of improving a-Ge:H solar cells involves the p-layer. layer. The stronger electric field in this case improves the red response
In our cells we used different types: a p + ·Si layer, a p-SiC layer and a and the fill factor of the light characteristic to a value of 47 %. The use of
double layer of both. Varying the thicknesses of the p + -Si and p-SiC a p-SiCip + -Si double layer combines both advantages, namely, better
layers in relation to each other, we obtained either a high open-circuit transparancy towards blue light and stronger electric field across the
voltage Uoc or a better short-circuit current density and optimum values i-layer. Indeed, the better a-Ge:H solar cells were made by using this
in between. We compared the two extremes, i.e. solar cells having either double layer for the front contact and show an efficiency of about 25 %
a p + -Si or a p-SiC front contact. The corresponding results for spectral (Uoc=307 mV, Ioc = 16.4 mA/cm2, FF = 49.2 %).
response and light characteristics under AM1 illumination are shown in Finally we want to demonstrate the effect of ion bombardment
F'I!S. 5 and 6. The use of just a p-SiC front contact increases the on solar cell behavior. This has already been demonstrated in the case of
open-circuit voltage. However, as can be seen in Fig. 5, the origin of the a-Ge:H material 12,3/. Fig. 7 is a plot of the short-circuit external
photoeffect is confined to the vicinity of the pi-junction. The blue collection efficiency versus wavelength for two a-Ge:H solar cells. The
response is quite good, whereas the red efficiency is rather poor, because dashed curve represents the spectral response of a solar cell deposited
of the low electric filed across the i-layer. In addition, the light under low ion bombardment. To prepare this cell, we mounted the
characteristic shows almost no curvature, resulting in a fill factor of substrate on the grounded electrode. This electrode is the larger one.
almost 25 %. With a thick p + -Si front contact on the other hand, the blue 'The bombardment energy was as low as 50 eV. It can be clearly seen in
response of the cell is small, because the light is strongly absorbed in this Fig. 7, that this cell, which has the same structure as cells discussed

30

r
O r------,-------.---r---.~--__.
I fI
A
.., , 1/
-., cm 2 20
~ -0.005t-----1t----1I--1V- - - 1 t - - --1
~ II
c .,'"
c:
10 I.
a
~
C
);
-
C
!!1 ~ ~rk
:s ·0.010 I-----+--,."'-~'---+----+------i
Ij
:> 0
<.) U

---====v
weak ion bombardment-y

-
· 10
light
·0.Q15~..-"'----t_---t_---t_--___l
·20
-f--
... -k-:---
?'
r- 'strong Ion bombardment

-0 .020 ~---'-------,-'-------:-'--------'
·30 I
o 0.1 0.2 0.3 V 0.4 -1.0 -0.5 o V 0.5
Voltage U Voltage U

Fig. 6: AMllight-characteristics of a-Ge:H solar cells deposited with a Fig. 8: Dark- and light-characteristics of a-Ge:H solar cells deposited
p + -Si and a p-SiC front contact versus wavelength under weak and strong ion bombardment versus wavelength

218
previously, shows poor sensitivity. The solid curve in F"tg. 7 represents the 9.1016 cm-3 as measured by PDS) and also to the Fermi level position,
spectral response of an a-Ge:H cell deposited on the rf-electrode, where which is about 0.1 eV ahove midgap. To improve the hole properties we
the bombardment energy reaches values of 200 e V. The i-layer material lightly doped the "i-layer" by adding diborane to the germane gas and
for this deposition mode is of better quality, as expressed by a product obtained an improved ""~ product for holes that is closer to that for
1/,,"~ =5.10-7 cm2N, and consequently the sensitivity of this cell is electrons. A first solar cell made with this slightly doped active layer
enhanced, especially towards red light. The collection efficiency for showed an efficiency value of 3.2 % (Uoc =312mV, Isc =20.3 mNcm2,
illumination of 900 nm wavelength reaches 25 % under shorl-circuit FF=49.8 %). We will report on these investigations in the near future.
conditions and can be enhanced to over 40 % when a reverse bias voltage
of -1 V is applied. In contrast to this, high-quality a-Si:H solar cells show
no response in this wavelength region. In F"tg. 8, which reproduces the ACKNOWLEDGEMENTS
dark and light characteristics of the corresponding cells, it can be seen
that strong ion bombardment improves not only the short-circuit current We gratefully acknowledge the experimental assistence of
density but also the open-circuit voltage. We attribute this to a lower B. Freienstein and the stimulating discussions with Dr. H. Pfleiderer.
midgap defect density. This work was supported by the Federal Ministry of Research and
Development of Germany.

4. SUMMARY AND OUTLOOK

REFERENCES
/11 T. Shimada, S. Matsubara, H. Itob, S. Muramatsu, Proc. of the 9th
Our results for the deposition of pure a-Ge:H solar cells reveal
E.C. PV Solar Energy Conf.(1989) 81
deposition parameters that completely contrast with those used to
121 F.H. Karg. H. Bohm, K, Pierz, Journ. of Non-Cryst. Solids (1989) 477
prepare high-quality a-Si:H solar cells. A substrate temperature of
13/ W. Kusian, H. Pfleiderer, G. Scheinbacher,W. Kasper, Proc. of the
T s = 150 ·C, strong ion bombardment with energy of more than 200 eV, a
Int'l PVSEC-5 (1990) %5
p-SiClp + -Si front contact and an i-layer thickness of about 0.3 jl.m results
/4/ R.D. Plattner, E. Giinzel, G. Scheinbacher, B. Schroder, Proc. of the
in an efficiency of 2.S % under AM11ight.
Int'l Meeting on Stability of Amorphous Silicon Materials and Solar
Although our a-Ge:H material possesses good properties, i.e. an
Cells (1991) in press
Urbach energy of E o =52mV and a photoluminescence line width of
/51 W. Kusian, E. Giinzel, R.D. Plattner, Proc. of the Int'l PVSEC-S
0.2 e V at 0.79 e V peak energy, the minority carrier (hole) properties are
(1990) 631
still poor. This of course is related to the high defect density (currently
161 R. Saeng-udom, W. Kusian, B. Bullemer, this conI.

219
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON. PORTUGAL

DEUTERON MAGNETIC RESONANCE SPECfRAL COMPONENTES IN a-Si:H.D

Plinio B, Santos-Filho. ITEP, Funda~40 Instituto Tecnol6gico do Estado de

Pernambuco, Av. Prof. Luis Freire nO 700, Recife, PE. 50730, Brazil.

Jeff Bodart and Richard E. Norbe[8 Department of Physics, Washington

University, St. Louis, MO, 63130, USA

Deuteron magnetic resonance (DMR) in a-Si:D,H and a-Si:D films resolves 5 distinct spectral
components. For the bonded deuterium, there are the rigid deuteride Si-D configurations and the
rotating silyl SiDxH3-1 groups. These components appear as a 66 kHz and a 22 kHz quadrupolar
Pake doublet respectively. DMR shows a central spectral component associated with Dz and HD
molecules inside microvoids. Its linewidth is due to the distribution of surface-generated static
electric field gradients (EFG) from microvoids. A 76 kHz quadrupolar broadened Pake doublet is
present at low temperatures and is associated with p-Dz molecules whose rotational transition rates
fill are no more than Cald (nuclear spin-molecular coupling frequency) for molecules in the presence
of large EFG's. Deuteron multiple spin echoes uniquely identify o-Dz molecules whose rotational
ground states are admixed by the EFG. A para to ortho (p-o) conversion experiment of Dz done on a
perdeuterated a-Si sample at 4.2 K provides further resolution of the bonded D and molecular Dz
configurations. For the Dz molecule dynamics we measure a phase transition temperature and p-o
conversion constant that suggest the presence of standard vapor pressure Dz inside the microvoids.

I. INTRODUCTION

In this paper we will present and discuss the are due to spin diffusion to molecular "relaxation
results of the studies done on deuterated amorphous centers" , RC, or paramagnetic impurities (1). In
silicon samples. The various spectral components will perdeuterated samples as XW895 the RC are para-Dz
be treated separately so that a clearer picture can be molecules subjected to large surface related electric
made of the individual physical processes and specific field gradients; in samples containing both D and H as
characteristics of each of the lines. DMR has been XP1, the RC's would be para-Dz and a small number of
performed at 30 MHz in a Chemagnetics CMX 3001200 J-l HD molecules under similar environmental
Dual Solenoid Spectrometer with a soleinodal field of conditions.
4.7 T. Line shapes, echoes, and relaxation times have
been measured between 1.5 K. and room temperature. 2.2 - SHyl Rotors
The DMR samples have been 50 to 100mg films Deuteron magnetic resonance detects rotating
plasma-deposited onto warm Al substrates, SiDxH3-x groups in a-Si:D,H films. The deuterated sHyl
rotors (DSR) give rise to a characteristic 22 kHz
2. RESULTS AND DISCUSSION quadrupolar Pake doublet with no noticeable
asymmetry parameter. Fig. 2 shows a schematic
2.1 - 66 KHz Doublet drawing of a sHyl rotor and three spectra for sample
Deuterons bonded to silicons give rise to two XPI at different temperatures. The silyl rotor signal
configurations : static and rotational. The rigid was present at all temperatures studied; from 300 K
configuration is associated to a 66 KHz quadrupolar down to 1.6 K for XPI. Similar results have been found
DMR doublet. This component arises from the signal for sample XW895.
produced by deuterons that make single covalent This lack of "freezing out" of the sHyl rotors is in
bonds to silicon (Si-D). Fig.l shows the monodeuteride sharp contrast to the results of similar observations of
"Pake doublet" powder pattern characteristic of a rotating methyl groups in a-C:D,H (2). The 38 kHz
single axial electric field gradient (EFG) averaged over inner doublet from methyl rotors freezes out and
all angles (e,.). The splitting between the peaks vanishes as the temperatUre is reduced to 77 K. The
corresponds to a coupling constant for this type of methyl rotor groups lie within the amorphous carbon
deuteron of eZqQ/h - 4/3Vq - 88 ± 1.3 kHz where Vq is bonded network, and their rotations are hindered. In
the doublet splitting of 66 ± I kHz. This coupling the amorphous silicon films, the fact that the 22 kHz
constant was shown to agree very well with the doublet signal persists down to 1.6 K shows that their
infrared stretching mode at 1460 cm- I (I). rotational motions are not as hindered as the methyl
The monodeuteride line presents very long TI's rotor. The sHyl rotors must extend out of bonded Si
(spin lattice relaxation times) and no apparent TI network, either into microvoids, or from surfaces of
minimum as a function of temperature. This supports microcrystalline inclusions.
the molecular relantion model proposed by Fedders.
The long relantion times for the 66 kHz component

220
2.3 - The Central Component the signal, are shown on the right. By pulsing ever
This component is associated with the existence of faster, the spectral components saturate at different
molecules of Dz and HD inside microvoids. These times, according to their individual Tl relaxation
molecules relax to RC that are very possibly surface processes and this can be followed by looking at Fig. 4
molecules inside the same void space. Fig. I shows the from top to bottom. The first component to saturate
central component of the a-Si:D sample XW895. The due to the repetition rate of the quadrupolar-echo
line broadening which occurs between from 77 K to 4.2 pulse sequence in the 66 kHz doublet. The next one to
K is clearly seen. Molecular para-Dz and ortho-Dz are saturate is the central component and the last is the
responsible for the signal of this component. rectangular shaped line associated with para-Dz
The width of the central line is due to the molecules. The molecules in the rectangular line shape
distribution of surface-generated EFG·s in voids of r
have had their rotational transition rates m slowed
different morphologies. It is important to realiZe that if down to the order of Old, the nuclear spin-molecular
a molecule is physisorbed on a void surface, or trapped coupling frequency (5). Due to a distribution of EFG's
inside a void, it will "sense" an appreciable average with different strengths and symmetries a distribution
electric field gradient. If on the other hand, if the of r m·s is also expected. Such a distribution of
molecule is inside a void large in all dimensions and is molecular rotation rates in the SMR can be appreciated
well away from the surface, it will experience only a in Fig. 5. From the temperature of 20 K down to 2.4 K.
small gradient. A molecule near the surface of a void These spectra are not to scale. They were chosen as
would be more likely to see an "axial" EFG. A molecule representatives of the line shape type at their
in the middle of the void would sense a gradient of temperatures and were selected at a pulse repetition
lower symmetry. rate that clearly separated the EFG aligned para-Dz
The central line in this sample has two T1 molecular signals from the other spectral component
components and these clearly correspond to molecular Signals, in this case, the central and the 66 kHz doublet.
populations of 73' and 27, at very low (T < 4 K) and The well defined cusps of the Pike doublet can also be
high temperatures (T > 100 K). The Tz data have a seen at 4.2 K in Fig. 4 at 1 second recovery time.
discontinuity at 18 K. This region corresponds to the With the onset of the Pike doublet at a temperature
motional diffusion region for normal (33') solid para- around 45 [, there starts a division of the central line
Dz during the solid to liquid transition. The curved line into two componenls, one being the 76 kHz related line
drawn over the Tl(27,) population in Fig. 3 is meant to and a broader central line that apparently has a
be a guide to the eye. Now, the straight line that passes constant Tz below the diffusion region for svp-Dz. Fig. 6
through the discontinuity of the Tl(27,) population shows the signal fractions relative to the division of the
line is an extrapolation to the Tz data taken in the central line. It is clearly seen that when the formation
diffusion region (from 9 to 17 K) for para-Dz by of the 76 kHz component starts the central line
Weinhaus et. al. (3). These authors have shown that the decreases. When the plateau is established below 10 [,
rate of molecular jumps varied some 7 orders of the signal fraction ratio corresponds to .. 4.85. Now, the
magnitude in this temperature region. This implies that DMR signal is a function of (21+ l)I(1+ 1) and for 1-2
in this sample (XWS95) (samples with large voids), and 1=1 the ratio is 30:6 - 5. It seems that the central
there is normal Dz (n-Dz) with SVP characteristics I This component at low temperatures (6), where the 76 kHz
is quite contrary to IR deduced results that suggest component eXists, is ortho-Dz with 1-2 and polariZed
that Hz (Dz) at room temperature would be under high J-O.
pressure of the order of .. 2 kbar (4).
2.5 Ortho-Dz Multiple Echoes
2.4 - The 76 kHz Doublet We have discovered Solomon-like deuteron spin
In Dz there usually are two nuclear spin systems echoes from ortho-Dz molecules at liquid He and higher
that can be observed by DMR: the deuterons can be in temperatures (7). Fig. 7 shows multiple echoes from 0-
the nuclear state 1-1 with J-l odd molecular state Dz molecules in sample XWS95. The echoes formed
(para molecules) or 1-2 primarily with J-O at low following a quadrupolar pulse sequence 90:l-1-26y at a
temperatures (ortho molecules). The experimental repetition interval of 30 seconds and temperature of
spectrum of interest in this section is the quadrupolar- 4.2 [. The three echoes are labeled by order of
broadened low asymmetry parameter Pike type appearance simply as El, EZ and E3. Echoes El and E3
are o-Dz molecule specific and echo Ez has signals from
doublet with a splitting of llv = 76 ± I kHz. In our Dz
both para and ortho-Dz molecular components. Also
microvoid systems, Dz is present as normal-Dz, 1/3 present in the figure is the 66 kHz oscillation pattern
para and 213 ortho, and at this concentration of para- that in the Fourier transform domain will give rise to
Dz molecules no long range ordering is possible. Also, the 66 kHz monodeuteride doublet; at faster pulse
the EQQ interaction between some of the Dz molecules repetition intervals (1 sec) one measures the 76 kHz
has been quenched (5) by electric field gradients oscillations characteristic of the p-Dz doublet. Our
(EFG's) arising from the void surface. echoes are at k- 6'1-(1 +3bl2a)/(1-bl2a), 1, and (t-
In Fig. 4 we show the dynamics of the various bl2a)/(l +3bl2a). Substituting the values a-33.6S1
spectral DMR components of sample XWS95 at 4.2 ((. kHz and b-2.719 kHz from molecular beam data (S)
Vertical axis scale factors are presented on the left of yield predicted echoes with k-0.S55, 1, and 1.169.
the spectra and observation times, for the recovery of

221
These values agree with the observations to within dynamics of the hydrogens within the a-Si: H,D
.. 21. The remarkable echoes EI and E3 provide an network. This work has been suported in part by
unambiguous signature of ortho-Dz molecules. The k .. FACEPE and CNPq (Brazilian Researsh Funding
1 echoes in o-Dz cannot occur in the spherically Agencies), the NFS Division of Materials Researsh under
symmetric JaO state, because a and b both would Grant numbers DMR 87-01515 and 88-01260, and
vanish. Since the echoes are observed at low SERI under subcontract No. XB-7-06055-1 of Contract
temperatures (well below the J-O and J-I splitting), DE-AC02-86CHI0093 with the U. S. Department of
the molecules are in their rotational ground states. The Energy.
ground states must have admilture of higher J states.
Thus, we infer that the molecules are located in sites REFERENCES
with substantial EFG in the amorphous silicon. The (1) D. Leopold, Ph. D. Thesis, Washington University in
Fourier transform of the k .. I echoes has a full width St. Louis, University Microfilms in Michigan( 1983).
(2) V. Bork, Ph. D. Thesis, Washington University in St.
at half muimum of 12.5 ± 1 kHz and this value is
Louis, University Microfilms in Michigan, (1986).
related to the width of the central component line at
(3) F. Weinhaus, et. aI., Phys. Rev. B 7, 2948 (1973).
4.2 K: (Fig.l) that was 12 kHz. This is further evidence
(4) Y. Chabal, et. al., Rev. Mod. Phys. 59, 835, (1987).
that the k .. I echoes are from ortho-Dz molecules. (5) P. A. Fedders, Phys. Rev. B 36, 2107 (t 987).
(6) P. Santos-Filho, Ph. D Thesis, Wasington University
2.6 Para to Ortho Conversion of Dz in S1. Louis, University Microfilms, (1989)
We performed a 75 day para to ortho (p-o) (7) M. Vou, P. Santos-Filho et. aI., Phys. Rev. Lett. 63,
conversionelperiment where perdeuterated sample 2582 (t 989).
XW895 sat at the temperature of liquid He (4.2 n The (8) N. Ramsey, Molecular Bellm$; OIford Press (1956).
purpose of the study was to determine the dependence (9) I. Silvera, Rev. Mod. Phys. 52, 393 (1980).
of the various deuteron signal components upon the
conversion of relalltion center (RC) p-Dz molecules and amp c XW8%
Repcln ion 700 sec
also to possibly increase the signal of the ortho-Dz fJ.(C

multiple echo molecular population by p-Dz conversion. Cen,~l

Because of the Pauli principle, the transition from

the triplet rotational state to the singlet rotational state
in Dz also requires a change of the nuclear spin state. A
Single-molecule conversion process is given by J - 1 - 1
-+ J - 1 - O. An isolated molecule is stable and will not
convert. The nuclear spin transition, as represented by ·1.2 K 4.
the eq. above, requires a magnetic field gradient that Cenu.l
must vary over the dimension of the molecule. This
creates a torque on the deuteron magnetic moments
and causes reorientation of the nuclear spins. In order
.~7 ~~.....I:. . . .__._~'
66 kHz doubl«
to convert, para-Dz can interact with the dipolar
magnetic field of another para-Dz, with the 1-2 state of
an ortho-Dz (j-O,I-2) and it can also interact, via its -------------:~.

spin-l deuteron nuclear quadrupole moment, with the Fia. I 250 kill
electric field gradients from the molecular electric
quadrupoles of nearby molecules.
Sample XPI
Dz obeys the differential rate equation dCo/dt--Ceo a·Si:D ,H
that has an elponential solution. C- 0.060 ± 0.003 IIh Centr:ll
is the currently accepted value for the rate constant of ,/
Dz at zero pressure and 4.2 I( (9). Our Cvalue for the
conversion of the para-Dz doublet molecules is
1(·0.06l±0.005 I/h. The para-Dz doublet component,
again, gives signs of belonging to normal Dz and near
svp. Also, dangling bonds must not elist inside the
microvoids because otherwise they would accelerate
the conversion process; the surfaces of the microviods,
we thus assume, must be fully decorated with Si-D
bonds. The complet behavior of the other spectral
components in the para to ortho conversion
elperiment, will be presented elsewhere.

3. CONCLUSIONS
DMR signals from bounded and deuteron molecular
species provide a useful characterization of deuterated
amorphous silicon samples. In the samples studied we
came upon five distinct spectral components that give
.
250 kHz
unique information on morphology and on the Pig. 2

222
 Cenlral
Component
10 It Sample XW895

• 4.2 K '\
66 kHz doublel

10 3

x 1000 •
10 2
2x 100 •

U
CII
III
101 8x s•
,...
t-
I s
10 0

10- 1
/
27% Central
128x 0.03 s
10 - 2
_1_28_x_ _~_ _ _ _ _...-......~_ _ _ _ _ _
0._00_5_s_
1 10 100 1000
Temperature (K)
250 kHz
Pia. 3 Pia· S
..
Sample XW895 1.0 o
o
00 o central
0.8
c
~
u
l! 0.6
u.
<II
:;, 0.4
VI
p-D2
0.2
•• • • ••
0.0 L - - ' - - ' - - ' - - ' - - ' - '.......
•
..I----'--"'--"'--''-'-"'""'":l~O 0
1 10
Temperature (K)

Pit. 6

fiB . 7
amp le XW895
4.2 K 1800".ec
~ ~
30 ,ec rep. rale

EI . E3 : Orlho-Dz echoe,
E2: pat. and orlho-D2 component'

250 kHz
66 kHz
Pia. of
~ / d~~blet

$I I. , , L ..... f d .........if
~
2p 0 .8'19"t "t 1.193"t

223
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DETERMINATION OF THE DENSITY OF STATES IN a-Si:H OVER A BROAD ENERGY RANGE BY

THE THERMALLY STIMULATED CONDUCTIVITY TECHNIQUE

M.B. von der Linden, G. Landweer I, R.E.I. Schropp,

J. Bezemer and W.F. van der Weg
Utrecht University, Debye Institute, Department of Atomic and Interface Physics,
P.O.Box 80.000, NL-3508 TA Utrecht, The Netherlands

ABSTRACT. The density of gap states (DOS) distribution is an important pa-

rameter in the performance of amorphous silicon solar cells. With the thermally
stimulated conductivity (TSC) technique the DOS in a part of the band gap can
be determined. We have refined the TSC technique in order to allow the determi-
nation of the DOS over a wider energy range. Results on the influence of different
deposition temperatures and hydrogen dilutions in the gas phase are presented.
The lowest value found for the DOS was 7x1015 cm- 3 ey- l . Furthermore we will
show that with TSC the onset of the conduction band tail can be determined. A
slope of the conduction band tail of 43 meY is found.

1 INTRODUCTION 2 EXPERIMENTAL

The thermally stimulated conductivity (TSC) technique 2.1 SET-UP

has been succesfully used for the determination of the DOS The TSC experiments are performed in a high vacuum
in a-Si:H [1..3). TSC is performed by cooling an undoped chamber (10- 6 mbar). For the measurements an electric
amorphous silicon layer after which electrons are excited by field of 2000 Ycm- l is used. The current between the two
means of illumination. This results in filling of gap states. coplanar electrodes is measured with a Keithley 617 elec-
Because of the low temperature the electrons do not have trometer. The temperature of the sample is directly mea-
enough thermal energy to escape from these states. By sured with a thermocouple. After cooling with liquid ni-
heating the amorphous silicon layer at a constant rate these trogen the sample temperature reaches 95 Kelvin. Before
'frozen in' electrons are thermally emitted. The increase of the TSC experiment the sample is illuminated for 5 seconds
the electrical conductivity at increasing temperatures is now with a light flux of 1.2 mWcm- 3 • After a waiting time of
a U.easure for the number of emitted electrons and thus for 30 minutes the experiment is started using a heating rate of
the density of gap states g(E) at the corresponding energy 3.5 or 5 °Cmin-l.
level separated by kT from the conduction band edge. This
energy level is gradually moving down towards the Fermi 1 Present address: Delft University of Technology, Electrical Mate-

level with increasing temperature. rials Laboratory, Mekelweg 4, 2628 CD Delft, The Netherlands

ISamplenr. A B C D E F z
Tdep (OC) 230 300 350
Gasflow (% SiH 4 ) 30 50 70 30 40 60 50
Power (W) 20 20 7.5
Dep. rate (nm.min- l ) 8.9 11.3 16.0 lOA 11.1 11.7 4.3
H (at-%) 16.2 18.2 25.5 14.3 14.8 15.0 11.5
Bandgap (eY, Tauc) 1.83 1.83 1.83 1.78 1.80 1.82 1.79

Table 1: Deposition conditions and properties of the samples.

224
 as a function of the light intensity and temperature [1,2].
The determination of I'T(T) in this work is achieved in a
much faster way by measuring the currents generated in
excess of the TSC due to short monochromatic (~ =637

-
nm) light pulses applied at specific temperature intervals
[4]. With a Si photo diode the number of photons reaching
the sample during these pulses is determinated. A typi-
I
>
ff)~
cal value for the number of photons of the light pulse is
1.4xIQll cm- 2seC 1 . From the integrated areas under the
I superimposed current transients we obtain the excess charge
8(,) 1017 generated due to these light pulses. Using the appropriate

"" absorption coefficient at 637 nm and assuming a quantum

efficiency 1/ = 1, this yields the values for the I'T-product.
The TSC curve itself is not affected by these low intensity
light pulses.

3.2 RESULTS
We have compared the results obtained for the device
quality sample (Z) from the light pulse method and those
from the steady state photoconductivity method, see fig-
-0.8 -0.6 -0.4 -0.2 o ure 1. In the latter case the photoconductivity as function
E - Ec (eV) of the light intensity at different temperatures is measured.
These experiments are time consuming but give also infor-
Figure 1: DOS from the • light pulse method
and * photoconductivity experiments of sam- mation about the nonlinearity ratio i between the photo-
ple Z. conductivity and the light intensity. If i differs significantly
from unity, the DOS from the photoconductivity method
can be different as compared to the DOS from the light
2.2 SAMPLE PROPERTIES pulse method. However, if very low light intensity pulses
We have investigated the DOS of different undoped a- are used, the error induced by not taking into account a i
Si:H layers, which were deposited on Corning 7059 glass < 1 can be neglected.
by PECVD in our UHV multichambe~ reactor, ASTER [5]. For both methods an increase of the DOS towards the con-
In this series the deposition temperature Tdep (substrate duction band is shown (figure 1). The slope of the con-
temperature) and the hydrogen to silane ratio in the gas duction band tail from the photoconductivity method is 43
phase were varied. A rf power of 20 Watt is used for most meV.
depositions. The deposition conditions are given in table 1.
One sample prepared under the conditions normally used
4 RESULTS
for the i-layer in solar cells, is also included in this study.
The effective attempt-to-escape frequency Veil' is extracted 4.1 H2 DILUTION
from the data at the temperature at which the TSC and In figure 2 the results for the I'T-product and DOS of the
dark conductivity become equal and varies from 6.6 x 1011 series with Tdep = 230 ·C are shown using the light pulse
to 1.3x1013 sec-I, which indicates that retrapping can be method. For the DOS and I'T-product a clear dependence
neglected. The thickness of the samples was in the range on the SiH4 /H2 ratio in the gas phase is found. A higher H2
from 0.349 to 0.570 I'm. The H contents in the layers were dilution gives higher values for the I'T-product and a lower
measured by infrared absorption. DOS.
Taking into consideration the data in table 1. it is shown
that a higher hydrogen dilution results in a lower deposition
3 LIGHT PULSE METHOD
rate and a decreasing H incorporation into the sample. We
3.1 DESCRIPTION suggest that H-enhanced etching prevents the incorporation
To deduce the DOS from the TSC curve the mobility- of undesirable radicals (SiH 2, SiH3 , •• ). This mechanism is
lifetime product, I'T(T), has to be known. Usually I'T(T) thought to reduce the density of dangling bonds associated
is obtained by separately measuring the photoconductivity with the poly-hydrides.

225
468 - 0.8 - 0.6 - 0.4 - 0.2 0
1000/T (K- 1) E - Eo (eV)
Figure 2: The a) flT-product and b) DOS for the series with Tdep = 230 ·C
using the light pulse method. The samples shown are • A, • B and • c.

10 19 1:'T"T"T""T"T~"""TIr-r-T""T""T""T"""""~

-I
>II)
(II)
I

- 810 17
()

4 6 8 -0.8 -0.6 -0.4 -0.2 0

1000/T (K- )
1 E - Eo (eV)
Figure 3: The a) jlT-product and b) DOS for the series with Tdep = 300 ·C
lIsing the light pulse method. The samples shown are 0 D, t:. E and 0 F .

226
 For the DOS distribution of the series deposited at 300 5 CONCLUSIONS
°c this gas phase dependence is not found, see figure 3. A
possible explanation is that for the higher deposition tem- As we have shown, the TSC technique is a powerful
method for the determination of the DOS and IlT-product
perature the radicals have enough thermal energy to move
along the surface and to incorporate at stable positions with in a-Si:H layers. The light pulse method has the advantage
a minimum of dagling bonds. In this case the hydrogen ratio that the determination of the IlT-product is achieved during
the measurements of the thermally stimulated conductivity.
in the gas phase will not significantly affect the DOS. How-
However, the nonlinearity of the photoconductivity versus
ever, the IlT-product does increase with increasing hydrogen
dilution and is higher compared to the samples deposited at the light intensity is ignored, which may cause an overesti-
mation of the DOS value of a factor two at the most. With
230°C. Usually it is found that a higher DOS gives a lower
TSC the onset of the conduction band tail can be measured.
IlT-product, see also figure 2. The results of the Tdep = 300
We found a value of 43 meV using the photoconductivity
°c series demonstrate that the IlT-product is not just a func-
method.
tion of the DOS in the middle of the band gap.
The H2 dilution in the gas phase does affect the DOS
and the IlT-product. At a deposition temperature of Tdep
= 230 ·C the DOS can be decreased by increasing the H2
dilution. This also results in a lower H incorporation and
lower deposition rate. The IlT-product is increased with
a higher H2 dilution. These results can be explained by
..........
>-Q) etching of undesirable incorporated radicals on the surface
by the hydrogen. At higher deposition temperatures the
~ 10 16 thermal energy of the radicals is large enough to incorporate
o
I at more stable positions so the hydrogen ratio is not affecting
~ 5 10 15 to the DOS anymore.
~
Deposition rate (nm.sec- ) The electronic quality of a-Si:H layers appears to deteri-
orate at increasing deposition rate.
Figure 4: The DOS at E = -0.6 eV versus
the deposition rate for the samples A to F.
ACKNOWLEDGEMENT

4.2 DEPOSITION RATE This work was financially supported by the Netherlands
The above results concerning the DOS can be summa- Agency for Energy and Environment (NOVEM).
rized by plotting the DOS at E = -0.6 eV versus the depo-
sition rate, see figure 4.
A lower deposition rate generally results in a decreasing REFERENCES
DOS for the range of deposition parameters used here. The
[1] M. Zhu and H. Fritzsche, Density of states and mobilily-
IlT-product at 200 K is plotted as function of the hydrogen
lifetime product in hydrogenated amorphous silicon,
content in the samples, see figure 5. We interpret this as
from thermostimulated conductivity and photoconductiv-
follows. The H incorporation influences the disorder in the
ity measurements, Phil. Mag. B 53 (1986), 4l.
material, which in its turn affects the carrier transport.
.-., [2] S. Dzdemir and D. Dktii, Light induced changes in the
I
density of states of a-Si:H, J. Non-Cryst. Solids 107
.J> 10-4 (1989), 289
E
U
'-' [3] C. Longeaud, J.P. Kleider, D. Mencaraglia, T.
g Mohammad-Brahim and A. Rahal, Study of band tails
010- 6 in a-SiGe:H alloys from the time of flight and thermally
o
N stimulated current experiments, Mat. Res. Soc. Symp.
II Proc. Vol. 192, 1990.
!:10- e ~~~~~~~~~~~ [4] G. Landweer and J. Bezemer, Advances in disorded
I-
~ 10 15 20 25 30 semiconductors 1, "Amorhous silicon and related ma-
Hydrogen content (at-%) terials", Ed. H. Fritzsche, World Scientific, Singapore
Figure 5: The IlT-product at 200 K as func-
1989.
tion of the hydrogen contents for the sam-
ples A to F. [5] H. Meiling, .. this conference.

227
HYfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DIODE RECOVERY MEASUREMENT OF a-Si:C:H DIODE USING

DIFFERENT SILANE-PROPANE RATIO FOR TANDEM SOLAR CELLS

K. Ibrahim
School of Physics
University of Science of Malaysia
11800 Penang, Malaysia

ABSTRACT. Diode recovery measurement was made on hydrogenated amorphous

silicon carbide (a-Si:C:H) diode, with silane to propane gas ratio of 1
to 10. An increase in propane gas ratio results in an increase in
lifetime and bandtails. Based on just the lifetime results, a-Si,:C:H
prepared from silane-propane gas mixture is a promiSing material for
the window layer of tandem solar cells but with the bandtails results
which attributed to an incirease in the disorde.r in a-Si:C:H with
an increase in propane ratio, an optimum ratio was found at silane
to propane of 20:4.

INTRODUCTION The rf power was switched off every time the shutter
moved. Glass coated with tin oxide (with traces of
~ wide bandgap hydrogenated amorphous silicon fluorine) was used as substrate. Aluminium dots
carbide alloys (A-Si:C:H) are suitable materials were vacuum evaporated for the opposing contacts.
for the window layer of a 'tandem solar cell.
A-Si:C:H can be used as a single p-layer or as a Diode Recovery Measured
whole pin diode. Much studies have been done on The lifetime is determined from the charge
the single p-layer but not much on the pin diode. stored, Qs' in the i-region due to a known forward
This paper will concentrate on the measurement current (2). If the switching event and the
of charge carrier lifetime and bandtail of a-Si:C:H
(1)
pin diode with variation in the silane-propane
ratio. Charge carrier lifetime has been shown to extraction of the carriers occur at a time fast
be important in characterizing the properties of enough that carrier losses are avoided, Qs can be
crystalline solar cells. In amorphous materials, determined from the time integral of the reverse
the presence of density of localised s·tates in the current. But due to the sample capacitance, the
bandgap should further enhance the importance of switching of the electric field is accompanied by
the carrier lifetime. a large displacement current (3). To extract
The technique used to determine the carrier the recovered charge, the reverse charging current
lifetime was the diode recovery method. In this of the device without a previous forward bias is
technique lifetime is determined from the recovery substracted from the reverse voltage after forward
phase after biassing the diode in the forward bias,
direction and suddenly reversing the current. The switching process was done ,by using a
CMOS diode which has a switching response of less
sample Preparation than 50 nanoseconds.
Samples were prepared using a glow' discharge The tail states was determined by measuring
system, using rf power at 13.56 MHz of 10 Wand the the recovered charge, Qs' as a function of forward
deposition chamber pressure was 0.1 torr. The current at a high forward bias (4). This is made
subs tate temperature was maintained around 216 0 C. possible because forward current in the diode
Table I shows the deposition conditions for samples. recovery method caus'es similar behaviour to the
I-layer charge carriers as illumination does in photocon-
Gas Mix Dopant Deposition ductivity. Thus the expression of Hack et al (5),
Sample SiH.:CgHa Br2H PHg Time Temp.
(mins) (oC) BT/TZ
(scem: sccm) (sccm) (sccm) ]It ;; const If .... un
8461 20:2 20 10 5 219
can be written in the form of (4),
8462 20:4 20 10 5 218
(1-BT/T2) T/T2
8463 20:5 20 10 5 216 Qs = const If •••• (3)
8464 15:5 15 10 7.5 215 where B is given by T/(Tl+T2), Tl is the charac-
8465 10:5 10 10 10 215 teristic width of the conduction bandtail, and T2
is the characteristic width of the valence band-
8466 5:5 5 10 20 218 tail.
Table I: Deposition conditions for samples used
in minority carrier lifetime measurements. Results and Discussions
Figure 1 shows a typical lifetime dependence
The samples used for the lifetime measurements were on the injection level. The lifetime decreased
prepared using the shutter system that has been with an increased in injection level. A similar
described elsewhere (1). The a-Si:C,H alloys were result has been observed by other researchers for
deposited in a single run so that the deposition a-Si:H (6).
chamber history did not influence the outcome of Figure 2 shows a plot of lifetime dependence
the comparative study. Low carbon concentration on the silane-propane ratio for forward current
samples were first deposited followed by an increa- density of 3.33 mA/cm 2 . The lifetime increases
sing amount of carbon proportion. Thus the first with carbon incorporation. A higher lifetime will
deposition will not affect the next deposition. increase the probability of the excess minority

228
carriers to be "swept" across the junction, giving regimes. Evaluating the slopes (51 and 52) and
rise to a photocurrent. using equation 3 gives the characteristic widths of
the bandtails and this is shown in Table II. From
the table we can see that for a-Si:C:H samples,

I~~--------------------------------------------~ Sample 51 52 Tl(K) T2(K) Tl (meV) T2(meV)

-
8461 0.925 0.226 677.3 1227.9 58 106
8462 0.852 0.183 135.4 1369.7 11.7 120
8463 0.515 0.158 -110.2 977.8 -9.5 84
o ••
8464 0.402 0.131 -136.5 920.6 -11.8 79

..;;:
... ,0 8465 0.189 0.089 -125.8 637.1 -10.8 55
Table II: Characteristics widths of the bandtail.
...'":;
equation 3 is valid only for propane-silane gas
ratio of greater than 4 that is the composition
range for the bandtail can be described by an
exponential function. Studies on Urbach edge for
high carbon concentration have also been shown to
"
FORWARD CURRENT DtNSIT.,. CmA/etrt"'ZI
100 000
indicate non exponential behaviour (7).

Conclusion
Figure 1: Lifetime dependence of the forward Diode recovery measurements on a-Si:C:H showed
current density. that an increase in propane gas ratio resulted in
an increase in lifetime and bandtails. A high
lifetime will increase the chances of creating a
10 photocurrent in the tandem solar cells but an

.
increase in the bandtail means an increase in
disorder so to obtain a high lifetime and a rea-
sonable bandtail width the silane to propane ratio
should be 20:4.

~ 10 Acknowledgement
The author will like to thank Heriot Watt
!I;;:
..
University for using the facilities during his
...'"
:l
stay there .

REFERENCES
10
(1) K. Ibrahim, U.I. Eicker, A. Qayyum, J.I.B.
Wilson, Proc. 6th Int. Solar Forum, Berlin,
OGS-Sonnenenergie Verlags GmbH Munich, Vol.l
• 0 10 II
(1988) 481.

SJH4/C3H8 (2) A. Hoffmann, K. Schuster, Solid State Elect.,

7 (1964) 717.
(3) K. Venkateswaran, D.J. Roulston, Electronics.
Figure 2: Lifetime of a-Si:C:H diode for different Lett.,6 (1970) 681.
silane-propane ratio at current density
3.33 rnA cm. (4) R. Konenkamp, J. Non-Cryst. Solids, 77 & 78
(1985) 643.
A typic~l plot of the variation of storage charge (5) M. Hack, M. Shur, J. Appl. Phys. 55 (1984)
with forward current, figure 3, shows two distinct 2967.
(6) A.J. Snell, W.E. Spear, P.G. LeComber,
phil. Mag. B, 43 (1981).
10~'r----------------------------------------------, (7) A. Skumanich, A. Frova, N.M. Amer., Solid
State Corom., 54 (1985) 597.

. - -.. - _.. . .
100

" ... ~

..... ~.a ;.. 4 ~-.· ..

.
.,It ..

,. '
10

Figure 3: variation of storage charges with

respect to forward current.

229
 SESSION 02

High Efficiency Crystalline Silicon Solar Cells

 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE
S-12 APRll. 1991 LISBON, PORTUGAL

REVIEW ON HIGH EFFICCIENCY SILICON SOLAR CELLS

Antonio Luque
Instituto de Energia Solar
ETSI Telecomunicaci6n
Universidad Polit~cnica de Madrid
E-28040 Madrid - Spain

ABSTRACT . The highest efficiencies so far achieved in one sun and

concentrating cells are reported and the causes of their efficiency are
discussed. In particular we discuss the reduction of recombination, the
enhancement of the light confinement, the high injection operation the
reduction of series resistance and the reduction of shadowing losses. It is
stressed the strong link between modeling and experimental progress. We make
some final comments on prospective achievements and economic aspects.

1. INTRODUCTION 2. THE FLAT MODULE CELL

Silicon high efficiency solar cells have A schematic drawing of the PERL cell, that
recently achieved efficiencies that nobody expect~d achieved the 24% eff iciency at one sun [1], is
before: 24% at one sun standard conditi~ns (25 C presented in Fig 1. It will help us to analyze a
spectrum AMl.5G, corrected to 0.1 W/cm, Sandia number of features; mainly the ways and worth of
confirmed) [1], and 28% under concentration (at 100 reducing recombinations and also the important
suns) [2]. These high efficiencies have constituted advantages of light confinement.
an unprecedented achievement and Si is viewed today
[3] as one of the most promising candidates for
flat panel and concentration approaches, and
certainly, as the most solid one.
The role of photovoltaics in remote
applications, sometimes in industrialized
countries, but mostly in the third World, is well
recognized. With these new high efficiencies,
linked to some cost reducing technologies already
in development, photovoltaics can be used for grid
connected houses, at costs not much higher than the
market ones. In concentrators, where climatic
conditions are favorable, it can be used for
electricity production in centrals, also at OBCI< metallaya,
reasonable costs. In all cases the environmental
impact is very low. We have never been so close to
provide a photovoltaic solution to the problem of Fig 1. The PERL cell. The ohmic contacts are made
the electricity production. through small holes located on the top of heavy
If only the unavoidable radiative diffusions, to avoid recombinations. A back mirror
recombination is considered a material with the with a pro-reflecting dielectric layer and an
bandgap of the silicon Cl.l ev) would present the inverted pyramid texture provide light confinement.
highest efficiency. However Si cells are disfavored This texture and an anti-reflecting layer provide
by the ~navoidable presence of the Auger excellent transmission.
recombination, and taking for its coefficient the
value found by Sinton and Swanson [4] for high 2.l Reducing recombinations
injection cells, that under the ideal conditions Recombinations affect to the open circuit
considered here give the top efficiency [5], we voltage and also to the short circuit current. It
find 35.6% (6) the silicon cell efficiency limit may also affect to the curve factor. Before
under AMl.5D spectrum. There is some uncertainty on studying the PERL cell let us consider a
whether the Auger coefficient is smaller for low conventional cell. Such a cell has a phosphorus
injection cells, in which case the top efficiency emi tter fabricated in a single diffusion step. As
might be higher . It has to be noted that , in theory phosphorus diffuses faster at high dopings, it
[5], this efficiency limit is independent from the results that a typical conventional profile has a
concentration level provided that advanced schemes flat region of a thickness of about 0.14 pm with a
of light confinement are used. surface density 4xl 020cm-3, followed of an erfc
This paper is based on two case studies: the profile of surface density lx10 2 0cm- 3 tailored in
PERL cell developed by Green and coworkers at the such a way that the Junction Is located at some 0.5
Uni versity of New South Wales (UNSW) and the PC /lm. The location of recombinations in such a cell
cell by Swanson and coworkers at the Stanford is shown in Fig 2. I t results apparent that the
University. Based on them we shall study the recombination (in a FZ wafer) is mainly caused at
following topics: (a) recombination reduction, (b) the front emitter. This recombination reduces
light confinement, (c) high injection, (d) series strongly the open circuit voltage but also the
resistance and (e) total light capture. Special short circuit current mainly by reduction of the UV
emphasis will be placed on the strong linkage we photo-response, as observed in Fig 3. As matter of
place between the modeling and the technology fact the surface layer acts as a dead layer in
progress that leads the work at our Institute, as which all the electron-hole pairs generated near
well as on some newer ideas in the light-cell the surface recombine there. Note that minor
coupling we are developing. changes in processing, that result in different

233
emitter tailoring lead to difference in t.he cell now some 645 mY, to compare with the 600 mY in the
performance. However efficiency with such cells is preceding case. In addition to that we find in Fig
usually limited to about 17~ .. 3 that the UV response is excellent, so that this
cell fresents a short circuit current around
Generstion-rllcomOinstion (mAlcrn2) 3mA/cm higher than in the preceding case . In Fig 2
48 we can observe that the emitter recombination is
not any more the dominating recombination, but now
40. it is the recombination at the rear metallization
corrmon
32~~------------~;
contact.
DE The next step is now clear; it is one of the
DERL characteristics of the PERL cell. It consists in
DE
drilling holes through which a localized high-low
16 junction is fabricated . At this stage we call our
DERL cells DERL (deep emitter rear locally-diffused)
8 cells.
Already the extended Al high-low junction
0. 0. 50. 100 150. 200 250 300 played an important role in reducing the
DiSISTICII from mil front (microns) recombination at the metal, leading to an effective
surface recombination velocity of 1700 cm/s . The
use of this point dotted contacting reduces it to
Fig 2. Cumulative generation and recombination 100 cm/s or less , but leaves uncovered a great
profile, in open circuit, for different cell extent of silicon-oxide interface . The surface
structures . The step to the left represents the recombination velocity there might be 100 or less,
emitter recombination. To the right we observe the although it is strongly affected by the positive
recombination in the back emitter. The vertical electric charge in the back oxide that can set the
lines at the right , ending in a cross, represent surface in inversion if p-bases are used, leading
the back metal recombination. The difference to strong recombination. If this problem is avoided
between G and R at the right is the front metal and a surface recombination velocity of, 100 cm/s
recombination, that is considered in our model as is achieved we can add it to that point shaped
an external diode. For the same base the common emitter and model the back surface with 200 cm/s so
cell has an open circuit voltage of 60.0. mV, the DE that we calculate an open circuit voltage of 666
cell of 645 mV and the DERL cell of 666 mV . mY . The location of the recombinations. also given
in Fig 2 , shows now that recombinations of the same
A first action is to tailor the emitter order are produced at the front emitter, at the
profile in a more adequate way. This has been done base and at the metal-silicon interface beneath the
by A Cuevas and collaborators at our Institute (7), grid. With this cell we expect to achieve
to produce the so called deep emitter cell, that efficiencies around 22~.
gave over 19~. The following two refinements, that are found
in the PERL cell, are intended to reduce the
remaining sources of recombination.
Y...:{.,;:l1i::.)_ _ _ _ _ _ _- ,
120 ;:.'n::;llIf:::.,::n:::81:.,.QU::!:::8::;n::;I<.m=..:II::;II::;IC::;18fIC=:c

5O~CU~r'~e~n~I~~~~~II~Y~(~m~A~)_ _ _ _ _ _ _ _ _ _--,
100

30

20.

10. common DE

DL-~--~~--~~~~--~~~~
-<>- DERL PERL
0.
0.3 0. 4 0.5 0.6 0..1 0.8 o.g I. I 12
Wavelengln (microns)
-10.
-0.2 0. 0.2 0.4 06 08
voll8ge (VOIIS)

Fig 3 . Internal quantum efficiency for several cell

structures .
Fig 4. JV curves for several cell structures. A
As a general trend, what it must be achieved common cell with an efficiency of 16.6%, a DE cell
is a reduction of the surface doping density, with 19 . 8'7., (20 . 1% achieved experimentally at our
because the Auger recombination associated to it is Institute (8), and and a DERL with 22 . 1'7., now
too spoiling. Witr~u~3the bandgap shrinkage effects under development (not yet achieved).
-absent below 10 cm - the reduction of majority
carrier density leads to an increase of the The metal-contact recombination is reduced by
minority carriers that balances out any increase of contacting through vias under the grid. Thus the
the Auger lifetime, this is why no particular contacting area is different, and smaller, from the
advantage is to be found in reducing the surface shaded area. in addition deep diffusions are made
doping below this value (and we can find problems under the metal-silicon interface to isolate this
of contacting). In presence of bandgap shrinkage highly recombinant region from the quasi-Fermi
effects, above 10 1B cm- 3 , no noticeable reduction of levels splitting. Underneath the uncovered surface
the minori ty carrier densi ty is achieved by area, which is passivated, a shallower Junction
increasing the majority carrier density, so that must be formed, whose depth must be selected so as
only a reduction of the lifetime is found resulting to obtain an adequate sheet resistance. As already
in an increased recombination for a given mentioned, efficiencies of 24~ have been achieved
quasi-Fermi level splitting. Our deep emitter cell so far with the PERL cell with an open circuit
(DEJ has the emi t ter doping surface densi ty around voltage of 696 mY.
10 1 cm- 3 ,so that to keep a reasonable emitter sheet The reduction of the base recombination, when
resistance (around 90 0), the emitter must go it is already limited by unavoidable Auger
rather deep, to 1.3 ~m. The open circuit voltage is mechanism, is achieved by reducing the cell

234
thickness. The confinement of the light, one of the Other surface treatments can also give good
techniques we are going to describe later, allows results in reducing reflection. We are using a
for substantial reductions of the thickness with classical random texture, with a AR Ta20s layer,
small reduction of the photo-generated current. giving reflectivities of 1%, at 0.8 pm, in the grid
uncovered area. As for the light confinement of
2.2 The light confinement this structure we think it is also very good.
Another interesting feature in the PERL cell Regularly spaced pyramids have been used in the
is the complex surface treatment with inverted past by the group of Green, and are being used by
pyramids engraved with an anisotropic etch. This the Swanson's one, also with good results. As the
structure has a double purpose. One is that the overall photo-response of the PERL cell is better
rays incident on a pyramid wall, reflected than that of any other cell (42.8 mAlcm2 ) we might
partially, are cast again on another pyramid wall. presume that its scheme of texturing is better, but
If an AR coating is put on the silicon, so that the other factors certainly influence this good result.
reflectance is of the order of, say, 10%, the Lacking of a good model to assess the performance
double reflection is about O.lxO.l=O.Ol; as it is of a texturing model we remain inconclusive with
approximately found experimentally for some regard to this point.
wavelegths. We present in Fig 5 the absorption of light
We want to note here that while for normal (in terms of generation current) vs the cell
light incidence this structure gives two incidences thickness with and without textured surfaces, and
per ray, it is not so when the light comes from wi th a conf inement achieved wi th a perfect back
other directions. A theoretical upper bound, mirror. In all cases the transmission of the light
possibly not reached, for isotropic illumination into the cell has been considered perfect.
has been found [9] to be v3= 1.73. The light confinement can be increased if the
In addition, this pyramid structure, Jointly etendue of the escaping (and incoming) bundle of
with the cell back mirror, is intended to produce rays is reduced.
light confinement. The weakly absorbed photons A ray is defined by the coordinates and the
having entered the cell go to the rear face; there optical direction cosines (x,y,z,p,q,r) of one
they are reflected and go back to the front face point of its trajectory; note that there is a link
where they have good chances to be reflected by p2+q2+r2=n2 . The etendue is defined in its
total internal reflection (TIR). Under some ideal differential form by d3f = dxdydpdq + dydzdqdr +
conditions the 2average path of the rays in the Si dzdxdrdp, and is an invariant in an optical system.
cell is I = 4n W, that in the Si it is about SOW. On the xy plane the etendue takes the simpler
This give a good chance to the long wavelength expression d3f = dxdydpdq. The power carried an
photons to be absorbed. However i t must be noted elementary bundle of rays of radiance ~ and etendue
than the reflectivity of the back mirror is a key d3f is dW = ~d3f. The etendue of an bundle of rays
factor for the good operation of the light falling on an area 'A within a cone of semiaxis r is
confining structure. Typical experimental value of 3f = ItAsin2r.
the reflectivity of an Al layer deposited on the
back cell face 2is around 72% (assuming ideal
reflectance I-1/n at the front face) and this is lignt
not satisfactory. A pro-reflecting (PR) layer of
low index of refraction, with a thickness of about
1/4 of wavelength must be located between the
silicon and the metal. This is done in the PERL
cell, and also in the cells under development at
our Institute, to improve this reflectivity. This
reflectivity is about 98% in the PERL cell. This
layer also plays a role in reducing recombination,
as discussed before.
We can see, in Fig 3, how the quantum
efficiency in the infrared is enhanced, in the DERL
cell, by the light confinement with an improved
mirror.

GonOllJltHi cuoenl dans (mAIcm21

Fig 6 Light confining integrating sphere. The light

average path length in a cell 100 tJ.I1l thick with
Ac/A. = 80 is, ideally, 40 cm.
- F
The maximum average path length of the rays in
-- T an structure of light absorbing volume V, in which
T!M
~
the rays can escape through a bundle of etendue 3f
TAile is given by [10]

100 200 300 400 500 (1)

C811 ItuC}(nflSS {mIcrons}

that for the case of isotropic escaping and volume

AW is the well known expression 4n~. There are
many ways to reduce the etendue of the escaping
Fig 5. Photo-generated current (no recombination, bundle [10]. We shall mention here only one that is
ideal AR surface) vs. the cell thickness for easy to implement. It consists in putting the cells
several light confinement schemes. F means flat inside a cavity like the one represented in Fig 6.
surface, T textured surface, in both cases with no In this case the etendue to use is the one escaping
reflection from the rear face. TSl1 means that a from the cavity, itA., (A. is the cavity entry
perfect mirror has been put at the rear face. Tl1SC aperture). The volume of the absorbing material is
means that the TSl1 cell is located in a perfect AeW (Ae refers to the area of all the cells inside
cavity of entry aperture 80 times smaller than the the cavity). The maximum average path length is in
cell area. this case

235
 in general, and of the PC cell in particular.
(2) Cells of conventional structure, with base
resistivities of 10 Clcm or more operate in high
We present also in Fig 5 the photo-generaked injection, at open circuit conditions, already a
current for the case TM&C, with (Ac/Ae) = 80 (1 ~ one sun . When the irradiance increases this
4000 W) . Note that a current improvement of 3 behavior is extended to MPP conditions. In short
mAlcm2 can be achieved . circuit the low voltage imposed externally makes it
difficult in most cases to continue in high
3 THE CONCENTRATION CELL injection . This point will be analyzed latter.
The most important feature of the high
Cells contacted bifacially, that is of injection operation is the appearance of a voltage
conventional structure, adapted for concentration in the high-low junction, only present in case of
operation concur with a revolut i onary structure, high injection there. This forces to make this
contacted only on the rear side -the Point Contact junction carefully to prevent its short circuit and
(PC) cell- and developed by Swanson and coworkers the loss of this voltage. The voltages present in
at Stanford University. A schematic drawing of this the pn and high low Junctions for a n-type base are
cell is presented in Fig 7 . It has achieved the
highest efficiency obtained so far with silicon, Vnp+= (kT/ql1n(pNB/nl 2 ) (3a)
28Y. at a concentration of some 100-150 suns [2] . Vnn+= (kT/ql1n [ (p+NB)/NB] (3b)
The analysis of this and other cells , and our
modeling work, will help us to analyze a number of where NB is the base doping , nI is the intrinsic
features ; mainly the high injection behavior and carrier density, and (kT/q) is the thermodynamic
the effect of the series resistance and how to voltage . The carrier density p refers in both
reduce it . equations to its value at the base side of the'np+
and nn+ junctions.
flBg8tlVfJ grid A Dember vol tage also appears when there are
gradients in the profiles . It is not as important
as the voltages at the np+ and nn+ junctions but it
must be taken into account as it adds up to the
ce~l voltage when t~e p+ face is illuminated and
thlS help to make p nn cell more efficient than
n+pp+ cells .
Ohmic drops, also take place across the.base.
An assessment on whether high injection or low
injection can give higher efficiency cannot be done
today with certitude . We know many of the
conditions of what we need to achieve the best Si
cell, but there are still some unknowns, in
particular related to the understanding of the
Auger recombination .
InCirJent I/gnt
Furthermore, for practical cells of a given
technology we cannot give general rules on the
Fig 7 . Schematic of the PC cell . p+ and n+ emitters optimal base resistively ; in fact every case must
are diffused through dot-shaped windows and located be investigated carefully. For instance with our DE
at the cell rear face where they are contacted. No cells we have found almost the same efficiency for
shadow is produced by front face contacting. high and low injection operation at one sun,
al though achieved wi th higher currents and lower
The most original feature of the PC cell is voltages in the high injection case.
the fact that it has all the contacts taken on the We are going to analyze, through calculations,
rear side. To our knowledge it was Schwartz and some of the features of our DE cells under
coworkers [11], in 1977, at Purdue University, who irradiance, but we must say that we do not have, as
first proposed a structure of this characteristics, yet, real practical experience with them in
but highest efficiency cells were obtained only in concentration.
1986 by the Swanson's group at Stanford .
Additiqnal features of this cell are the point
shaped emitters to reduce the recombination in the 8 Ox'8nt aens,ty (AI
heavily doped regions. To our knowledge the
Stanford group is the first one to have proposed
such point shaped emitters .
Other factor that contributes to the high
4 n'pp' 02 Olvn~rcm
efficiency of the PC cells is the use of highly
resistive bases, operating in high injection, that n'()p' 5 Ol'nlS'cm
2 ____
allow for base lifetimes in the range of n·np· 100 Ohms'cm
milliseconds . When a highly resistive base ends in
a surface, and there are excess carriers in its o~--------------------------~~
vicini ty, the surface recombination is strong due
to the abundant minority carriers ; in consequence -2L-----~--~~--~----~----~
this surface must be very effectively passivated . -02 o 02 04 06 08
VoltB~ (VOlts)
They have done it with a dry oxide, achieving
values below 10 cm/s [12] (2-3 cmls in some cases).
If such low surface recombination velocities are
not achieved the point shaped emitters are not Fig 8, Several DE cells at 20 Wlcm2 (AnI,5G, 25·C),
worthwhile.
For all the cells in our analysis the
3.1 High injection operation ; comparison with low structure is similar: n+xp+; x can be p or n .
injection operation Illumination is only possible on the nT-face (and
A complete study of the high injection this is not the best) because of the difficulty of
operation of the solar cell can be found in the fully removing the alloyed aluminum from the p+
literature [13]. Here we . are only going to give face . The JV curves are presented in Fig 8 under
some characteristics of the high injection cells, illumination of 20 W/cm 2 (AMl.5G, 25·C) . Note that

236
for both, the n·pp· and the n+pp+ high injection carrier densities outside them are increased and so
structures, the efficiency and open circuit happen to the density profile in the
voltages are very similar (the base behaves as one-dimensional region of the cell . Therefore the
intrinsic) . However near short circuit the behavior recombination increases decreasing the extracted
is rather different. When short circuit conditions current .
are imposed to the n+pp+cell the back junction Also the sloppy profile in the one-dimensional
comes to low injection, and a strong ohmic drop is region of the cell causes strong Auger
produced, so that the front cell region is forward recombination as compared to the voltage extracted.
biased and may still be in high injection. The fact this effect being worse than the profile increases
that a sUbstantial recombination can remain in this due to the valleys at the emitters vicinity if
conditions gives to the JV curve the appearance of these are small and dense. This makes this cell
having a shunt resistance that in fact is rather very sensitive to the cell thickness, due to the
caused by a variable series resistance. This strong nonlinear character of the Auger
behavior was experimentally observed and recombination. For a cell in which all the
interpreted first, as far as we know, by JM Huiz, recombinations except the Auger one has been
of our Institute [13). suppressed the JV equation (near the MPP) is
We also present a low injection n·pp+ cell approximately given by [15)
with optimal doping. Contrarily to what it has been
observed in low concentration, here the low
injection cell is better. This can be mainly
attributed to the higher Auger recombination
coefficient for high injection (and its (W)
- 1.87x10-15 (qG) -0- exp
.01
2 --
25.7
(V )
supra-linear behavior).
At this point it also worth to mention that
the highest short circuit current so far achieved (W )
- 3 . 94x10 -10 (qG) 2 0.01 exp
3 (V )
51 . 4 (4)
at our Institute [8) , of 41.8 rnA/cm 2 at one sun
(approximate AM1.sG spectrum), is with a n+np+
structure in a cell of 1121 Ocm base resistivity where the coefficients have been calculated at
illuminated on the n+side . The operation at one sun 25°C; for a calculation at other temperature the
leads to less Auger recombination so explaining main influence appear in the ~ubtrahend~,

thi s result. proportional respectively to nl, nl and n ! ,

respectively and the denominators in the
3.2 PC cell operation exponentials, that are proportional to the absolute
In Fig 9 we present the calculated [14) temperature . In these equations the voltage is in
carrier profile of a PC cell in MPP conditions at millivolts, the current density as well as the
an irradiance of 10 W/cm 2 , presenting an efficiency integrated photogeneration, qG are given in A/cm2
of only 23.2% (no texturing, no light confinement). and the cell thickness is in cm.
We observe an e:ssentially one dimensional profile Note the strong dependence of the terms, in
that decreases from the front to the back of the particular of the short circui t current term on
cell . When approaching to the point shaped emitters the thickness. A good light confinement insid~ an
(relatively wide in this cell) we can observe the integrating sphere , as described before, permitting
appearance of concentration valleys necessary to shallow cells, can greatly help to the good
drive the electron flux towards the n+ region and behavior of this structure. As matter of fact the
the hole flux towards the p+ region. For a given strong Auger recombination, is the main limitation
coverage fraction of the rear surface with a type for increasing the concentration at which this cell
of emitters (n+ or p+), say 10%, the depth of these operates efficiently.
valleys increase with the emitter size, so that It must be stressed that this cells relies in
when the emitters are very small, b~t very closely a strong reduction of the surface recombination at
spaced, so as to keep the coverage fraction the front surface. Associated to the generation of
constant, the wells become negligible. If the highly energetic electron-hole carriers, the low
coverage fraction is decreased, then the wells are surface recombination velocity at the front surface
increased in depth as well as if the densi ty of degrades, as reported by the Stanford group. A
current is increased ego by increasing the floating junction on the front face has reduced
irradiance on the cell . this instability to negligible values, but the
efficiency, as far as we know, has remained so far
reduced, to about 24 . 5% [16) .

3.3 Bifacially contacted concentration cells

One of the highest efficiencies with
bifacially contacted cells in concentration, 26.3%
at 50 suns (AMl . sD , 25°C, Sandia confirmed) has
been achieved by A Cuevas et al [17] at Stanford
p. n University (Cuevas was on sabbatical leave from our
(cm-3)
Institute). Their cells have the following
features: (a) the base is highly resistive (n type
100 Qcm) as t he one in the PC cell; (b) the
emitters are point contacts, being located: the p.
a~ the front face (under the metal fingers) and the
n at the rear face; (c) the surfaces are well
passivated; (d) the metal fingers are deposited on
angular ridges in such a way that the light falling
20 40 60 80 100 120 depth (microns) normally on them is sent to the silicon again ; (e)
Fig 9. Carriers density ~rofile in a PC cell at MPP the surface is covered of inverted pyramids.
conditions under 10 W/cm . Cuevas undertook an investigation to assess
whether high or low injection is to be preferred;
The voltages at the np+ and pp regions are wi th the preceding technology high injection cell
given by the carrier concentrations at the bottom gave higher current, lower voltage an lower fill
of these valleys, in accordance with Eqs (3), so factor, as it is to be expected; however the
that, at a given voltage, that determines the overall efficiency was about one point percent
concentrations at the bottom of the wells , the higher .

237
 More recently the Green's group has presented Cuevas et al. high injection p+nn+cell is a good
[18] a low injection cell with the PERL structure example of it. However in high injection cells the
using 15% grid coverage and a prismatic cover [19], carrier-impurity scattering is changed into
that gave 26.7% at 10 mW/cm2 (AM1.5D, 25°C, Sandia inter-carrier scattering, in accordance with a
conf i rmed) . recent investigation by Kane an Swanson [21], that
All this proves that the contest between low introduces a residual resistivity of 0.02 Qcm,
and high injection is still open. unavoidable in cells with electrons and holes
flowing in opposite directions (as it is the case
3.4 The series resistance and novel schemes of in the bifacially contacted cells).
light coupling to the cell In bifacially contacted cells the vertical
For cells under concentration the series series resistance can be reduced by making the
resistance is a serious problem. As matter of fact cells thinner; this may imply to use some advanced
while the efficiency of a resistance-free solar scheme of light confinement as the use of cavities
cell increases with the irradiance, by increase of repeatedly mentioned.
Yoc and of the fill factor, for a cell of specific In the case of PC cells, as no current flows
res~stance rs (resistance of a unity area cell in across them (from top to bottom) no vertical series
Qcm) the highest efficiency is achieved, resistance exists. The grid series resistance does
approximately, at the irradiance giving a short exist, but as the cell coverage can be very high
circui t current such that Jrs ; Yth [20], Vth, (the grids for both contacts are located at the
being the thermodynamic voltage, usually kT/q cell rear) is-value is rather small.
(sometimes 2kT/q or 2kT/3q, depending on the With respect to the emitter series resistance
dominating recombination mechanism). it is substituted in PC cells by the resistance of
In bifacially contacted cells, of classical the current from the n· to the p + emitters. This
structure, the series resistance has three resistance has two terms: the spread resistance at
components: the lateral emitter resistance caused the p+ emitter and the spread resistance at the n+.
by the flow of the current to reach the metal These resistances are proportional to the to the
fingers, the grid resistance caused by the flow depth of the density valleys near the emitters, and
along the fingers to the busbars and the vertical therefore it disappears when they are closely
resistance along the base thickness. spaced. With well designed cells this resistance
Being F the metal shaded area fraction, the can be very small. The consequence is that the PC
two first terms are inversely proportional to F2 cell presents very small series resistance.
and to F respectively. The emitter series It is concluded that the limitation for
resistance is rendered negligible for fingers very operation in concentration of PC cells is rather
narrow (while keeping F constant). The grid the strong supralinear Auger recombination that the
resistance by making the cell small. series resistanc~.
The classical design of concentrating solar
cells is based in optimizing the trade off between 4 THE USE OF LIGHT CONFINING CAVITIES
the grid shading, that reduces proportionally the
photogeneration, and the series resistance that Light confining cavities have been mentioned
decreases with the shading. The PC cell is along this paper as a a mean for (a) increasing the
primarily intended to get rid of this optimization. light confinement and (b) permitting high grid
However there are other methods today to achieve coverage. Both considerations are associated to the
similar results. One is the use of an, already ability of the cavity to permit multiple incidence
mentioned, prismatic cover [19] in the fingers that of the light on the cells. An increased absorbance
deflects the rays that are directed to the fingers (reduced reflectance) is thus produced.
and make these rays to fallon the unshaded silicon Cavi ties with spheric shape as the one shown
areas. The maximum concentration at which this in Fig 6 present a reflectance
solution can operate is limited theoretically [10]
and technologically and today this cover has p' ; pAe/[pAe+(l-p)Ac] (5)
operated with effectiveness, as far as we know, at
some 200x. where p is the cell reflectance, Ac its area and Ae
With similar results and problems than the the area of the entry aperture; roughly speaking
prismatic cover is the use of tilted metal fingers reflectance is reduced in the factor of Ae/Ac and
that reflect the light on the cell, as in the this is the reduction pr';duced in the apparent' grid
example of the cell developed by Cuevas et al. at coverage of the cell.
Stanford, and also in some UNSW cells. More practical cavities are the ones described
Another solution for the reduction of the in Fig 10 where only the light reflected towards
effect of the finger shadowing is to put the cells the concentrator primary can escape.
inside cavities, as the one described already, so
that the rays falling on the metal fingers are
reflected by them, injected into the cavity, and
have additional chances to be absorbed by the
silicon. Again this cannot be done, theoretically,
without loss of the concentration on the cell [10],
but we think that this method is promising.
Note that with all these methods the grid
coverage can be substantially increased, resulting
in a reduction of the emitter and grid series
resistance.
With respect to the vertical series
resistance, i t is worthwhile to mention that, in
silicon cells to operate at very high irradiance,
it is often the dominating component of the series
resistance. The use of low injection p bases (n+pp+
cells) leads to verti~al.resistances higher than if
n bases are used (p nn cells). It must not be
thought that the high injection cells are too Fig 10. Light confining elliptic cavity. The mirror
resistive for concentration: in fact the outline is an ellipse of foci in FF' .
photogenerated carriers provide enough conductivity
and suffer less from impurity scattering. The The reflectance of this cavity is given by

238
 p' = pAe/[pAe+(1-p)Atl (6} REFERENCES

where At is the area of the ellipsoid equatorial [1] MA Green, in Tech. Dig. IntI. PVSEC-5, Kyoto,
plane. It is of interest to note that in this (1990), 603-5.
cavity the concentration on the cell is not to be [2] RA Sinton, Y Kwark, JY Gan and RM Swanson. IEEE
reduced with respect to that at the entry apertur~, Elec. Dev. Lett. EDL-7, (1986), 567-9.
as it occurs in the integrating sphere but, indeed [3] EA DeMeo, FR Goodman, TM Peterson and JC
the concentration at the cell is limited by the S";,,,,,fer, in Proc 21th Photov. Spec. Conf. IEEE,
angular spread limitation of the incoming beam New York, (1990), 16-23.
necessary to enter through the entry aperture. The [4] RA Sinton and RM Swanson, in Proc 7th EC PVSEC,
apparent grid coverage is thus roughly reduced by Reidel, Dordrecht, (1987), 917-21.
Ae/At. [5] A Luque, The requirements of High Efficiency
In this cavi ty the average path of the rays Solar Cells, in A Luque and GL Araujo, Eds.
for light confinement is given by PhYSical Limitations to Photovoltaic Energy
A 2
Conversion, Adam Hilger, Bristol, (1990)
1 = 4n (At/Ae)W [6] GL Araujo, Limits to Efficiency of Solar Cells,
in A Luque and GL Araujo, Eds. op. cit (1990)
The behavior of solar cells under such Limitations to Photovoltaic Energy Conversion, Adam
cavities is being investigated at our Institute Hilger, Bristol, (1990).
under the sponsorship of the European Communities. [7] A Cuevas and M Balbuena, in Proc 8th EC PVSEC,
We have already obtained an experimental increase Kluwer Academic, Dodrecht, (1988), 1186-91.
of the current density from 38 to 40.1 mA/cm2 by [8] J Alonso, I Tobias, A Luque, G Sala and A
locating a cell in a cavity of such nature. Cuevas, Progress in high-efficiency silicon solar
cells. This confer.ence.
5 CONCLUSIONS AND FURTHER COMMENTS [9] JC Minano, Optical Confinement in
Photovoltaics, in A Luque and GL Araujo Eds. op.
The limit efficiency of a Si solar cell is cit. (1990).
around 35.6%, at any level of concentration [10] A Luque, in Tech. Dig. IntI. PVSEC-5, Kyoto,
provided that the proper advanced light confinement (1990), 595-600.
scheme is used. However we are still far from this [11] MD Lammert and RJ Schwartz, IEEE Proc. Elec.
value. One sun cells have achieved 24% and Dev. ED-24, 337-42, (1977).
concentrating cells 28% although with some [12] A Luque, Solar Cells and Optics for
instability, so that stable cells are expected to Photovol taic Concentration, Chap. 5, Adam Hilger,
reach an efficiency of 27%. In a recent workshop, Bristol, (1989).
held in Mishima, Japan, in november 1990, it was [13] GL Araujo a~d JM Ruiz, Chap 2, in A Luque Ed,
concluded that 26% at one sun will be achievable in Solar Cells and Optics for Photovol taic
the mid term. Concentration, Adam Hilger, Bristol, (1989).
The use of high injection vs low injection is [14] A Luque, Solid St. Electr. 31, 65-79, (1988).
still an open question. While for one sun the best [15] A Luque, Solar Cells and Optics for
cells so far operate at low injection, high Photovoltaic Concentration, Chap. 6, Adam Hilger,
injection have given better results in Bristol, (1989).
concentration. Yet for stable cells it seems that [16] PE Gruenbaum, JY Yan, RR King and RM Swanson,
low injection have given better results, so far. in Proc 21th Photov. Spec. Conf. IEEE, New York,
With respect to bifacially contacted and PC (1990 l. 16-23.
cells we can say that both seem to have excellent [17] A Cuevas, RA Sinton, NE Midkiff and RM
prospects, without anyone of them clearly winning Swanson, in Proc 9th EC PVSEC, Kluwer Academic,
the contest. Dordrecht, 761-4, (1989)
Present high efficiency cells follow rather [18] J Zhao, A Wang and MA Green, in Proc 21th
well the models so that we can say that this topic PVSC, IEEE, New York, (1990).
consti tutes already a science, rather than an art [19] MJ O'Neill, US Letters Patent 4069812, (1978).
(this is true even for the material growth [22]), [20] E Sanchez and GL Araujo, Solar Cells, ~, 263-7,
in which sound modeling can produce further (1984) .
improvements. While we have effective and friendly [21] DE Kane and RM Swanson, in Proc 20th PVSC,
codes for one-dimensional analyses of solar cells IEEE, New York, (1989), 512-517.
[23], powerful two-dimensional or three-dimensional [22] TF Ciszec, in Proc 20th PVSC, (1989), 31-38.
codes are still lacking (or if existing they are [23] DT Rover, P Basore and GM Thorson, in Proc
not well trimmed to this application). Their 18th PVSC, IEEE, New York, (1985), 703-9.
development should result in considerable help for
progress.
The economic interest of crystalline silicon,
even FZ single crystal, is certainly linked to high
efficiency. A panel of experts conveyed by EPRI in
june 1989 [3] determined that crystalline silicon
is one of the most serious contenders for
production of electricity in centrals at $0.08/kWh.
In particular in concentration (useful in Europe
only in southern countries) the probabilities of
achieving the required module cost and efficiency
were both considered highly probable, and only
slight doubts (probability medium to high) remained
on the reliability requirements. For flat panels
the doubts for re I iabil i ty did not exist but the
probability of achieving the cost target was found
to be medium. No other flat module technology was
found more promising than that of wafered cells
(some were found equally promising). For grid
connected houses I think that crystalline silicon
flat modules are highly promising even in northern
European climates.

239
l{JfH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORl1JGAL

PROGRESS IN THE MANUFACTURING OF PRODUCTION-TYPE CRYSTALLINE

SILICON SOLAR CELLS

R.P. Mertens
Interuniversity Microelectronics Center
Kapeldreef 75, B-3001 Leuven, Belgium

ABSTRACT. This paper reviews the different developments in the field

of multicrystalline silicon solar cell fabrication. The trends in the
manufacturing steps are discussed. The importance of achieving high
cell conversion efficiencies with a limited number of high throughput
and environmentally benign operations is emphasized. Large area
industrial silicon cells with efficiencies in the 16 to 17 % range can be
expected in the year 2000. Manufacturing cost projections for that year
vary considerably between the different manufacturers with
estimations going from 0.75 to 2 ecu per Watt. About 30 % of the costs
are caused by the cell fabrication. It can be expected that different
processing technologies will be used in parallel.

1. INTRODUCTION wafers using micro-electronic techniques. An

industrial cell, on the other hand, is made on
This paper reviews the different industrial cheap substrate material and has typical
developments in the field of high efficiency dimensions of of 10 x 10 cm 2 . Throughout this
silicon cells with emphasis on fabrication paper it will be assumed that future industrial
processes starting from sliced cells will be made on an ingot-type
multicrystalline ingot wafers. Input for this multicrystalline silicon material. An
paper comes from different European cell important observation is the fact that the
manufacturers : Telefunken System Technik, BP efficiency of industrial cells is considerably
Solar, Italsolar, Photowatt, ENE and Soltech. smaller than that of laboratory cells. The
It is generally accepted now that the reasons for this efficiency gap are as follows:
conversion efficiency is the key solar cell - the differences in material quality and the
parameter. Further cost reductions of silicon cell dimensions
based photovoltaics will only be possible by - the necessity to use
combining cheaper materials with high production methods with high
throughput manufacturing techniques and by throughput
the simultaneous achievement of higher cell - the difficulties to implement
conversion efficiencies. The impact of cell efficient texturing schemes in an industrial
efficiency is application dependent. The environ ment
importance of cell efficiency grows with the - the achievable linewidths that
installed power of the system due to the tend to be larger.
increasing effect of the balance-of-system Any action to close this efficiency gap, must
cost for larger installations. It can therefore take the above considerations into account.
be expected that two silicon solar cell
technologies will co-exist. The first is the 25
lower cost option with efficiencies in the 12 D

to 13 % range, predominantly used in

20
applications such as village power and roof Lab
type installations. For central utility type of ~ •
systems considerably higher efficiencies are
>,
0 15
c
0
necessary, due to the larger balance-of-system '0 Indusl1y
cost. For these applications a high efficiency
t::
' ,- 10 IOOcm2 '
option with efficiencies of 16 to 17 % will be
required. 5
An analysis of published data shows (Fig.1) 1975 1980 1985 1990
that the efficiency of "laboratory-type" silicon Year
cells has increased considerably over the
years. A "laboratory-type" cell is a cell for Figure Efficiency of laboratory and
which the efficiency is maximized without industrial silicon solar cells.
cost consideration; it has typical dimensions
of 2 x 2 cm 2 and is made on high quality silicon

240
2. THE BASELINE PROCESS accaunt. Ideally, a cell fabricatian pracess
should be limited ta 10 operations; each
Let us cansider the pracess-flaw .of a aperatian taking .only 1 ta 2 secands per cell.
typical baseline pracess as it is used in Alsa the material cast shauld be as law as
industry taday (Fig.2), yielding efficiencies of passible. Mareaver, enviranmental cancerns,
11 ta 13 % on multicrystalline 10 x 10 cm 2 related ta waste dispasal, are receiving more
silicon substrates. attentian.
In the next sectians these pracessing
trends and changes fram the base line pracess
are described.

3. TRENDS IN THE INDUSTRIAL

PROCESSING OF SILICON
SOLAR CELLS

3.1 Multicrystalline substrates

In recent years the quality .of
multicrystalline silican ingat materials has
impraved cansiderably as diffusian lengths
have increased from typically 60 ~m in 1980
ta 150 ~m in 1990 with a scape far further
impravement up ta mare than 200 ~m. This
enhancement in diffusian length results fram
the aptimizatian .of the ingot farmatian
pracess .or fram .optimized gettering schemes.
Althaugh the standard multicrystalline
substrate is still 10 x 10 cm 2 , there is a clear
trend ta larger cell sizes. Efficient cells .of 15
x 15 cm 2 and even .of 20 x 20 cm 2 (1,2) have
been reparted. The driving farce tawards these
larger cell sizes results fram the fact that the
Figure 2 : Simplified base line pracess, 0 cell manufacturing and the madule assembling
refers ta aptianal steps. casts show little area dependence and that
therefare the cast/Watt ratia decreases with
The pracess starts from a multicrystaliine cell size. On the other hand the .optimum cell
wafer of typically 200 ta 350 ~m thick, 10 x size is limited by series resistance and by the
10 cm 2 large and with a typical diffusian limitatian .of madule size due ta handling, wind
length .of 80 to 120 ~m. The wafers are laads an the madule and transpartatian.
etched ta remave the sawing damage. Therefare cell sizes larger than 20 x 20 cm'
Depending an the manufacturing pracess the are excluded.
wafers are gettered by a heavy phaspharaus Due ta the success .of develapments in
diffusian which is remaved from the backside multiwire slicing (3,4), virgin wafer
and etched back at the frantside ta .obtain the thicknesses .of 200 micrans or smaller are
desired sheet resistance of the emitter. If the bec.oming a reality, corresp.onding t.o final cell
wafers are nat "gettered" the emitter diffusian thicknesses .of 180 ~m .or lawer. This allaws
itself can be cansidered ta induce same an impartant material saving but at the same
gettering actian. In a third passibility the time thinner cells carrespand mare ta the
emitter is farmed after gettering and after an .optimum value if efficient light trapping and
etching step at bath sides. Again depending an surface passivati.on are passible (5).
the pracess aptian a back-surface-field is An impartant paint is the cleaning after
farmed. In mast cases this .occurs by allaying sawing. As the material quality impraves mare
an aluminium thick film, printed at the back attentian shauld be paid t.o the cleanliness .of
side .of the wafer. Then the frant and back the sliced multicrystalline substrates in .order
electrades are farmed by screen printing. The ta avaid metallic c.ontartlinatian at the surface
final pracess steps are the depasitian of an at the start .of the cell manufacturing pracess.
antireflectian caating and a hydragen
passivatian step which is aptianal. In view .of 3.2 Optical canfinement-Texturing
the aim ta realize higher efficiencies and An impartant reasan far the lawer
lawer manufacturing casts the base line efficiency an multicrystalline silican is
pracess fallaws the trends,discussed in caused by the higher reflectian lasses due ta
chapter 3. the less efficient texturing than an single
It is impartant ta nate that all process crystalline silican. Texturing .of
madifications ta .obtain a higher efficiency multicrystalline silican can be dane
must take growing cast requirements inta anisatropically .or isatrapically.

241
 In the anisotropic texturing method an and clusters. Other experimental results have
etchant with an orientation dependent etching indicated that high quality semicrystalline
rate such as NaOH is normally used. When such silicon substrates with initial diffusion
an etchant is used on multicrystalline silicon lengths exceeding 150 J.l.m do not improve
two problems can occur. First the texturing is substantially by a phosphorous gettering
inefficient because of random grain treatment.On the other hand, it has been shown
orientation. A second problem is the that phosphorous gettering and subsequent
possibility that planarity of the surface is lost gettering by aluminium or boron are additive
because of the highly selective etching along (9).
particular orientations. In such case
subsequent processing becomes very difficult. 3.4 Junction formation
It turns out that by increasing the preferential Advanced silicon cells need an optimized
orientation through an optimized ingot emitter. Detailed studies (10) have indicated
formation process and by a better control of that for cells with evaporated contacts the
the etching speed, typically 50 % of the grains optimum emitter is a rather deep, moderately
can be efficiently texturized without too much doped profile having surface passivation.
loss in planarity. Surface passivation is normally obtained by
A second possibility is to use isotropic growing a dry oxide on the surface. Since
etchants in combination with a patterning oxide growth is done prior to metallization,
technique. An example is given in Fig.3. A care should be taken that the interface
sandwich layer of Si02 and a protective resist properties can be maintained during the
layer is deposited on the silicon substrate. metallization but also that ohmic contacts can
be made.
In the case of screen printed contacts a
much more heavily doped surface is necessary
/ Re is( in order to obtain an acceptable contact
resistance with the metal fingers. In such
" .... . . .1-- i02 case a selective emitter is an attractive
'~' solution. Such emitter is heavily doped in the
ilicon contact regions and only moderately in the
active parts of the cell (Fig.4). Ideally the
structure is made in a self-aligned process
Figure 3 : Texturing using isotropic etchants scheme.
(after 6).

After definition of a structure an etchant (e.g.

an HF/HN03 mixture dilluted in H20) with an
optimized ratio of the respective etching rates
of Si and Si02 is used. Under these conditions
a pseudo-V-groove type structure can be p-Si
obtained (6). This procedure normally involves
photolithography and is therefore probably too
expensive in production. However, if an
appropriate resist layer can be found that can
be screenprinted with small linewidths (60
Figure 4 : Selective emitter concept.
J.l.m) this technique may become economical.
A third possibility is that of laser Wenham et al. (11) have proposed to use the
texturing, combined with chemical etching (7). screen printed metal grid as a mask in a dry
etching step in which the emitter in the active
3.3 Gettering regions is etched back. Another possibility is
Narayanan et al. (8) have shown that a to print a polymeric protective paste over the
phosporous pretreatment can increase the silver pattern and to etch-back the active
diffusion length in multicrystalline substrates. emitter part using wet etching. Alignment is
Martinuzzi (9) has studied in detail such easily obtained as follows : silver is printed on
phoshorous gettering techniques and has found a batch of cells and the screen is subsequently
that an optimum temperature for gettering cleaned without removing it from the printer;
exists. This optimum temperature, lying the polymeric layer is then printed on the cells
around 900 DC results from the competition maintaining the same cell sequence. The
between external gettering, extracting protective layer can be burned off in a
impurities, and internal gettering, decorating conveyor furnace. Application of this concept
the grain boundaries and increasing the in the standard process adds 0.8 % to the
recombination at defects in the material. The efficiency. Further work in this area is needed
efficiency of gettering steps depends on the in order to find ways of surface passivation,
method used and phosphorous gettering is not compatible with this self-aligned emitter
able to remove impurities from precipitates concept. The processing of Si3N4 plasma

242
deposited coatings is actually being optimized height- to-width ratio of the finger
in order to reach acceptable levels of effective metallization. This is a very important
front side surface recombination velocity. advantage when fabricating large area cells, as
it allows top contact shading losses below 3%.
3.5 Back-surface-field and back electrode A second advantage is the fine line width
As future cells tend to be thinner and will (typically 20 J,lm) which can be obtained. This
have longer diffusion lengths, the surface allows the first diffusion, forming the active
recombination of the rear surface must be part of the emitter, to be very light without
reduced. This can be achieved by the creating an excessive lateral series
incorporation of a pp+back-surface field. The resistance.
most efficient pp+ junctions are obtained by
boron diffusion rather than by the more
industrial aluminium alloying process. This is
probably due to the higher solid solubility of Si02
boron in silicon.
A second way to reduce the back surface
recombination is the use of a gridded back
electrode with passivated non-contacted areas
to reduce the surface recombination velocity. Figure 5 : Buried contact approach-Laser
The excellent results with the new generation Grooving (after 14)
of PERC and PERL cells at the university of
New South Wales (12) have demonstrated that The fact that very high efficiency modules
an oxidized p-type back surface with selective can be obtained with this technology has
contact openings yields higher open circuit recently been demonstrated since a solar
voltages than obtained with a homogeneous pp+ electric car, equipped with these cells, was
back surface field. In such cells the interface the winner at the 1990 contest for solar
at the back side is probably inverted such that electric cars in Australia.
the recombination at the backside is controlled The main possible disadvantage related to
by the hole capture rate with cross section ap this technology is an environmental issue. In a
rather than by the electron capture with cross recent paper (15) Wald states very clearly that
solar technologies have to be environmentally
section an. Since an is one to two orders of
benign. Therefore the external costs to meet
magnitude larger than ap, surface environmental specifications with processes
recombination on p-type surfaces tends, for producing enormous amounts of rinse water
equal surface doping levels, to be considerably containing nickel and copper must be taken into
larger than for n-type surfaces. Under specific account. This is not . an important issue in the
conditions lower back-surface recombination development or preproduction phase but it will
velocities can therefore be obtained with be a major point for future 100 MW plants.
floating n-type islands than with a pp+ The second route, followed by several
diffused region (13). Moreover such gridded manufacturers, is based on a further
back electrode allows the implementation of improvement of the screen printing technology
optical confinement schemes in a more flexible used in most production lines. Several
way. companies have shown that screen printed
finger widths smaller than 100 J.lm can be
3.6 Front electrode formation obtained with the lowest limits estimated to
High efficiency, large area cells require be in the 60 to 80 J.lm range. Achievement of
front electrodes with low series resistance such linewidths in production will require
and low coverage. In order to meet these novel screens and pastes and at the same time
requirements further developments are a tighter control of the production environment
necessary. In this respect silicon solar cell to maintain constant process parameters.
man ufacturers are following two different Screen printing has the advantage to be easily
routes. amenable to automation and to be a dry process
The first route is the buried contact with negligible production of waste materials.
approach (Fig.5), invented at the University of The disadvantage is the width of the fingers
New South Wales (14) and now in its pre- which is 3 to 4 times larger than with the
production phase at different manufacturing laser buried contact approach. The way to
plants in Europe. In this process, grooves are further reduce the coverage factor with such
made in a lightly prediffused and nitride (or line widths is to decrease the number of
oxide) covered wafer by high speed laser fingers. However, since high efficiency cells
scribing. After cleaning, the grooves are require emitter sheet resistances of 70 Q per
subjected to a second, very heavy diffusion. square or more, problems with series
The metallisation is then obtained in a self- resistance caused by lateral voltage drops in
aligned way by electroless deposition of the diffused layer occur. Grooves
nickel, followed by copper. The major perpendicular to the fingers provide a solution
advantage of this technique is thp. very large for this problem (16) as the cross section of

243
me current pam In me emitter towaras me of oxide passivated emitters. For these cells
fingers can effectively double (Fig.6). Ano~her the plasma power must be adjusted to avoid
possibility is the use of a conductive damage to the oxide during plasma treatment.
anti reflective coating. If the plasma power is too high hydrogenation
may actually result in an efficiency decrease.
It has also been found that, in the case of cells
finger ~~~ which are covered by nitride on both sides, the
cell efficiency drastically improves after a
heat treatment at 700 0C (1). Heat treatments
after hydrogen implantation recover the
n+ degradation of the silicon/silicon - dioxide
Figure 6 : Grooving perpendicular to the metal interface which occurs during the electrode
fingers (after 16) firing (16). The effect of recovering of the
surface passivation and of the bulk lifetime
3.7 Antireflection coating makes that some form of hydrogenation will
Today Ti02 , deposited by chemical vapour remain necessary in most future high
deposition, is the most widely used anti- efficiency cell processing lines even if the
reflection coating for industrial initial diffusion lenghts will exceed the wafer
multicrystalline silicon solar cells. The thickness. Moreover, as for gettering, hydrogen
material has a refractive index of 2.3, passivation results in a narrower efficiency
corresponding to the optimum value for distribution in production.
encapsulated silicon cells. Somewhat lower
values can be obtained by a plasma deposited
Si3N4 layer. 4. EVOLUTION OF COST AND
The use of double anti reflective coatings EFFICIENCY
in production is probably not justified since
most future high efficiency multicrystalline The different solar cell manufacturers
cells will have some kind of texturing or agree that the efficiency of industrial cells
microgrooving. It has been demonstrated (17) will continue to increase. Table I lists the
that very low reflectance is possible for V- efficiency projection of four major European
grooved silicon even with a single non-ideal manufacturers for the year 2000. On the other
anti reflective coating such as Si0 2 . It can hand projections, of the same companies, for
therefore be expected that future industrial cost reduction are strongly different. The
cells will continue to have a single second column in the table gives the
anti reflective coating (Ti02 or Si3 N 4) in estimation of the direct manufacturing costs,
combination with a thin passivating « 20 nm) excluding investment depreciation costs, of
Si0 2 layer. multicrystalline silicon modules. The cell
On the other hand, if screen printing processing part corresponds to 30 % of the
continues to be used to deposit the front side total manufacturing cost.
electrode a renewed interest in conductive
oxides such as ZnO may emerge in order to EFFICIENCY AND DIRECT COSTS IN 2000
decrease the coverage factor on lightly doped,
passivated emitters without loss in lateral Manufacturer 11 direct cost
series resistance. (%) (ecu/w)

3.8 Hydrogen passivation A 17 0.75

It is well known that hydrogen atoms B 15 2
introduced into the silicon substrate act as a C 17 1.5
terminator to the dangling bonds associated D 16 1.5
with the crystal defects in multicrystalline
silicon. Three ways to introduce atomic Table I projection of the efficiency and direct
hydrogen have been proposed : ion implantation, costs of production type cells in the year 2000
plasma hydrogenation in a hydrogen glow
discharge and hydrogenation using a plasma 5. LONGER TERM TRENDS
deposited silicon nitride layer. Hydrogenation
can be performed through the front and On top of the more evolutionary type of
backside. In the literature contradicting ongoing developments to increase the
results concerning the efficiency of frontside efficiency of production type cells, discussed
hydrogenation by hydrogen plasma and silicon in the previous sections, longer term options
nitride have been published (1,18). Whereas exist.
some groups have reported important A drastic change in cell production will
improvements in cell efficiency resulting from occur when ribbons will become economically
frontside hydrogenation other groups found no competitive. Ribbon technology allows further
improvements. It is generally agreed upon that cell processing simplification such as in situ
frontside hydrogenation is critical in the case junction formation during ribbon growth (15).

244
On the other hand special metallization (6) P. Verlinden, O. Evrard, F. Mazy and A.
techn iques may be required (19). Crahay, to be published in Solar cells.
Simplified back contact solar cell designs (7) S. Narayanan, J. Zolper, F. Yun, S. Wenham, A.
(20) also have some long term potential for one Sproul, C. Chong and M. Green, 21th PVSC,
sun operation because of the possibility to Orlando, 1990, 678.
decouple the electrical and optical design of (8) S. Narayanan, S. Wenham and M. Green, Appl.
the cell and due to the potential for a Phys. Lett. 48 (13), 1986, 873.
simplified interconnection on the module level. (9) S. Martinuzzi, H. EI Ghitani, D. Sarti and P.
This approach is only viable if sufficiently Torchio, 20th PVSC, Las Vegas, 1988, 1575.
high ratios of diffusion length to wafer (10) R. King, R. Sinton and R. Swanson, IEEE T-
thickness can be achieved at an acceptable ED, 1990, 365.
cost. (11) S. Wenham, M. Willison, S. Narayanan and
MIS cells (21) can be a viable competitor M. Green, 18th PVSC, Las Vegas, 1985, 1008.
due to the low thermal budget of the (12) A. Wang, J. Zhao and M. Green, Appl. Phys.
fabrication process, provided that long term Lett. 57 (6), 1990, 602.
stability can be demonstrated. (13) M. Ghannam, to be published.
These different long term alternatives can (14) M.A. Green, High Efficiency Silicon Solar
only become economically viable if a high Cells, Trans. Tech. publications,1987.
enough throughput can be achieved in (15) F. Wald, 5th Int. PVSEC, Kyoto, 1990, 191.
production. (16) H. Vagi, S. Kokunai, S. Suzuki, Y. Saegusa, K.
Matsukuma, K. Nishinoiri, T. Saitoh, R.
6. CONCLUSIONS Shimokawa, Y. Hayashi, K. Morita and M.
Nakatani, 4th Int. PVSEC, Sydney, 1989, 129.
An almost general consensus exists that the (17) O. Kamataki, S. Lida and T. Saitoh, 5th Int.
efficiency of industrial silicon cells will reach PVSEC, Kyoto, 1990, 497.
the 16 to 17 % range in the year 2000. It is (18) J. MulJer, E. Hussian and P. Siffert, 9th EC
believed that different processing schemes PVSEC, Freiburg, 1989, 407.
(e.g. dry versus galvanic steps for the (19) J. Grabmaier, H. Urbach and K. Eisenrith,
metallization) will be used. In order to reach 5th Int. PVSEC, Kyoto, 1990, 509.
low enough cost levels, the cell processing (20) R. Sinton and R. Swanson, IEEE T-ED, 1990,
sequence must consist of a limited number of 348.
high throughput steps. Moreover the (21) W. Bauch and R. Hezel, 9th EC PVSEC,
fabrication must meet future environmental Freiburg, 1989, 390.
requirement. Finally, the manufacturing cost
projections of different silicon solar cell
suppliers differ by more than a factor of two.

7. ACKNOWLEDGEMENT

The author would like to thank the

companies Telefunken System Technik, BP
Solar, Italsolar, Photowatt, CNRS (Paris and
Strasbourg), University of Marseille, ENE and
Soltech for the fruitful discussions on the
topic.
He appreciates the help from J. Nijs, M.
Ghannam, J. Coppye and E. Demesmaeker in the
preparation of the manuscript. Part of this
study was sponsored by the EC, Joule Program :
Contract Number JOUR-0036.

REFERENCES

(1) H. Watanabe, K. Shirasawa, K. Masuri, K.

Okada, M. Takayama, K. Fukui and H. Yamashita,
Optoelectronics-Devices and Technologies, Vol.
5, 1990, 223.
(2) ENE, 9th EC PVSEC, Freiburg, 1989, Late
news.
(3) G. Smekens, J. Vialart and C. Hauser, EC
Contractors Meeting, Stuttgart, 1990.
(4) C. Hauser, 9th EC PVSEC, Freiburg, 1989,
437.
(5) T. Tiedje, E. Yablanovitch, G. Cody and B.
Brooks, IEEE T-ED, 1984, 711.

245
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

High-Efficiency Backside Contact Solar Cells with a Self-Aligned Process

and New Texturization Technique for Silicon

P. Verlinden·, B. Lafontaine, O. Evrard, E. Mazy and A. Crahay

Universite Catholique de Louvain, Laboratoire de Microelectronique

Place du Levant, 3 B-1348 Louvain-Ia-Neuve BELGIUM

·present address: SunPower Corporation

625 Ellis Street, Suite 209 Mountain View, CA 94043 USA

ABSTRACT. A self-aligned process for Back Side Contact solar cells has been demonstrated. It requires only
one photolithography step without alignment, instead of 4 to 6. As a consequence, the fabrication cost of BSC
solar cell has been highly reduced. Also, a texturization method using the HF-HN03 couple in an isotropic
etching has been investigated. This method allows to obtain a surface of periodic V-grooves with controllable
sidewall angles. This new texturization technique, which may be used with multicrystalline silicon as well,
opens the way to new light trapping structure.

1. INTRODUCTION texturization. However. the application of texturization has

generally been limited to monocrys!!llIine substrates because
1.1 Self-aligned process for Back Side Contact it uses a anisotropic etchant like KOH. Also, the design of light
trapping structures has been limited to a fixed angle (54.7°)
The Interdigitated Back Contact (IBC) design presented determined by the crystallographic planes. Although a double
by R.J. Schwartz et al. in 1975 (8) has proven to be an side textured solar cell is known to give a better light trapping
excellent cell design for concentration application. This paper than single side texture, this design is difficult to apply on
presented for the first time a new photovoltaic device in which high efficiency BSC since it also brings several problems
the electrical optimization and the optical optimization might during photolithography steps. Finally, it was found that such
be completely decoupled. For example, in a IBC cell there is no texturization, with a large angle, may give non-stable BSe
trade-off between the series resistance and the collection concentrator cells due to a highly stressed front surface [6). A
efficiency unlike most double side contacted solar cells. Since new texturization method using the HF-HN03 couple in an
that date, the Back Side Contact (BSC) design has been isotropic etching has been investigated. This method, which is
improved and several research groups have contributed to reported in the second part of this paper, has allowed us to
understand new high efficiency concepts, silicon efficiency obtain a surface of periodic V-grooves with controllable
limits and process related defects. A particular case of the BSC sidewall angles, from 0° to 40°, with a very good
design is the Point-Contact cell developed by R.M. Swanson at reproducibility.
Stanford University [1). His team presently owns the world
efficiency record for silicon solar cells: 28.5% under 15 2. SELF-ALIGNED PROCESS
W/cm 2 at 25°C. The calculations however show that the
efficiency limits of IBC and BSC cells do not differ in a decisive 2.1 The self-aligned process sequence
manner, though the Point-Contact design has a marginal
advantage (9). Still today, the BSC solar cell is considered by We have recently proposed a self-aligned process
several PV specialists as the best design for concentrator cells. sequence for the fabrication of BSe solar cells [5). Self-
The disadvantage of BSC cells is the requirement for a aligned means that only one photolithography without
sophisticated processing sequence, making use of high quality alignment is required to define the total solar cell structure
materials (e.g. FZ silicon), high purity chemical products, (N+ emitter, P+ emitter, contacts and interconnections). The
extremely clean processing techniques in an environment process flow (Figure 1) has b!)en described in (5). After the
similar to that used for integrated circuit processing and , at deposition and densification of a thin phosphorous doped oxide
least, 4 photolithography steps. This number could even be and a non-doped oxide as capping layer (Fig. I-a), a
increased to 6 for double level metallization when it is photolithography step defines the geometry of the BSe cell. The
necessary (3). In the fabrication cost calculation, the oxide layer is then etched in BOE (Fig. I-b) and deep grooves
photolithography is. among the most important components. are formed in the silicon substrate by using a HN03 - H F
For this reason. some workers prefer to develop conventional solution that etches silicon at least 100 times faster than the
solar cells with metal contacts on both sides of the cell, with oxide layer (Fig. I-c). At this step, oxide cantilevers are
the use of a prismatic cover for concentration application formed and overhang the silicon trenches. These cantilevers,
[10). However, the fabrication of conventional solar cells. used as spacers in order to separate the emitter regions, are
with a high conversion efficiency under concentrated sunlight, supported by narrow mesas which will be N+ doped. The length
should be not so cheap that it may first appear. Their of the cantilevers and the depth of the trenches are about 8
processing sequence also includes several photolithography 11m, I.e. about twice the desired value of the emitter junction
steps. depth. The sidewalls of the grooves are then oxidized in dry
After demonstrating high efficiency BSe silicon solar oxygen with 1% TCA to form a 100 nm thick passivating layer.
cells of small [2) and large area [3), our subsequent research It is important to notice that, during this high temperature
work has been to demonstrate a new process flow that presents oxidation, as well as during all the high temperature steps, the
low fabrication cost and high efficiency capabilities. In a first cantilevers do not bend. AlSO nm thick silicon nitride layer
part, this paper reports on the recent results of BSC cells is then deposited (Fig. I-d). Using Reactive Ion Etching
fabricated with a low-cost self-aligned process. (RIE), which is an anisotropic plasma etch, a window is
opened in the nitride and oxide sandwich on the bottom of the
1.2 Variable angle texturization of Silicon grooves (Fig. I-e). These windows are self-aligned on the
cantilever gaps and can be as small as 2 11m in width. Also
The bulk recombination effects (such as the Auger effect, during this step, the nitride layer disappears from the top of
for instance) that prevent the collection of photogenerated the cantilevers and the mesas. A boron predeposition is then
carriers are determining the present tendency of designing performed and both emitters are diffused in oxidizing ambiant
thinner cells. Such an evolution requires light trapping
(Fig. I-f). The N+ emitters are located in the mesas and the
methods in order to keep the average light path to good values.
Light trapping is usually obtained through surface P+ emitters underneath the grooves, without formation of

246
 Undoped OKlda

PSG

t) Oeposi. phosphorous
doped (PSG) and undopcd
o.ide sandwich On Si
wlfen.

5) RCICtive Ion Elching (RIE)

ornilride o.ide layer.

2) Phototllhognphy and
o.ide.",h.

6) Boron ptedcposition.
oxicWion and <!rio.·in.

3) Ocq> groo... ore ronncd I

silieon by isotropic Clelling.
As. • resul •• long cantileven
overhang lh<: SJ<Xlv", Aluminum

Nor P SIlicon

OK lde
7) Oxide etch and aluminum eVlporation.

Figure 1: Sequence of the self-aligned process

Figure 2: SEM micrograph of a silicon trench

Figure 3: SEM micrograph of a silicon trench

247
compensated regions. All the oxide layers. except those that 3
are protected by the nitride layer on the sidewalls of the
grooves and the front ARC protected by photoresist. are etched
in BOE. A thick aluminum layer is then deposited by vacuum
evaporation which features an extremely poor step coverage rt~ m~
iii

i\.
(Fig. 1-g). The cells are finally annealed in forming gas.
The principal reason in developing such a self-aligned
process is to suppress the compensated regions appearing in a

1\
2
p~viously reported self-aligned process (4). Although A.A.
Sinton showed a lack of detrimental effects due to the
compensated regions [11). they reduce drastically the
breackdown voltage of the diode. and the optimization of the :
emitter resistivity has to take this parameter into account.
=
Isc 2.54 Amp
The length of the cantilevers. approximately equal to the deep
=
Voc 0.756 Volt
of the grooves. is obviously an Important feature . It
determines the distance between opposite type emitters. the - =
Pmax 1.294 Watt
emitter coverage fraction and. for a fixed pitch. the size of the =
FF 67.4 "10
N+ region confined in the mesa. Finally. it is important to =
Jsc 10.16 Alcm2
notice that the self-aligned process allows to produce a certain
texturization of the backside of the cell. It forms a regular
=
Sublin 0.93
array of 8 oem deep grooves on a small pitch (34 oem in our
case) which. along with the randomly texturized front surface
o IIII I II I I
and a 100% backside reflector. performs a very good light
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
trapping structure.
Voltage (V)
2.2 Results
Figure 4: I-V curve of a BSe solar cell under
5 mm square BSe solar cells were made with this new concentrated light
self-aligned process. The starting material was FZ. 10ncm.
n-type. high lifetime silicon wafers. The structure of the cell
1.1

-
was a IBe design with tapered stripe emitters. The unit cell is
34 Ilm in width. with a windows width (the gap between J"1!IA
¥ .... ttIOIo.. ID-m
1.0
cantilevers) varying from 1 to 17 Ilm. With a trench depth of
8 tJ.ffi. it gives 1 tJ.ffi to 17 tJ.ffi wide N+ mesas. 0.9 1 ~

Figure 4 shows an I-V curve of one of these BSe cells

under highly concentrater;j light (around 30 W/cm2) and O.S
..... a- il!!
Figure 5 presents its shdrt-circuit current sublinearity vs.
the incident power density. The low value of the fill factor Is ~ 0.7
due to the fact that the P metal fingers are much narrower .~
than the N type fingers (the cell layout has been designed for a Q) 0.6
previously developed process). .S:
:0
::J 0.5
3. VARIABLE ANGLE TEXTURIZATION C/)

-
0.4
The surface texturization fulfills two Important
functions: the first is to reduce the loss of incoming light by 0.3 ---, Sublinearity
recollecting the reflected light. The second role is to trap the
light inside the cell by internal reflection. 0.2
The surface configuration that we have decided to produce
is a periodic sawtooth function whose sidewall angle is Included 0.1
between O· and 40·. A way to make V-grooves in silicon is to
grow or deposit Si02 on the silicon surface. to deposit a mask 0.0
with a very narrow window (2.5 Ilm) on that oxide. and to o 10 20 30 40 50 60
etch the set with a HF-HN03 solution. The oxide will be
Pine (W/em2)
laterally dissolved by HF and will uncover surfaces of silicon
to the etching solution (Fig. 6). The ratios of the etching rates
Figure 5: Short-circuit current sublinearity
of Si and of Si02 • Er(Si) and Er(Si02). will determine the
tangeant of the sidewall angle a :
photoresist is expected to also give good result. The dependency
a = arctg { Er(Si) / Er(Si02)} of the V-grooves sidewall angle on the percentage of the HN03
solution was studied. By adding HN03 to the solution. the
In order to have a periodic sawtooth function. a periodic etching rate of silicon should increase and the V-groove
array of parallel vertical trenches are created In the mask. We sidewalls should become steeper_ Figure 7 shows the
chose a pitch of 22 tJ.ffi. In our experiments. we used a silicon increaSing sidewall angle with the nitric acid concentration.
nitride layer as the mask. but a thick UV hardened and singed For a fixed HN03 : HF mixture (1 .6 % : 98.4 %). the observed
saN4 canlileYor

_ _ SaN4 2000.4

_ _ Si02 2000.4

Figure 6: The etching of Si and Si02 by HN03-HF

248
 average texturization angle is 19.3°, with a 2.3° standard [3) P. Verllnden et aI., 'Multilevel Metallization for Large
deviation and a 5.3° interval of 95% confidence. The details Area Point-Contact Solar Cells', 20th IEEE Photovoltaic
about this technique are published elsewhere (7). This Spec. Conf., Las Vegas 1988, pp. 532·537
technique leads to interesting applications for the [4) R.A. Sinton et aI., 'Development Efforts in Silicon
texturization of polycrystalline silicon which cannot be Backside-Contact Solar Cells', 8th European Photovoltaic
treated with anisotropic methods, for designing new light Solar Energy ConI., Florence 1988, pp. 1472·1476
trapping structures, and for lower stress textured surfaces. (5) P. Verlinden et aI., 'Super Self-Aligned Technology for
Backside Contact Solar Cells: A Route to Low Cost and High
4. OPTICAL MODEUNG Efficiency', 21rst IEEE Photovoltaic Spec. Conf.,
Kissimmee 1990,pp. 257-262
A ray tracing program has been developed in order to [6) P. E. Gruenbaum, 'Photoinjected hot-electron damage at
calculate average paths before escape of long wavelenght rays the silicon/silicon dioxide interface in Point-Contact solar
with an air-silicon refraction index ratio of 3.5. V-grooves cells', Ph.D. Thesis, Stanford University, April 1990
surface configurations with a period of 22 11m have been (7) P. Verlinden, et aI., 'The surface texturization of solar
modeled. Parallel directions of V-grooves on the front and back cells: A new method using V·grooves with controllable
surface were considered. The sidewall angles of V-grooves sidewall angle', to be published.
belonging to the front surface have been set to 54.7 0, which is [8) R.J. Schwartz et aI., 'Silicon Solar Cells for Hlgh-
the angle formed by the (111) planes, resulting from a front Concentration Application', IEEE International Electron
texturization with ethylene diamine or KOH. The average path Devices Meeting, Washington DC, December 1975.
before escape is shown on Figure 8 as a function of the angle of [9) J.E. Parrott et aI., 'Comparison of the Predicted
back surface texturization, the results are reported for cell Performance of IBC and Point-Contact Solar Cells', 19th
thicknesses of 50 11m and 100 11m. An efficient light trapping IEEE Photovoltaic Spec. Conf., New Orleans 1987
occurs at small sidewall angles, in the 2° to 10° range: the [10) J. Zhao, A. Wang and M.A. Green, '24% efficiency PERL
light path is 8 times the device thickness. The second function structure Silicon Solar Cells', 21 rst IEEE Photovoltaic
of the ray tracing program is to calculate generated Spec. Conf., Kissimmee 1990, pp. 333-335
photocurrents in the light trapping structures. A solar AM 1.5 [11) A.A. Sinton et al., 'Simplified Backside-Contact Silicon
spectrum with an integrated intensity of 1000 W/m 2 has been Solar Cells Designs for One·sun and Concentrator
used and the cell is assumed without any recombination Applications', 9th European Photovoltaic Solar Energy
mechanism. The program takes 10 different rays meeting a V· Conf., Freiburg 1989, pp.324-327
groove with a vertical incidence, calculates mean paths before so
escape for 40 wavelengths for every incident ray . Figure 9
shows the global photocurrents for texturized wafers with a
front angle of 54.7°, and thicknesses of 25 11m, 50 11m and
100 11m. A significant gain In photocurrent (from 2.5 to 5 %)
can be earned when passing from a planar back side to a back
side texturlzed with sidewall angles varying from 2 ° to 7°.
Those small angles would accept a photolithography process
and thus will allow a back side contact (BSC) technology.

5. CONCLUSlQ'.JS
A new self-aligned process for BSC solar cell has been
proposed. The early results showed promising performances,
though some improvements are still needed, I.e. the
replacement of the silicon nitride layer by another material. o ~~--~----~----~----~----~~--~
The self-aligned concept is however demonstrated as a route o % HN03 2
for making less expensive high efficiency concentrator solar Figure 7: Sidewall angle vs. HN03 concentration
E
cells. This self-aligned process presents several interesting
features. It does not form compensated regions between .
..3
,.
1528

..."
0.
opposite type emitters, the emitter coverage fraction and the t 1;JB
u 11
metal contact area are reduced compared to the total area and l'ea
the silicon surface is passivated with thermally grown oxide. u loao
We have developed a new texturization method that
. 9ao
~

produces V-grooves with controllable sidewall angles. This .:! Bao

method involves the lateral etching of an oxide layer by HF that .D 72
GaO
progressively uncovers surfaces of silicon which will be seo
isotropically etched by the HF-HN03 solution. The sidewall -=" .ae
"-
angle is controlled by the percentage in volume of the HN03 lao
solution. This technique leads to interesting applications for
the texturization of polycrystalline silicon solar cells which ~
.."
c 100
lae
e
cannot be treated with anisotropic etching methods, for 8 18 \5 12 2S 32 Angle [oJ
designing new light trapping structures, and for getting lower
Figure 8: Average path length before escape for a 54 .7°
stress textured surfaces. A ray tracing program has shown
(front) - variable angle (backside) texturization
that a back side texturization with small sidewall angles .5
associated with the traditional front side texturization that
produces sidewall angles of 54 .7° can increase the
photocurrent by a few percents. Backside texture with small ::;-
E
sidewall angles would still be compatible with BSC designs. ~

.. -<
.. E
ACKNOWLEDGMENT ..,~

The process developments and cell fabrications have been

carried out with the help of P. Proesmans, B. Katschmarskyj,
P. Loumaye and B. Heren!. This work is supported by the C.E.C ••
under contract JOUR/0040/C (MONO-CHESS).

REFERENCES
[1) R.A. Sinton et aI., '27.5 % silicon concentrator solar
cells', IEEE Electron Device Letters, EDL-7, nOlO, 1986
[2) P. Verlinden et aI., 'High efficiency interdigitated back
35
contact silicon solar cells', 19th IEEE Photovoltaic Spec.
e 5 18 \5 18 25 38 Angle [oJ
Conf., New Orleans 1987, p. 405-410
Figure 9: Maximum short·circuit current vs . sidewall angle

249
lOTI! EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

RECENT ADVANCES IN SILICON SOLAR CELL PERFORMANCE

Martin A. Green
Centre for Photovoltaic Devices and Systems
University of New South Wales
Kensington, Australia, 2033

ABSTRACT Recent years have seen marked improvements in silicon solar cell performance.
Standardization of past measurements shows there has been a 42% improvement between 1983
and 1991. The cell structure responsible for the most recent performance increase is
described, as is the potential for further improvement. Accelerated UV and 1 MeV electron
testing of such structures is also briefly described. Recent progress is outlined with the laser
grooved, buried contact cell approach, which appears capable of transferring these
improvements into low cost commercial sequences.

1. INTRODUCTION As a result of early SERI initiatives in the high-

efficiency silicon cell area, several key cell results
The last decade has produced substantial were measured at SERI in the early 1980's. Prior to
improvements in silicon cell performance. Most of 1985, these were at 28'C relative to NASA-
the key results in this development have been calibrated reference cell N265 under a Spectrolab
independently measured at recognized testing X-25 simulator, without spectral mismatch correction.
centres. However, over the same period there has In 1984, SERI began calibrating its own reference
also been refinement and standardization of cell cells using a technique incorporating spectral
measurement techniques. A unified history of the correction (3). A large difference from the NASA
evolution of silicon cell performance is presented, calibration was noted, as recently reported (5).
normalized to present reporting conditions.
From May to September, 1985, SERI reported
The highest performance to date has been measurements under the NASA DN spectrum
demonstrated by the PERL cell (passivated emitter, relative to both NASA and SERI calibrated cells, as
rear locally-diffused cell). This structure is well we under the ASTM DN and global spectra.
described as is scope for improvement. Accelerated From October, 1985, all measurements were relative
UV testing shows that UV stability is not an issue for to the ASTM global AM1.5 spectrum at 25'C
non-concentrating systems. Preliminary results referenced to SERI calibrated cells.
under 1 MeV electron testing are also outlined.
To refer earlier measurements to this standard
The laser grooved, buried contact cell offers spectrum, several factors have to be taken into
scope for transferring recent efficiency account (2). Measurements on UNSW cells in 1985
improvements into commercial production. A show that an average 5.3% reduction in current was
double-sided laset grooved structure, a potentially required to incorporate spectral mismatch correction
low-cost analogue of the PERL cell, is described. and the difference from the NASA calibration.
Conversion from the NASA DN to the ASTM global
2. NORMALIZED EFFICIENCY EVOLUTION spectrum produced a 0.6% current increase for good
cells. A 0.2% decrease would result from the
Over the last decade, both silicon cell reduced measurement temperature, giving an
efficiency and terrestrial cell measurement overall 4.9% decrease in current. Voe would
techniques have evolved considerably. The latter increase a nett 5 mV with fill factor unaltered to
have reached the stage where independent measurement accuracy resulting in approximately
laboratories measure the same result for standard 3.7% reduction in efficiency. Uncertainty in this
silicon cells within 1-2% (1). However, many key conversion process is estimated as 1%, additional to
results have been reported under conditions normal measurement errors.
different from present standards. This section
summarizes a more detailed analysis (2) of these key JPL initially conducted module measurements
results and their normalization to current using reference cells calibrated outdoors. It
international standards. subsequently developed LAPSS simulators with
900d spectral match to the ASTM specified spectra
In the US, NASA took the early lead in cell (6). In the mid 1980's, JPL measurements upon high
measurements. Techniques for AMO cell calibration efficiency silicon cells under the global spectrum
to an estimated accuracy of 1% have been were reported (2). These gave currents about 2%
established for some time (3). Although since shown higher than present standards (2).
that such AMO reference cells can give accurate
terrestrial measurements with appropriate spectral
corrections (3), NASA chose an independent Sandia was next to play an active role in high
efficiency silicon cell measurements. Prior to 1987,
technique for terrestrial cell calibration under a
direct normal (ON) AM1.5 spectrum (4). In this these were under the direct spectrum, reflecting
Sandia's involvement with concentrating
methodology, NASA was designated as the sole photovoltaics. Measurements were referenced to
source of reference cells and supplied cells to many NASA calibrated cell Y49 under a XT-10 simulator
groups, both in the US and internationally. without spectral mismatch corrections. After these
Unfortunately, many years were to pass before these corrections were incorporated, measurements were
terrestrial calibrations were shown to be inconsistent reported relative to either the ASTM DN or global
with the AMO calibrations (5). spectrum, with spectral correction to account for the
different calibration spectrum of the NASA reference

250
cell. Measurements under the global spectrum were finger
generally 2-3% less conservative than earlier "Inverted" pyramids
measurements under the NASA DN spectrum,
depending upon the cell's spectral response.
A recent interchange of UNSW cells between
Sandia and SERI brought to light a much larger
measurement difference than had earlier such
interchanges. To resolve this difference, Sandia
developed its own primary cell calibration
technique. Round robin comparisons of cells
calibrated with this technique have shown no
systematic differences from SERI calibrated cells,
despite the difference in calibration aooroach.
CallbraUon rauo
r'la ' con tact
r-

Figure 2: PERL cell.

II
1.05 SERII Sandia 2
Efficiency above 25% is feasible with this
structure. The fill factor gives the most scope for
I-
near-term improvements. The cells demonstrate the
Sandia I highest level of light-trapping to date with optical
f- path enhancement factors (Z) of up to 26 measured.
This can be improved, for example, by tilted surface

1.00
l-
I , I II, I I SEru 2 , Sand la3
pyramids (8). There is also scope for reduced
surface reflection since only an oxide antireflection
coating is used. Further improvement in Voe is also
1982 1984 1986 1988 1990 anticipated.
Figure 1: Calibration factor allowing conversion of These cells were fabricated solely to
past short-circuit current measurements to present demonstrate improved performance. Nevertheless,
standards. they are surprisingly rugged. Concerns have been
expressed over the UV stability of high-efficiency
Based on the above discussion, Fig. 1 shows silicon cells (9). A PERL cell of efficiency close to
the factor by which past reported currents must be 23% was exposed to 600 hours of 254 nm light from
divided to convert to present standards. The shaded a mercury lamp. Using the acceleration factor
areas indicate uncertainty in this transformation due deduced by Stanford University in similar testing (9),
to different spectral responses. Normal measurement
errors must be superimposed. Table 1 shows the this corresponds to over 1,000 years damage in the
evolution of silicon cell efficiency standardized on terrestrial environment! Cell Voe dropped from 696
the basis of this discussion (or further refined where mV to 615 mV, Ise dropped by 20% fill factor
additional information is available). This table changoed by 2~o, with over~II power output reduced
documents a massive 42% increase in silicon cell by 27 Yo. Heating for 30 minutes at circa 400'C in a
efficiency over the last 8 years. 4% H2/Ar mixture restored performance to
essentially its former level, « 1% reduction in
3. PERL CELLS output). These results are conSistent with a
rever.sible .change in top surface recombination
Figure 2 shows a PERL (passivated emitter, velo<?lty which reduces the blue response of the cell
rear locally-diffused) cell, the highest efficiency and Increases the contribution of the surface to the
silicon cell to date (7). Revised output parameters total recombination current. Under 200 suns
are 695 mV, 41.3 mNcm 2 , 81.0% and 23.3% for Voe, concentration, this would correspond to about 6
Isc, fill factor and efficiency respectively. ye~rs exposure and suggests that stability is more of
an Issue In concentrating systems (9).

Table 1: Reported history of silicon cell improvement (> 1 cm 2 area) normalized to present standard test conditions
(25'C, 1000 W/m 2 , IEC 904-2 global or ASTM E892-87 global spectrum).

DATE REPORTED TEST CONDITIONS CORRECTED CELL DESCRIPTION

EFFICIENCY EFFICIENCY
5/83 17.1% SERI1 16.4% ASEC
8/83 17.1% SERI1 16.4% Westinghouse
9/83 18.0% SERI1 17.3% Spire textured
18.7% SERI1 18.0% UNSWMINP
12/83 19.1% SERI1 18.3% UNSWPESC
5/85 19.8% SERI1 19.0% UNSW PESC
10/85 20.0% SERI2 20.0% UNSW I1g PESC
7/86 20.6% SERI2 20.6% UNSW I1g PESC
4/88 21.4% SANDIA 2 20.8% UNSW I1g PESC
9/88 22.3% SERI2 22.3% Stanford
6/89 23.2% SANDIA 2 22.3% UNSWPERC
12/89 23.0% SERI2 23.0% UNSWPERL
12/90 24.2% SANDIA 2 23.3% UNSWPERL

251
 This overlooks the fact that the above results
were for a bare cell. Cells would not be exposed to
such a high UV level in the field. For example, the
acrylic lenses used in concentrators typically have
a wavelength cut-off of about 360 nm. This would
protect cells from most of the high energy photons
reported to cause surface damage. The low-iron
glass used in terrestrial modules has a cut-off of
about 320 nm. Additionally, the EVA material used
in modules is subject to UV enhanced oxidation if
exposed to photons below 360 nm (10). To prevent
such damage, additives are incorporated which
would similarly shield cells.
Considering there has been no optimization
for this purpose, the cells are also reasonably
rugged under the more severe 1 MeV electron
radiation used in space cell testing. Table 2
compares a high performance PERL cell to
published data for a commercial "high efficiency"
silicon cell (which has been so optimized). The Figure 4: Double-sided laser grooved, buried
PERL cell retains higher power output over the contact solar cell.
whole operating life.
To aid this development, a hybrid cell
Table 2: AMO efficiency of a PERL cell at structure has also been investigated with a laser-
"beginning-ot-lite" (BOl) and "end-of-life" (EOl) grooved front surface and PERC or PERL rear
compared to a typical high efficiency space cell. surface. Vee above 690 mV has been demonstrated
Measurements of PERL cell courtesy of Applied with this structure. High Vee is the most important
Solar Energy CorRoration (EOl determined by prerequisite to demonstrating high efficiency. This
exposure to 10 15/cm 2 1 MeV electrons). result shows that the cruder emitter treatment
compared to PERL cells will not constrain
BOl EOl achievable efficiencies.
CEll
EFFICIENCY EFFICIENCY Work also continues on a combined laser-
PERL(10cm) 20.3% 9.8% textured, laser-grooved sequence for polycrystalline
BSF/BSR (100 cm)" 14.3% 9.8% substrates. Good results have been obtained on
polycrystalline material produced by a range of
• From Reference (11) commercial techniques (14).

4. LASER GROOVED CEllS 5. CONCLUSION

The laser grooved structure of Fig. 3 Silicon cell performance has improved 42%
since 1983, when past measurements are converted
incorporates most features of the UNSW mg PESC to present standards. Refinement of the present
cells of Table 1. Similar efficiencies in the 19-20% PERL cell is expected to result in further near-term
range have been demonstrated for large cells with efficiency increase.
sequences which appear suitable for large-scale
commercial production (12,13). Such a sequence UV stability is not an issue for PERL cells
was used by Telefunken to fabricate the cells for the under non-concentrated sunlight. Even without
Swiss car "Spirit of Biel/Bienne" which won the optimization, acceptable rerformance was
recent World Solar Challenge, a solar car race from demonstrated after hundreds 0 years of equivalent
Darwin to Adelaide in Australia, by over 400 km. field exposure of bare cells. Actual cells would be
The array efficiency on this car was a remarkable further protected by UV absorption in the low-iron
17%. glass used in modules or by the UV absorbers which
protect encapsulants. Again, without optimization,
the cells are reasonably tolerant to 1 MeV electrons.
laser grooved, buried contact cells, now
being evaluated commercially, hold the promise of
transferring recent efficiency improvements into low
cost commercial production. A double-sided
grooved structure is expected to produce similar
efficiencies to the best silicon cells to date at much
lower cost.
6. ACKNOWLEDGEMENTS
The author acknowledges the contributions of
other members of the Centre for Photovoltaic
Devices and Systems. In particular, he
acknowledges contributions of Aihua Wang and
Jianhua Zhao in the fabrication of PERL cells, of
metal Adele Milne, Mike Willison and Jianhua Zhao in the
Figure 3: laser grooved, buried contact solar cell. UV and MeV electron testing, of Chee Mun Chong,
Fuzu Zhang, Jianhua Zhao and Fei Yun in the
Present work is directed at incorporating into refinement of laser grooved cells and Stuart Wenham
the structure the improvements to the rear of cells for contributions across all areas. He thanks Carl
demonstrated in UNSW PERC and PERL sequences. Osterwald and Keith Emery of SERI and David King
The cell structure being investigated is shown in Fig. of Sandia for cell measurements and for information
4. This has now been implemented with consequent relevant to the normalization scheme described. The
Vee and Jsc advantages confirmed. support of the Energy Research and Development

252
Corporation of Australia and Sandia National
Laboratories is gratefully acknowledged. The
Centre for Photovoltaic Devices and Systems is
supported by the Electricity Commission of New
South Wales and the Australian Research Council
under the Special Research Centres Scheme.
7. REFERENCES
1. J. Metzdorf et alia, "Objectives and Results of
the PEP '87 Round-Robin Calibration of
Reference Solar Cells and Modules", Conf.
Record, 21 st IEEE Photovoltaic Specialists
Conference, Orlando, 1990, p. 952.
2. M.A. Green, in preparation.

3. K.A. Emery et alia, "A Comparison of

Photovo Itaic Calibratio n Methods", 9th
European Photovoltaic Solar Energy
Conference, Freiberg, Sept., 1989, p. 648.
4. "Terrestrial Photovoltaic Measurement
Procedures", Report ERDA/NASA/1022-
77/16, June, 1977.
5. C.A. Osterwald, K.A. Emery, D.A. Myers and
R.E. Hart, "Primary Reference Cell
Calibrations at SERI: History and Methods",
Cont. Record, 21 st IEEE Photovoltaic
Specialists Conf., Orlando, 1990, p. 1062.
6. Flat-Plate Solar Array Project Final Report,
Volume VI: Engineering Sciences and
Reliability, JPL Publication 86-31, October,
1986, p. 36.
7. A. Wang, J. Zhao and M.A. Green, "24%
Efficient Silicon Solar Cells", Appl. Phys.
Lett., Vol. 57, pp. 602-604, 1990.
8. P. Campbell, S.A. Wenham and M.A. Green,
"Light Trapping and Reflection Control with
Tilted Pyramids and Grooves", Conference
Record, 20th IEEE Photovoltaic SpeCialists
Conf., Las Vegas, Sept. 1988, p. 713.
9. P.E. Greunbaum, R.A. Sinton and A.M.
Swanson, "Stability Problems in Point Contact
Solar Cells", Conf. Record, 20th IEEE
Photovoltaic Specialists Conf., Las Vegas,
1986.
10. Flat-Plate Solar Array Project Final Report,
Volume VII: Module Encapsulation, JPL
Publication 86-31, October, 1986, pp. 47-52.
11. Applied Solar Energy Corporation, Space
Products Data Sheet, 4 pages plus cover,
undated.
12. H.W. Boller and W. Ebner, "Transfer of the
BCSC-Concept into an Industrial Production
Line", 9th European Photovoltaic Solar
Energy Conference, Freiberg, Sept., 1989, p.
411
13. N. Mason et alia, 10th European Photovoltaic
Conference, Lisbon, April, 1991.
14. J.C. Zolper, S. Narayanan, S.A. Wenham and
M.A. Green, "16.7% Efficiency, Laser
Textured, Buried Contact Polycrystalline
Silicon Solar Cell", Appl. Phys. Lett., Vol. 55,
pp. 2363-2365, 1989.

253
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

HIGH-EFFICIENCY AND LOW-COST OPTION IN THE DEVELOPMENT OF CRYSTALLINE SILICON SOLAR CELLS

Yutaka Hayashit. Tadashi Sai tohU. Kunio Hane***. Ryuichi Shimokawa*

and Toshihide Takeshita****

Electrotechnical Lab .. 1-1-4 Umezono. Tsukuba-shi. Ibaraki 305. Japan

** Tokyo Univ. of Agric. and Ind .. 2-24-6 naka-machi. Koganei-shi. Tokyo 184. Japan
*** Keio Univ .. 3-14-1 Hiyoshi, kohoku-ku. Yokohama-shi. Kanagawa 223. Japan
**** Technova Inc .. 2-2-2 Uchisaiwai-cho. Chiyoda-ku. Tokyo 100. Japan

ABSTRACT. Activities of Japanese Crystalline Solar Cell Committee on the

development of crystalline silicon solar cells under the "Sunshine Project"
are summarized. Elemental technological requirements for higher efficiency
are extracted. and approaches and results obtained for them are introduced.

1. I NTRODUCT I ON point-contacted back electrode (i terns

7.9 and 10) on the efficiency was demonstrated by
To support scientifically and technically na- Prof. M. A. Green by his 24.2% PERL cell (I).
tional programs on the development of crystalline In addition to the above elemental cell-design
solar cells. the Japanese Crystalline Solar Cell and fabrication technologies for high efficiency.
Committee was founded in Technova Inc. in 1986 by low cost silicon material. crystallization and
the financial support of New Energy Development Or- slicing technologies were investigated.
ganization (NEDO) under 'the Japanese "Sunshine
Project". The Committee initially covered silicon
material. crystallization and cell fabrication 3. OBTAINED RESULTS
technologies. and the area of compound semiconduc-
tor solar cells was included later under the survey Obtained results described below are not all
of the Committee. Three subcommittees for silicon due to committee activities but rather mainly due
materials. silicon cells and compound semiconductor to individual efforts of the committee member and
cells. respectively. were made under the main com- his colleagues in his company. university or in-
mittee as shown in Fig. 1 to facilitate further stitution financiallY supported by NEDO under the
detailed investigations and discussions on ex- Sunshine Project. However. scientific insights and
perimental results. In this paper. activities. and technologies obtained with reference to the com-
obtained results on silicon solar cell technologies mittee activity will be introduced here with out
in the Committee and Sub-committees are introduced. respectively mentioning the above situation.

3.1 Low Cost Technology with Higher Efficiency Op-

2. ELEMENTAL CELL TECHNOLOGIES EXTRACTED tion
Electromagnetic(EM) casting of poly-
In the DOE/JPL-MITI/NEDO High Efficiency Semi- crystalline silicon (fig. 3) revealed to be a hope-
nar. which had been originally scheduled as one of ful low cost technology. This technology eliminates
activities for the Joint Coordination Meeting of waste of an expensive crucible and enables growth
the Photovoltaic Energy Project originated by the of far longer ingots than conventional cast tech-
Versailles Summit and was jointly held at 7th nologies. resulting reduction of crystallization
European Photovoltaic Solar Energy Conference and cost to a half of that of the conventional cast.
Exhibition. early phase activities of the Committee Reduction of wafer thickness from conventional 450
and future prospects for higher efficiency were in- p. m to 200p. m by improving a wire saw technology
troduced using a schematic diagram shown in Fig. 2. further reduces wafer cost down to about a half of
The cell model shown there for the efficiency the conventional one and results economy of silicon
higher than 20% includes: material which is still expensive. Development of
1. front surface texture. these two new technologies are now under way (3)
2. front surface passivation. and will reduce the cost of crystallization and
3. optimized emitter surface concentration. slicing from about 200 yen/W near to 50 yen/Was is
4. fine front electrode. shown in Fig. 4 (2).
5. front electrode passivation by EM casting has advantage of no impurity incor-
a. point contact poration from the crucible but suffers from neces-
b. deep and/or high surface cone. junction un- sity of careful control of cooling and temperature
der the contact gradient to obtain a ingot with less crystal
c. hetero-face or junction (common to item defects and stress. If we succeed in this heat con-
2. ). trol. a higher Qual i ty ingot is obtained. because
6. thin base (far thinner than a minority carrier remarkable reduction of impurities as shown in Pig.
diffusion length). 5 was confirmed by recent analysis of impurities in
7. back surface passivation. the EM cast silicon (3). MBIC (4) and EBIC (5)
8. and/or back surface field diffusion. diagnosis on EM cast cells showed less grain bound-
9. back electrode passivation by ary effects. Reduction in cell thickness also has
a. point contact an option of improvement in the conversion ef-
b. back-surface diffusion under the contact ficiency. if the cell is properly designed. Device
c. hetero-face. physics suggests higher conversion efficiency with
10. back reflector. a thinner cell with optical confinement as shown in
11. back surface texture (option) Fig. 6 (6). Thus both the EM casting and thin slic-
Up to that time. items 1. 2 and were employed in ing gives us potential high efficiency options.
high efficiency cells. Recently. the effect of a

254
3.2 Elemental Cell Technologies important parameter. It rules short wave length
3.2.1 Surface texturing and back Reflection photo-current over wide efficiency range and open
The conventional surface texturing technology circuit voltage in high carrier lifetime or thin
widely applied to single crystal silicon cells base cells. Optimum design on emitter junction
utilizes anisotropic etching with crystal orienta- depth is necessary for the high front surface
tion dependence. However only a part of the surface recombination case as shown in Fig. 13.
of the cast silicon wafer can be textured by this c) When a BSF process is appl ied to the cell. back
technology. because usually. crystal orientation of surface passivation is necessary for the cells with
each grain in poly-silicon material is rather ran- a carrier lifetime higher than several hundred
dom. Texturing of the cast silicon surface by micro-second.
mechanically grooving and chemical etching as shown d) Electrode recombination should be considered
in Fig. 7 (7) or by photolithographical patterning only for super high efficiency cells with more-
and etching as shown in Fig. 8 (8) showed effec- than-several-hundred micro-second carrier life time
tiveness in the reduction of optical reflection and as shown in Fig. 14.
improvement of the conversion efficiency. The next
step is to seek lower cost fabrication tech-
nologieshigher through-put processes. One of the 4. SUPER HIGH EFFICIENCY POSSIBILITY
candidaf~s might be a textured anti-reflection Further investigation on the EM cast technol-
coating such as textured S~02 or Z~O as shown in ogy indicated a low cost FZ or cold crucible CZ
Fig. 9 (6). The figure shows a passivated textured single crystal technology. And also results from
anti-reflection coating and a passivated back the computer simulators indicated possible paths
reflection structure for possibly improving opti- for super high efficiency up to 28% as shown in
cally and electrically the conversion efficiency. A Fig. 12. Some new design principles to break a
conventional back electrode directly contacting and limit shown in Fig. 12 are born among the Committee
alloying silicon back surface loses it's back members but it is not yet the time to disclose
reflectivity. Principle shown in Fig. 9 is also ap- them. These results will be helpful for the super
plicable high efficiency single crystalline solar high efficiency project under the Sunshine Project
cells. to started last year.

3.2.2 Surface Recombination and Passivation

EarlY phase discussion in the Committee sug- 5. ACKNOWLEDGMENT
gested the necessity of detailed investigation on The paper is summarized out of results ob-
surface recombination velocity with reference to tained through the activities in the Japanese Crys-
surface states densi ty as it' s origin and surface taline Solar Cell Committee. Authors are also in-
potential. A numerical computer simulator for the debted for their financial support and guidance to
calculation of the surface recombination velocity Director T. Goto and his staffs in the Sunshine
from a given surface state density vs. energy Project Promotion Head Quarter. AIS!. MIT!. Japan.
profile was developed and predicted quantitatively and Mr. T. Ohno. Director General and Mr. T. Ohyu.
relation between the surface recombination velocity director of Solar Energy Division of NEDO. Two (Y.
and the surface state density with surface poten- Hayashi and R. Shimokawa) of the authors are grate-
tial and/or surface impurity concentration as shown ful to Drs. K. Sugusaki and T. Tsurushima for their
in Fig. 10 (9). A dual mercury probe method has support and guidance.
been developed to measure bulk lifetime and the
surface recombination velocity of cells or wafers
separately. Fig. 11 shows the measured surface REFERENCE
recombination velocity of (100) n-type silicon sur- (1)J. Zhao. A. Wang and M.A. Green. "24% Efficient
face as a function of gate bias of transparent gate PERL Structure Silicon Solar Cells". The Conference
MIS structures (10). Suggested b,y these theoretical record. 21st IEEE Photovoltaic Specialist Con-
and analytical works. surface passivation by ions ference. Kissimmee. Fla .. 1990. p.333.
embedded in a surface insulator is propoSedll1J. (2)T. Kobayashi. Y. Hayashi. N. Yui and K.
Yoshimura. "Japanese Photovoltaic R&D Program under
3.2.3 Thin Cell Structures the Sunshine Project". Proc .• Nineth E. C. Photovol-
Substrate thinner than 200-S00Jl m could not be taic Solar Energy Conference. Freiburg. 1989. p.
handled in conventional automated cell fabrication 1206.
lines. Contractors of the Sunshine Project are (3)K. Kaneko. T. Misawa and K. Tabata. "Cold
using the mechanical supporting frame structures Crucible Induction Casting of Semiconductor Silicon
shown in Figs. 7-9. for Solar Cell". The Conference record. 21st IEEE
Photovoltaic Specialist Conference. Kissimmee.
S. 2.4 Fine Front Electrode Fla .. 1991. p. 674.
Fine front electrode technology is effective (')H. Yagi. R. Shimokawa. T. Saegusa. k. Matsukuma
for reducing shadowing effect in the cast cells and and S. Kokunai. "High Performance. Cold Crucible
also important for increasing open circuit voltage Cast Silicon Solar Cells Using Hydrogen Ion
in a super high efficiency design. A fine electrode Implantation". Technical Digest. International
technology is being developed for the electrode PVSEC-5. Kyoto. 1990 p.311.
with less than 50Jl m by using a new printing tech- (5)K. Yasutake. A. Takeuchi. K. Yoshii and H.
nology. Kawabe. "Electrical Activity of Grain Boundaries in
Polycrystalline Silicon". ibid. p.307.
3. 2. 5 Computer Simulators (6) Y. Hayashi. K. Hane. R. Shimokawa and T.
ID (12) and 2D (13) computer simulators were Saitoh. " Design Option for a Crystalline Silicon
developed for the Committee to see what is an im- Solar Cell". ibid. p. 323.
portant technology to realize a respective level of O)T. Nunoi. S. Okamoto. K. Nakaj ima. S. Tanaka. N.
the efficiency and diagnose if the device and Shibuya. K. Okamoto. T. Nanori and H. I toh. "Cast
process designs are reasonable. The programs are Polycrystalline Silicon Solar Cell with Grooved
open to the members of the Committee. Surface". The Conference record. 21 th IEEE
General implications obtained by the 2D Photovoltaic Specialist Conference. Kissimmee.
simulator are as follows: Fla .. 1990. p. 664.
a) Bulk recombination lifetime is the paramount (8)K. Shirasawa. H. Yamashita. K. Fukui. M.
parameter ruling the conversion efficiency at a Takayama. K. Okada. K. Masuri and H. Watanabe.
fixed cell structure as shown in Fig. 12. "Large Area High Efficiency Multicrystalline
b) The front surface recombination is the second Silicon Solar Cell". ibid. p. 668.

255
 (9)T. Saitoh and H. Hasegawa, "A Computer Simula-
tion of the Recombination Process at Semiconductor 1~'1.
200
Surfaces", Extended Abstracts, 1990 International
Conference on Solid State Devices and Materials,
Sendai, Japan, 1990, p. 155.
(JO)E. Suzuki, H. Takata, K. Ishii and Y. Hayashi,
~150
~
--_
---
~50 m 15'1.

"Measurement of the Surface Recombination Velocity § 100

S at the Si-Si02 Interface by the Dual-Mercury
Probe Method", ibid , p. 287. 50
E. Suzuki, H. Takata, K. Ishii and Y. Hayashi, auclble-lu5,20'I.
"Evaluation of the Si-Si02 Interface by the
10 50 100 1000
Measurement of the Surface Recombination Velocity S producllon scole 1MWp/Y I
by the Dual-Mercury Probe Method" , J JAP , Vol. 29,
12, 1990, pp. L2300. Fig.4 Scenario of cost reduction in crystallization
(Jt)T. li'arabisako, T. Uematsu, K. Kanda and S. and slicing process by the electro_agnetic casting
Iida, "Field Effect Passivation of Silicon Surface and thin slicing technologies.
by Photochemical Si02 CVD", Technical Digest, In-
ternational PVSEC-5, Kyoto, 1990, p. 583.
(12)C. Amano, A. Sugiura, M. Yamaguchi and K. Hane,
" Fabrication and Numerical Analysis of AlGaAs/GaAs I 0' .
Tandem Solar Cells with Tunnel Interconnections",
IEEE Trans. Electron Devices, Vol. 36, No.6, pp.
L- t -
1989. .1. ~
(J3)K. Hane, M. Ida, T. Uematsu, T. Saito and Y.
Hayashi, "Performance Simulation of High Efficiency
Silicon Solar Cells with a Buried Channel Emitter",
u
Proc., 4th International Photovol taic Science and tOil
Engineering Conference, Sydney, 1989, p. 713.
o
I
Main Coromi ttee
I 0 It I
(chair: Y . Hayashi) o 1.1

(secretary: R.Shlmokawa)
'"Ol P.llla. Rut (mm/m In)

Silicon Mater1al Cell Technology Compound Cell

Sub-committee Sub-comml ttee Sub-comml ttee Fig.5 Carbon and Olygen concentration in solid
(chalr:N,Sano) (chair:T. Saltoh) (chair :M.Konagal) silicon vs. ingot ~uliing rate.

Fig.! Organization of the Japanese Crystalline

Solar Cell Committee.

,-....
f'OlNT COifTAt'
'''K!'AC'
'ASSIV"TIIC* lA'rUf5J - 25
.,e
0:-

li-
e:
'"
~ 20
c
o
.§

~ perfect optical confinement

..
f'OtIIr CClNfACf ·
'" N,SSI"hOIll ·
nrolNlO£
15 randomly textured surface

Fig.2 Schematic diagram of a high efficiency solar 10 20 30 4050 100 200

cell (originally presented at 7th E.C. photovoltaic cell thicl<ness , d (I'm)
Solar Energy Conference and Exhibition).

Fig.6 Efficiency versus cell thickness, assu.ing

front and back surface reco.binatlon velocity S-O
or 10"c~/s and Sr=10"cm/s, and 100~ back o~ti~~1
reflection.

Soil dl fl eel lnaol

v~ Valve

Fig.7 Schematic diagra. of a .echanically grooved

Fig.3. Schematic Cross sectional diagram of an polycrystalline silicon solar cell with a support-
electromagnetic casting equipment. ing frame structure .

256
 tII-· ..... '0
~~~~~~~~~iiii~~; or.HT'JlIlPUCTIOfII "IUI
, .. - LAl'U ''''':a''tN)
r n ..... UPTIlATI:
.. - U.TU

~2i~~~~~~~~~~~~~~.'- "~ cell thickness' 50JLm

XI" O.5JLm, Xl' ' 3JLm

15~--~--~--~--~
Fig . 8 Sche.atlc diagral of a polycrystalline 2JLs 20JLs 200JLs 2ms 20ms
silicon large area cell with photollthographically
patterned grooves and a supporting frame structure. minorily corrler lifetlme, T

Fig . 12 Efficiency vs. bulk carrier lifetime with

surface recombination velocities as paraleters.

r--------------25

'11
..........................................
24

supporting 45 700 23
frame JSCI
E44
Fig.9 Proposed thin cell structure with a pas-
...
~

.5 43 .5
> VOCI
------------- ---

. . . ... . !
sivated textured ARC and a passivated back reflect- ...,:J! '12
ing electrode and also a supportin.g frame struc- ........ :;,-::-.:- _--- 22 i-
--42 .;
ture. ~ ""
.2
~4t ~
:; ."§
u o
'54 '5600 T -2001'S 21 .~

~39 ...
c: d • 50l'm
'" 0 1 : 5, -0,5, -104 cm/s §
2 : 5, - 5, • 10· cm/s
38
0.1 0.2 0.3 0.5 1.020
emit'er junellon dep'h, x" (I'm)

Fig . 13 Efficiency dependence on emitter junction

depth . High front surface reco.binat ion needs care-
ful oPtimizati~n of the junction depth.

~ 10·', - - - - - - -- - -_ _ ,

-3 -2 -1 o I
0 .. ( X to"~m~ )
2 3 4 i
I/)
• Ate£ALED
IN 0 ,
'. - ANNEALB>
\, IN H,

Fig . IO Calculated surface reco.blnatlon velocity

using a surface state density distribution accord-
.
ing to the DIGS model.
I
....
~ ),I Ot----===:::--====_ T - 201'5 .
.:
c: ~ T-200p5
g.t!..-20 T"2ms
,
C!S' d-:iOpm T'20ms
-=1
g._ -40 5,-5,-0 ,
-15 -10 ·5 0 5 10 15
.,g :~
....... GATE BI....S VOlJ'AGE Vg ( V)
0.00075 0.0024 0.0074 0.024 0.074
eleclrode widlh rolio Fig.l1 Measured surface recombination velocity as a
function of surface field for high and low surface
Fig.14 Influence of electrode recombination on Voc. state density samples.

257
 POSTER SESSION 2A

Crystalline Silicon - Materials and Devices

 WTH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-12 APRll. 1991 LISBON, PORTUGAL

ELABORATION OF PHOTO VOLTAIC SILICON BY A

PLASMA PROCESS AND ITS CHARACTERIZATION
P. HUMBERT, R. COMBES, J. ERIN, N. MADIGOU, D. MORVAN, J.
AMOUROUX
L1boraloirc des rCacleurs chimiques en phase plasma
E.N.S.C.P., 11 rue Pierre el Marie CURIE
75231 Paris Cedex 05 France

ABS1RAcr
This paper presents the results of the thermal plasma process which makes possible the purification of
metalurgical grade silicon and off grade electronic material ("n" and "p" type) coming from electronic and photovoltai'c
industry (1).
The pilot of the laboratory operating at 25 kW was made in order to determine the energetic efficiency of an
industrial plasma plant (100 kW). The heat balance of the reactor and particularly those of the metallic water cooled
tnrch was made in function of the different e"perimental parameters.
In order to improve the mechanism of the purification, we investigate systematically and e"perimentaly the
efficiency of the alkaline fluoride when used as synthetic slag (CaF2. MgF2. BaF2). A thermochemical analysis takes
into account the electrochemical reactions of the impurities (B. P) with fluoride and oxygen of the plasma at a given
temperature and gives us a better understanding of the elimination of those dopants in the form of molecular species
(BOF2. BOF. POF) ... ).
The photovoltai'c properties of the end material are measured in connection with Photowatt Company. The
physical datas (Jsc. Ld. N. efficiency) of the silicon indicated that even if the dislocation content is very high (more
than )()6cm -2) the diffusion length is very high (until 70 !lm). This is due to the fact that the silicon obtained il an
ultra pure one and perfcctly passivated by the use of an argon thermal plasma containing 1% of hydrogen.

1. SCALE-UP OF II IE PROCESS (25 kW> :

TIIC cxperimental torch we used previously for plasma

j(cneratinn consists of a quartz tube and thus, such a
---
gas supply

technology is not adaptcd for high power requirements.

TIII:n the scale-up of the process at 100 kW requires the
use of II water-cooled copper finger torch. Nevertheless.
some difficulties have been encountered with this new
apparatus. In one hand. it has been necessary to realize a
device for starting the plasma and on the other hand, we
tried to optimize the torch design and the gas flowrates in
order to decrease the heat losses in the torch and thus to
obtain a good energetic efficiency.

1.1 The pilot reactor: crucible

Figure 1 shows a shematic diagram of the new reactor.
silicon bar
The apparatus consists of the cooled torch in the upper
part and of a water-cooled cylindric vessel in the lower
\,3l:UUIll pump
part. This vessel is connected to a vacuum pump so as to
decrease the pressure to 1 mbar during the plasma starting
step. As soon as a "cold plasma" is generated, the
prcssure is increascd and tlle low prcssure plasma leads to
a thermal plasma at atmospheric pressure.
The torch can be operated with one or two gas
supplies (an inner gas for plasma generation and an outer Figure I : plasma pilot installation for silicon
gas for limiting tllermal losses on the wall of the water- purification (25 kW)
cooled torch). The chemical composition and the fioWfate
of each gas supply can be choosed independantly (Figure
2). A water-cooled injector can be added in the torch if the _ TIle water cooled metallic torch can be operated without
case of powder treatment. damage at low gas flowrates but the efficiency is rather
Because of the power limitation of our R.F. poor. The increase of gas flowrate pro~ides a ~etler
generator, we performed e"periments at 25 kW. but the efficiency but it has been shown that there IS an oplimum
pilot reactor has been tested previously at higher power gas flowf8le (Figure). This optimum flowrate depends
levels (HO kW). on several parameters (the power of the plasma. the gas
composition and the gas velocity repartition inside the
1.2 Energetic efficiency of the installation: torch).
TIle aim of this study is to improve the energetic _ The addition of hydrogen in the outer gas reduces the
heat losses. Moreover. since the thermal conductivity of
efficiency TJ (ratio of heat transfered to the silicon bar lhe plasma is known to be very sensitive to the hydrogen
under tllC plasma to the total electrical power delivered by contcnt, we studied tlle influence of the composition of
the network). Thus. we carried out calorimetric lhe plasma gas on the melting kinetic of the silicon bar.
measuremements of the heat loss in the cooled torch and The addition of 5% of hydrogen in argon allows the
of the heat content of the plasma versus operating
conditions. The main results arc the following:
mcllin~ of bars of section up to 10 cm 2 .

261
 The introduction of a fluoride in this medium contributes
to the dissolving of ionic compounds. An electric field
appears between the plasma which is at a floating bias
inner ga~ and the water cooled copper crucible where the sample is
laid.
The purification occured in two different ways:
oUler gas dissolution into the slag or evaporation if the compound
obtained is a gas.
water The simultaneous action of fluoride and oxygen
-+
coming from the plasma makes possible the formation of
oxyfluoride whose stability depends on the activity of
oxygen in the liquid drop and the temperature.
The redox reactions between reagents in our silicon
(0, P, B, F) produce oxides, fluorides and oxifluoride. The
thermochemical data makes it possible to calculate the
exact conditions under which they arc stable. With this
information we can determine what chemical species are
necessary to eliminate boron and phosphorus.

1.2 ANALYSIS OF THERMOCHEMICAL DIAGRAMS

inner ga~ Plasma oxygen acts as an oxidant and a base. In our
medium. it appears as the anion oxide 0 2 - and leads to
Figure 2: diagram of the water-cooled torch the oxidation of fluoride slags (CaF2' BaF2 and MgF2)
which become BaO, MgO, CaO. But the formation of
- The efficiency of the water cooled torch is around 50%. these oxides depends on the bias of the medium and on
This is quite a good result if we compare it to quartz torch the smount of oxygen.
efficiency. The medium we study is very complex. We have to
consider not only the diffusion of oxygen and reaction
Efficiency " products in the fonn of gas in the liquid, but also the
effect of the kinetics reactions and the correspondig
0,5 themodynamic data. At this point we have limited our
research to the study of the thermochemical
characteristics of the medium. We have choosen to show
what happens ar 2500 K with one fluoride (CaF2, BaF2 or
MgF2)' We plolled the relation-ship between the bias of
ionic medium and the oxidant concentration expressed as
p(RaO)=colog (activity of BaO).
The limits of the diagrams are determined by 02/0 2-,
• 5000 V
0.2 •a 6000 V
7000 V
MF2/F2 and M2+/M (M is the cation of the fluoride)
which are equilibrium straight lines. This means that no
• 750QV reaction nceding 0 2 - could take place above the 02/0 2 -
0.1 cquilibrium straight line and that no reaction needing
M F 2 occurs outside of the stability range of this
coumpound.
We havc considered the following compounds for
boron: B02, nOF, BOF2 and BF3.
P.
Figure 3: energetic efficiency of the torch versus M
plale voltage
(one flow, gas: argon)

In conclusion, the cooled-copper fingers torch is an

effective tool for silicon purification in a pilot
installation because its lifetime is quasi-unliinited and it
can operate at high power level with a quite good
efficiency.

BOF 6F~
2. THERMOCHEMICAL ANALYSIS OF THE
EXTRACDON OF BORON AND PHOSPHORUS OXIDES.
aUORlDE AND OXYFLUORIDE
B
12) " , .

1.1 INTRODUcnON Figure 4: Stability diagram of boron species in MgF2

Previous experimental studies (2) pointed out that the at 2500 K. Activities (except a(MgO» are taken equal to 1
usc of fluorides (CaF2, BaF2 or MgF2) could improve and partial pressurcs are equal to 1 atmosphere. Spotted
plasma purification of silicon and, particularly, the lines, equilibrium lines of fluorine F-/F2 (F- in the form
extraction of doping elements (B, P). At the surface of ionized MgF2: Mg2+ +2P- in the liquid drop and F2 in
temperature of the molten silicon ingot (2500 K), the
the form of gas, above the liquid drop).
viscmitv of silicon is the same as that of water (10- 3 PI).

262
 Il nOF is the species that we have to looks at in order
M to measure the efficiency of the slag fOT silicon
purification. We notice that in MgF2 , the stability range
of ROF is smaller than in RaF2.
We have shown diagrams obtained for the
phosphorus species in BaF2, CaF2 and MgF2 in the same
way we have done for oxides. fluorides and oxinuoride for
boron.

----
M

B - ..
~
2 J.' 61.
Figure 5: Stability diagram of boron species in BaF2
at 2500 K. Activities (except a(BaO» are taken equal to 1
and partial pressures are equal to 1 atmosph~re. Spotted
line~, equilibrium lines of fluorine F'/F2 (F' in the form
PO
of ionized RaF2: Ra2 + +2F' in the liquid drop and F2 in
the form of ga~, above the liquid drop).

We noticed that the stability range of B02 and BOF2 Figure 7: Stability diagram of phosphorus species in
arc above the equilibrium straight line of 02/02-. B02 MgF2 at 2500 K. Activities (except a(MgO» are taken
equal to I and plU'tial pressures are equal to 1 atrnosph~re.
needs 0 2 , for its formation and this compound has to be
Spotted lines. equilibrium lines of fluorine F-/F2 (F- in
disregarded in the interpretatioil. But BOF2 can be formed
from ROF. the form of ionized MgF2: Mg2+ +2F' in the liquid drop
It seems clear that it's necessary io create BOF in and F2 in a foml of gas. above the liquid drop).
order to make possible the detection of boron as an
atomic entity by means of emission spectrometry (3) I!
during the silicon ingot treatment. nlis assumption is (VI

confirmed by J. Radic-Peric (4) who worked on air

plasma~ in the presence of boron and fluorine. She V..

----
calculated the variations of boron OXygc;fl and nuorine
species contents in function of the temperature. Bo F~
POF~

-6

-8
p
- 0 BF p ( &a.o)
BFO
-12
Figure 8: Stability diagram of phosphorus species in
-11. BaF2 at 2500 K. Activities (except a(BaO» are taken
BF2 equal to I and partial pressures are equal to I atrnosph~e.
~ -16 Spotted lines. equilibrium lines of fluorine F'IF2 (F- in
.2 BFi
the fonn of ionized BaF2: Ba2 ++2F- in the liquid drop and
-18 BFiO F2 in a form of gas. above the liquid drop).
Phosphorus has a higher number of valencies than
-20 boron. This leads to the formation of a higher number of
oxides. nuorides and oxynuorides. The 02/0 2 -
-22 equilibrium straight line is. for MgF2. below the
stability ranges of POF3. P02 and P401O. So. those
-24 compounds couldn't be formed.
TIle oxide P4010 couldn't be formed because its
-26 stability range is above the 02/0 2- equilibrium straight
line.
2000 4000 6000 In comparing the phosphorus diagram in MgF2 and
T (K )
in BaF2. we noticed that the stability range of POF3
Figure 6: Complex equilibrium calculation in air shifted towards lower oxygen concentration (higher
plasma in the presence of boron and nuorine. p(MO». So. we have to adjust the oxygen concentration
in silicon and in the plasma. The compounds we formed
with our experimental conditions are listed in the table

263
(Ar/l % 1Iz/0,1 %02 plasma, p(MO)>4, temperature of the determine the nature of the compounds which are produced
surface ingot: 2500 K). in the liquid drop. This clearly explains the influence of
Oxlooll, nuorkkt and
oxygen and fluorine on the elimination of boron through
the formation of stable BOF molecule, Consequently. the
~
o,,"yOuorideJ Oxides, Duorldes and
or Jito.phorus oxyOuorldo of bar...
SilK
best conditions of oxygen and fluorine contents could be
p PO 11 0, I'Ol Pri Pl's POI~ B IIOz BOP BOPz DIS d)()()sen for a given boron contents.
RaP, X X X X X X X X X X X
3 CIIARACrERIZATJON OF TREATED SILICON.
c.~ X X X X X X X X X X X
/¥IgP, X X X X X X X X X 3.1 Introduction
Table I: Boron and phosphorus specIes that can be Minority carrier lifetime evaluation is important for
formed in our plasma conditions (Ar/! %H2/0,l %02, photovoltalc properties determination particularly for
surface ingot temperature:2500 K) the efriciency of photocells elaborated from this
material. It could be calculated (5) and thc calculation
2.3 EXPERIMENTAL STUDY OF PHOSPHORUS AND shows that it incre:lses when dopant and recombination
BORON ELIMINATION centres decreases. In fact, it depends on doping element
(6) and cristal elaboration As we have shown previously
IngoL~ have been treated with a plasma (30 I.mn-I Ar (3) the plasma process reduces the dopant content. In
+ 1% H2 (0,3 I.n1O- I ) + 0,1% 02 (30.to- 3 I. mn- I ). The order to determine the specificity of plasma samplcs. we
power was 7 kW. decided to measure the resistivity. the dislocation density
Experiments have been done on monocristaline photovoltaYc properties and diffusion length.
silicon "p" type (r=18 ou 31.to- 1 W.cm) or "n" type (r=
O,56()W.cm). All of the sample "p" type came from the 3.2 Experimental results
same monocristaline bar. Our aim was to compare the We chose to measure the evolution of two raw
effects of each slag (BaF2' CaF2 and MgF2) on the materials:
purification of "p" type silicon. The results are listed on N-type material (P doped at a concentration
the table. >10 16 al/cm3 )
Sil icon p type P-typc material (B doped at a conccntration
Monoa1Jtaline .ilicon > 10 18 al/cm 3 ).
Nature ufibe 11".- polycriJlalinc monoailtaline
n type p=O.562 a.an
1"'3J.10 3 n.cm p~18.103a.cm The photovoltai'c efficiency of the photocells (without
p~. ~ ....an
Without ~oO.I-O.25. poO.04-o. t 1. ptype antireflecting layer)
aan a.an photo efficiency 7,91 % increases with the resistivity (Fig.9).
MRl'z poO,03-0,04S.
nan
c.~ p~-().oz8-o,075
n.an
lI'lj p"O,38"'{).S3. ~,l-O,4.
ll.cm nan

Table 2: Resistivity measured before and after plasma

treatment on "p" and "n" type silicon ingots

Final resistivities are highcr whcn we us cd BaF2 for

thc slag. So, between the three slags. BaF2 is the beller
one. Anlysis of thermochemical diagrams showed us that
BOr is at the range of 1'(02 -)=1,42; 2.54; 3.85 () 2 3 4
respectively in MgF2, CaF2 BaF2. So, we can assume Resistivity (Ohm.em)
that in the conditions unde; which we performed our Figure 9: Efficiency versus resistivity for samples
experimcnts. in MgF2. we couldn't form BOF. However. tented by plasma
with CaF2, we are in a boundary zone. which makes the
elimination bf a certain quantity of boron possible in the Wathevcr is the nature of !lIe raw material. the optical
form of BOr. With BaF2, a much large quantity of boron microscopic observation after plasma treatment revealed
is eliminated. also in thc form of BOF. So, the p(02-) twins, grain boundaries and high dislocation density
area in the liquid drop is between 2,54 and 3,85. close to 106 _10 7 at/cm-2 in dislocated grain. (photo I).
The purificntion of "n" type silicon, phosphorus
doped is very efficient, We observe a type inversion and
a good photovoltai'c efficiency. This lead us to believe
that all the phosphorus was eliminated and no boron was
left.
In the absence of fluoride, we can only obtain oxides.
Nevertheless, phosphorus leads to the formation of a
large number of oxides (PO, P406, P02 and P40to) than
boron (B02) docs.
According to the thermodynamie calculations, the
elimination of phosphorus will be much easier than that
of boron,

24 CONCLUSION
In the condition under which we performed the
experiments. the thermochemical anlysis help us to
Photo I: Etch piL~ pointed out dislocation density
(Ndis= 5.106 cm-2 )

264
 110
We can see that whatever is the density of the plasma
sample, the dislocation density of the material is quite
•
the same (Figure 10).

107 M
30 •
E 20
. :5
......
.-1
E!
~. ~
E
';:;'10
- :6

~
v ~ ....
II>

0 O+-~~~~T-T-T-T-~
'Vi 10f
cQJ o 2 4 6 8 1012 14 16 18 20
'0 Efficiency (%)
C
0
.~
<'$
Figure 12: Comparison of variation of the current
~ density for cells elaborated from different
"Vi materials.
i:5 I) Samples trcated by plasma (Ndis= 106 _107 cm- 2 )
I (t'i
2) Polix samples (Ndis= 104 _toScm- 2)
o I 2 3 4 3) Silicon claboratcd by static unidircctional method
Resistivity (Ohm.cm) (Ndis= 101I_IOScm- 2 )(8)
Figure tol Dislocation density for samples treated by
4) Photowatl standars cells (Lu= 100 11m,
plasma for different resistivity reached.
Ndis= 104 _ 105cm-2 )(9)
This value can be compared to those obtained by 5) Monocrystalline commercial cell(10)
6) Best resull~ achieved (Green et a11983)
controled crystallisation processes which gives N=104 -
](15 cm- 2 with a recombination rate at 104 -105 cm/ s The variation of current density versus diffusion length
(Figurc 11)(7) for Polix material (11) and plasma samples (2) arc plotted
L, effWn) in figure S. We can sec that for a given Isc. for example
100 . , . - - - - - - - - - - - - - . , 23 mA/cm 2 , classical material has a Ld close to 20 11m,

80 ------
-----
I
2
but 73 11m in case of plasma sample.
~30r---~~==~====~1l
...... ./> ....... 3
:;;.....
-.0-- 4
ta 20
Vi
c
'Q
...
C 10
t
u"

u 20 40 60 80 100 120
Disloc'ltion density
(/cm2) ])lffuslun len!:!h (11 m )
~ : Measures realised for samples purified by
Dependance of Leff on dislocation density for
plasma.
different Sd valucs ( El Ghitani et al).
Figure 13: CUTTent density versus diffusion length for
Photow81l standard(II). Measures of samples
I: Sd= 103cms- 1 2: Sd= 104 cms- 1 treated by plasma have been set.
3: Sd= 105Cn1S-1 4: Sd= 106cms -l
Figure 11: The hatching section stands for the Leff area 3.3 inlc[Drctaljon and conclusjons
for samples treated by plasma if Sd is about
104 - to5 _cllls- 1 The only way to explain this rcsult in front of those
The problelll is to explain how such a high indicated on the figure II is to consider. like other
dislocation density can give nervertheless a photovoltaYc authors (12). that the clectrical activity of cristal defects
efficiency close to 10%7 is not due to the cristal defects themselves. but rather by
Photocell efficiency depends on current density and impurities decorating the defects and dangling bounds in
the comparison between plasma samples and other silicon. In the previous work. we have underline (ref) the
crystallised silicon shows that this electrical information high purity of the material obtained. Recently an
cannot explain a difference between these kind of analysis of the plasma material in connection with CRN
material. of Strasbourg by ERDA technique indicated that the final
content of hydrogen can rise up to 9%. So, even if the
dislocation density is very high (106 -10 7 cm- 2 ). the
hydrogen of the thermal plasma passivate the defects. and
the high purity of the silicon. leads to a material having a
photovoltai'c efficiency close to 10%.

265
Acknowledgments: A.F.M.E. and C.E.E. arc greatfully
acknowledged for their financial support.

REFERENCE
(1) 1. AMOUROUX, D. MORVAN, H. APOSTOLIDOU, F.
SLOOTMAN, P.V. HUONG, J.L. VUOTTO, M.
FEDOROFF, J.C. ROUCHAUD, Proceedings of the 6th
Photovoltaic Solar Energy Conference, London, U.K. ,
15-19 April 1985.
(2) N. MADIGOU, Thesis of University P. and Marie
Curie, January 1990.
(3) D. MORVAN, P. HUMBERT, N. MADIGOU, J.
AMOUROUX, S. CA VV ADIAS, SuppIemcnt au Journal de
Physique, FASC 18, tome 51, Colloque C5, Odeillo,
France. 15 Septembre 1990, Les Editions de Physique.
(4) 1. RADlC-PERIC, Proceedings of the 9th International
Symposium on Plasma Chemistry, Pugnochuiso, Italy,
September, 4-9, 1989.
(5) Y. PAULEAU, Principes de base et principaux
parametres de fonctionnement d'une cellule solaire,
L'Actualitc Chimique, Nov. 1981.
(6) T. F. CISZEK, T. WANG, T. SHUYLER, A. ROHALGI,
J. of Electrochemistry Society, Vol. 136, nOl, 1989.
(7) S. MARTINUZZI, A. RAUBER, 9th E. C. Phot. Solar
Energy Cone. Sept. 1989, Freiburg, RFA.
(8) E. SCAFE, A. PARRATA, L. OJ RESE, V.
ADONCECCHI, D. MAGADONNA, V. PARAGGIO,
Proceedings of the 6 th Photovoltaic Solar Energy
Conference. London. U.K. , 15-19 April 1985.
(9) H. LAUVRAY, P. LAY, G. NOUET, 1. DONON,
Proceedings of the 6 th Photovoltaic Solar Energy
Conference, London, U.K. , 15"19 April 1985.
(10) R. J. VAN OVERSTRAETEN, R. P. MERTENS,
Physics, Technology and use of photovolta"ics, Adam
Hilger Ltd, Bristol and Boston, 1986.
(11) 1. FALLY, C. GUIGNOT, L. GOEFFON, Proceedings
of the 6 th Photovolta"ie Solar Energy Conference,
London, U.K. , 15-19 April 1985.
(12) S. PIZZINI, M. RUSTIONI, Proceedings of the 6th
Photovoltalc Solar Energy Conference, London, U.K. ,
15-19 April 1985.

266
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CLASSICAL AND RAPID THERMAL PROCESS INDUCED GETTERING IN MULTI CRYSTALLINE SILICON

B. HARTITI, J.C. MULLER, R. STUCK and P. SIFFERT

Centre de Recherches Nucltlaires (IN2P3)
Laboratoire PHASE (UPR du CNRS n0292)
B.P.20
F-67037 STRASBOURG CEDEX (France)
D. SARTI
PHOTOWATT, 6 rue de la Girafe, F-14043 CAEN CEDEX (France)

ABSTRACT. In this study, we have examined the effects of different doping sources (including
unanalyzed ion implantation and doped silica-film deposition followed by classical or rapid thermal
annealing) on bulk minority carrier diffusion length measured by the Surface PhotoVoltage Spy
technique. The gettering efficiencies of these doping sources have been compared to those obtained
by classical POCla diffusion process.

1. INTRODUCTION in a rapid thermal cold wall furnace FAV4 model from

JIPELEC (France) with temperatures governed by an infrared
As the metallic impurities are known to act as efficient pyrometer.
recombination centers in silicon [1] and more particularly in The minority carrier diffusion length (1D) was determi-
multicrystalline materials, the importance of forming emitter ned with a Spy set up [4]. This method has the advantage that
regions (n+) or back- surface fields (p+) which maintain or there is no need for permanent contacts. It allows studying the
increase the bulk recombination lifetime has been clearly effect of RTA in the bulk without concern for the surface and ap-
established as necessary for the fabrication of efficient silicon plies to doped samples as well as to virgin silicon. Furthermore,
solar cells. Increasing of the minority carrier diffusion length in only energy levels acting as recombination centers are detected,
the base region can be achieved by swipping the metallic those, however, in very low concentrations compared to other
impurities out of the active region of the cell by a gettering techniques. The most important drawback of this technique is
process [2]. Many different types of diffusion sources are the choice of the value of the optical absorption coefficient a()..).
available for performing the n+ or p+ diffusion for each dopant We have used the a()..) values reported by Runyan [5] for stress
species. These include ion implantation, CVD-oxides, pastes, relieved single crystal. Perichaud and al. [6] have recently
liquids or gases. Each of them can affect the device quality in demonstrated that they are also the best values for Polyx multi-
different ways because of the presence of different chemicals crystalline silicon.
and contaminants. Althrough various doping sources have been
used by different research groups, few studies have been done to
determine which source is superior in terms of gettering 3. RESULTS
efficiency.
In this study, we examine the effect of different diffusion The gettering efficiencies of these two different doping
sources, including unanalysed ion implantation and doped techniques on the bulk minority carrier diffusion length have
silica-fUm deposition followed by classical (CTA) or rapid ther- been compared to those resulting from the standard diffusion
mal annealing (RTA) on bulk minority carrier recombination li- process used by PHOTOWATT.
fetimes in Polyx multicrystalline silicon sheets provided by
PHOTOWATT (France). The gettering will be compared to that 3.1 PO Cia diffusion: reference
obtained by a classical POCla diffusion process. The efficiency of the POCI3 gettering is illustrated on
For devices like solar cells, it seems very appropriate to Table 1 which gives the values and mean values of ~ as
study the efficiency of the different gettering processes directly measured by the SPY technique before and after diffusion at
on the diffusion length of the photocarriers in the base region by 850°C on Polyx multicrystalline sheets. This improvement will
using the Surface PhotoVoltage (SPy) method. be used as a reference to test the efficiency of the other doping
processes.

2. EXPERIMENTAL PROCEDURE POLYX

eg
Position ofthe
Our experiments have been performed on 2.5 x 2.5 em 2 SPV measurement
Virgin POC1 3
multicrystaIline samples cut in standard 10 x 10 em 2 Polyx
sheets provided by PHOTOWATT and on reference P-type CZ 36 46
<100> monocrystalline wafers of 10 a.em resistivity.
The molecular ion implantation have been performed on ~ X 40 67
an equipment without mass separation. A detailed description '; X 39 67
of this process can be found in [3]. 8 X 35 65
The direct acceleration of the molecular species towards
the sample amorphizes the crystal surface more rapidly than a Mean value 37.5 61.1
conventional ion implantation. For the spin-on doping techni·
que, we have used boron doped silica-films from Emulsitone Co.
(USA) doped at two levels (1 x 1020 and 1 x 1021 cm-3). The resi· Table 1 : Gettering effect of POCl3 analysed by the Spy technique
dual contaminants present in the two sources have been (the diffusion length LD values are given in pm).
QIlalysed by the Atomic Absorption technique. In general, 2000
A thick silica-film layers have been deposited by spin-on at a 3.2 Ion implantation
3000 rpm velocity. The film is then dryed in a steamroom at Two sets of wafers (p-type CZ <100> 10 a.cm and Polyx
250°C during 20 min. sheets) including for each run one virgin and one BF. implanted
These two doping processes have been followed by a c1as· sample (at an energy of 15 keV and a dosis of 5 x 10 10 cm· 3 ) have
sical or a rapid thermal annealing step at different been annealed together in the classical or rapid furnace unit in a
temperatures between 600 and 1000°C for durations of 15 min Ar + O2 gas ambiant for temperatures between 600 and 1000°C
and 25 sec, respectively. The first annealing has been obtained and respective times of 15 min and 25 sec.
.in an open tube furnace under gas flow of argon or (Ar + 20 % Figure 1 gi:ves .the mea~ diffusion length 1" v~ues obtained
O2), The second one is performed under the same gas ambiant after BFa Ion Implantation followed by a cfasslcal or a rapid

267
 thermal annealing on the monocrystalline wafers having an The samples have been processed as before with a virgin
initial Ln value of 190 ~. We can observe that the initial Ln and two boron doped samples (one at the lower, the other at the
value is conserved after CTA, whereas after RTA the Ln values higher level), for each classical or rapid thermal run, but under
decrease rapidly with increasing temperatures. an argon gas flow. We can point out, at first (by comparing re-
On Polyx sheets, the Ln values reported after RTA show the sults of Figures 1 and 3) that the Lo values obtained on the ther-
same evolution. However, the Lo values on Polyx after eTA mally heated reference samples are higher if an Ar + O2 am-
processing can exceed considerably the initial value of the virgin biant is used during the process. Secondly, we can notice that
material and can be as high as those reported after POCla diffu- the degradation of the ~ values is very important after CTA
sion. These results, confirming that an effective gettering can (Figure 3) for the lower dOped source. At least, we observe, after
occur after ion implantation, have been reported on Figure 4 to- CTA (as before for BFa.implantation), a gettering only for the hi-
gether with those observed after silica-film deposition. ghly doped source, Wltl1 Lp values higher than those reported
~ r-----------------------------------,
after thermal processing 0 the virigin wafers and equal to the
peZSI initial value reported for the unprocessed starting material. As
before, after RTA, the Ln values decrease rapidly with in-
creasing temperatures.
On Figure 4, we have summarized the main results obtai-
ned for polycrystalline Polyx wafers after BF ion implantation
or boron silica-film deposition followed by erA, together with
those reported before for the POCIa diffusion at 850°C which is
used as reference. It should be nored that a gettering efficiency
as high as reported after POCla diffusion can be obtained after
boron doped silica-film depoSItion for samples annealed at
100 700 lOa 150 .00 ISO 1000 850 0 e whereas for BF ion implantation, the maximum is
PROCESSING TEMPERATURE ('C ) around 950°C. The exis1ence of an optimum temperature de-
Figure 1 :Spy l1U!asurel1U!nts on BFJ implanted samples for p- pending on the process has been previously observed by Rang
type CZ <100> 10.Q:m silicon. Initial value ofLD : 190 JUTI. and Schroder [2].
~ r---------------------~------------'
3.3 Silica-film deposition P CZ SI + CTA •
...
VIRGIN
For boron doped silica-film we have used Emulsitone Co. ~
•
•
BORON tE20 an"
BORON tE21 an"
sources doped at a level of 1 x 1020 and 1 x 1021 em.:!, respecti-
vely. The residual contaminants present in this source have ~ >CO
been analysed by the Atomic Absorption technique. The concen- -' '50
tration in ppm of the observed contaminants are given in Table
2. The highly doped source seems to have the same level of con- ~ 100
tamination as the lower doped one. ~

..
is

...
50

Dopant level
(em")
B AI
Detected elements (in ppm)

Fe Cu er Mn Ni V
... ,
PROCESSING TEMPEAA1UAE ('C)
I x 1()21 19 000 0.08 0.9 0.05 0.1 0.16 O.OS SO.1
Figure 3 : Spy l1U!asurements on borosilicafilm samples after
I x 10'" 2000 0.06 0.8 0.4 SO.OS 0.004 O.OS SO.1 cl!lBsical thermal annealing (CTA). Initial LD values of the p-CZ
SI, 10 .!2cm : LD = 190 JUTI.
Table 2 : Atomic absorption analysis - Borosilicate film Emulsi- 'oo r-------------~
P~O~
L~YX
~---------
.-~------.
tone Co. (USA). • SI\JCAFLW
UI",
CTA • (8)

We have reported on Figure 2 the boron profiles obtained ra PIXU

by Secondary Ion Mass Spectrometry (SIMS). measurements.

These profiles, performed on the lower doped (1 x 1020 cm.:!) sili-
ca-film deposited layers after thermal annealing at 950°C, have
shown that subsequent boron diffusion occurs for the two ther-
mal processes, with a similar profile. The RTA treated sample
exhibits a slightly higher surface concentration (about 1.5 x 102°
cm.:!) and a more abrupt shape than the CTA processed wafer.
... ,..
PROCE~NQT EMPEAATUAE ('C)
n.

SIMS ANALYSIS
Spln-on SlIIe. FUm
Figure 4 : Spy l1U!asurements on Polyx sheet after various do·
Boron doped : , . 1020 em·)
ping processes followed by CTA Initial LD value of Polyx :
• RTP I50 C / 2.5see
LD = 50 JUTI•

4. DISCUSSION
The origin of the getteringobserved after POCl3 diffusion
can be deduced from the doping profile characterization. The
comparison of the total and active fraction of the dopant profile
[7] reveals the presence of dopant precipitates in the front region
so that gettering can occur by the presence of a great amount of
lO" silicon self interstitials.
The strong decrease of the diffusion length reported for the ion
implanted samples after the RTA process can be related to an
external contamination. In fact, Eichhammer et al. [8] have re-
cently shown, for classical as well as for molecular ion implanta-
o 100 200 300 400
tion, that the bulk recombination centers which degrade the
DEPTH <n.> bulk diffusion length after rapid thermal processing of the im- ·
planted silicon wafers, are related to metals which are co-im-
Figure 2 : Second4ry Ion Mass Spectrometry (SIMS) analysis of planted with the main implanted species as a consequence of
the low doped boron-silica source. sputtering in the implantation system.

268
 The Ln degradation observed after RTA annealing of the
silica-film can be also attributed to an external contamination as
metallic f,ecies are present in the source at a concentration bet-
ween 10- and 10-1 ppm.
At the present state of our investigations, as the activation of
impurities is the dominant phenomenon which characterizes
the RTA, it seems that the purity of the doping source is the
major parameter to be controlled. In fact, by using a high purity
high doped phosphorus-silica-film source, we have recently
shown [9) that the Ln values can be maintained and even more
increased in silicon alter a RTA cycle.
By classical furnace annealing with a slow cooling rate «
0.1 °Clsec), effective gettering can occur as the impurities are
mostly in a precipitated inactive form [10), so that the improve-
ment level of the classical POCl..a diffusion can be reached after
boron-silica-mm deposition or B1<'3 ion implantation.
The fact that we have no gettering for the low doped silica-mm
source can be related to the surface concentration which is at
least 1.5 x 102° cm-3. This value, which is too far away from the
solubility limit, will not enhance the gettering effect [21.

5. CONCLUSION
We have found that the well known gettering phenome-
na resulting from the classical POCladiffusion process used to
form the n+ contact of the cell can also be obtained with other do-
ping procedures including molecular ion implantation and
doped silica-film deposition. The most important point that
emerges from our study of the effects of the doping process on
the bulk diffusion length (measured by the Surface PhotoVolta-
ge technique) is the fact that metallic impurities co-implanted or
present in the solution can limit the efficiency ofthe gettering in
case of rapid thermal annealing (RTA). For classical thermal an-
nealing (CTA), with a slow cooling rate « O.I°C/sec), effective
gettering can occur as the impurities are mostly in a precipita-
ted inactive form, so that the level of the classical POCl3 diffu-
sion ca be reached after boron doped silica-film deposition of BF3
ion implantation.

ACKNOWLEDGMENTS
This work was funded by the PIRSEM in a program of
the AFME (Agence Fran~se pour la Maitrise de l'Energie) and
the CNRS (Centre National de la Recherche Scientifique) and
by the CEC in the JOULE MULTI-CHESS contract Nr.
JOURI0036.

REFERENCES
[1) E.R. Weber: Appl. Phys. 30, 1 (1983).
[2) J.S. Kang and D.K Schroder: J. Appl. Phys. 65, 2974
(1989).
[3] J.C. Muller : In Solar Energy R and D in the European
Community, Series C, ed. by D. Reidel, Proc. of the EC
contractors meeting, Brussels (Belgium), Nov. 16-17
(1982), Vol. 3, p. 67 (D. Reidel, Dordrecht, 1983).
[4] AM. Goodman: J. Appl. Phys. 32, 2550 (1961)
[5) W.R. Runyan: In NASA CR 93153 National Technical
Information Service, N68, 16510 (1968).
[6) I. Perichaud, S. Martinuzzi, G. Mathian, M. Pasquinelli
and J. Gervais: In 21th IEEE Photovoltaic Specialists
Conference (IEEE, New York, 1990) (in press).
[71 J.C. Muller and R. Stuck: Annual AFME contractor re-
port (1990) unpublished.
[8) W. Eichhammer, M. Hage-Ali, R. Stuck and P. Siffert :
Appl. Phys. A50, 405 (1990).
(9) B. Hartiti et al : to be published.
(10) B. Hartiti. W. Eichhammer. J.C. Muller and P. Siffert : In
ImJlurities, Defects and Diffusion in Semiconductors :
Bulk and Layered Structures, Mat. Res. Soc. Symp. Proc.
(MRS, Pittsburgh, 1990) (in press).

269
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

COMPARATIVE EFFICIENCY OF HYDROGEN PASSIVATION PROCEDURES FOR COST EFFECTIVE MULTICRYSTALLINE

SILICON SOLAR CELLS

B. HARTITI, J.P. SCHUNCK, J.C. MULLER, J.P. STOQUERT, E. HUSSIAN and P. SIFFERT
Centre de Recherches NucIeaires (IN2P3)
Laboratoire PHASE (UPR du CNRS n0292)
B.P.20
F-67037 STRASBOURG CEDEX (France)
D. SARTI
PHOTOWATT, 6 rue de la Girafe, F-14043 CAEN CEDEX (France)

ABSTRACT. We will show that a si~cant bulk passivation of recombination centers can be ob-
tained on 170 j.IlIl thick commercial1y available multicrystalline silicon solar cells (POLYX from
PHOTOWATT), after isothermal hydrogen ion implantation at a temperature of 380°C during 8
min. Improvement of the bulk diffusion length 10 values can be obtained for implantation perfor-
med through the front rectifying n+ contact or thrOugh the back side of a bifacial grid metallized cell.
Both sides hydrogenation gives the highest improvement. Similar improvement of the 10 values
can be achieved by ShN deposition by RF plasma on virgin p-type material or on n+p cells but
only in case of back side ~eposition. Our results show that the presence of a highly doped n+ layer
strongly reduces the hydrogen diffusion, even if pure hydrogen plasma is performed in the same
reactor.

1. INTRODUCTION same Polyx ingot. All cells were manufactured by using the nor-
mal industrial process of PHOTO WATT including POCla diffu-
As a majority of the solar cell manufacturers are presently sion of the n+ layer and screen printed metallized grids on both
selling multicrystalline silicon solar cells for terrestrial sides of the wafers. One oftwo consecutive cells in the ingot were
applications, the course in the improvement ofthe performances cut in 5 x 5 cm 2 in order to have four samples for hydrogenation,
will be the most significant contribution in the future price re- the other 10 x 10 cm 2 sample was the reference. This opera-
duction. tion was repeated for tl)e whole usefull part of the ingot.
This improvement results from a better understanding Prior to the hydrogenation, the cells were characterized
and control of the crystal growth mechanism and of device tech- by measuring their dark and illuminated I-V characteristics
nology. A significant improvement of the photogenerated carrier and spectral response.
collection has been achieved by reducing recombination of cen-
ters in the bulk and at device surfaces. Hydrogenation techniques
Several techniques have been developed to minimize the The implantation procedure for hydrogenation must have
recombination of the minority carriers. They consist in surface a total beam current of around 100 rnA in order to achieve high
passivation and in swipping out or neutralization of the defects production rates over large areas and to be close to industrial ap-
and impurities in the bulk by external gettering or by hydrogen plicability. This can be easily obtained using a Kaufman-type or
introduction [1-5]. our multiple-beam ion source based on the same principle than
Several models have been developed to explain hydrogen the former one, but equipped with a post-acceleration [1]. The
introduction into silicon and hydrogen interaction with the de- post-acceleration voltage can be set between 0.5 and 5 kV and a
fects and impurities [6, 71. Many techniques have been used for total current of 100 rnA can be obtained. The correspqnding
the hydrogen introduction as low energy ion implantation per- beam flux exceeds 1 rnA cm- 2 and the irradiance 5W cm· 2. This
formed with Kaufman-type ion sources or similar equipmcnts leads to a significant hcat-up of the sam pIc during the ion bom-
[I, 2, 4, 8-10], annealing in hydrogen gas flow [41, rJ. hydrogen bardment which can be controlled and stabilized in order to per-
plasma [5, 11] and hydrogenated silicon nitride deposition [12, form isothermal hydrogenation. By this way we have found the
13]. Two technological ways are in competition: the ion implan- optimal temperature which results in a maximum neutraliza-
tation which is the most efficient for the bulk and the silicon ni- tion of inter- and intragrain defects for an irradiation performed
tride deposition which can play also as surface passivation and through the n+ front layer of a metallized cell (1).
as antireflective coating. The plasma hydrogenation was performed introducing
The first approach to reduce bulk recombination uses low the cells inside a rJ. hydrogen glow discharge reactor (13.56
energy broad beam ion implantation of hydrogen with a Kauf- MHz) at different substrate temperatures (from 150 to 380°C).
man type ion source or similar equipment. In particular, we The available area for samples in the reactor is 100 cm 2• The hy-
have demonstrated [1] that a precise control of the temperature drogen pressure was changed from 0.8 torr to 1.2 torr and the
of the processed sample itself under the hydrogen beam is ne- treatment time was varied from 4 to 10 min in order to optimize
cessary. Isothermal hydrogen ion implantation at an optimal the hydrogen penetration.
temperature around 350°C results in a maximum neutraliza- The hydrogenated silicon nitride layer (SiaN ) has been
tion of inter- and intra-grain defects. By using this optimal con- obtained in the same plasma reactor by a decomposition of SiH4
dition for multicrystalline Polyx cells from PHOTOWATT we andNH.a.
have shown [2] that the major parameter which increases is the Alter the hydrogenation the cells were again characteri·
bulk diffusion length Ln. zed by means of dark and illuminated I-V characteristics and
But even if the ion implantation seems to be the most effi- spectral response.
cient technillue to improve the bulk diffusion length for laborato-
ry cells, technological and economical considerations for cost ef- 2.2 Results
fective cells are more in favour of a hydrogenated film deposition As we have previously reported [9], the screen printed
technique. For this reason the aim of the present work is to com- contacts are more degradated by the hydrogenation process
pare the efficiency of these two hydrogenation processes. than classical Ti-Pd-Ag contacts. About 10 % reduction of the fill
factor (FF) can be observed, which impeeds an efficiency impro-
vement due to the hydrogenation.
2. RESULTS For this reason, we will focus our study on the two parameters
which are directly proportional to the reduction of the concentra-
2.1 Experimental procedure a
tion of bulk recombination centers, the short-circuit current se )
~ and the minority carrier diffusion length (10) in the base region.
We worked on 170 J.1Ill thick slices originating from the

270
 Djrect hydrogenation the more expressive for the following plasma conditions: power
• Ion implantation process density about 300 mW/cm 2 (which corresponds to a total power
The main results, obtained in a llfevious work (2) after S of 100 W) under a gas pressure around O.S torr. For these condi-
min isothermal hydrogenation at 350°C under an irradiance of tions, it can be observed an increasing of the Ln values of about
0.6 W/cm 2, have shown that short-circuit current Ise and bulk 70 % in the virgin p-type Polyx sheets. However, no improve-
diffusion length Ln improvement can be observed even if the hy- ment of the !'D values has been reported if the plasma treat-
drogenation is performed through a highly doped n+ front ment is performed through the n+ phosphorus doped front
contact; secondly, that the Iso iI!tprovement is conserved after lever
antireflective coating with a titanium oxide layer.
The most important point that was emerging from this
study was the large improvement (more than 50 %) of the mean
.. r------------------------------, Hplasme
30ClnWlcm

value of the diffusion length after hydrogenation, with some 38O"CI3Omn

points equal to the sample thickness (170 IJ.III). This is illustra-
ted by the representation of Figure 1 which shows the evolution
of the Ise values as a function of the diffusion length.
N
E
POLYX SOLAR CELLS ( .. 1701ffil thic k ) v
30 ~
E ~+---------~--------~----~--_r--~

.....
0.' 0.' l.O
'.'
z HYOROGEN PRESSURE ( TORR )
.....
200:: Figure 3 : Relative variation of the diffusion length LD versus
0:: hydrogen plasma pressure.
:;)
L.J

Uncoate d solar cells .....

• Hydrogenated W
:;) Undirect hYdrogenation by means of a silicon nitride layer
10 :i The SiN layer has been obtained for the same plasma
Virgin V u conditions as before but in presence of SiHA,and NH3. The ana-
~ Some other bad ce II s ..... lysis of the SiN layer by ERDA technique lr;lastic Recoil Detec-
0::
0 tion Analysis) has shown that the H content of the layer de-
::r: pends on the process temperature. We have found about 12 %
V>

200 150 100 50 0

H for a plasma at 150°C and 6 % at 3S0°C (see Figure 4).
DI FFUSION LENG TH (~m J
Figure 1 : Evolution of the short-ciruit values as a function of the
diffusion length values. Note that after hydrogenation LD can Energy (MeV)
0.5 1.0 I.S
reach the sampk thickness (dashed vertical line). .
Z
PLASMA SIH4 .NHl IIJ

• Hydrogen plasma 400

H in Si)N..
1) 1SO - C 8a:
Far this hydrogenation procedure, only used in order to 21380-<: c
>-
test and optimize the plasma equipment before performing sili- :z:
~
con nitride deposition, it was very appropriate to study the effi- JOO
ciency of the hydrogen neutralization of recombination centers C"
directly on the minority carrier diffusion length in the base re- :>
0
gion, by using the Surface PhotoVoltage (SPY) method. For this u ~
200 ~
purpose, we have used p-type Polyx sheets without junction.
The first analysed parameter was the plasma power density
which was varied from 150 to 600 mW/cm 2• The plasma was 100 ~
done at a pressure of O.S torr during a time of 30 min at a tem- '"
perature of 3S0°C.
The relative variation of the diffusion length 1" divided by 0
the value of Ln before the treatment is shown on FIgure 2. In 0 SO 100 150 200 250 JOO
order to compare the effect of the different treatment conditions, Channel
only cells with similar initial values of 1" were considered, due
to the fact that hydrogenation is more efficient for low efficiency
cells [91. Figure 4 : Elastic Recoil Detection Analysis (ERDA) of the hydro-
.. ,-------------------------~
H pI••m•
0.8 tOfT
gen content of a silicon nitride layer obtained by RF plasma de-
composition of SiH4 and NH:r

The optical properties of the 7oo-S00 Adeposited layer is

comparable to the standard Ti0 2 antireflective coating used by
PHOTOWATI.
In order to enhance the diffusion of hydrogen, we have in-
creased the deposition temperature from 15Q to 3BO°C and
reduced the deposition velocity from 240 to 100 Almin.
For SiH4 + NHa plasma, no improvement of the 1" and
..
~+---~,,~o------~.o--.------~-o------~~.--~ •• Isevalues have been reported after the treatment perfOrmed
PlASMA POWER DENSrTY ( mWlcm') through the n+ front layer, indicating that the diffusion of hydro-
Figure 2 : Relative variation of the minority carrier diffusion gen is strongly reduced by the presence of a highly doped n+
kngth versus hydrogen plasma power density. layer, as reported before for hydrogen plasma.
The same experiments performed on unprocessed p-type
Polyx materials show more than 60 % improvement of the bulk
A maximum improvement of the Lnvalues seems to ap- diffusion length values, as illustrated on Table 1.
pear for plasma power densities around 3(10 mW/cm 2• Figure 3 This result confirms that diffusion of hydrogen occurs in
shows the relative increase of the diffusion length Ln as a func- both plasma procedures (silicon nitride deposition or hydrogen
tion of the hydrogen pressure. Here also a gas pressure of O.S plasma) only in case of operating on p-type sheets or on the
torr seems to be the best choice. The improvement seems to be backside of a n+p structure. This improvement disappears if a

271
post-annealing treabnent is perfonned at 650°C/10 min or at REFERENCES
700°Cl1 inin after SiN or H plasma. This last condition corres-
ponds to the nonnal thennal cycle used for the electrode firing. [1] J.C. Muller, Vu-Thuong-Quat, P. Siffert, H. Amzil, A. Ba-
This result is in agreement with the stability of all known asso- rhdadi and N. M'Gafad, Solar Cells, 25, 109 (1988).
ciations of hydrogen with crystallographic defects or metallic im- [2] J.C. Muller, E. Hussian, P. Siffert and D. Sarti, Proc. 9th
purities [14] which is at least 500°C. But recent papers of Kyoce- Photovoltaic Solar Energy Conference, Freiburg (FRG)
ra [12] or from the FOM Institute [13] do not reflect this result (Kluwer Publishers, Dordrecht, 1989) p. 407.
as !'D improvements are reported after 700°C annealing of the [3] S. Martinuzzi, D. Sarti, P. Torchio, T. Buonomo, J. Ger-
hydrogenated cells. vais, G. Mathian and I. Perichaud, Proc. 9th Photovoltaic
p-type Polyx sheets Solar Energy Conference, Freiburg (FRG) (Kluwer
Publishers, Dordrecht, 1989) p. 458.
Loinilial Loflnal Alogaln [4] S. Martinuzzi, H. El Ghuitani, L. Ammor, M. Pasquinelli
40 60 20 and M. Poitevin, Proc. 19th IEEE Photovoltaic Spec.
Sl3N4
(bIIaces)
30
35
53
55
23
20
Conf., New Orleans (USA) (1987) p. 1069.
[5] M.A.G. Soler, I. Pereyra and AM. Andrade, Proc. 9th
22 53 31 Photovoltaic Solar Energy Conference, Freiburg (FRG)
28 42 14 (Kluwer Publishers, Dordrecht, 1989) p. 398.
22
30
39
52
58 19
[6] J.W. Corbett, J.L. Lindstrom, S.J. Pearton and A.J. Ta-
38 56 18 vendale, Solar Cells, 24, 127 (1989).
[7] M. Capizzi and A. Mittiga, Physica 146 B, 19 (1987).
34 35 +1 [8] N. Lewalski, R. Schindler and B. Voss, Proc. 19th IEEE
Si3N4
(blla...)
45
42
34
47
·11
+5
Photovoltaic Spec. Conf., New Orleans (USA) (1987) p.
CTA 35 38 +3
1059.
650°C 10 mn 28 34 +6 [9] J.C. Muller, Y. Ababou, A Barhdadi, E. Courcelle, S.
Unamuno, D. Salles, P. Siffert and J. Fally, Solar Cells,
40 43 +3 17,201 (1986).
RTP 27 ·30 .+3
700"C 1 mn .35 .20 .·15
[10] H. Yagi, K. Matsukuma, S. Kokunai, Y. Kida, N. Kawa-
kami, T. Nunoi and T. Inoguchi, Proc. 20th IEEE Photo-
voltaic Spec. Conf., Las Vegas (USA) (1988).
[11] M.A.G. Soler, I. Pereyra, F.J. Fonseca and AM. Andrade,
Table 1 : Hydrogen and SiIV4 plasma performed at a power Proc. 21th IEEE Photovoltaic Spec. Conf., Kissimmee
density of 300 mWI cm2, a pressure of 0.8 torr and a temperature (USA) (1990) p. 650.
of 380 'C. The deposition velocity of the Si:l'l4 layers on the two [12] H. Watanabe, Technical Digest of the 4th International
faces was 200 A/min and the duration 2 x "4 min. PVSEC Conf., Sydney <Australia) (1989).
The hydrogen plasma was only performed on one face during 30 [13] P.P. Michiels, LA Verhoef, J.C. Stroom, W.C. Sinke,
min. The L1). measurements by SPY were obtained on the hydro- R.J.C. Van Zolingen, C.M.M. Denisse, M. Hendriks,
genated sUrface. Proc. 21th IEEE Photovoltaic Spec. Conf., Kissimmee
(USA) (1990) p. 638.
[14] S.J. Pearton, J.W. Corbett and T.S. Shi, Appl. Phys. A43,
By using bifacial_phosphorus diffused cells, covered on 153 (1987).
both sides with a SiN(H) layer, we have recently shown [15] [15] J.C. Muller and al., to be published.
that the Ln improvement starts only at .. 700°C, so that it can-
not be reI8ted to hydrogen effect (presence of the double n+
layer), but rather more to a gettering effect which begins to ope-
rate at 700°C.

3. CONCLUSION
We can confinn, by using the thinnest multicrystalline si-
licon solar cells on the market (Polyx from PHOTOWATT),
that a maximum bulk passivation of the recombination centers
can be obtained after optimum hydrogen ion implantation. In
particular, we have shown that the bulk diffusion length values
increase up to the sample thickness (170 J.Un) even if the hydro-
gen bombardment is perfonned through the rectifying contacts.
For plasma process the hydrogenation must be perfor-
med on the p-type substrate to be effective. The duration of the
process is in general also greater than the time needed for ion
implantation.
For silicon nitride deposition, it seems that only backside
hydrogenation improves significantly the bulk diffusion length
values, whereas the front silicon nitride plays as antireflective
coating and surface passivation.
The thennal stability of all the hydrogen associations with
metallic impurities or crystallographic defects are not higher
than 500°C, so that the effects observed after a 700°C thennal
treabnent ean not be explained in tenn of neutralization but ra-
ther more by a gettering effect.

ACKNOWLEDGMENTS
This work is funded by the PIRSEM in a program of the
AFME (Agence Fran¢se pour la Maitrise de i'Energie) and the
CNRS (Centre National de la Recherche Scientifique) and by
the CEC in the JOULE MULTI-CHESS contract Nr.
JOUR/0036.

272
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THERMODYNAMICS FOR REMOVAL OF BORON FROM METALLURGICAL

SILICON BY FLUX TREATMENT

K. SUZUKI and N. SANO

Department of Metallurgy, Faculty of Engineering, The University of Tokyo,
Bunkyo-ku, Tokyo, Japan.

ABSTRACT:The thermodynamics of the removal of boron, which is most difficult

among various impurity element, was investigated by applying the flux treatment in
order to develop a new production technique of solar-grade silicon (SOG-Si) by refining
inexpensive metallurgical-grade silicon(MG-Si).
The equilibrium distributions of boron between silicon and various fluxes were
determined from 1723 to 1823K in a CO or Ar atmosphere. The distribution of boron
has a maximum with increasing basicity and increases with raising temperature for all
fluxes studied. The maximum value was 2.0 for the CaO-BaO-Si02 system.
The reaction of the removal of boron from silicon to flux is expressed as follows:

On the basis of those findings, the mechanism of boron removal is discussed.

1. INTRODUCTION [B)in Si + 23 0 2- + 3 (3-

402 = B03 ) (1)

A large scale solar cell fabrication has been expected as As seen in Eq.(l), the boron removal by flux treatment is
a new energy source for a long time. One of the prerequi- influenced by the activity of an oxide ion, ao'-, which is
sites for the large-scale photovoltaic energy conversion using closely associated with the basicity of the flux and the pre-
silicon solar cell is a substantial reduction in the manufac- vailing oxygen partial pressure.
turing cost of these devices. The purpose of this sturdy is to The data obtained in two different atmospheres (Ar and
develop a new process to produce SOG-Si by refining MG- Ar+12.5%C0 2) using the silicon of 80 ppm B and the flux of
Si. For this purpose, the thermodynamics of the removal 36%CaO-64%Si02 are shown in Table 1. The value of LB is
of boron, which is most difficult among those of various im- not affected by gas composition, indicating that the oxygen
purity elements, has been investigated by applying the flux partial pressure of the system is controlled by the Si-Si O2
treatments. equilibrium. Therefore, boron removal may be described by
the following reaction, which does not contain any gaseous
2. EXPERIMENTAL species:

The equilibrium distribution ratio ofboron(L B) between (2)

various fluxes and silicon melts was measured from 1725 to
1825K using an electric furnace with heating elements of
molybdenum disilicide. The initial concentration of boron Table 1. Boron removal by 36%CaO-64%Si0 2 at different
in silicon melts was controlled beforehand in the range of 30 atmospheres.
to 90 ppm. Four flux systems, namely CaO-Si0 2 , CaO-MgO-
Si0 2 , CaO-BaO-Si0 2 and CaO-CaF r Si0 2 systems, were used. [B) in (B) in
Ten grams of silicon were eqnilibrated with 109 of the flux Exp. Atmosphere Temperature metal Slag 41
in a graphite crucible( 30mm J.D. and 50mm depth) for 2 No. (K) (mass (mass
hours in a CO atmosphere or Ar-C0 2 mixture. ppm) ppm)
1 Ar 1723 54 22 0.41
3. RESULTS AND DISCUSSION 2 Ar+C0 2 1723 55 20 0.36
3 Ar 1773 45 21 0.47
3.1 Effect of the surrounding atmosphere on LB. 4 Ar+C0 2 1773 55 17 0.31
The boron in MG-Si is expected to be oxidized and 5 Ar 1823 28 26 0.92
transferred to a molten flux in an oxidizing atmosphere in 6 Ar+C0 2 1823 27 31 1.15
accordance with the following reaction:
..
lDltlal B content:80 ppm; fiux:36%CaO-64%SI02

273
3.2 Effect of flux composition on the LB'

I.
Figures 1 to 3 show the relationship between the LB 2.0
and the basicity of flux for the CaO-BaO-Si0 2 (-CaF 2 ), CaO-

r-r~
MgO-Si0 2 (-CaF 2) and CaO-CaF r Si0 2 systems, respectively.
In all cases, the value of LB increases with increasing basic- 1,5
ity, reaching the maximum, and then decreases. The largest
LB of 2.0 was obtained when the CaO-BaO-Si0 2 system was
used. This agrees with the findings of Dietl et al/ I /, who used lO •
5kg of MG-Si with 30ppm B and 7kg of 30%CaO-40%CaFr j
30%Si02 for industrial application of the flux treatment.
0.5

2.5 und. r CO Qtrno.plllrt

o
o 2.0 4.0 6.0

.
2.0 C.O/SIOz
•
~ Fig.3 Relationship between La and CaO/Si0 2 for the
1.5 -L
• ~~
I
0
CaO-30%CaF 2-SiO. system at 1723 K.

LO
I 0
The equilibrium constant of Eq.(2) is given by Eq.(3):

• C.o- BoO- SIOr K = aBOl.5 . ( aSi ?/4. = 'YBOI.5 . Xaol.5 . ( aSi )3/4 (3)
o C.O-c.F.-BoO-SIOli aa aSiO, 'YB XB aSiO,
where a; is the activity of component i, "Yi is the activity
0.5
0.3 0.8 1.3 l8 coefficient of component i, and Xi is the mole fraction of
C.O/SIOz component i. In this study, the boron in silicon melts may
be assumed to obey Henry's law because its concentration
Fig.1 Relationship between LB and CaO/Si0 2 for the is very low. The value of K, therefore, is constant at a con-
CaO-1O%BaO-Si0 2 (-CaF 2 ) system at 1723K in a stant temperature. The value of asi is nearly equal to 1, and
CO atmosphere. XBOl.5/XB can be converted to LB. Therefore, Eq.(3) can be

2.0 1.0

l5 r--/!~~
.. .........
0

-1.0
." .",
~

J
•
~
/ 0
lO
><
0 - 2.0

~
co
0.5 2
• CoO-MoO-SIOr -3.0
o CoO-c.FI-MoO-SIOt

o l8 to 2.0 3.0
0.3 0.8
CoO/SI02 C.O/SI02

Fig.2 Relationship between La and CaO/Si0 2 for the Fig.4 Relationship between (log C· 'YBOl..) and CaO /Si0 2
CaO-1O%MgO-Si0 2 (-CaF 2 ) system at 1723 K in for the CaO-CaF r Si0 2 system at 1723 K.
a CO atmosphere.
rewritten as Eq.( 4).

3.3 Thermodynamic considerations on the relationship be- 3/4

a Si0 2
1;- -
_

tween LB and flux basicity. C . 'YBOl.5 (4)

The behavior of LB in connection with basicity of flux
where C is a temperature dependent constant. The values of
can be explained by thermodynamic consideration on the
LB are experimentally determined, so that the value of the
reaction of Eq.( 2) as follows.

274
left-hand side of Eq.( 4) can be calculated ifthe value of aSiO, 4. CONCLUSIONS
is known from the literature, for example, for the CaO-CaFT The maximum distribution coefficient of boron was ob-
Si0 2 system/ 2 /. Figure 4 shows the change of C . /'Bo .. with served at the fixed composition for all fluxes used. It is
the basicity for this system. It can be seen that the value of approximately 2.0 for the CaO-BaO-Si02 system.
C ·/'BOl.6 decreases monotonously with increasing CaO/Si0 2 • The boron removal by flux treatment can be described
This indicates that an increase in 1B in the range of lower by the following reaction:
CaO-Si0 2 in Figure 3 is due to a decrease in aSiO,. and
that a decrease in the range of higher CaO /Si0 2 is due to
a decrease in Po, at the slag-metal interface as a result of a
decrease in aSiO,. The optimum flux composition is determined by the balance
between the basicity of the flax and the oxygen potential of
3.4 Effect of temperature on the 1 B. the system, which is controlled by the Si-Si0 2 equilibrium.
Figure 5 shows the temperature dependence of LB for
the CaO-CaF 2-Si0 2 and CaO-MgO-Si0 2 systems. The lin- REFERENCES
ear relationship between log 1B and reciprocal temperature (1) U .S.Patent 4304763, Dec., 8(1981).
can be seen for both fluxes. The removal of boron is enhanced (2) Slag Atlas: Verlag Stahleisen M. B. H., Diisseldorf,
when temperature is high. The heat of reaction of Eq.(2), (1981), 122.
llffO, is obtained from the slope of straight line in Fig.5. The (3) E.T.Turkdogan: Physical Chemistry of High Temper-
values of llffO are 95.6 kJ/mol for the 54%CaO-30%CaF 2- ature Technology, Academic Press, (1980), 5.
16%Si0 2 and 153 kJ /mol for the 29%CaO-20%MgO~51 %Si0 2 ,
respectively. The Gibbs energy change of Eq.(5) in reference
to pure substances is given by Eq.(6):

B(l) + ~Si02(1) = ~B203(1) + ~Si(l) (5)

llGo = 45,490 + 24.1T (J/mol)/3/ (6)

The difference between heats of dissolution of Si0 2 and B20 3

in the flux, (1/211H B,o, - 3/411H siO ,), is obtained by sub-
tracting 45.5kJ in Eq.(6) from the above values. The values
are 50.1kJ/mol for the CaO-CaF T Si0 2 system and 107.5
kJ/mol for the CaO-MgO-Si02 system, respectively.

0.50

r __t __!
0.2~
I----o~

I 0-r-t0~I
..
~
~
0 6~
I 6

-0.2&
o ,."c.o-lOllc.F.- 1.1I51Oo
to 2'Jlc:.o-20JlMoQ-OI1l5lC)o
..., •• r CO ItllftOlpf\tn
-0,&0
6.4 0.7 0.8 0.'

Fig.5 Temperature dependence of 1B for the 54%CaO-

30%CaF2-16%Si02 and 29%CaO-20%MgO-51%Si0 2
fluxes.

275
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

AN ANALYSIS OF HIGH EFFICIENCY Si PROCESSING

S. Hogan, O. Darkazalli, and R. Wolfson

Spire Corporation
One Patriots Park
Bedford, MA 01730-2396 USA

,ABSTRACT. There have been many recent advances in cell processing to improve the efficiencies of cells and
modules. This paper addresses the issue of cost effectiveness of different techniques. While improvements in
efficiency are important, researchers must not overlook the overall cost impact associated with these improvements.

1. INTRODUCTION What also has become clear, and is the most significant
conclusion of this analysis, is that the novel techniques of
Spire Corporation has contributed extensively to the extending the base efficiencies beyond present levels are intro-
development of high-efficiency cell processing over the last 10 ducing significant costs from increased capital equipment re-
years. Analysis of the production costs 9f high-efficiency pro- quirements and lower yield associated with the higher-complexi-
cessing has always been of primary concern. Our analytical ty processes. To demonstrate this effect, we present six case
methods, while initially based only on the JPL-originated SAM- studies of silicon cell processing. These case studies are shown
ICS (I), have been refmed to incorporate SAMICS, the EPRl/ schematically in Figs. 2 through 4 and are summarized in the
RTI originated IMCAP (2), and Spire's own internal fmancial following sections. The specific cases were chosen to represent
models. The degree of correlation among all methods is well a broad range of cell processing: from the basic processing of
within the error expected from varied inputs. Results of the today's production to novel, high-efficiency concepts demon-
several analytical methods are so strongly driven by input strated in laboratories.
variations that the trends are significant.
The original concept of sensitivity analysis for this work:
was presented over five years ago (3) and is summarized in
p
Fig. 1 with updated analysis values. This figure shows how the
base cost is driven by several factors, given that all other vari-
ables are fixed. It is clear from this figure that, up to the base Thin an d texture etch front
efficiency (14.5%) and slighdy beyond, efficiency is the most
significant factor. 11ms the analysis justifies the importance P diffusion from POCI 3 source
attached to the development of higher efficiencies over the last p

ten years.
~..... ~ ..... ~
..... _ .... :::::: n'
- l ~G\as.
Glass etch and edge etch

COST ($ ) PER WATT

$6 ~--~---------------------------. p
Thermally oxidize iTont and
screen print AI back contact
$5

Fire ink
$4

$3 Screen print Ag front

and back metallization
p
$2

~.Wji.i~C- p'
$1 • • •• AI
Ag
Fire ink
$o L------L______L-____ ~ ______~____~

o o~

VARIANCE FROM BASE LINE

1~ 2 2.5
p!
....~
.....~
.....~
.....~
.... r-=- SiC,
n'

vAg
p
Cell Eft 14 .6" -+- Prod 'n Volume .OMW
Coat of Pol y S.6/kg
.. .. ·p.. ·t-
Capital Coo. S.6.2M
.. ~
~ .. · rt.G
"· F "·~
· ·~ ·v.z p'

• • • . _ Ag AI

Figure 1: Baseline Cost Sensitivity Analysis. Figure 2: Baseline Process Sequence (Case I).

276
 ____
p p

L-_I===~ Thin and texture etch Iront

1---- -- - - - - Thin and texture etch front

p
POCl 3 spin on and
P dil1usion
~,~
", ~
, ,~ . .~.. ~
, ~ .. .~
. , I-""GlasS
- n' POCl3 spin on and
p P diffusion
If -_______ Glass etch and
I edge etch
~~~5:q'.
1='= -::::- SlO,
n'
1-_ __ _ _ _ _ Glass etch and
edge etch
p
Thermally oxidize Iront and back . .. . ..• ~SIOt
n'
I===;=:===:l - 510. p Thermally oxidize Iront
If
I -_ _ _ _ _ _ _ Photo resist spin on (back)

~-------- Laser groove

~
.' s:" p
Photolithographic delin~ion
and photo resist strip E~~ C:;:::::;:::;:-SoO,

~~
--rr
p Oebris etch

Evaporate AI back contact;

1 - - - - - - - - photo resist spin on (Iront) 1 - - - - - - - - - - - P diffUSion

~
. .~
. .~1. P=;=I-n~'
p Photolithographic delin~ion "V_ n·
p Evaporate AI back contact

1 - - - - - - - - Evaporate Ag and 1---------- Diffuse Al

Ag photo resist strip; sinter
. ,..:....
~::f1;~s:e:~ 510 2 . . .. . - 5 0 0 ,
n'
I ......
p
n'
~
p
n'

Figure 4: Laser Grooved Process (Case VI).

Figure 3: High Efficiency PERC Process (Case IV).

Table I. Manufacturing Cost Analysis.
The financial analysis of all cases is presented in CASE CASE CASE CASE CASE CASE
Table I, with common factors listed. The analysis only takes I II IU IV V V1
into account capital cost and cost of material. While we have CELL EFFICIENCY % 14.5 16 16 18 18 16
seen significant reduction in yield in many advanced processing EQUIPMENT COST $M 18.2 17.9 28.8 32.5 26.8 24.3
steps, this has been ignored for the purposes of this study. OPERATORS 160 148 249 276 224 202
REQUIRED AREA Ft' 18,100 13,800 26.300 31,000 22.800 21,300
Fig. 5 provides a graphical summary of Table I.
PRODUCTION COST/W $3.05 $2.86 $4.05 $4.35 $3.69 $3.54
1.1 Case I: Baseline Processing: Today's Mass Production CONTRIBUTION
EQUIPMENT $1.08 $1.06 $1.70 $1.92 $1.58 $1.44
This case demonstrates the general status of processing AREA SO.25 $0.20 $0.36 SO.42 SO.31 $0.28
technology today. It is comprised of a texture etching process, LABOR $0.27 $0.25 $0.41 $0.46 $0.37 $0.34
diffusion from a POCI3 source, surface cleaning (etching), front- OVERHEAD $0.27 $0.25 $0.41 $0.46 $0.37 $0.34
MATERIALS $1.08 $0.99 $1.06 SO.98 $0.95 $1.03
surface oxidation, screen printing and firing contacts, and AR UTILITIES SO.II $0.11 $0.11 $0.11 SO.11 $0.11
coating. The wafers are squared from 5 inch diameter CZ in-
gots, and the process achieves 14.5% efficient cells. The fman-
COMMON ASSUMPTIONS FOR ALL CASES
cial results of this analysis are presented in Table I.
10 MW Production Facility, 3 Shift Operation
1.2 Case II: Proposed Low-Cost Process: Utilization ofIon 5" grown CZ, Slabbed to 96,6 cm2
Implantation 12 mil thick wafers, 8 mil kerf loss
This case study is siriillar to the baseline but utilizes the Factory yield 84%, not including kerf loss
process of ion implantation for the junction formation. The justi- Materials cost: Poly Si @ $15/kilo
fication for implementing this process is the flexible doping Cell materials @ $O.40fwafer
profiles that are achieved with ion implantation. This allows a Module materials @ $13.50/module (36 cell)
better match of junction profile for the screen printed contacts, Consumables are 5% of equipment cost
without loss from the deeper junctions required for printed Utilities are $90K/MW
contacts. The process is the same as above otherwise, but yields Overhead is @ 100%, G&A is @ 21 %
a 16% efficient cell. Labor is @ $9.oo/hour

277
 COST PER WATT
$5 1.7 Some Other Issues
Two issues mentioned above require further clarification.
The first of these is the question of the cost of raw silicon. Our
$4 analysis, for all cases, assumed a cost of $15/kilo for the raw
material. While this is the purchase price which several compa-
$3 nies have contracted from silicon suppliers, in our opinion it is
an artificial price. The important consideration is that all cases
were treated the same, and that the trend will not be changed.
$2 To demonstrate the cost component ofraw silicon, Fig. 6 shows
the sensitivity for Cases I and III of module cost to silicon
$1 costing from $15/kilo to $60/kilo. The influence is obvious, but
does not change the basic conclusions stated above.

$0 CASE I CASE II CASE III CASE IV CASE V CASE VI

CASE ANALYSIS COST (S) PEA WATT
$5 r--------------------------------,
[ _ EOPT ~ AREA o LABOR
il!il!lI OH EAO P;;] MTLS n UTiL
S4

Figure 5: Manufacturing Cost Summary. $3

$2
1.3 Case III: Standard High Efficiency Process: Photolith-
ography
SI
This process, like the ftrst case, uses a conventional
diffusion process, but now adds photolithography and evaporat-
ed contacts. The reduction in surface coverage and shallower $O ~---L--~----~--~----~--~--~
junction pennit a 16% efficient cell to be produced. However, o 10 20 30 40 50 60 70
the added cost in capital equipment and labor, as shown in COST OF POLYSi, $/KILO
Table I, far outweighs the advantage gained of higher efficiency.
- CASE I -+- CASE III
1.4 Case IV: High Efficiency Perc Process: Back-side Photo-
lithography (4)
This case is similar to the previous one, except for the Figure 6: Poly Silicon Cost; Effect on Cases I & II.
addition of photolithographic processing of the back contact. It
is also capable of achieving 18% cell efftciency, but again with
the addition of prohibitively high capital and labor costs. Even The second issue is the question of the systems cost, and
at the achieved efficiency, it is the most costly process of those its reduction with higher efftciency modules. We have used the
studied. Only significant advances in low-cost photolithographic EPRljDOE analysis fonnat (7) to look at this. The results,
processing techniques will allow this technology to become shown in Fig. 7 illustrates that the cost per generated kWhour
cost-competitive. is not reduced by the gains in efficiency at the higher cost.

1.5 Case V: Spire/JPL Block V Process: Demonstrated High-

Efficiency (5)
This case demonstrates the technology used in 1986 to COST PER GENERATED K WH
develop the high-efftciency cell for the JPL Block-V program.
$0.4
It includes advanced ion-implantation equipment and techniques
successfully demonstrated at Spire, and it pennits the fabrication /'

of 18% efftcient cells. This case is similar in other respects to

Case III, including the use of a one-step photolithographic
processing. The 10% cost reduction is achieved primarily
through the incorporation of ion-implantation.

1.6 Case VI: Laser Grooved Process: Front Surface Groov-

ing (6)
This case examines the proposed new technology that
incorporates a laser grooved front surface. This process has the
advantage of not requiring photolithography and still achieving
a 16% efficient process. TIle present problems with this technol-
ogy are the cost of the lasers and the processing time. Ongoing
work at several labs may reduce this problem, but until laser
equipment cost is reduced and scan rates are improved, the
process will be costly. This process also has the possibility of
CASE I C ASE II CASE III C....SE IV CASE V C.... SE VI
achieving higher efficiencies. If a 21 % cell is achieved, the cost
ofthe product decreases to $2.82/W (keeping all other variables
fixed). Figure 7: Cost of Energy for Six Cases.

278
2. CONCLUSIONS

This study has shown the importance of high-efficiency

cell processing to a cost-effective photovoltaic product. We have
also demonstrated the impact that increased capital equipment
costs make on the product cost. These are clearly issues of
manufacturing technology, which should be considered with as
much importance as high-efficiency techniques. While not
specifically addressed in this study, the effect of process yield
must also be considered, since the role it plays is significant.
Researchers must remember that processes introduced for higher
efficiency purposes will serve little use if they erode the cost
advantage by disproportionately increasing capital cost and
reducing product yield.

3. REFERENCES

(I) R.G. Chamberlain, Solar Array Manufacturing Industrv

Costing Standards SAMICS Workbook, JPL Document
5101-15, Jet Propulsion Laboratory, Pasadena, CA (Feb-
15, 1977; revised Sept. 30, 1977).

(2) Research Triangle Institute, Interactive Manufacturing

Cost Analysis Program, (lMCAP), Engineering Sciences
Program, Research Triangle, N.C. (November, 1985).

(3) R.G. Little and S.J. Hogan, "Economic Analysis of a

30 MW/year Silicon Production Facility," Proc. 6th
European Photovoltaic Solar Energy Conference (1985).

(4) M.A Green, AW. Blakers and S.R. Wenham, "Improve-

ments in Silicon Solar Cell Perfonnance," 4th Interna-
tional Photovoltaic Science and Engineering Conference,
p 159, 1989.

(5) M.B. Spitzer, L.M. Geoffroy, M.J. Nowlan, and

C.J. Keavney, "Perfonnance Improvements in Silicon
Flat-Plate Cells and Modules," Solar Cells, 20 (1987)
333.

(6) M.A. Green, S.R. Wenham, C.M. Chong, S. Narayanan,

Z. Fuzu, J.C. Zolper, A.B. Sproul and AW. Blakers,
"Laser Grooved and Polycrystalline Silicon Solar Cells
Research," Sandia Contract Report, SAND89-7035,
August, 1989.

(7) U.S. Department of Energy, Five Year Research Plan,

1987-1991, Photovoltaics; USA's Energy Opportunity,
National Photovoltaics Program, DOE/CHI0093-7.

279
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A HIGH EFFICIENCY SILICON SOLAR CELL PRODUCTION TECHNOLOGY

N B Mason and 0 Jordan

BP Solar Espana, Valportillo 1-5, 28100 Madrid, Spain
J G Summers
BP Solar, Bridge Street, Leatherhead KT22 8BZ, UK

ABSTRACT
BP Solar have developed a cost-effective production technology for
the manufacture of high efficiency laser grooved buried grid
(LGBG) crystalline silicon solar cells. The process has
demonstrated 17-18% photovoltaic conversion efficiency on a new
demonstration plant recently commissioned at BP Solar Espana in
Madrid. This paper outlines the technology of cell fabrication
and details the performance characteristics of the resulting cells
and modules. with low cost polycrystalline substrates the
technology is expected to lead to a significant reduction in the
fabrication cost per watt of silicon solar cells.

1. INTRODUCTION

The Laser Grooved Buried Grid (LGBG) solar required to operate modern semiconductor
cell was invented by S R Wenham and M A processing equipment. Furthermore, its
Green and patented by Unisearch /1/ in proximity to the BP Research Centre
March 1985. Since that time Green and enabled vital support to be provided
others at the University of New South during a decisive phase in the technology
Wales, Australia, have reported record appraisal. Construction of the plant
photovoltaic efficiencies for this type of commenced in January 1990 and was
device, most notably 20.2% for 49 cm2 area completed on schedule by July of that
cells produced in their laboratory using year. This was followed by an equipment
0.5 ohm.cm floating zone (FZ) silicon commissioning and technology proving phase
wafers /2/. which has tested the reproducibility of
the process and provided statistical data
In 1985/6 BP Solar reviewed this on process yield and performance.
technology and judged it to have the
potential to produce cost-effective high This paper gives a brief description of
efficiency solar cells. The invention was the technology of cell fabrication and
licensed by BP Solar in 1985 and a project summarises the performance of cells and
was established at their Australian modules produced.
facility (Sydney) to investigate methods
of fabricating the LGBG solar cell which
could be used as the basis for a viable 2. THE TECHNOLOGY
cost-effective production technology. The
scope of the development work was wide The design and construction of the LGBG
ranging and encompassed different methods cell has been widely reported in the
and/or materials for the following key scientific literature and has been
steps in cell fabrication; reviewed by Green /3/. The production
sequence evaluated by BP Solar is
* machining grooves into silicon illustrated in figure 1 and can be
* p-n junction formation summarised as follows:
* surface dielectric film deposition
* rear surface treatments a) A thin (10 nm) phosphorus pentoxide
* metallisation film followed by a thicker (80 nm)
silicon nitride layer are deposited by
By late 1988 a technology had been conventional chemical vapour deposition
identified which enabled cells to be (CVO) techniques onto the pyramidal
fabricated in small quantities with surface texture of boron doped, (100)
efficiencies of 18% on 100 mm orientation, Czochralski grown (CZ)
pseudo-square silicon wafers. The silicon wafers.
technology showed sufficient potential to
be considered for the next stage of b) The front grid pattern is fabricated by
development, namely, a detailed evaluation cutting V-shaped grooves of width 15
of the fabrication cost and process yield micron and depth 40 micron into the
using production equipment. surface of the silicon using a high
speed, split beam, pulsed Nd:YAG laser
To verify the performance-yield-cost micromachining system.
characteristics of the technology a
demonstration plant was established in c) The p-n junction is then formed in the
Spain, annexed to BP Solar Espana's front surface at elevated temperatures
existing production facility. Being in in a phosphorus oxychloride (POCI3)
Europe, this plant is well supported by vapour atmosphere. The pre-deposited
the engineering and processing skills P205 film is the dopant source for the

280
 a) [SSSSSSSS\\~l

b)r:v='1
b<lCk cont<!ct
c)

Figure 2. Schematic cross-section of

the LGBG cell.

A schematic cross-section view of the cell

is given in figure 2.

3. CELL AND MODULE PERFORMANCE

3.1 Cells
Since late 1990 process data has been
compiled whilst operational experience is
gained in cell processing and operating
conditions are fine-tuned for optimum
yield. The first three months experience
has been very encouraging with typical 100
mm pseudo-square cell efficiencies of
Figure 1. Schematic showing production 17.5% being realised (Fig. 3).
sequence for cell fabrication: a) P205
and Si3N4 film deposition, b) laser 3.2 Modules
groove, c) p-n junction formation, d)
rear surface treatment and e) Cells are being incorporated into a
metallisation. conventional BP Solar module package which
has demonstrated reliable resistance to
outdoor exposure for over 8 years. The
active silicon surface whilst POCL3 36-cell series connected module has the
vapour is used to dope the exposed
silicon in the grooves. In this way,
the surface is lightly doped for
optimum current collection over the
entire solar spectrum, whilst the
groove regions are more heavily doped r:
for low grid resistance characteristics 30
and metal junction formation. The
resulting sheet resistance of the
n-type surface and groove regions are
nominally 100 and 10 ohm/square, c
respectively. e 20
d) The rear surface of the cell is coated I
with a thin film of aluminium by vacuum I C.
deposition followed by a high s
temperature sinter. 10
c
(%) f:.
e) The co-metallisation of the front and
rear surface is achieved by a sequence
of electroless plating processes
involving nickel, copper and silver,
with intermediate metal sintering. o 98ftfI
j j j
~

A reflow soldering technique is used to 16 17 18

connect electro-tin plated copper strip to Efficiency (%)
the metallised cell surfaces to affect
series interconnection of the cells. The
module construction is standard vacuum
lamination of a high transmission tempered
glass superstrate, ethylene/vinyl acetate Figure 3. Efficiency distribution of
copolymer (EVA) pottant and tedlar LGBG cells fabricated on BP Solar's
/polyester/tedlar trilaminate rear seal. demonstration plant.

281
following typical electrical
characteristics measured using a Spire a) 100 mm Monocrystalline Cell
Solar Simulator (temperature 25 C, AM1.5
insolation of intensity 1000 Wm-2). cost/Wp wafer
Pmax 60 W
Isc 3.5 A
Voc 22 V
Vmp 18 V
A prototype 60-cell series connected
module is being evaluated for grid
connected applications. With a rated
output of 100 Wp, the power density of royalties
this module (159 Wp/m2) is higher than any
known commercially available flat plate indirect
solar module.

". ENVIRONMENTAL TESTING olher materials

Modules made from the demonstration plant labour energy

have been tested under a range of
accelerated environmental conditions
including thermal cycling between
temperatures of -20 C and 80 C, high b) 125 mm Polycrystalline Cell
humidity and intense illumination. All
modules exhibit the high stability
experienced from crystalline silicon cell cost/Wp saving
technologies.
Modules made during the development
programme in Australia have now undergone
12 months continuous outdoor exposure (in
spain) with no sign of degradation within
measurement error (Table 1).
1Ih;; :-----. royalties
TYPE DATE Isc/A Voc/V FF(%) wafer indirect

Mono- 13-03-90 1.68 21.8 79

cryst- 06-06-90 1.66 21.8 79
al1ine 13-12-90 1.66 22.0 79
01-04-91 1.66 21.9 79 olher materials labour

Poly- 13-03-90 1.46 20.6 77

cryst- 06-06-90 1.45 20.7 77
alline 13-12-90 1.45 20.7 77
01-04-90 1.46 20.7 76 Figure 4. Pie charts showing the cost
break down of BP Solar's LGBG cellon a)
100 mm pseudo-square monocrystalline
Table 1. output of monocrystalline and wafers and b) projected cost on 125 mm
polycrystalline LGBG cell modules during square polycrystalline wafers. The
12 months outdoor exposure trials. polycrystalline cell shows SUbstantial
cost savings which would result in
module fabrication costs of below
Currently, modules are undergoing BP Solar $3.00/Wp.
Espana's extensive test programme for new
product qualification. This subjects 50
modules to a combination of thermal the USA. The combination of low cost
cycling, high humidity and high polycrystalline wafers and high
illUmination, and outdoor exposure over a performance LGBG cell technology offers
minimum period of 3 months. Interim the potential for reducing the fabrication
results indicate no signs of degradation. cost of solar modules significantly below
$3.00/Wp using present-day technology with
proven reliability.
5. A LOW-COST COMBINATION

The highest photovoltaic conversion 6. CONCLUSIONS

efficiency for this technology is achieved
using monocrystalline silicon substrates. BP Solar have developed a cost-effective
However, as seen from figure 4(a), the 100 production technology for the fabrication
mm pseudo-square wafer cost dominates the of laser grooved buried grid solar cells.
overall cost of cell fabrication. Photovoltaic efficiencies of 17-18% have
Polycrystalline wafers (125 mm square) on been achieved on a newly commissioned
the other hand offer the potential for demonstration plant in Madrid, Spain.
significant cost reduction as can be seen
from figure 4(b). Development work is Prototype solar modules with rated outputs
underway to optimise process conditions on of 60 and 100 watt have been produced
polycrystalline silicon wafers produced at which, at a power density of around 159
BP Solar's recently commissioned plant in Wp/m2, are believed to be the highest

282
performance high-power flat plate solar
modules developed to date. Environmental
testing of these modules is well advanced
with over 12 months rooftop exposure
having been achieved with no loss in
performance. This demonstrates the
stability of BP Solar's high efficiency
laser buried grid technology.

7• REFERENCES

/1/ Australia Patent 570309, Buried

Contact Solar Cell, SR Wenham and MA
Green, 26 March 1985.
/2/ MA Green, PVSEC5, Kyoto 1990, p603-5.
/3/ MA Green, High Efficiency Silicon
Solar Cells, Trans Tech Publ., p170-5.

8. ACKNOWLEDGEMENTS

BP Solar gratefully acknowledge financial

support during the development of this
technology from; Centro para el Desarrollo
Tecnologico Industrial (CDTI)- Espana, The
Spanish Ministry of Industry, Commerce &
Tourism and the Department of Energy of
the New South Wales Government, Australia.

283
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DEVEWPMENI' OF SOIAR CELL Ml\NUFACI'URlN3 TECHliOLCGY UNDER THE SUNSHINE PROJECI' OF JAPAN

C. Yokoyama, K. Kuno, S. Sawadaishi, N. Chyu, T. Kuroda, H. Suginoto

Solar Energy Department

New Energy and Industrial Technology Developrent Organization (NEJX))
Sunshine 60 Building, 3-1-1 Higashi-Ikebukuro, 'I'Oshiroa-ku, Tokyo 170
Japan

ABSTRAcr

NE!X) has been praroting a photovoltaic R&D program entitled "TOCHNICAL DEVEIOPMENT FCR THE
PRACI'ICAL USE OF THE PBJrOVOLTAIC PCWER GENERATIN:; SYSTEM" under the Sunshine project of the
Ministry of Internatiooal Trade and IndUStry(MITI) since 1980.
In this paper, recent activities related to solar cell manufacturing technology and new
targets of the phot~ltaic R&D program are described.

1. THIN SUBSl'RATE OOIJ\R FABRICATICN 'I.'EX::tlNJI(X film )Xllycrystalline Si was undertaken. And tandem cell
were fabricated. So far, a 13.5% efficiency has been
Thin substrate solar cell manufacturing technology obtained with a cell area of 25mn'.
consists of development of silicon refining, especially In order to produce amorphous/CCJIIlXXlIId semiCCllductor
of boron removal to produce high quality silicon; tandem cells, study of fabrication for large area
development of electro-magnetic casting process to tandem cells with 30 CID X 40cm size was curried rut. And
produce high p.rrity substrate;and developnent of a cast a trial module of a CdS/CdTe cell was fabricated as a
ribbcn prooess to produce high speed substrate cells and first step of a practical module. So far, a 8.1%
thin substrate cells and large area cells to fabricate efficiency has beeri obtained with a cell area of 1200 cnf.
high efficiency solar cells.
Romoval of boron in silicon is fundamentally 3. AM::RPIOJS SI OOIJ\R CELL FABRICATICN TEOlNJ.[(X;Y
investigated by DC arc treatment and plasma arc
treatment. Boron =ncentration was decreased fran 5.7 IIJrprphous solar cell fabrication technology consists
to under 0.5ppmw by 1.5 hour plasma arc treatment on of improvement of the thin film de)Xlsition process to
600g silicon. This result is able to attain removal of produce high-quality solar cells; simulation of light-
boron. induced degardation to produce high-reliablity solar
As for the electro-magnetic casting process, cells; and developnent of thin film deposition process
esrtablishment of electro-magnetic casting method by and patterning process to produce large-area solar
the =ld crucible was undertaken as to confirm the ingot cells.
casting of 75rnn square cross section using No.1 electro- Regarding high-quality manufacturing production
magnetic casting furnace installed last year, and the technology, research is being pursured into the
crystalline properties of the ingots were evaluated. In technology of p-Iayer quality, the super chamber to
addition, the designing and the installation of the No.2 improve the quality of the i-layer and the p/i buffer
furnace for the production of a 100mm X 100mm square layer, and the technology of the deposition technology
ingot were performed. So for, 13% to 13.5% efficiencies of repeating thermal CVD/H:. plasma eX)Xlsure. So far, a
were obtained which were equal to conventional mold 10.2% efficiency has been obtained with a cell area of
using cast silicon. 100cnf •
As for cast ribbon process, in order to carry out As for sumilation of high-induced degardation,
basic experiments for producing thin and large silicon research is being pursued into the technOlogy of
sheets, basic experimental apparatus for producing progressive cycles of light-pulse treatment and anneal
200rnn width silicen sheets was constructed. As a result, recovery treatment for tandem structure cells, and
it was confirmed that silicon sheets which have uniform clarifing the relation between the light-indnced
thicJmess were constantly obtained. degardation effects and i layer thickness for triple-
With respect to thin substrate solar cells, research layer tandem structure cells.
is being pursued into the new technology of grooved With respect to process technologies such as
surface, buried contact and backside rib structure of =ncerning large area production and patterning, Voc
preventing a breakage. Thin substrate solar cells were lowering was suppressed by improving the deposition
also fabricated. So far, a 16.4% efficiency has been condition for a-Si films applying a textured snO, with
obtained with a cell area of 100 cnf. high reflectivity and steep valleys an the surface.
with respect to large area solar cells, research is Also, examination regarding laser patterning
being pursued into the technology of uniform junction technologies and applying wide unit cells as trial
and optimizing the electrode pattern of the front manufacturing process of large area tandem structure
surface. So far, a 15.6% efficiency has been obtained sub-irodules was carried out.
with a cell area of 225 cnf.
4. SUPER IllGH EFFICIENCY OOIJ\R CELL DE.VEIa'MENI'
2. TANDEM OOIJ\R CELL FABRICATICN TEOIIDICGY
The above technology developrent is being undertaken
Tandem solar cell manufacturing technology consists with the objective of reducing the cost of solar cells
of developnent of cell design and high p.rrity thin film to~100-200/wp by FY2000. This work is expected to
deposition technology to produce amorphous/)Xlly-Si and p:lpUlarize solar cells for caunercial IJOIler swrces.
amorphous/canpound semiconductor tandem cells. HOwever, much higher efficiency and cost reduction
In order to produce amorphous /poly-Si tandem is necessary for a really wide popularization.
cells,study of fabrication techniques for thin film NEDO is therefore planning to pranote technology
)Xllycrystalline si Solution growth technique of thin development of an even higher efficiency solar cells

284
fram FY1990, as an addition to the four year reseach 5.A~
program started in 1989.
'!'his technology developnent is a lcng range research We >O.1.ld like to thank the people ~o were concerned
program for FY2010, and takes aim at extremely high with the development of solar cell manufacturing
efficiency. The target is 30% efficiency for technology under the SUnshine Project of Japan.
crystalline silicon solar cells, 20 ~40% efficiency for
compound semi=nductor solar cells and 20% efficiency
for amorphous solar cells. CUrrent research targets
(for 1992) consists of a basic examination for high
efficiency cell design, high fabrication process, new
sanicooductor' materials, etc.

~Y1992 Targets for Solar Cell Manufacturing Technology

Develcprent ltan Target ( FY 1992 )
....
00 High p.1I'ity substrate Drp.Irity cxn:altrati.en 'lhickness of substrate :
.:: fabricatien technology Oxygen : 2Xl0" atm/cnt 200.um (cell efficiency: 18% (10Q:nf »
0
c:: ~ CarIx:.n : 5X10' 6 atm/cnt Diffusien length : 150 .urn
.., '"....
-
..c:: Ingot size : 20 cmX20cm
Q)
e-
....en_
. 0 OJ
High-speed substrate Substrate size : 2OcmX2OcmXO.lrrm
.....-
c:: =>u
0 en
.... fabricatien technology Profu::ticn sp;!E!d : 1 sheet/min
Cast ribbcn netOOd Cell efficiency: 14%(10Q:nf)
..,'" .=~

-.-
..c:: 0
e-en
..c New cell design ana Single crystalline cell efficiency : 23% (.bf )
'"
0=- fabricatien technology Poly-aystalline cell efficiency: 18%(10Q:nf)' 17%(225:n/)
--
u
w

....
-a;
u
Amorphous/poly-Si
tarrlem cell technology Gell efficiency : 14%(10Q:nf)

-'" -....'"
0
ell
..., 0
en Amorphous/c:x:np::uOO
..,w aw sanicx:n:luctor tandan Cell efficiency: 13%(1,20Q:nf)
c:: ...,
...,'"
cell technology
...,
> c::
e-'"
High quality Gell efficiency : 12% (10Q:nf ), 13%(1cnl)
....
-- - 00 D;gradatien ratio : less than 10%/year
0
w 0
large area cell Cell efficiency: 10%(1,20Q:nf)
u c::
.... .., ..c:: D;gradatien ratio : less than 15%/year
-'"
o
en
OJ
e-
c:: High reliability D;gradaticn ratio : less than 5%/year
0 Initial efficiency: 1l%(10Q:nf) ,12%(1cnl)
en
=> ....
..,'"
.-....
0
..c:: Basic technology TraJ1SFeIency : 85%, Sheet resistivity: 5Q/O, Haze ratio : ±10%
Co
.... ..c a)Crng;llex transprrent Substrate size : 1,20Q:nf, rurability against plasrra : ±10%
..., '"
0
e anhJctive film Cost : ¥50 or less per 10Q:nf(in case of lOMN d81\3Ild)
0=-
............ _-_ .................... ....... _-- ...... _-.- ........ - ... _----_ ........... __ ... ........... -- ............ --_ ........ _-_ ..............
b )NarrGl barxigap Cell efficiency : 4% (1cnl )
I!B.terial (using an a-Si film of 5000A thickness as a filter)

....., -;:;
w
Crystalline silicen
solar cell research
Basic M3nUfacturing Techrnlogy
Investigate high prrity substrate ana cell design
c: e !leN
.-..,
w c-

.... ->
0
w
.... ~ sania:n:luct:or Basic M3nUfacturing Techrnlogy
=
OJ

-- solar cell research Investigate l1E!Vl senia:ulnct:or materials

0

..c::
.-eo
..c:: u
w
.... .... l\noqix:A.Is solar cell Basic M3nUfacturing Techrnlogy
'"=> '"0
Co

en
-
CJ:l
research Investigate high fabricatien process ana !leN cell design

285
10TH EUROPEAN PHOTOYOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

REMOVAL OF BORON FROM MOLTEN SILICON BY ARGON-PLASMA MIXED WITH WATER VAPOR

H. Baba, N. Yuge, Y. Sakaguchi, M. Fukai, F. Aratani and Y. Habu

High-Technology Research Laboratories, Technical Research Division,
Kawasaki Steel Corporation
1 Kawasaki-cho, Chiba, 260, Japan

ABSTRACT
Removal of boron from molten silicon was investigated using Ar plasma
mixed with water vapor in view of developing a new process for solar-grade
silicon (SOG-Si). The raw silicon was induction-melted and blown with a plasma
flame of Ar gas with water vapor. It was concluded that the concentration of
boron in silicon was decreased from 20 ppmw down to 0.1 ppmw or less by the
plasma treatment for 300 minutes, and the rate of removing boron is
proportional to both of the concentration of H20 in Ar plasma and the reaction
area on the molten surface formed by blowing of the plasma. Presumably, the
boron atoms were removed as volatile boron oxides. The kinetic analysis
revealed the first order character of the reaction.

1. INTRODUCTION 3. EXPERIMENTAL PROCEDURE

Kawasaki Steel Corporation and Nippon Sheet In the experiments, the silicon was melted in
Grass Company Limited have been jointly developing silica crucible, Ar plasma contended H20
a process for producing solar-grade silicon (oxidizing agent) was blown on its surface, and
(SOG-Si) by a carbothermic reduction of silica reacted boron in molten silicon and evaporated it
under a sponsorship of New Energy and Industrial as volatile boron oxides. Silicon used in the
Technology Development Organization (NEDO). This experiments was adjusted by mixing
process consists of the production of pure raw semiconductor-grade silicon (SEG-Si) with MG-Si
materials (silica and the carbon reductant), the which included of boron about 30 ppmw.
reduction of silica and the refinements of The experimental apparatus is shown in Fig.2.
silicon. The purity of the raw materials should The plasma torch was a non-transfer type, 30kW at
be as high as possible, since almost all the maximum output. The distance from the torch to
impurities in raw materials remain in the reduced the surface of molten silicon was kept 60mm
silicon. In particular, the concentration of constant. Ar or Ar-23%He gas was used as plasma
boron has to be enough low in raw materials. gas and H20(g) was added in it as oxidizing agent.
However, the purification of the raw materials The concentration of boron in molten silicon
costs too much to produce SOG-Si at low cost. during the experiments was measured by ICP-AES
Therefore, a process for removing boron from the analysis with pin samples which were pulled out at
refined silicon at low cost has been anticipated. regular intervals. The temperature of molten
As for the removal of boron from silicon, it silicon was measured by soaking a themocouple in
was reported by Theuerer 2 ) in 1956 that boron was the bath after the sampling and stopping the blow
eliminated by using zone melting process in an of plasma and H20. It rose at the begining of the
atmosphere of Ar with H2/H20 and that the experiments and attained the constant range. Flux
resistivity of the silicon could be adjusted. (CaO/CaF 2=1:1) was put into the molten silicon
Other methods employing a treatment of CaO-Si02 after the silicon melted down in case of
system slag was tried 3 ,4). The results were experiments for evaluating the effect of the flux.
unsatisfactry because of low partition ratio of The experimental conditions are shown in Table 1.
boron between the slag and silicon. Madigou et
al. 5 ) showed in 1989 that the boron in metallic 4. RESULTS AND EXAMINATIONS
silicon (MG-Si) was eliminated to a concentration (1) Removal of boron from silicon
as low as the one required for SOG-Si by the zone The experimental results are shown in Fig.3
melting of solid silicon in an atomosphere of Ar and Fig.4. The concentration of boron in molten
using plasma, with a high frequency, which was silicon was decreased with the time of plasma
added small amounts of H2 and 02' treatment and became to 0.1 ppmw or less of the
Recently, Suzuki et al. 6 ) carried out a detection limit of analysis. In addition the more
similar experiment in which solid silicon was the concentration of H20 added is, the faster the
heated with high-frequency plasma of Ar/02 or rate to remove boron is. Boron was removed
Ar/C0 2 . One of the result was that the removal of smoothly from 0 to 7.2% with the concentration of
boron accompanied a significant vaporization loss H20. However when H20 was added 7.2% or more the
of silicon. reaction removing boron stagnated because the film
The purpose of the present investigation is of silica formed on the molten silicon.
to establish a mass-productive process for The experimental datum are plotted in Fig.5.
removing boron from molten silicon by the In this figure (a) means the datum with only H20
treatment of plasma of Ar/H20. in plasma and (b) means the datum on the other
conditions. It shows that the experimental datum
2. THERMODYNAMIC CONSIDERATIONS distributes near the linear equation. The
The equilibrated vapor pressure of boron reaction with the removing of boron is expressed
oxides and hydrides are shown in Fig.1. The as Eq. (1),
values in the thermodynamical data table of JANAF - d [BJ / dt = k [BJ ,
are used for the calculation. The equilibrated In [BJ = - k t + In [BoJ , ..... (1)
vapor pressure of boron oxides or hydrides are [BJ,[BoJ boron concentration in silicon
lower than that of SiO to be 4000K. The higher (Bo : initial) ppmw,
the temperature is, the higher the pressure is. t treatment time sec,
The pressure of BO is higher than that of other k rate constant of the reaction
boron conpounds. Therefore, it will be very removing boron sec-I.
useful for the removal of boron from molten The relationship between the concentration of
silicon to use plasma supplying oxidizing agents H20 and the rate constant k with the reaction
to molten silicon. Because plasma process can get removing boron is shown in Fig.6. The rate of
high temperature easily. removing boron was in directly proportional to the

286
concentration of H20 to 7.2%. As for this, unidirectional solidification, Ti and Ca are
Theuerer 2 ) reported different effect of H20, that eliminated by the treatments of high temperature,
the rate was in proportional to square root of the and C is decarburized by treatments of the
concentration of H20 and its addition was limited filtration 7 ) and the oxidation8 ).
to 1.2%. The figure shows that the experimental Therefore, the present way with the
datum distributes near the same linear equation above-mentioned purification will make it possible
with only H20 in plasma in spite of using Ar-He to produce SOG-Si from MG-Si of suitable grade.
plasma or graphite crucible. It was regarded as
effect of supplying oxygen from silica crucible 6. CONCLUSIONS
that boron was removed with no addition of H20. Removal of boron from molten silicon by the
Therefore, the Eq.(l) is renewed as Eq.(2) by the treatment of Ar plasma mixed with water vapor has
results of Fig.6, been investigated and the following conclusions
In ([B] / [Bo]) = - k' (%H 20) t, ..... (2) are obtained.
[B] ,[B 2 ] boron concentration in molten (1) Boron in silicon is able to be removed to
silicon (Bo : initial) ppmw, 0.1 ppmw or less by blowing with a plasma
k' rate constant of the reaction of Ar/H 20.
removing boron ~ec-1, (2) The concentration of boron in the molten
t treatment time sec,.., silicon is decreased obeying the first
%H 20 the concentration of H20 in order reaction.
the plasma %. (3) The rate of removing boron is proportional
The rate constant k' was IxlO- 4 sec.- 1 read to both of the concentration of H20 in Ar
from the incline of linear equation in case of plasma and the reaction area on the molten
600g silicon was melted with no addition of the surface formed by blowing of the plasma.
flux. (4) Vaporization loss of the molten silicon
The addition of the flux made k larger and during the treatment remained less than 5
slightly promoted the reaction removing boron from mass %/hr.
silicon. To summarize, the present investigation has
Consequently, the way of adding flux will not be established a way to produce SOG-Si from MG-Si of
suitable for practical manufacturing process of suitable grade at low cost.
SOG-Si, considering contamination from the flux.
To examine where was the reacting interface, 7. ACKNOWLEDGEMENTS
the relationship between volume of the molten We express our thanks to NEDO for financial
silicon and rate constant k of the reaction is support and technical discussion. We also express
shown in Fig.7. Two size of silica crucible were our sincere gratitude to Prof. N. Sano of the
used for this examination, i.e. 70mm of inside University of Tokyo for his enlightening
diameter was used for 260cm3 (600g) of silicon and discussion and encouragements.
120mm used for 1043cm3 (2400g). In order to keep
the area of depression blowed by plasma constant, 8. REFERENCES
the distance between the plasma torch and the 1) F. Aratani et al., Proc. 9th EC
surface of molten silicon was kept constant in Photovoltaic Solar Energy Conf., Freiburg,
this experiments. Proportionality of k to l/V (1989), 462
suggests that removal of boron reacts at the 2) H. C. Theuerer, J. of Metals,
depression. October(1956), 1316
In addition, there was no contamination from 3) J.Dietl, Proc. 8th EC Photovoltaic Solar
cupper plasma torch (Table 2). Energy Conf., (1988), 599
4) R. M.ellstrom et a1., PTC/SE88/00469
(2) Silicon loss (Kemanord Industrikemi AB)
The relationship between the concentration of 5) N. Madigou et al., Proc. 9th EC
H20 in plasma and the loss of molten silicon is Photovoltaic Solar Energy Conf., (1989),
shown in Fig.8. The vaporization loss of the 773
molten silicon during the treatment remained less 6) K. Suzuki et al., J. of the Japan ins. of
than 5 mass %/hr at 5.0% concentration of H20 in metals, vol. 54, No.2 (1990)168
plasma, caused by the generation of SiO. It was 7) S. Suhara et al., Current Advances in
verified that the loss of the molten silicon Marerials and Processes, Japan, (1989), 365
caused by the treatment was fairly small compared 8) N. Yuge et al., Current Advances in
with the thermodynamical considerations or the Materials and processes, Japan, (1989), 366
report of Suzuki et al. 6 ). This treatment became
a high yield process, because small amount of SiO
evoluted from the molten silicon surface except
the reacting interfacial area blowed by high
temperatured plasma gas.
Figure 9 shows a model of removing boron from
molten silicon, based on the above-mentioned
results. Removal of boron from molten silicon
occurs at the depression which is formed by
blowing plasma of Ar/H20 on the molten silicon
surface. Presumably, boron atoms were removed as
boron oxides.
It was regarded that both of the transfer of
boron through the molten silicon to the reacting
interface and supplying H20 from plasma to there
controlled the rate of reaction.

5. APPLY TO PRODUCING SOG-Si

There are comparatibity high-pured MG-Si in
the market. Table 3 shows concentration of
elements in the MG-Si. These elements except for
boron are removed from silicon by the purificating
technique which has been developed up to now. For
example, Fe and Al are removed by the

287
 a ll = I
[ a. =If' O' /p1I)
P t. ~ Si(U I I,{I! = SiI,{s.1l
P,.

,,-- -' --
-- --
-1.11 (III) _ - - - - - - 510

51

'" "
0-
00
o
......

-s
30 60 90
Time (min)
BH
Fig. 3 Experimental results
3100 4000 1.0 .--,.......,......,...,.~~--r--,-...,..."T"""T""..-,.-,.-r-;
Temperature (K) 0.8

Fig.1 Equilibarated vapor pressure of

boron oxides and hydrides

§: 0.2

0.1 --i----o- .
\ ~tecrion
. runit. of the an~ySis
.
0.05 I..-L-J'--I--I.--'---'---'--'--'--'--'--'--'--'-..........
o 60 120 180 240
Si02 crucible Time (min)

Fig_ 4 Experimental result

Graphite
plate

100 ~--r;;}~~:::o:;::-;:r:::!:::~
: 600s. AI+ He
t::. : 2AOOJ •AI
}sWca
crocible
50 0 : 600s. AWl
. : 600s Gnphitc crucible

t Gas outlct

Fig. 2 Experimental apparatus I

..:; 5.
Table 1 Experimental conditions
§:
2.
[B]initi:d 2 - 28.6 ppmw

Silicon 600 or 2400 g

Crucible Silica or Graphite 30 60 90 120

Time (min) Time (min)
Treatment time 30 - 300 min. (a) Datum with only H20 in plasma 3'" (b) Datum on the other conditions

Plasma gas Ar • Ar/l'le : 15 litre/min.

Fig. 5 Experimental results
11,0 0-7.2 %

Flux SlI'7uCal\ - 50'7.CaO : IOOg

288
 8~~=c~~~~-r-C~~~1
oOft. :: A,A, • He 1Sili.. ~
o : Artf1ux ~ 6
6 .• : A, G..pllile ....... 5
g o Si: 600g
.~
~ 4
'0 4 .§'"
'0 3
...
E
o Si : 2400g 6

S
'::l
.~ 0 L..-."--"--"--"--,,--,,--,,--,,--,---,

H20 contcnt (%)

10 ..
>
8. 0 2.0 4.0 6.0 8.0
H2O in the plasma gas (%)
10,0

Fig. 6 Relationship between the concentration of H20

Fig. 8 Relationship between
and the rate constant k
the concentration of H20 in plasma gas
8
and vaporization loss of the molten silicon
H2O : 4.4-5.3 %
6
o Silica Q'Ucible
':' • Graphilc Q'Uciblc
~
...0 4
.......
.>0:
2

°0~--~--~--~---L--~5
2 3 4 Plasma name
lO'/V (em')

Fig.7 Relationship between volume of the

molten silicon and the rate constant k

Table 2 Cupper contamination from plasma torch (ppmw) The

Raw elememts H2O Cupper concentration reacting Stirring of
Silicon contents Initial ' Final interfac molten
SEG-Si 0% <0.1 <0.1 silicon
Cu
(9-N) 2.8% <0.1 <0.1

Silica
Table 3 Concentration of impurities in the MG-Si (ppmw) crucible Graphite plate
B Al Fe Ti Cu Ca Fig. 9 A model of removing boron
30-40 500-900 260-400 10 2 60 from the molten silicon

289
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PAssrVATED SEXrCRYSTALLrNE SrLrCON SOLAR CELLS:THE EFFECT OF HYDROGEN

PLASMA PARAMETERS

M.A.G. Soler and A.M. Andrade

Laboratorio de Microeletronica - Escola Politecnica
Universidade de Sao Paulo
Phone: (11) 8159322 ext. 3256 - Fax: (11) 2114308
Caixa Postal 8174 - 05508 - Sao Paulo - Brazil

.ABSTRACT. This work presents recent improvements in the hydrogen

passivation on semicrystalline silicon solar cells. Hydrogen was
incorporated into the substrate from the back side of the solar cell
by r.f. plasma, in a capacitively coupled plasma enhanced remote CVD
reactor. The effects. of the plasma parameters on the hydrogen
penetration are analised and a correlation with the performance of the
solar cells is done. The r. f. power and the hydrogenation treatment
time were varied.
Hydrogen introduction effects in the solar cells parameters were
evaluated from their characterization before and after the treatments.
The results indicate that treatments carried out on thin cells during
4 hours is sufficient to passivate all the defects able to be
pasivated using this technique. After this hydrogen plasma treatments
the short circuit current density showed relative increases between
20% and SO% for thin base s~lar cells wit~ initial photocurrent
values varying between 15 mA/cm and 21 mA/cm • The study inVOlving
variations on the r.f. power density showed that this hydrogen
passivation time could be further diminished.
Solar cells photocurrent and photovoltage were increased after
hydrogenation whithout generation of front side defects, which reduce
the fill factor, and, consequently, the conversion efficiency was
improved. The results showed that back side hydrogenation for
semicrystalline silicon solar cells is very effective.

1. INTRODUCTION the minority eleqt"on diffusion lenghts.

The. hydrogen· pressure was changed from
The technical feasibility of many 200 mtorr to 1000 mtorr and the treatment
solar cell plants installed for time was .varied from 1 to 6 hours for
terrestrial applications fabricated from .tnick cells and from 1 to 4 hours for thin
crystalline silicon material has been cells; The effect of the hydrogen pressure
demonstrated. Economic reasons however was also analysed. ~hin and thick cells
have not permited the widespread use of were treated during 4 hOYs at a r. f .•
solar cells as means of generating bulk power density of 500 mW/cm for hydrogen
electricity. The recent progresses in the pressure varying from 200 to 1000 mtorr.
development of smicrystalline silicon The results from the relative increase in
solar cells (1) shows that this material the photo current as a function of the
could replace the single crystal material hydrogen pressure for thick and thin cells
in the efforts to diminish the· costs of treated during 4 hours showed that the
solar cells. Commercial ribbons and cast improvement is more expressive for thin
wafers are typically grown at high rates cells and for higher hydrogen pressures.
resulting in high defect densities. In It was verified that the backside
order to improve the material quality, hydrogenation utilizing r.f. hydrogen
several techniques have been developed to plasma is very effective specially for
minimize the effects of the recombination cells fabricated on solar grade
of minority carriers at the structural substrates. with the objective of
defects such as dislocations and grain achieving the optimum conditions for
boundaries. practical applications on the solar cell
Hydrogen introduction for defect fabrication manufacturing process, other
passivation in photovoltaic devices has hydrogen plasma process parameters were
been used but its behavior in silicon is analysed. Process variables were the r. f.
still . being studied (2). Several power and the treatment time.
techniques have been used for the hydrogen
introduction as low energy ion 2. EXPERIMENTAL PROCEDURES
implantation (3,4) or r.f. hydrogen plasma
(5) • The samples were prepared on large
In a previous work (5), bulk grain p-type semicrystalline material
passivation treatments through backside (produced by Heliodinamica S.A. Brasil and
hydrogenation were done on semicrystalline Wacker Chemitronic GmbH, Germany) and
silicon solar cells by r.f. plasma using a also on solar grade silicon (grain size
capacitively coupled glow discharge 1mm) substrates with resistivities in the
reactor. The short circuit current density range of O.S-to ~.o ohm.cm. Groups of nine
showed relative increases of over 40% and cells of 5xS mm area each, 300 ",m or
it was verified an improvement in the 200 ",m thick were fabricated on substrates
substrate quality by the enhancement of of 25x2S mm2 area. A n+/p junction, about

290
0.4 ~m deep was formed by conventional be passivated with the hydrogen
thermal phosphorus diffusion. The introduction were passivated.
aluminium grid back contact, which defines Observed improvements in the solar cell
the actual hydrogenation area, was parameters treated with the same
prepared by the lift-off photolithografic hydrogenation conditions but with a fixed
process. The front contact is the time of 4 hours are shown in table 2.
conventional Ti-Pd-Ag, also defined by the TABLE 2 - Changes in the solar cells
lift-off process. The cells were prepared photocurrent before (b) and
wi thout any antireflection coating. Prior after (a) the hydrogenation.
to the hydrogenation, the cells were Plasma conditions: r.f. power
characterized by measuring their dark and density 500 mWjcm2~ H2
illuminated I-V characteristics and pressure 1000 mtorr~
spectral response. substrate temperature = 220oC~
treatment time = 4 hours.
Hydrogenation
CELL Jsc(b) Jsc(a) JjJsc(b)
Bulk passivation treatments by backside NUMBER (mAjcm2) (mAjcm2) (%)
hydrogenation were done on thin (200 ~m)
semicrystalline silicon solar cells. The 1 19.5 23.4 20
hydrogenation was done introducing the 2 18.0 24.0 33
cells inside a capacitively coupled plasma 3 17 24.4 43
en~anced CVD reactor (6) (13.56 MHz), 400 4 20.8 24.8 18
cm sample holder area at a substrate 5* 15.2 23.4 54
temperature of 220 C.o The r.f. ~ower
density wa~ changed from 500 mWjcm to * (Solar cell fabricated on solar grade
1000 mWjcm and the treatment time was silicon substrate) •
varied from 2 to 6 hours.
3. RESULTS The results showed also that the
photovoltage had an improvement around 2%
The solar cells parameters were after hydrogen treatments and the fill
obtained from their dark and illuminated factor remained practically unchanged.
I-V characteri~tics using an ELH-type lamp In order to further improve the results
at 100 mWjcm. The temperature of the and optmize the hydrogen penetration and
samples was kept at 25 ±l o C during the the treatment time, the r. f. power was
measurements. varied. The effect of the r.f. power was
The first analysed parameter was the analysed on solar cells treated during 2
treatment time. The plasma was done at a hours at a hydrogen pressure of 1000 mtorr
pressure of 1000 mtorr and a r.f. plasma for r. f. pow'Sr density varying from 500
power density of 500 mWjcm 2 • The values to 1000 mWjcm .
obtained for the short circuit current
density before and after hydrogenation
treatments and the relative variation of
photocurrent, defined as the increase in
the photocurrent divided by the value of
the photocurrent before the treatment, are
shown in table 1. In order to compare the 15
effect of the different treatment times, 0
only cells with similar values of ~
~ 0
photocurrent were considered, due to the
.0
fact that hydrogenation is more efficient
in low photo current cells. U
10
.....'"
J 0
u
J
<3
'"
TABLE 1 - Changes in the solar cells
photocurrent before Jsc(b) and 5
after Jsc(a) hydrogenation as a
function of passivation time.
Plasma conditions: r.f. power
density = 500 mWjcm 2 , H2
pressure = 1000 mtorr, substrate
temperature = 220 o C. 500 750 1000 mW
cm2

Fig. 1. Relative variation of the

TIME Jsc(b) Jsc(a) JjJsc(b) photocurrent versus r.f. power density.
(HOUR) (mAjcm2) (mAjcm2) (%) Plasma conditions: time = 2 hr. T = 220 o C,
H2 pressure = 1000 mtorr.

4 17.0 24.4 44 Figure 1 shows the relative increase in

6 17.0 24.1 41 the photocurrent as a function of the r.f.
power density for thin cells treated
It can be observed in the table 1 that during 2 hours. It can be seen that, for
the increase in the cell parameter values this temperature and hydrogen pressure,
is about the same for thin cells treated the improvement is more expressive for
during 4 and 6 hours. These results higher r.f. power density. The results
indicate that all the defects that could also suggest that the treatment time could

291
be even diminished thus contributing to
develop a hydrogenation processes
compatible with low cost solar cell
fabrication processes.
As a conclusion, it is possible to say
that backside hydrogenation utilizing a:
r. f. hydrogen plasma is very effective.
Further research could be conducted to
determine the minimum treatment time,
together with an adequate substrate
temperature which could assure an
optimized attainable passivation.
AClWOLEDGENENTS

The autors wish to acknowledge

Heliodinamica S.A. for supplying
substrates.
This work received financial support
from CNPq and CAPES.

REFERENCES

(1) S. Narayanan, et al., Proc. 21st IEEE

Photovoltaic Specialists Conf., 1990,
p. 678.
(2) J.W. Corbet, et al., Solar Cells,
Holland (24): 127-133, 1989.
(3) B.L. Sopori, et .al., Proc. 21st IEEE
Photovoltaic Speciliats Conf., 1990,
p. 644.
(4) J.C. Muller, et al., Proc. 9th E.
Photovoltaic Solar Energy Conf.,
1989, p. 407.
(5) M.A.G. Soler, et al., Proc. 21st IEEE
Photovoltaic Speciliats conf., 1990,
p. 650.
(6) R. Martins, et al., Proc. 5th E.C.
Photovoltaic Solar Energy Conf.,
1983, p. 778.

292
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

TEXTURE ETCHING OF MULTICRYSTALLINE SILICON

U. Kaiser, M. Kaiser and R. Schindler

Fraunhofer Institut fUr Solare Energiesysteme, OltmannsstraBe 22, D-7800 Freiburg i. Brsg., Germany

ABSTRACT

Random texturing of monocrystalline silicon sheets by means of alkaline etches is a standard method in
solar cell manufacturing. In contrast, polycrystalline silicon cannot readily be texturized by means of a
simple anisotropic chemical etch due to different grain orientations. Therefore, plasma etches and laser
texturizing are being considered.

This paper presents approaches for wet chemical etching of multicrystalline silicon slices in order to
achieve substantial reduction of light reflection. The results obtained so far are comparable to those
achieved by plasma etching and exceed those reported so far in the literature for wet chemical etching.

1. INTRODUCTION etch ants used for Si consist of an oxidizing compound which

forms a surface oxide and a solvent which dissolves the oxide
One of the disadvantages of low cost processing of consecutively. The· oxide nucleation is enhanced at
multicrystalline silicon for solar cells as compared to dislocation sites, and therefore ensures a favourable
monocrystalline silicon is the difficulty of successfully dissolution along dislocation lines by lowering the surface
texturing wafers or sheets for reduced reflectivity. energy and thus increasing the rate of dissolution /1/. So the
Monocrystalline wafers, especially {100} oriented wafers, nucleation of etch pits is not only correlated to surface
can be easily texture etched by inexpensive alkaline etches roughness but also to inherently present nuclei. As in
like KOH or NaOH. These etches work anisotropically and multicrystalline silicon dislocations always are present at a
lead to the well known random pyramids. This scheme does density of about 106 cm- 2. These can be used as nucleation
not apply satisfactorily to multi crystalline material, as the sites for the etch attack to achieve a useful texture. This
crystal orientation varies from grain to grain. Therefore structural etching of dislocations possibly causes well defined
different methods need to be applied in order to reduce light etch pits which with proper shape and etch depth reduce
reflection in multi crystalline material. This paper reports on light reflection substantially.
improved texture etching of polycrystalline silicon material.
A number of different structural etches as well as the
combination of preferential and non-preferential etches has
2. EXPERIMENTAL APPROACH been investigated for texturing of multicrystalline material.
After a cleaning procedure consisting of HN03' water rinse,
Texture etching of monocrystalline silicon relies on a HF, water rinse of the as cut SILSO wafers, they were
distinct feature: the anisotropy of the etch attack due to exposed for different times to one or more etches. Total
different etch speeds on different crystallographic lattice reflection was measured with a spectrometer utilizing an
planes. The etch pits created by this kind of preferential etch Ulbricht sphere.
have crystallographically well defined walls. For the case of
caustic etches the walls of the etch pits are {lI1} planes. Cr03 based etchants, which are preferential structural
The nucleation of the etch attack on the surface is closely etch ants generally, are widely used in defect delineation
connected to surface roughness, minute surface damage and etches. For the single etch step a structural, but non-
to the formation of gas bubbles and their masking very small preferential etchant /2/ is applied which etches all
areas of surface. For {100} oriented Si wafers random crystallographic lattice planes uniformly. This etchant of the
square pyramids are formed with {111} oriented walls. HF-HN03-CH3COOH system etches faster along
dislocation lines and is superior to Cr03 based etchants. The
For our experiments concerning multicrystalline silicon absence of metallic components moreover is regarded as a
material (SILSO) a different approach is favoured. Most beneficial property of any texture etch for further high

293
temperature processing of the material. A comparison with other experiments e.g. /3/ shows
that this improved wet etch is better than most others
reported. In fact, the results are comparable to those
obtained by plasma etching, which is a process requiring
lithography. According to our knowledge only texturing by
3. RESULTS AND DISCUSSION
laser scribing up to now results in lower reflection.

The best results in reducing light reflection yielded pure

defect etches of the HF-HN03-CH3COOH system. Figure 1
gives an example of the structure of the surface. 4. CONCLUSION

A difference in etch behaviour between surface Changing from structural preferential etches to non-
damaged and damage free etched material suggests the preferential etches allows further reduction of light
necessity for sufficient numbers of nuclei at the surface for reflection of multicrystalline Si. The experiments so far
the etch attack. Besides lattice dislocations surface damage yielded reduced reflection of light to values well below 15 %
contributes to good results. depending on the etch condition and defect density of the
material.
The dependence of the reflectivity reduction on the etch
time is not straightforward. Short etch times tend to yield
better results than prolonged etching (figure 2).
REFERENCES
The best results obtained so far are given in figure 3.
Two distinct types of reflection curves are given: Graph 1 is /1/ R.B. Heimann: Crystals, growths, properties, and
fairly constant over a wide range of wavelength. For 500 nm applications, Vol. 8 edited by J. Grabmaier, Springer
to 1050 nm the reflection is below 14 %. Graph 2 has . a Berlin 1982
lower reflectance for short wavelengths. For 400 nm to 900 /2/ A.F. Bogenschiitz: Atzpraxis fUr Halbleiter Carl Hanser
nm the reflectivity is below 14 % with a broad minimum of Verlag Miinchen 1967
10 %. 'the reflectivity increases with increasing wavelenghts. /3/ S. Narayanan, S.R. Wenham, and M.A. Green,
PVSEC 4, p.l11, 1989
Both types of reflectance curves are regularly observed.
~'~'r-----------------------------------
The difference is not yet completely understood. It may be
argued that the difference is due to different dislocation
4O.0Cl
densities yielding a higher density of very small etch pits,
sized in the order of about 1 em. This would allow positive
deflection of the light for shorter wavelengths, while for
increasing wavelengths the light is less favourably deflected. 2'

10.00'---------------------------------'
)ISO:IOO 850 1150

Fig. 2: Reflectivity versus wavelength: effect of etch time.

"·~'r-__ ------------------------------Tn
..... (3)
s.

(1)

,...~-------------------------------~
»0 55CI ~ ~ 11500-
Il.&VI\.lHGITH (n-.J

Fig. 1: SEM picture of a SILSO surface structure, treated Fig. 3: Reflection vs. wavelength: (1) type 1, (2) type 2
with a structural, but non-preferential etchant. (3) multicrystalline silicon without any treating.

294
lOTI-! EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

WCALIZATION OF THE SOLID/LIQUID INTERFACE DURING DIRECTIONAL

SOLIDIFICATION OF SILICON BY A PULSE-ECHO ULTRASONIC TECHNIQUE

A Eyer, F. Haas, and P. Schatzle

Fraunhofer-Institut fUr Solare Energiesysteme, Oltmannsstr. 22
D-7800 Freiburg, Germany, Phone: 49-761-4014-0; Fax: 49-761-4014-250
M.Paul
Fraunhofer-Institut fUr Zerstorungsfreie Priifverfahren, Universitat, Gebiiude 37, D-6600 Saarbriicken, Germany
R. M. Knobel
Heliotronic GmbH, P.O. Box 1129, D-8263 Burghausen, Germany
ABSTRACT
An ultrasonic system is suited to detect the solid/liquid interface in a directional solidification process of
silicon. Such processes are used to produce large multicrystalline ingots from which silicon wafers for low
cost solar cells for terrestrial power applications are sliced. Several ultrasonic techniques are discussed in
this paper.
A pulse-echo system has been built and attached to a commercial type solidification machine. The pulses
of a piezoelectric transducer are coupled into the melt via a transmission rod made of an appropiate
material that withstands the temperature and the reactivity of liquid silicon. Several ceramic rods have
been considered, SiC and AL,0.l.. were found to be suited best. Clear echoes from the liquid/solid
interface .have been detected:' ·lhese signals· allowed to calculate the depth of the melt and the
solidification velocity. For comparison the depth was measured with an ~03 sounding rod. Good
agreement was found.

1. INTRODUCTION
For PV power applications cry~talline silicon is the most
widely used material. Besides mono-crystalline wafers from
CZ-crystals, multi crystalline wafers are becoming more and Ultrasonic
more important. They are sliced from large ingots (up to ~ ---- Excllaloon
Transducer Laser
more than 100 kg) which are produced from silicon melts I
by a directional solidification process in appropriate
..
I
crucibles. There are different producers of such material I
and their respective processes differ slightly regarding the Transmission I ~- Inlerferometer
crucible material, the heating configuration and the I I
Rod I I
performance of the solidification process. I I
The quality of the material is significantly influenced by I I CrUCible
these process parameters. Best results are expected from a I
planar or slightly convex solid/liquid interface which travels
at constant crystallization rate from the bottom to the top Silicon
of the crucible. The shape of the interface is s1?ecified by Solid/Liquid
Melt
the thermal environment of the melt which, In turn, is Inter ace
determined by the crucible, the isolation, the heating and
the cooling conditions. Finding the best configuration is a
complex task.
In order to control the solidification velocity the 'problem
is to localize the solid/liquid interface. liqUId silIcon is a
chemically very aggressive high temperature substance and
not penetrable by any kind of electromagnetic radiation ransmlSSlon Rod
besides x-rays (1). Cryslallized
As x-ray radiography is not applicable in a production
line, only an ultrasonic method seems to be suited to detect Silicon
Ultrasonic Transducer
the solidification front.

2. AIM OF THE PROJECT Figure 1: Principle of the different ultrasonic techniques

The aim of the project is the realization of an ultrasonic
s¥.stem for the localization of the solid/liquid interface of
SIlicon and its integration into a standard directional Excitation can occur at the surface of the liquid either by
solidification process for large ingots. sending high power laser pulses onto the surface or by
dipping a piezoelectric transducer into the melt. As no
commercial transducer would withstand the temperature
3. ULTRASONIC TECHNIQUES and the chemical reactivity of liquid silicon, a transmission
rod has to be put between.
From the literature only basic experiments for the Excitation could also take place at the bottom of the
interface localization in different metallic melts and even crucible i.e. at the surface of the already crystallized part of
for silicon are known (2). They show that in the case of the ingot. Since the transducer (or a transmission rod) has
silicon the reflectivity of ultrasonic waves at the solid/liquid to be In contact with the silicon Itself and not only with the
interface is high enough to receive a detectable echo. crucible walls the sealin~ problem is severe at the
The main problem is to couple the ultrasonic waves into beginning of the solidification process. Another disadvan-
the solidifying material. Different techniques exist, a tage is strong scattering of the mcident ultrasonic waves at
schematicaI drawing of these is given in fig. 1. the grain boundaries of the multicrystalline silicon.

295
 For the detection of the signal there are different Several ceramic materials were tested as transmission
possibilities, too. It can be achieved by the same transducer rods in a real solidification process. The experimental set-
as excitation, if a reflected echo from the solid/liquid up is shown in fig. 2. Up to now, ~03 and SiC were found
interface is preferred or by a second transducer if the to be suited well. They are commercially available in
transmitted Signal is preferred. sufficient lengths, show low ultrasonic damping (1x10-3
The sound velocity changes by a factor of two from solid dB/m and 2.5xlO-3 dB/m resp.) and low thermal
to liquid silicon, therefore the transmission time strongly conductivity (6.0'W/mK and 45.0 W/mK resp. at 1000°C).
depends on the portion of silicon that is already solidified. In the case of ~03 the chemical reaction IS problematic
At the surface of the liquid the echo signal can also be because of contamination of the melt by aluminum. In the
received by an interferometer that detects the surface case of SiC different qualities were tested but only sintered
waves excited by the echo. Since in a real system the SiC (without free silicon) showed sufficient low damping.
surface may be unstable due to rotation of the crucible or The diameter of the rods was 19 mm and the length 430
gas flows in the system, the interferometrical technique is mm. A monocrystalline silicon rod would be the best
fairly difficult to realize. choice. It is under test.
Excitation by laser and detection by an interferometer, We used a commercially available digital ultrasonic
that means also by laser beam, is a contactless and measurement unit. A 5 Mhz piezoelectric transducer was
therefore a very ele~ant and clean method, but such a coupled to the watercooled upper end of the transmission
complex and expensive system has never been tested in rod. It is attached to the machine by a feed-through in
high temperature melts. order to adjust its hei~ht in respect to the melt surface. Its
lower end was dipped mto the melt for about 5 mm.
The incident ultrasonic waves are reflected strongly at
4. EXPERIMENTAL SET-UP the lower end of the rod and at the solid/liquid interface.
The actual position of the interface is calculated from time
After several prelimanry experiments using laser differences between the corresponding echoes. The quality
excitation and interferometrical detection, it was decided to of the received interface echo del?ends on the interface
realize a pulse-echo system with only one piewelectric roughness and the reflection coeffiCient. At the solid/liquid
transducer, the ultrasonic waves of which are coupled into interface of directional solidified high purity silicon there
the melt surface by a transmission rod. exists no "mushy-zone" which scatters the incident signal in
This rod, which is dil?ped into the melt at one end and is a non-directional way. The reflection coefficient (R=30 %)
bearing the piezoelectnc transducer at its other end, is the is much higher than in most metals (R=10 %). The well-
most critical part of the system. It has to fulfil the following known ultrasonic velocity of liquid silicon allows to
requirements: calculate the solidification velocity from the detected time
- no or low chemical reaction with liquid silicon differences with high accuracy.
- good ultrasonic coupling to liquid silicon (i.e. it has to As the interface travels from the bottom of the crucible
be wetted by liquid silicon) to the top, the time differences decrease during the
- low thermal conductivity and sufficient length solidification process. The received echoes can be dis'played
to keep the transducer at low temperatures (below on a monitor (see fig. 3) and used for further processmg.
SO°C)
- low ultrasonic damping
5. MEASUREMENTS
Ultrasonic measurements were performed in a real
solidification system which can produce ingots of multi-
crystalline silicon up to 24 kg. Transmission rods made of
SiC and AIP3 were used. After generation of the melt, the
rods were moved to the surface of the melt. When the rod
gets contact with the melt the intensity of the echo from the
lower end of the rod decreases due to the decrease in
reflectivity.
Fi~. 3 shows a sequence of echoes at different times of
solidification. The first picture shows the situation without
contact between the transmission rod and the melt. On the
left side the first echo from the lower end of the rod can be
seen and on the ri$ht side the second one. The echo from
the solid/liquid mterface appears between these two
echoes in picture two. In the following 'pictures the motion
of the interface in an upward direction IS obvious. The time
differences between the echoes from the lower end of the
rod and from the interface decrease continously. The solid-
ification velocity calculated from the detected time differ-
ences is shown in fig. 4. The calculated velocity from
sounding experiments with another Al.,03 rod (diameter
about 6 mm) is also shown in fig. 4. BotIi curves show good
agreement. At the beginning of crystallization, the velocity
is low because of the great heat capacity of the silicon melt.
With proceeding solidification the velocity increases.
The determination of the solidification velocity by such
an ultrasonic system offers the possibility to control the
crystallization rate during the whole process.

6. CONCLUSIONS

Figure 2: Schematical drawing of the experimental set-up We have demonstrated the localization of the
(1 ultrasonic transducer; 2 transmission rod; solid/liquid interface during directional solidification of
3 cooling tube; 4 sounding rod; 5 heater; silicon by a pulse-echo ultrasonic technique. The technique
6 silicon melt; 7 crucible; S heat insulation; is simple and cheal? and could be used to determine the
9 lowering equipment) solidification velOCity. Thus this techniques offers the
possibility to achieve a homogeneous solidification and
higher quality of multicrystalline ingots.

2%
 velocity (mm/min)
1,2r---~------------------------,

1,0

0,8

0,6

0,4

-- sounding rod -9- us-signal

0,2L-------L-------L-------L-----~

o 50 100 150 200

time (min)

Figure 4: Calculated solidification velocity

7. ACKNOWLEGEMENTS
The authors thank Dr. W. Arnold, FhG-IzfP, for fruitful
discussions and Th. ZOllner for the performance of the
solidification experiments.
This work has been supported by the German Federal
Ministry for Research and Technology. The authors alone
are responsible for the contents.

REFERENCES
(1) K Kakimoto, M. Eguchi, H. Watanabe, T. Hibiya,
J. Crystal Growth 99 (1990), 665
(2) R. C. Parker, J. R. Manning,
J. Crystal Growth 79 (1986), 341

Figure 3: Sequence of echoes at different times of

solIdification (t1-t4)
(>-<: echoes fiom the solid/liquid interface;
t1 = 0 min; 1" = 45 min; t3 = 90 min;
t4 = 150 min)

297
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Development of High-Quality P-Type Si Thin-Films by a New SPC

Method and its Application to Solar Cells

T. Matsuyama, M. Taguchi, M. Tanaka, T. Matsuoka, S. Tsuda,

M. Ohnishi, Y. Kishi, S. Nakano and Y. Kuwano
Functional Materials Research Center, Sanyo Electric Co., Ltd.
1-18-13, Hashiridani, Hirakata-City, Osaka 573 Japan
Phone: (720) 41-1261 FAX: (720) 41-0386

Abstract
In order to establish low-cost solar cell technology using crystalline
silicon, it is necessary to develop a low temperature p-o junction fabrication
process by which large-area solar cells can be produced. We have tried to
prepare the p-n junction by depositing high-quality p-type Si. Highly
conductive p-type Si thin-films were prepared at the low temperature of 650t,
using the new solid phase crystallization (SPC) method. In this SPC method,
amorphous silicon (a-Si) films deposited by plasma-CVD were used as a starting
material. These films have a conductivity of more than lXI0 3 (lIcm)-l on the
quartz substrate, which is the highest ever reported on p-type Si prepared by
a low-temperature process. Concerning the surface morphology of substrate, it
was found that the mobility of p-type Si on the textured substrate was 3~7
times higher than that of p-type Si on the flat substrate, where maximum
value was 32.2cm2/Vs at the carrier concentration of 2XI0 2°cm-S • These films
were applied to p-layers in solar cells using textured single crystalline
silicon (c-Si). Using 2cm2 sized p-n junction solar cells prepared by the SPC
method, a high conversion efficiency of 14.2% has been achieved. Therefore, p-
type Si thin-films prepared by the SPC method are of high-quality, and it is
considered that the SPC method has high potential from the viewpoint of low-
cost solar cell fabrication.
1. Introduction which consists of two processes. In the first
Recently, solar cells have gathered much process, boron (B) doped a-Si:H films were
attention as a clean energy source. In order to deposited by the plasma-CVD method on a quartz
make a solar cell for practical use in bulk power or n-type c-Si (,: 1.35~2.25 IIcm). In the second
generation, it is necessary to develop low-cost process, B-doped a-Si:H films were recrystallized
solar cells, keeping the high conversion to highly conductive p-type Si thin-films by
efficiency. thermal annealing in a vacuum at various
In general, solar cells using crystalline temperatures. In our SPC method, there are three
silicon(c-Si) are prepared by thermal diffusion ' ) features:
and ion implantation methods 2). These methods (II The impurity atom (B) is in-situ doped into
are high temperature processes( ~ 1OOOt). Until a-Si:H.
now, we have d~veloped highly conductive p- (2) This process is very simple.
type Si thin-films(> lX1 03 (11 cm)-l) using the solid III Highly conductive p-type Si thin-films with
phase crystallization (SPC) method 3 ). Because large area can be prepared.
this SPC method is a low temperature process of The p-type a-Si:H films were deposited by
600~700t and use a amorphous silicon (a-Si) plasma-CVD. Boron atoms were in-situ doped by
films deposited by the plasma-CVD method, large- mixing B2H6 with SiH .. as a source gas for the
area solar cells can be prepared in which dopant SPC process. The substrate temperature range
atoms are more uniformly distributed in plane was 200~350t. The flow rate of SiH", was 2.5-10
and in depth, compared with the thermal B doped SPC
diffusion and ion implantation methods. Highly-
conductive p-type Si thin-films prepared by the
a-Si -7 p-type Si
SPC method were applied to the p-layer in solar Quartz Quartz
cells using c-Si. This paper discusses the
Plasma CVD Thermal annealing
effectivity of p-type Si prepared by the SPC
method as a p-layer and the potential of the SPC
method as a fabrication technique for solar cells. 11:::','\\\\\\
:::::! .,: :::
Also, mentioned are solar cells with high-
photosensitivity in the long wavelength region.
\J" ,-' ,
I , I ,

L...-_ _ _ _
, \

'fi
' I , I , I

«
¢

2. Experimental (a) a-Si deposition (b) Thernial annealing

Figure 1 shows a diagram of the SPC method Figure 1 Diagram of the new SPC method.

298
sccm and that of B2H6 (1000 ppm diluted by Hz) Considering the reason why the dark
was 4.0 - 50 sccm. RF power density was 95.5 conductivity of the SPC-processed films is
mW/cm2 • Pressure was 0.1 Torr. Thermal maximum at the doping ratio of 5x lO- a for each
annealing was performed for 1 - 7 hours at SPC temperature, the crystallinity of the films
temperatures of 600 - 700'(; in a vacuum. After was analyzed by means of the absorption
the SPC of B-doped a-Si: H on a quartz coefficient. Figure 3 shows the absorption
substrate, the dark conductivity, carrier coefficient of the SPC-processed films as a
concentration, and Hall mobility were measured. function of the doping ratio, where SPC
In the Hall effect measurements,' the magnetic conditions are 650'(;, 3H ; and 700'(;, 3H. The
field intensity and temperature were 5000 Gauss absorption coefficient of the films is the
and 300 K, respectively. The electrode was van smallest at a doping ratio of 5X 10-a , and its
der Pauw configuration. The absorption coefficient value is the nearest to that of c-Si 5 ). So, the
of the films after the SPC process was measured p-type Si at a doping ratio of 5xlO-a shows the
for evaluation of crystallinity, using the highest-quality crystallization in these samples.
measurement method developed in our laboratory·). Therefore, the conductivity of p-type Si is
Transmittance electron microscopy (TEM) was used highest at the doping ratio of 5X lO- a .
for the structural analysis. p-type Siln-type
c-Si solar cells were fabricated by thermally
annealing p-type a-Si:H/n-type c-Si at a
temperature of 650'(; for 7 hours in a vacuum.
The ramp rate in the thermal annealing was 1-40
'(; Is. The collection efficiency of a p-type Sil
n-type c-Si solar cell was measured for
photosensitivity evaluation in the long wavelength
region above 800 run.

3. Results
3-1. Film properties on flat substrate
Figure 2 shows the dark conductivity before
and after SPC as a function of the doping ratio
(BzH61 SiH.). The SPC conditions were 600-700'(;
for 3 hours. The dark conductivity of highly
conductive p-type Si thin-films increases with
the SPC temperature for each doping ratio, and
the maximum value of 2X I0 a (Qcm)-1 was achieved
at a doping ratio of 5xlO-a. Furthermore, a
prominent feature is the increase in dark
conductivity after SPC by 9-11 orders at a c
o
:;:;
doping ratio of 5xlO- a , using the low dark 0-
conductivity film. '-
o
r------------------, en
104 ..D
~
Photon energy (eV)
Figure 3 Absorption coefficient of p-type Si
prepared by SPC method as a function of
doping ratio(B 2 Hs /SiH.).
The Hall effect of highly conductive p-type
Si thin-films prepared by the SPC method was
measured for evidence of high conductivity.

J
Figure 4 shows Hall mobility of the films after
SPC as a function of carrier ctmcentration,
where SPC temperature is 7001: and ramp rate is
changed from 1 'C Is to 40 'C Is. Hall mobility of
the films on the quartz substrate is in the
range of 2.3 - 13 cmz/Vs in a high carrier
concentration of 2X10 2°_2XI021 em-a. These Hall
mobilities are about 20% of that of c-Si in the
same carrier concentration. As mentioned above,
the new SPC method which we have developed is
Figure 2 Dark conductivity of films before and
more effective for the fabrication of highly
after the SPC process.
conductive p-type Si thin-films.

299
 "'-C-Si

•
• ••
-~ Aat 700'C
•
:0
o
~ • I:. Texture) 650'C ••
• Flat
. Aat 700'C
1020 1021 101~~__~~~~uu~~~~~
Carrier concentration (cm-3) 1020 1021
Carrier concentration (cm~
Figure 4 Hall mobility of p-type Si thin-films
prepared by SPC method as a function of Figure 5 Difference of Hall mobility of p-type
carrier concentration, compared with that Si thin-films on textured substrate and
of c-Si. flat substrate prepared' by SPC method,
3-2_ Film properties on textured substrate compared with that of c-Si.
For further improvement of film properties, SPC condition: 650'C, 7hours
a textured substrate was used in the preparation
of highly conductive p-type Si thin-films_ The
surface roughness resembles a range of sloping
mountains, with a peak to valley depth of 10 ~m.
Figure 5 shows the Hall mobility of the
films on the textured quartz substrate after SPC
as a function of carrier concentration, compared
with that on the flat substrate. The SPC
temperatures are 650t and 700t. The ramp rate
is in the range of 1 - 40 t/s_ The Hall mobility
~f the films on the textured substrate is 3-7 Substrate
times larger than that of the films on the flat
substrate, in the high carrier concentration of
2X10 20 - 2x10 21 cm-s . Most of the films on the (a) Textured substrate
textured substrate prepared at 650t have greater SPC condition: 650'C, 7hours
mobiiity than that of the films on the flat
substrate at 700t_ The maximum mobility is 32.2
cm 2 /Vs at a carrier concentration of 2x10 2 °cm-s
on the textured substrate. This value is 64% of
that of c-Si in the same carrier concentration.
In order to investigate the difference between
both subst"rates, structural analysis was
performed by cross-sectional TEM images_ Figure
6 shows cross-sectional TEM photographs of the
films on both substrates after SPC, where SPC
conditions were 650t for 7hours. The grain size Substrate
of the films on the textured substrate after SPC
is larger than that on the flat substrate. For
example, the maximum grain size is about 3 - 4 (b) Flat substrate
pm for the textured substrate, compared with 1- Figure 6 Cross-sectional TEM image of p-type Si
2 pm for the flat substrate. Therefore, it was prepared by the SPC method.
found that the textured substrate was effective
for the preparation of p-type Si with large grain of p-type Siln-type c-Si solar cells, compared
size. with the conventional solar cell. The ramp rates
3- 3. Solar cell performance are 1t/s, 10t/s and 40t/s and the temperature is
By using our SPC method, highly conductive 650t in the SPC process. At the peak, the ratio
p-type Si thin-films were fabricated at the low of 'eoll at IOOOnm" eoll increases with the ramp
temperature of 650t. So, tliese films were applied rate and shows a maxiinum value of 79.4%. This
to the p-Iayer in p-n junction solar cells. Figure ratio is almost equal to that of the conventional
7 shows the peak normalized collection efficiency

300
p-n junction solar cell, in which the p-n
junction was prepared by a high temperature 45.----------------------n.
process, such as the thermal diffusion or ion Texture,size:2.0cm2 2
implantation methods. A p-type SPC-Siln-type c- Isc: 42.0mAlcm
Si solar cell has short circuit current(Isc) of Voc:0.525V
30.0-34.3mA/cm 2 • Therefore, our SPC method is F.:0.642
7] '14.2%
effective for the fabrication of p-n junction
solar cells at the low temperatures. 1::u 30

SPC
't'
---------- Conventional
ramp rate
40·C/s
125
C
Flat,size:2.0cm2
Isc: 34.3mAlcm 2
con dI Ion _._._.- 10·C/s
~
650·C 7hours - . - . - 1·C/s ...~
::J
20 Voc:0.489V
c F.F.:0.577
Q)
'0
o 15 7] :9.7%
~ 0.8
10
j
c
0.6

'8 AM-1.5100mW/cm 2
"U
.~ 0.1 0.2 0.3 0.4 0.6
"iil Voltage (V)
E
o
z 600 800 tooo 1200
Wavelength (nm) Figure 8 I-V characteristics of p-type SPC-Sil
n-type c-Si solar cells prepared by SPC
Figure 7 Normalized collection efficiency of method, where n-type c-Si substrates
p-type SPC-Si/n-type c-Si solar cells, were flat and textured.
compared with that of conventional c-Si
solar cells. a low temperature of 650t. In the p-type SPC-
Siln-type c-Si(textured) solar cell prepared by
From the viewpoint of solar cell performance, the SPC method, a high conversion efficiency of
the texture of n-type c-Si substrate mainly 14.2% has been achieved with a size of 2cm2. So,
effects the increase of Isc. On the other hand, the SPC method which we have developed has
we have found that the p-type Si thin-films high potential, from the viewpoint of low
prepared on the textured substrate by the SPC temp'erature device fabrication.
method had large grains and showed high
mobility in the high carrier concentration
range of 2XI02°_2xl021cm-s. So, the p-type Sil
n-type c-Si(textured) solar cells prepared by Acknowledgement
SPC from p-type a-Si:H/n-type c-Si(textured) will This work is supported by NEOO (New Energy
be expected to have higher conversion efficiency. and Industrial Technology Development
Figure 8 shows the I-V characteristics of the Organization) as a part of the Sunshine Project
p-type SPC-Si/n-type c-Si (textured) solar cell, under the Ministry of International Trade and
compared with that of the p-type SPC-Siln-type Industry.
c-Si(flat) solar cell. In the p-type SPC-Si/n-
type c-Si( textured) solar cell, Isc, fill factor References
(F.F.) and open circuit voltage (Vod increase, 1) R. N. Hall, et al.:Phys. Rev., 80, 1950, p467.
compared with the p-type SPC-Siln-type c-Si 2) W. Shockley:Proc. IEEE, 56, 1968, p295.
(flat) solar cell. As a result, a high conversion 3) T. Matsuyama, et al.:Jpn. J. Appl. Phys. Vol.
efficiency of 14.2% has been achieved with a size 29, No. 12, 1990, p2690.
of 2 cm2 • 4) Y. Hishikawa, et al.:Proc. 9th E. C.
Photovoltaic Solar Energy Conf., 1989, p37.
4. Conclusions 5) W. C. Dash and R. Newman:Phys. Rev. 99, 1955,
The new SPC method was developed for the p1l51.
preparation of highly conductive p-type Si thin-
fi:lms with conductivity of more than lXI0 s
(Dcm)-l. It was found that the mobility of the
films prepared by the SPC method increased by
using textured substrates. The highest mobility
was 32.2cm21Vs at the carder concentration of
2x10 20 cm-s. This SPC method was suitable for
the fabrication of p-n junction solar cells at

301
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &--12 APRIL 1991 LISBON, PORTUGAL

Silicon-Film~ Product I: Initial Production Performance

A.M. Barnett", W.R. Botteriberg, J.A. Bragagnolo, D.S. Brooks,
J.C. Checchi, C.L. Kendall, P.G. Lasswell,
J.E. Brown, W.P. Mulligan, J.A. Rand and R.B. Hall

AstroPower, Inc. "Dept. Electrical Eng.

30 Lovett Avenue and Univ. of Delaware
Newark, DE 19711 USA Newark, DE 19711 USA
FAX: (302) 368-6474 TEL: (302) 366-0400

ABSTRACT. The initial Silicon-Film~ product is now being produced

using commercial-scale manufacturing. This Silicon-Film~ Product
I, which does not incorporate light-trapping, has achieved a
laboratory-scale, one cm2 solar cell conversion efficiency of
15.7%, and a commercial-scale, 100 cm2 conversion efficiency of
10.9%. A commercial-scale process has been designed and tested for
the manufacture of Product I wafers. Conversion efficiencies in
excess of 10% have been achieved with devices formed with this
material. Passivation and gettering methods have been used to
enhance performance. Two 20 kWp systems using Silicon-Film~
Product I solar cells have been designed and will be installed in
1991.

1. INTRODUCTION cell conversion efficiency of 15.7% (1),

and a commercial-scale, 100 cm2 solar cell
The Silicon-Film~ process is being conversion efficiency of 10.9% (2). The
developed with the objective of achieving a solar cell fabrication processing for this
high performance, low-cost solar cell for latter result is identical to that employed
terrestrial power appl ications., The on conventional wafers.
specific technical approach involves the The achievement of a 10.9% commercial-
deposition of a thin polycrystalline sized solar cell using a lab-scale process
silicon layer on a rigid low-cost (1. e. to fabricate the silicon-Film~ Product I
ceramic) substrate. The fully mature solar wafer, led to a focus on the development of
cell product is designed to achieve a the pilot-scale machine for the manufacture
conversion efficiency of 19%. It includes of these wafers. The pilot-scale machine
a thin 30 to 50 micron layer of silicon, is designed to produce Silicon-Film~
with a minority carrier diffusion length of Product I wafers at a rate of 120 per hour.
60 to 80 microns; a metallurgical barrier, Present material from the pilot-scale
with properties to trap light and passivate machine has demonstrated efficiencie.s in
the back silicon interface; and a low-cost, excess of 10%. This equipment is now being
conducting substrate. The silicon-FilmrM employed to produce wafers for the
solar cell without light trapping uses a production of solar cells to be
100 micron thick active silicon layer and incorporated in two arrays that are
is called silicon-Film~ Product I. The nominally 20 kWp each. These two arrays
light trapping version is called Silicon- are being manufactured for the Emerging
Film~ Product II. They are shown Module Technologies segment of the

.L--------S"'il con
schematically in Figure 1. Photovol taics for Utili ty Scale
Applications, PVUSA program (3,4). One
Contac:u. array is to be installed near Richmond,
- - Virginia for Virginia Power and the other
array will be installed in Davis,
100 lilT California for Pacific Gas and Electric.
ProdUCI 1
S U ~l r.ll't: This technology took 27 months to
mature from a 1 cm2 laboratory-scale to a
100 cm2 commercial-scale solar cell at 10%.
CO.UXH
The demonstration of pilot-scale production
was initiated in February 1991, 44 months
after the initial laboratory scale
demonstration.
This paper will describe the wafer
Prod uct II manufacturing process, the solar cell
process, and present solar cell performance
for the pilot-scale wafer machine. Design
C(ln"''CI~ parameters for the 20 kWp systems will be
presented.
Figure 1. Silicon-FilmrM Products I and II.
2. SILICON-FILMrM WAFER FABRICATION STEPS
An initial solar ce11 product that The first of four proprietary steps in
does not incorporate optimal light-trapping the fabrication of Silicon-Film~ wafers is
or back-surface passivation has been put the substrate formation. This step
into production. This product has a total involves the fabrication of a low cost,
thickness of 500 microns including the 100 electrically-conductive substrate with
micron thick active silicon layer. This suitable properties to support subsequent
structure, the Silicon-Film~ Product I, has processing, but to not deleteriously affect
achieved a laboratory-scale, 1 cm2 solar device operation.

302
 The second Silicon-Filmm wafer Table 1
fabrication step involves the formation of
a metallurgical barrier. A metallurgical silicon-Film™ Wafer and Solar Cell
barrier possessing the requisite properties Process Steps
is presently under development. As
previously indicated, the achievement of Wafer l'ormatioD Solar Cell l'ormatioD
light-trapping and back surface passivation
are required attributes for the 1) compound 9) Etch and clean
metallurgical barrier, and are necessary to materials
achieve high performance operation of a
5ilicon-Film™ solar cell. The 5ilicon- 2) Form substrate 10) Diffuse, strip,
Filmm Product I, presently in Pilot Scale plasma etch
production, does not employ a metallurgical 3) Fire substrate 11) Screen print,
barrier, and consequently does not enjoy fire back
the performance advantage of the light- contacts
trapped design. The eventual development
of a suitable metallurgical barrier which 4) Apply metal- 12) Screen print,
will be incorporated in Product II will lurgical barrier fire front
allow for higher performance operation, and contacts
the utilization of even lower cost
materials for the fully-integrated 5) Sinter 13) Antireflection
supporting layer. coat
The last two Silicon-Filmm wafer 6) Apply active 14) Burnish
fabrication steps involve the sequential layer contacts
application and formation of the
photovoltaic silicon layer. A silicon 7) Grow Silicon- 15) Test
solution is formed which covers the Filmm
substrate. The photovoltaic properties of
the grown layer are enhanced using a 8) Scribe and break
process employing a prescribed thermal
profile. A schematic of the silicon-Filmm
substrate manufacturing process is shown in The equipment necessary to carry out
Figure 2. the first step of the Silicon-Filmm wafer
fabrication process is commercially
available, although modifications to
accommodate the specific process are
required. Since at the present time the
metallurgical barrier component of the
fully-integrated Silicon-Filmm structure is
under active development, the specific
process, and accordingly the equipment to

Ii i
effect that process, has not been
I&l

Ii
established.

--
The equipment required to complete the
last two steps in producing Silicon-Filmm

-
i (!!!J wafer material (active layer application
and Silicon-Film™ growth) is not
c-a ~ and 0"" GrOWI'I
~- commercially available. The processes
Silicon-Film'" Substrate Manufacturing Process employed in these two steps were developed
to accomplish specific features of the
Figure 2. Schematic drawing of the Silicon-Film™ process; therefore, the
Silicon-Film™ substrate manufacturing equipment required development as well. An
process. analysis of the laboratory-scale process
that was employed to produce the
In the Product II structure, the commercial-scale Product I prototype solar
thickness of the active silicon layer is, cells has been carried out to formulate a
by design, in the range of 30 to 50 microns process specification. This process
thick. In the present Product I prototype, specification was used to design a
the active silicon layer is approximately commercial-scale process for the
100 microns thick. The overall thickness application and formation of the active
of the Silicon-Filmm wafer is 500 microns layer of the Product I Silicon-Film™ wafer.
for both the Product I and II structures. During the last six months an initial
These wafers are distinguished from manufacturing machine for the fabrication
conventional crystalline wafers in that of Product I Silicon-Film™ wafers has been
only the side with the thin designed, fabricated, and is presently
photovoltaically active silicon can be operating. The PV-layer application and
usefully employed for solar cell formation process has the design capability
fabrication. The solar cell fabrication of producing 120 10xlO cm2 wafers per hour.
process steps are identical to those used Silicon-Film™ sheets nominally 11 cm wide
for conventional wafers. and 45 cm long have been produced
continuously with this apparatus. A
3. SILICON-FILMm WAFER AND SOLAR CELL subsequent sizing step converts the silicon
The Silicon-Filmm Wafer and Solar Cell sheet material into four properly sized
Process steps are shown in Table 1. wafers suitable for solar cell fabrication.
A photograph of a group of Silicon-
Film solar cells is shown in Figure 3.

303
 recombination term of the dark I-V curve,
likely attributable to grain boundary
passivation.
Table 2
Effects of Passivation and Gettering
(I) (II)
Voc (mV) 582 600
Jsc (mA/cm2 ) 25.0 33.0
FF (t) 70.2 79.2
Eff. (t) 10.2 15.7
Area (cm2 ) 0.85 1.02
Ln (microns) 50 llO

Fiqure 3. Silicon-Film solar cells from Phosphorous gettering and hydrogen

the Pilot Production line. passivation are being used in the
commercial solar cells. The current-
voltage curve of a solar cell meeting the
4. SOLAR CELL PERFORMANCE specifications of the 20 kWp systems is
A summary of Product I results for shown in Figure 5.
laboratory size and commercial size devices
is shown in Figure 4. o 100 200 300 400 SOO 600
0

Voc (mV): S62.2

~.5 Isc: (A): 2.279
16.0
"'...... ,
Ubot' .OIT"ScaIt
Vmp (mV):
Imp (A):
468.8
2.017
-1.0
~ FF (%): 73.77
>. 14.0
0
c -1.5 Pmax (W): O.94S
u
';)
IE 12.0
\.Il -2.0

/
c

'2 -2.5
u
>
c
c
10.0
0
p,odI.Ict I
U Produdlol')
-3.0
8.0 ......... ,
COMmttdlr b~

6.0 -4-.---.---,--.---.---.- Figure 5. Characteristic curve for

Silicon-Filmm Product I solar cell.
a

Year 5. DESCRIPTION OF THE PVUSA AND VIRGINIA

POWER 20 KWP SYSTEM DESIGN
Fiqure 4. Performance of laboratory-size The system consists of 624 modules of
(one cm2 ) and commercial size (100 cm2 ) the AstroPower AP-M38 Silicon-Filmm design.
Silicon-Filmm solar cell without light Each module uses 36 SF-2101 solar cells
trapping •. with a nominal power of 1.05 wp each. The
first modules delivered to PVUSA will
Passivation and gettering techniques demonstrate 90t of the rated power, or 34.2
have been used to improve silicon solar wp. Further experience with the wafer
cell performance. When combined with other manufacturing process and development of
methods of solar cell processing such as the solar cell process for this material
deep diffusion and low temperature will lead to 1.2 watt solar cells.
annealing, the benefits of passivation and AstroPower's AP-M38 Silicon-Filmm PV
gettering are enhanced. Module consists of 4 strings of 9 SF-2101
Table 2 compares a Silicon-Filmm solar Silicon-Filmm solar cells each. The solar
cell fabricated using a standard processing cells are encapsulated in a laminated
sequence[I], and a cell fabricated using a structure of high-transmission tempered
specialized processing sequence[II] (5). glass (3 mm thick), EVA and a
The specialized processing sequence polyester/Tedlarm laminated backsheet. The
incorporated phosphorous gettering, modules are framed in aluminum. The package
hydrogen passivation, and low temperature is mechanically strong and moisture
annealing into the standard solar cell impervious. The module dimensions are 0.96
process. The overall energy conversion by 0.44 meters. The thickness, including
efficiency gain for a one cm2 laboratory the frame, is 3.5 centimeters. The modules
scale solar cell was 54% with an L,. increase have one bypass diode each.
of 100%. The majority of the improvement The modules are rated at 34.2 Wp under
is observed as increased short circuit Standard Test Conditions of 100 mW/cm2 ,
current, the remainder in improved fill AM1.5 illumination.
factor. The increase in fill factor The solar cell and module parameters
followed the decrease in the n=2 are summarized in Table 3.

304
 Table 3 Table 4
Silicon-Film~ Solar Cell and Silicon-Film~ Panel and Array Output
Module Rating Parameters at 27°C at AM1.5, 27°C
PARAMETER
PARAMETER Panel Array
Solar Cell Module
Voc [V] (+/-) 233 233
Voc [V] 0.54 19.4
Isc [A] 2.43 63.2
Isc [A] 2.43 2.43
Vmp [V] (+/-) 183 183
Vmp [V] 0.43 15.5
Imp [A] 2.21 57.4
Imp [A] 2.21 2.21
Pmp [Wp] 810 21,040
Pmp [Wp] 0.95 34.2
6. SUMMARY AND CONCLUSION
The modules are connected in series to Progress continues in the development
form a panel which is three modules across of thin polycrystalline Silicon-Film~ solar
and eight modules in height. A drawing of cell products. An initial commercial-scale
the panel is shown in Figure 6. solar cell, Silicon-Film~ Product I, is now
being successfully produced. Material
produced with this process has yielded
devices with conversion efficiencies over
10%. This product is expected to achieve
commercial device efficiencies as high as
14%. Post-wafer fabrication enhancement
methods, including passivation, gettering
and annealing are being employed to improve
device performance. The addition of light
trapping for Product II will lead to
commercial-size efficiencies over 17%.
The initial silicon-Film~ product will
be employed in a module comprised of 36
solar cells with nominal power of 1.05 Wp
each. The first modules will demons~rate
90% of the power or 34.2 wp. Two 20 kWp
systems will be delivered this year as part
of the Emerging Module Technologies segment
Figure 6. Silicon-Film~ panel with 24 of PVUSA.
modules.
REFERENCES
Each of these panels has an output of (1) A.M. Barnett, F.A. Domian, D.H. Ford,
810 Wp at AM1.5. The 24 modules in a panel C.L. Kendall, J.A. Rand, M.L. Rock and
are electrically connected in two circuits R.B. Hall, "Thin Silicon-Film Solar
of twelve series modules each, one positive Cells on Ceramic Substrates",
and one negative relative to a neutral Technical Digest of the International
center tap. The output voltage has both a PVSEC-4; Sydney, Australia (February,
positive and negative polarity of 183 1989), pp. 151-158.
volts. Thirteen panels are then connected (2) C.L. Kendall, J.C. Checchi, M.L. Rock,
in parallel to form a sub-array with an R.B. Hall and A.M. Barnett, "10%
output of 10.5 kWp. There are two sub- Efficient Commercial-Scale Silicon-
arrays leading to a total power generated Film Solar Cells", Proc. 21st IEEE
of 21 kWp at 27°C solar cell temperature. Photovoltaic Specialists Conference;
The 24 module panels are factory- Kissimmee, Florida (May, 1990), pp.
assembled using aluminum channels as the 604-607.
maj or structural element and fastened to (3) C.J. Weinberg, S.L. Hester and T.U.
the supplied structure. Each triple panel Townsend, "The Photovoltaics for
has a panel junction box with isolation Utility Scale Applications (PVUSA)
diodes and metal-oxide varistors (MOV's) Project in the United States",
connected to the array ground for transient Technical Digest of the International
protection. PVSEC-5; Kyoto, Japan (Nov., 1990),
Each sub-array has a junction box pp. 571-576.
providing (+) or (-) 183 V, relative to an (4) S.L. Hester, T.U. Townsend, W.T.
insulated neutral terminal strip, from a Clements and W.J. Stolte, "PVUSA:
parallel connection of 13 triple panels. Lessons Learned From Startup and Early
The sub-array junction boxes have panel Operation", Proc. 21st IEEE
fuses designed to isolate each triple panel Photovoltaics specialists Conference;
for ease of installation and maintenance. Kissimmee, Florida (May, 1990), pp.
The power from the two sub-arrays is 937-943.
combined at the interface junction box (5) M.L. Rock, D.W. Cunningham, C.L.
(IJB) which serves as the interface to the Kendall, R.B. Hall and A.M. Barnett,
utility. The IJB provides separate sub- "Process Induced Improvements in
array disconnect switches and the required Polycrystalline Silicon-Film Solar
fusing. Cells", Proc. 21st IEEE Photovoltaic
The IJB is divided into separate power Special ists Conference; Kissimmee,
and instrumentation sections. Transient Florida (May, 1990), pp. 634-637.
protection is provided by the MOV's at the
triple panel level.
The output parameters for the panel
and the array are shown in Table 4.

305
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Silicon-Fi~ Product II:

Initial Light Trapping Results
J.A. Rand, D.H. Ford, C. Bacon, A.E. Ingram,
T.R. Ruffins, R.B. Hall, A.M. Barnett*
AstroPower, Inc. *University of Delaware
30 Lovett Avenue Electrical Engineering Department
Newark, DE, USA 19711 Newark, DE, USA 19717

ABSTRACT. The Silicon-Film~ on ceramic device has been

designed to achieve high conversion efficiency with the use of
imperfect materials. The Product II designation refers to a thin
silicon layer incorporated into a light trapping structure.
Excellent optical characteristics have been measured on
structures that have silicon layers <200 ~m thick formed over a
metallurgical barrier on low cost ceramic substrates. The
reflectivity of the silicon - metallurgical barrier interface has
reached 70%. The optical path length has been estimated to be 9
times the silicon layer thickness. Device efficiencies of 8%
have been obtained in 1.0 cm2 devices. These efficiencies are
limited by low minority carrier diffusion lengths.
1. INTRODUCTION 2. DEVICE STRUCTURE
Light trapping must be incorporated A cross section of a Silicon-Film~
in thin crystalline silicon films to Product II structure is shown in Figure
enhance the absorption of long wave- 1. Pictured is the a 70 ~m thick
length light. Light in the wavelength crystalline silicon layer over a
range 0.9~m to 1.1~ is only weakly metallurgical barrier on a coarse
absorbed in a 50 ~m thick silicon film. ceramic substrate. Excellent film
Of the photons in the AM1.5G spectrum morphology has been obtained with an
with energy greater than the bandgap of aspect ratio (ratio of the lateral grain
silicon, 24% are in the wavelength range dimension to the film thickness)
0.9-1.1~m. Without light trapping, a exceeding 10 .
significant portion of these photons are
transmitted through the silicon film and
lost.
The Silicon-Film~ process involves
the deposition of a silicon layer on a
conducting ceramic substrate. The
ceramic has been coated with a
reflective metallurgical barrier layer.
The device structure has been designed
to incorporate light trapping, and
maximize the collection efficiency of
photogenerated carriers. Obtaining back
surface passivation at the silicon -
metallurgical barrier is a key element
in achieving high collection efficiency.
Device results for Product II
structures on ceramic are presented that
show the metallurgical barrier is up to
70% reflective, and significant levels
of light trapping are present. The long
wavelength spectral response and ,igure 1. Cross section of a Silicon-
reflectance measurements are analyzed to Filmm Product I I structure showing
estimate the level of light trapping silicon layer, metallurgical barrier,
achieved. Results for similar and ceramic substrate. The silicon
structures fabricated on single crystal layer is approximately 70~m thick.
silicon substrates are discussed.
Device efficiency has reached 8%, and is
limited by low minority carrier The ceramic is formulated from low
diffusion lengths. Significant levels cost materials in powder form, and is
of impurities have been detected in the formed into its 10 cm x 10 cm x 0.5 mm
silicon films and are believed to be form, then heat treated to remove
causing the low diffusion lengths. binders. Constituents of the ceramic
Contamination from handling is suspected have been carefully chosen to maintain
as the impurity source. strength at the high silicon deposition
temperatures, be expansion matched to
silicon, and be electrically conductive .
This work is supported by the Photovoltaic Energy Technology
Division, U.S. Department of Energy through Sandia National
Laboratories, contract DE-AC04-76DP00789

306
 The metallurgical barrier is obtain high levels of light trapping. A
deposited by production based, low cost range of back surface reflectivity
processes. Interaction of the barrier values can be computed from the
and silicon active layer should result reflection data taken on the complete
in a low recombination surface to structure [3]. Quantum efficiency
maximize efficiency [1]. High measurements indicate higher levels of
reflectivity, and low recombination have response for long wavelength light than
been key design elements of the predicted by a simple (non-light
metallurgical barrier development. trapping) model. Analysis of these
results reveals light trapping
The silicon active layer is properties.
deposited by the proprietary Silicon-
Film~ technique. Films 50 to 500 ~m The reflectance of a Silicon-Film~
thick are grown. Aspect ratios exceed on ceramic structure is shown in Figure
10, with grain dimensions on the order 3 with and without a barrier. The
of centimeters. Samples have been increased long wavelength reflectance of
fabricated with and without the sample with the barrier is due to
metallurgical barrier layers using the light being initially transmitted into
same Silicon-Film~ process. the silicon, internally reflected, and
then transmitted back out the front.
Electro-deposition techniques are The re-emitted subbandgap light is
also being used to deposit thin, high referred to as the escape reflectance.
performance, silicon films [2]. RESCAPE is defined as the difference
Electro-deposition is a current- between the total subbandgap reflectance
controlled, large area, solution growth (R"oTAL) and the external front surface
method. Silicon is being deposited by reflectance (R,.EI.
this technique on ceramic substrates and
poorly performing single crystal wafers.
A light trapping structure is (1)
incorporated to upgrade performance.
Test structures have been
fabricated on single crystal substrates 100
for evaluation of back surface
texturing. A cross section of such a
structure is shown in Figure 2. 80
Pictured is a 15 ~m thick silicon layer ,....
over a textured substrate. The barrier
layer was fabricated from a non- ..!.
60
conductive dielectric material. Vias Q)
0
were opened to allow electrical contact c:
to the substrate. .3 f
---------- - -
0

- - -,
R SSCAPE
~Q) -
a::
- - - - with barrier
~~.. ·:~i~~~·
- - - - - - - . without barrier
.~\.tl;·,,~!t 1 '~'.' ' ....
., J.. ' ~'"
600 1000 1100 1200
0ep0si1ed Si
Dielectric Barrier
(with Via Openings) ,igure 3. Normalized reflectivity of a
Silicon-Film m Product II structure.
Silicon Substrate Data is shown for samples fabricated
with and without a barrier.

The silicon-barrier reflectance

(RBAcKI can be calculated (for non-
absorbed light) based on the following
relationship [4]
,igure 2. Cross section of a Silicon-
Filmm Product II test structure
fabricated on a specially textured (2)
single crystal silicon substrate.

3. RESULTS where RFr is the internal front surface

reflectance. Solving (2) for RBACK gives
3.1 Light Trapping Characteristics
The light trapping properties of a
structure can be deduced from the
measurements of reflectivity and quantum
efficiency. Reflectivity measurements
give an estimate of the back surface
(silicon-barrier interface) The internal front surface reflectance
reflectivity. This interface is (R,rl is an unknown quantity, although
required to be highly reflective to minimum and maximum values can be

307
calculated. The m1n1mum value is that (4)
of an uncoated silicon/air surface. The 1 =1+_1_
maximum is defined by the best case OEINl7lRNAL uLN
texturing scheme that would allow for
re-direction of the trapped light, and
total internal reflection at the
silicon/air surface. FOr this analysis This relationship predicts a linear
the Lambertian limit of 92% is used as region in the plot-of l/QE verses l/a.
the best case [5]. This best case These data are shown in Figure 5 for the
assumes the light reflected from the QE data from Figure 4. Two linear
back surface is distributed as the regions are shown. The first linear
cosine of the angle from the horizontal. region appears when l/a < 10~. The
The resulting range of barrier inverse slope of this line predicts a LN
reflectance values can be calculated. = 8 ~m, which can be interpreted as the
Table I shows that the presence of the effective minority carrier diffusion
barrier layer increases the net back length of the silicon film. The data
surface reflectivity to the range of 32% deviates from simple theory as l/a
to 80%. exceeds 10~. Higher levels of
collection efficiency are recorded than
predicted by the simple theory of
Table I. Analysis of long wavelength equation (4). The slope and intercept
reflectivity. of this line can reveal detailed
properties of the light trapping
characteristics [6]. For the device
RoSC»B R.. R., Rue. data in Figure 5, a Z of approximately 9
Estimated is indicated (where Z is the optical
With 17.5 29.5 29.5 (min) 31.9 path length given in multiples of the
Barrier 92.0 (max) 80.5 device thickness). The optical path
length of long wavelength light was
No 1.5 29.6 29.6 (min) 3.0 therefore 1.7 mm for this 190 ~m thick
Barrier 92.0 (max) 21.4
film. The low levels of QE measured at
the long wavelengths make the exact
evaluation of Z uncertain.
Quantum efficiency measurements were
made on structures with and without The high level of light trapping
barriers. The data for a silicon on seen in the QE data implies a high
ceramic sample with a barrier is shown degree of scattering by the back surface
in Figure 4. The silicon layer is 190 reflector. In this case the internal
~m thick for the sample tested. The red reflection of the front surface is
response is poor, indicative of material estimated to approach the Lambertian
with low minority carrier diffusion limit of 92%, and silicon/barrier
length (L N). interface has a reflectivity of
approximately 70% [6].

100 60
Non-li~ht Trapp~ng Model
Is/ope = ILIa
50
,/ Ln= 81Jm
W ~o
060 I
w
a
,30
Non-ideal
Long Wavelength Response
due to Light Trapping Effects

UO 7:.0 _ ;:.0 1050

Wovelength (nM) 500 1000 1500 2000 2500 3000
Absorption Length (urn)
Figure 4. Quantum efficiency of a
silicon on ceramic structure with a Figure 5. Absorption length versus the
metallurgical barrier layer. inverse of the internal quantum
efficiency for the silicon on ceramic
structure.
Based on non-light trapping theory,
a measurement of LN can be made using
internal quantum efficiency data. For a Similar analysis has been done on
device that has a diffusion length less test structures formed on single crystal
than the device thickness (W), the substrates (as shown in Figure 2).
following relationship exists for weakly Samples with textured barrier layers
absorbed light showed long wavelength internal quantum
efficiencies twice as high as the
controls formed with planar barriers.
High levels of light trapping and high
barrier reflection were seen (through

308
total internal reflections) [1). Optical lengths 9 times the device thickness.
path lengths could not be quantified due Device efficiencies of 8% have been
to the possibility of the collection of obtained in 1.0 cm 2 devices, limited by
photogenerated carriers from the low minority carrier diffusion lengths.
substrate (through the vias) . Improvements in diffusion length are
expected as material handling procedures
are improved. Device efficiencies are
3.2 I-V Characteristics expected to exceed 14% in the near term.
Silicon-Film~ on ceramic
structures have been fabricated into 1.0 5. ACKNOWLEDGEMENTS
cm 2 devices utilizing standard solar
cell processes. Samples were cleaned The assistance of Paul A. Basore of
after fabrication. No surface etching Sandia National Laboratories in both
or thinning was utilized. N-layers were measurements and analysis is greatly
formed by PH J diffusion. Front contact appreciated. Device fabrication and
was made by evaporated Ti-Pd-Ag, characterization was performed by Steve
followed by Ag plating. Back contact M. Lampo, Thomas H. Lampros, and William
was formed by silver paste, and an AR P. Mulligan. Much of the optical
coating was deposited. In results testing was assisted by Brian E.
verified by Sandia, an AM1.5G efficiency McCandless, Steven S. Hegedus, and
of 6.6% was measured. Specific results William N. Shafarman of the Institute of
are shown in Table II. Energy Conversion, University of
Delaware.
Table II. Measured I-V characteristics
on 1.0 cm 2 Silicon-Film~ Product II.
REFERENCES
voc J.c FF 1'1
(mV) (mA/em') (t)
[1) J.A.Rand, R.B.Hall, A.M. Barnett ,
Sandia 513 19.3 65.8 6.6 21st IEEE Photovoltaic Specialist
Verified Conference, (IEEE Press,New
Device
(9/90) York,NY,1990),pp.263-268.
[2) D.W.Cunningham, M.G.Mauk,
Recent 546 21.1* 70.2 8.1 J.P.Curran, and A.M.Barnett, 21st
Results IEEE Photovoltaic Specialist
\.,;orrec~ea or an~1-re~~eC~1on coa~1ng
Conference, (IEEE Press,New
York,NY,1990),pp.307-310.
Improvements in device processing, [3) P.A.Basore, Personal Communication.
and changes in doping levels have in- [4) P.A.Basore, IEEE Transactions on
creased Voc and FF in the most recent Electron Devices 37(2),337(1990).
samples. Overall J sc values remain low, (5) A. Goetzberger,15th Photovoltaic
limited by poor minority carrier Specialists Conference, (IEEE, New
diffusion length (LN) . Values for LN York, NY,1982),pp. 882-889.
have been in the range of 5 to 10 ~m. (6) J.A.Rand and P.A.Basore, to be
published.
Investigations of the cause of the
low minority carrier diffusion length
include analysis of stress induced
defects, grain boundary losses, and
impurities. Through the use of test
structures and defect etching, stress
has been found to playa small role.
Defect densities (measured after Secco
etchinq) have been found to be as low as
3 x 10 5 cm- 2 • Residual stresses have
been low,·as no warping or cracking has
occurred over the 50 cm2 areas presently
under development. Grain boundary
recombination effects are also
insignificant as columnar grains are
formed with aspect ratios exceeding 10.
Chemical analysis has shown significant
levels of electronically active
impurities. The source of these
impurities is presently under analysis.
Contamination from handling procedures
is suspected.

4. CONCLUSIONS
The Silicon-Film~ on ceramic
structure has been designed to generate
high efficiency utilizing imperfect
materials. This design involves thin
silicon layers in light trapping
structures. Excellent optical
characteristics have been measured, with
silicon/barrier back surface
reflectivities of 70%, and optical path

309
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORWGAL

CHARACTERIZATION OF COMMERCIAL
SINGLE CRYSTAL Si CELL TECHNOLOGY

K, W. Mitchell, D. L. Aldrich, J. R. Walle, J. M. Gee and J. D. McBrayer

D. G. Gretlein, K. L. Pauls, and R. Probst Sandia National Laboratories
Siemens Solar Industries 1515 Eubank SE
P.O. Box 6032 Albuquerque, New Mexico 87185 USA
Camarillo, California 93011 USA

ABSTRACT. Improvements in commercial single crystal Czochralski (Cz) Si

module performance accrue from reducing cell-to-cell variations and
improving individual cell performance. Slip dislocations in the bottom
region of Cz ingots with "lost structure" impact performance yields.
Total area cell efficiencies of 15.6% can be improved by increasing
minority carrier lifetimes and reducing shadowing losses. Current cell
processing can maintain minority carrier lifetimes above 1 msec.

1. INTRODUCTION

Although several laboratories have reported high

effiCiency Si cells (1-4), commercial multi-megawatt
production of high performance modules requires
reliable, high yield processing and handling with
minimal cell-to-cell variations. This paper de-
scribes the characterization of Siemens Solar com-
mercial Cz Si cells and their related processing.

2. CELL MANUFACTURING

Single crystal Si photovoltaic (PV) manufacturing

is fully integrated at Siemens Solar Industries,
namely, polysilicon and other materials enter the
facility and are fabricated into completed PV cells
and modules. The steps in this process are:

1. Grow <100> p-type Si ingots using

Czochralski growers.
2. Slice ingots into 0.05 cm thick
wafers.
3. Chemically etch to remove saw dam-
age, followed by a texture etch.
4. Form junctions by.thermal diffusion. Fig. 1. Photo of a Siemens Solar commercial
5. Electrically isolate the front junc- Cz Si cell.
tion frDm the back of the cells.
6. Deposit antireflection (AR) coating. was characterized as a function of location in the
7. Screen print and anneal front and ingot under different growth conditions.
back electrical contacts. Normal conditions result in growth of "full term"
ingots, which at the end of growth are properly
The Si ingots are boron-doped in the 1 ohm-cm resis- tapered as they are removed from the Si melt.
tivity range. Figure 2, a plot of cell short-circuit current (Ise)
as a function of wafer location in the ingot, shows
3. CELL DESIGN uniform performance throughout the ingot. The left
to right direction represents moving from the bottom
A representative Si cell, shown in Fig. I, is towards the top of the ingot. About 12 wafers are
10.2 x 10.2 cm square, cut from a 13.4 cm diameter obtained per cm of ingot length. The other cell
wafer. The 86.7 cm2 active cell area is 83.7% of the parameters, Voe and fill factor, show similar trends.
total cell area of 103.5 cm2 . Grid shadowing from 53 A second condition termed "lost structure," can
1

screen printed grids 0.2 mm wide spaced 2 mm apart result from a perturbation during growth such as an
account for most (61%) of this loss. Two parallel 2 oxide flake in the melt colliding with the growing
mm wide bus bars spaced 2.5 em from the cell edges ingot or an interruption of furnace power. Loss of
account for an additional 21% of the area loss. The crystal structure typically results. Figure 3,
remaining area loss results from inactive area similar in layout to Fig. 2, shows that for a "lost
around the perimeter of the cell. structure" ingot, Ise decreases in the bottom portion
of the ingot. Similar losses occur in Voc and fill
4. EVALUATION OF INGOT GROWTH factor. The primary cause of this performance
decrease is the presence of slip dislocations. In
One cause of reduced cell performance yield is the graph, the origin for the wafer location is the
the generation of slip dislocations in the bottom point of initial occurrence of lost structure in the
portions of ingots caused by "lost structure" as ingot. The data thus indicate that the slip dislo-
described below. In this study, cell performance cations propagate back into the grown ingot. Opti-

3\0
 -::;- 3.4 f--------.~;:;r...~·
g 3.4 1-- ,-""--.-. .-.¥:-'-':---
":':"\~;"-_1.......
-_",,.':-:;..,
'-~-:"!"'~:r"'~
.""S'>''''''"_~
..-...,I"l't~~:":",,:,._.......;••~
!\r;;;",-;_,~"""",,,,_...,-......"\-
r!
•• -t
....
-' ,,',,. ..
~:r:-. '.
.'
C ,... c~ .-r..:. - - - - - - - - - - - - - - 1
l!! ~ 3.2 f------=,I~,
~ 3. 2 ~--------~-----------------------------t
o o ~ •.
·s -
·S I---.~-------------------------------,-----
u ,"
2
(3 3.0 I------------------------~ <3 3.0 f---..~------------~----

1:
1: II,':
o o
~
~ 2.8 00 2.8f----------------- .-

o 50 100 150 200 250 o 50 100

--'----
150 200
-'--- 1 ..1
250
Bottom - - -- - - - - - -- Top Bottom - - - - - - - - -- Top
Wafer Location
Wafer Localion

Fig .. 2. I.e versus wafer position for Fig. 3. Ioe versus wafer position for
a "full term" Cz ingot. a IIlost structure" ingot.

Fig. 4. OBIC images of cell areas showing slip dislocations.

cal beam induced current (OBIC) images of cell areas The dark I-V characteristics of Type B cells as
with slip dislocations are presented in Fig. 4. The shown in Fig. 7 are similar but shifted to higher
dark cross-hatched lines are areas of reduced re- current densities than Type A cells.
sponse associated with the slip dislocations. In the absence of slip dislocations, the minority
The light current-voltage (I-V) curves (under 100 carrier lifetime of the Cz wafers depends on the
mW/cm2 ASTM 1.5 global spectrum) for Type A (no slip nature of the electronic defects (5). Fourier
dislocations) and Type B (slip dislocations present) transform infrared (FTIR) absorption analysis indi-
are compared in Fig. 5. The Type A cell is 15.6% cates oxygen concentrations of 30 ppm or 1.5 x 10 18
efficient with a 32.9 mA/cm2 Joe' 621 mV Voc ' and cm· 3 . These high oxygen contents correlate with
0.763 fill factor based on a 103.5 cm2 total area. reduced minority carrier lifetimes (6). Low minori-
In contrast, the Type B cell is 14.0% efficient with ty carrier lifetimes of 15 to 20 microseconds are
a 30.6 mA/cm2 J. e , 607 mV Voe ' and 0.754 fill factor. confirmed by photoconductivity decay measurements at
The comparison of the spectral response curves for Sandia Laboratories and calculations based on the
Type A and Type B cells in Fig. 6 shows the loss of diode reverse saturation current J o (7). Work is in
long wavelength response for cells with slip dislo- progress to improve the minority carrier lifetime in
cations present, implying a reduced effective minor- the Cz wafers.
ity carrier diffusion length for the Type B cells.

311
 5. EVALUATION OF CELL PROCESSING

The achievement of high efficiencies in manufac-

30 ----------- .... tured cells depends ultimately on the ability of the

",\
cell processes to maintain high minority carrier
lifetimes. In order to evaluate this, wafers with
initial minority carrier lifetimes greater than 1
N \ msec were processed. Both n-type float zone (FZ)
E \ and magnetic Cz (MCz) wafers were used for compari-
20 \
~ \ son. Wafers were evaluated post damage removal,
post diffusion, post AR coating, and post contact
§. \
\ annealing (without metal). Lifetime measurements
..
'EGl
:;
I
\
performed at Sandia Laboratories (7) demonstrated
that all wafers maintained lifetimes greater than 1
,,
I
0 10 \ msec.
Type A
TypeS
, I
6. DISCUSSION

Current Siemens Solar Cz Si cells have diffusion

lengths of approximately 0.024 cm compared to cell
0 200 400 600 thicknesses of 0.043 cm. As the minority carrier
Voltage (mV) lifetime and thus diffusion length increases, more
minori ty carriers reach the back surface where,
without proper passivation, they can recombine,
Fig. 5. Light I-V curves for Type A and limiting the gains in long wavelength spectral
Type B Si cells. response expected with increased diffusion lengths.
Further analysis of surface recombination velocities
and the relative contributions to J o from the bulk,
emitter, surfaces, and contact areas is necessary to
leverage performance gains from improvements in
minority carrier lifetime.
100 As described in Section 4, commercial Siemens
Solar Cz Si cells have total area cell efficiencies
'--- ...
~ 80 .......... ..... in the 15% range. On the basis of active area,
these cell efficiencies translate to the 18% range .
>-
u
i ""\. Improved grid design and reduction in shadowing
losses will enhance total area cell performance.
60
,,
\. Modest improvements in minority carrier lifetimes
U
::W I \.

,
\. with appropriate cell design will result in commer-
\
40 \
cial 20% active area Cz Si cell efficiencies.
E \.
~ I
'E Type A \ ACKNOWLEDGMENTS
\
(II
:J
20 Type S \
0 .... The authors would like to acknowledge their
............ colleagues at Siemens Solar and Sandia Laboratories
~.3 0.5 0.7 0.9 1.1 1.3
who assisted in the preparation and analysis of the
samples. Wes Chesarek deserves special recognition
Wavelength (Ilffl) for the FTIR measurements and analysis.

REFERENCES
Fig. 6. Spectral response for Type A and
Type B Si cells. 1. M.A. Green, S.R. Wenham, J. Zhao, J. Zolper, and
A.W. Blakers, Proe. 21st IEEE Photovoltaie
Specialists Conf., Kissimimee, Florida, pp. 207-
210. (1990) .

2. S.R. Wenham, F. Zhang, C.M. Chong and M.A.

Green, Proc. 21st IEEE Photovoltaic Specialists
10 0 Conf., Kissimimee, Florida, pp. 323-326 (1990).

3. J. Zhao, A. Wang and M.A. Green, Proc. 21st IEEE

Photovoltaic Specialists Conf., Kissimimee,
10-2 Florida, pp. 333-335 (1990).
N
E 4. R.R. King, P. E. Gruenbaum, R. A. Sinton, and R.
! M. Swanson, Proc. 21st IEEE Photovoltaic Spe-
cialists Conf., Kissimimee, Florida, pp. 227-232
i: 10-4 (1990).
!!
:; 5. T.F. Ciszek, Proc. 20th IEEE Photovoltaic Spe-
0
cialists Conf., Las Vegas, Nevada, pp. 31-38
Type A (1988).
10-0 TypeS
6. B.L. Sopori, Proc. 20th IEEE Photovoltaic Spe-
200 400 600 cialists Conf., Las Vegas, Nevada, pp. 591-596
0
(1988).
Voltage (mV)
7. P.A. Basore and B.lt. Hansen, Proc. 21st IEEE
Fig. 7. Dark I-V curves for Type A and Photovoltaie Specialists Conf. , Kissimimee,
Type B Si cells. Florida, pp. 374-379 (1990).

312
10TIf EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

RECRYSTALLIZATION METHODS IN PRODUCTION OF PHOTOVOLTAIC DEVICES

N.B. Urli, B. Vlahovic, and M. Persin

Rudjer Boskovic Institute, P.O.Box 1016,
41001 Zagreb, Croatia, Yugoslavia

ABSTRACT. In a search for a low-cost thin film polycrystalline silicon solar

cell production, several two-step procedures have been proposed and tested
experimentally such as: deposition of a silicon film onto an inexpensive
substrate by plasma spraying, SSP method, PECVD, and dc-magnetron sputtering
followed by recrystallization by YAG, argon, or UV-pulsed lasers, by zone
melting with focused incoherent light, or by pulsed electron gun.Efficiencies
of various procedures and techniques have been estimated by comparing results
of structural, optical and electrical transport measurements before and after
recrystallization of originally deposited thin films. Some recommendations are
given on the possible choice of low cost solar cell technologies.

1. INTRODUCTION 2.1 Plasma Spraying

Silicon powder (CERAC 5N, 325 mesh) was
In overall efforts of finding low-cost and inject-ed into a D.C. plasma flame made of 90\ A
reliable techniques for solar cell manufacturing, and 10% H2 of a commercially available METCO MBN
thin film polycrystalline silicon solar cells plasma torch and sprayed on various substrates
and/or their tandems with amorphous silicon (stainless steel, alumina, SnOx/glass, Mo,
components may represent a serious alternative to graphite, tile), depositing 30 - 180 pm thick poly-
other thin film devices made from different 2
Si layers on an area of 16 cm in less than one
materials. minute. The quality of the as-deposited layers and
Properly designed 20-50 /-lm thick multicrystalline their adherence to the substrates depend on various
silicon solar cells with light trapping and grown factors such as heating-up and cooling-down
on low-cost ceramic substrates may outperform con- sequences, hydrogen content, differences in the
ventional, much thicker solar cells. Requirements coefficients of thermal expansion, torch-to-
are that the oriented single-crystal grains are at substrate distance, etc. Graphite and alumina
least twice as wide as the thickness of the silicon substrates showed superior adherence to the
layer with electrically inactive grain boundaries, deposited layers while the layers easily peel off
and that the minority carrier diffusion length is the metal substrates. Because of porosity and poor
at least two times the thickness (1). The last con- electrical properties, the plasma deposited silicon
dition relaxes somewhat requirements on materials layers require further thermal treatment
purity enabling utilization of less expensive (recrystallization).
solargrade meltstock silicon. As the production
cost of thin silicon ribbons obtained, for 2.2 SSP Method
instance, by ESP (2), EFG (3), or RAD (4) Several kinds of silicon powder such as CERAC
techniques is still relatively high, some new undoped (80 /-lm grain size), Wacker p-type (25 /-lm
approaches are needed. In this paper, several two- grain size), and Ethyl Corp. undoped spherical
step processes have been proposed and tested: granulates (0.15 to 0.5 rnm in diameter) were placed
deposition of silicon films with appropriate on quartz or graphite plates as reusable or perma-
thickness onto inexpensive SUbstrates followed by nent substrates. In order to reduce excessive
one of recrystallization methods in order to generation of dislocations and grain subboundaries
improve their optoelectronic properties. because of interface stresses, substrates were
In deposition processes, plasma spraying of coated initially by borosilicate or phosphoro-
silicon powder or a simple feeding of powder for silicate glass which provided a liquid support and
the SSP (Silicon Sheets from Powder) technique (5) acted as a source of dopants for the back contact
were applied. In addition, PECVD and DC-magnetron of solar cells while heating and melting the powder
sputtering methods for deposition of amorphous in the optical furnace.
silicon thin films were utilized, too. Actually, in
order to achieve appropriate film thickness in a 2.3 Plasma-Enhanced Chemical Vapor Deposition
short enough time, some high-deposition rate (~l TEGAL-PL 100 rf (13,56 MHz) PECVD reactor was
/-lm/min) CVD techniques should be favored, instead. utilized for deposition of thin amorphous silicon
Thin deposited films were recrystallized by YAG, films on glass or crystalline silicon substrates
argon, and UV pulsed lasers, by zone melting with from the gas mixture of 5% silane in argon at a
focussed incoherent light, or by pulsed electron- pressure of 1 Torr. The films were depOSited under
gun. Efficiencies of various recrystallization different gas flow rates (10 - 80 sccm), and rf
techniques have been estimating by comparing power levels (15 - 250 W) on substrates heated at
results of structural, optical, and electrical 150 or 300 0 C. The deposition rates varied from 1 to
transport measurements before and after 8 A/s.
recrystallization of originally deposited silicon
films. 2.4 D.c. Magnetron sputtering
A horne made sputtering system was used for
deposition of a-Si thin films in argon with 5 to
2. THIN-FILM DEPOSITION METHODS 15% of hydrogen, at gas pressures of 5 and 25
mTorr, current densities of 2.5 to 20.5 rnA/cm 2 , and
ion energies of about 100 eV.

313
3. RECRYSTALLIZATION TECHNIQUES 3.4 UV-Pu1sed Laser
A highly absorbed light of 20 kW LG-21 pulsed
3.1 Zone Melting by Focussed Incoherent Light laser at the wavelength of 0.2 I'm, pulse duration
An optical furnace consisting of high inten- of 10 ns, and repetition frequency of 200 Hz was
sity tungsten filaments placed in the focuses of also used in the a-Si crystallization. The energy
upper elliptical, and lower, parabolic mirrors density of 2.3 J/cm2 was sufficient for this
mounted on a common frame was constructed. A flat process.
rectangular quartz tube has been inserted between
those mirrors and supported at its ends by carriers 3.5 Pulsed Electron Gun
on an optical bench. Silicon powder on a substrate A home made electron gun, with indirectly
was placed in a quartz tube and positioned in a heated cathode, acceleration potential of 20 kV and
focal plane of the upper furnace. The heat produced current up to 1.5 A, with focusing lenses, deflec-
by incoherent light concentrated to the 2 mm wide ting coils and plates was designed to operate with
focal line with 136 watts/cm was sufficient to melt pulse duration from 10- 2 to 10- 4 seconds. It has
silicon, supported by the lower furnace with a much been possible to change the beam diameter at the
wider heating zone, maintaining a favorable thermal target from 2 to 20 mm. For an area of 45 mm 2 , a
gradient across the sample. The whole furnace was complete crystallization of a-Si was obtained in 5
scanned over the stationary sample at about 2 seconds at 13 kV and lA, pulse duration of 1.2 ms,
em/min, performing zone melting, refining, and and with pulse repetition frequency of 40 Hz.
recrystallization. Different heating and cooling
sequences, thermal gradients, and single and
multiple scans have been tested in order to find 4. RESULTS AND DISCUSSION
optimal melting conditions.
4.1 X-Ray Diffraction
3.2 YAG Laser Philips counter diffractometer with Cu Ka mo-
Coherent light of 1.06 I'm was utilized in nochromatized radiation was used in determination
crystallization of 0.5 I'm thick a-Si films on glass of the degree of crystallinity and perfection of
substrates at the power density of 4.4 kw/cm2 at an various samples. Diffraction line profiles of the
area of 4.5 mm2 , or 3.3 kw/cm 2 on 6.0 mm2 • The prominent reflections Ill, 220, and 311 indicate
crystallization time was 2.5 and 2.8 seconds, that the plasma sprayed silicon layers on stainless
respectively. It showed to be a critical and very steel and alumina substrates have the crystalline
sensitive parameter as the temperature of the film grain sizes of the order. of 10 I'm in accordance
rises steeply leading to the evaporation of thin with the previous SEM measurements (6). Similar
films. results have been obtained for recrystallized
Wacker powder by the SSP method. Much larger grains
3.3 Argon Laser (of the order of 100 I'm) were grown from the Ethyl
The most thorough investigation of crystalli- granulates by the same method.
zation processes on a-Si films has been done by cw
argon laser at the wavelength of 514.5 run, 4.2 Electrical Measurements
belonging to Z24-DILOR Raman spectrometer, by 4-point probe measurements on the plasma de-
changing the power up to 6 W, and a diameter of the posited films, taken immediately after spraying of
laser beam down to 0.4 I'm. In such a case, it was undoped silicon powder in air on various
possible to monitor a progress in the crystalline substrates, have shown resistivity values in the
grain growth by the Raman spectra at low enough range from 15 to 45 ohm cm. TC probe measurements
light pOwer densities. For instance, for the radius revealed that they were of n-type conductivity.
of the beam of 1.5 I'm, crystallization of PECVO Their resistivity tends to increase with time,
a-Si starts at 283 kW cm- 2 and reaches the indicating a higher degree of compensation.
saturation threshold (with no further increase in Hall-effect and resistivity measurements from
the grain size) at about 1100 kW cm- 2 • More results 80 to 300 K on silicon samples grown from the
of this kind may be found in Table I. Wacker powder by the SSP method and zone melting by
focussed incoherent light in an oxygen ambient
indicate they are also n-type and of low
resistivity (0.05 ohm cm). Similarly as in the case
of layers grown from Ethyl granulates, we found the
Table I. Dependence of the saturation threshold Ec 0.08 eV energy level associated with the
for crystailization (5) on film thickness (d), formation of thermal donors (7).
hydrogen content, exposure time (t), and beam Some of the plasma sprayed samples were re-
radius (r) crystallized in the optical furnace and turned to
be p-type indicating an effective purification by
a-Si d H2 t r 5 zone melting.
(I'm) (aU) (s) (I'm) (kw/cm2) An increase in dark conductivity as well as in
photoconductivity by more than two orders of magni-
PECVO 0.11 8 0.5 50 4.6 tude is shown in Fig. 1 for PECVO a-si thin films
on glass 0.5 8 500 70 5.9 after the crystallization procedures.
0.5 10 60 25 10
0.5 10 60 1.5 1100 4.3 Optical Measurements
Raman spectra of thin a-Si films after de-
DC Magn. 0.2 15 5 0.4 1450 position as well as those after various
Sputter. 0.2 5 5 0.4 1530 'crystallization processes have provided a whole
on glass 0.7 15 5 0.4 1700 series of data on energy thresholds and crystalline
0.7 5 5 0.4 2900 grain sizes as a function of power density,
exposure time, temperature etc. giving a better
DC Magn. 0.2 15 5 0.4 3750 insight in the crystallization mechanisms.
Sputter. 0.2 5 5 0.4 7330 PECVO and sputterred a-Si films showed four
on c-Si 0.7 15 5 0.4 2560 characteristic peaks at 190, 300, 380, and 472
0.7 5 5 0.4 4430 cm- 1 , associated with TA, LA LO, and TO phonon mo-
des, respectively. A new peak located at 517 cm- l
associated with the appearance of the crystalline
grains after the laser treatment with power density
greater than 5 kW/cm2 is shown in Fig. 2.

314
 In our experiments, we used more concentrated
-4 laser beams corresponding to much larger power
densities, and microcrystals were formed during
solidification from a liquid phase providing larger
grain sizes, as the TO peak approaches the single
-6 crys-tal value of 522.5 cm- l In this power den-
sity range microcrystals reach immediately the
maximal possible size which does not further
increase with the exposure time. The data are shown
in Fig.3.

CI)
.9 -10
o o o o
o o o
••
-12 ••
b
515
• o
~ 510
•
-14
100 150 200 250 300 Z
TEMPERA TURE [K] ~ 505
«
0::: 500 '--_'-----_'-----_'-----_-'-_-'-_ __
Fig. 1 Temperature dependence of the dark current
of a PECVD deposited sample: before (1), and after
o 50') HID 1500 2000 2 500
crystallization by high intensity incoherent light
(2), or by electron gun (3).
POWER DENSITY [kW em-2]
Fig.3 Raman TO peak shift as a function of the
argon laser power density; the beam diameter is 3
11m; a PECVD samples, 0 d.c. magn. sputt. samples.

2 •:.:.
\ :"2 -
......'
800
,.......
...... ~y -
N
-. go"
/f -
,, ,#
>QJ 000 f- // /
./
, IF
I
-.
N f-
J ../
'E I "·
u
400 I-
, ~
/ I

->
L..J
-.
N / 1/
/ 1/
200> 1500 1000 500 ..... /
200 1/ -
Raman Shift (em-I) ..c. /
tj
f- /
/
, It
It
-
Fig.2 Raman spectra of a-5i:H thin film deposited
by the PECVD method and -cfystallized for 1 minute o /
/
/ I

by argon laser (6.0 kW/ cm ). 1.2 1.4 1.6 1.8 2.0 2.2
E [eV]
Fig. 4 Optical absorption spectra of two a-5i
samples (No.2 and No.3), deposited originally by
the PECVD method, after different recrystallization
Actually, there is no well-defined power steps.
density threshold for crystallization of a-5i, as
this process proceeds also at much lower power
densities but taking much longer times to induce The TO peak widths of about 15 cm- 1 indicate
the same degree of crystallization, indicating it the remaining unidirectional stresses in poly-5L
is a thermally induced process. It has been found There is an increase in crystallized area with the
that the position of the TO peak shifts to larger increase in laser power density. We also have found
wave numbers for longer exposures to the laser that the recrystallized volume is proportional to
radiation (8). As the shift is proportional to the the sixth power of the laser power density (10).
size of microcrystals (9), larger and larger All thin film samples obtained by different
microcrystals could be formed by using larger techniques show the same Raman spectrum after a
exposure times for the same laser power density. complete crystallization.
This fact indicates an increase in temperature and By extrapolating the slopes of optical
a solid-phase bulk-induced crystallization process absorption curves, it is possible to determine the
with microcrystal lites of sizes less than 30 ~ forbidden energy gaps of the silicon samples after
(corre~~onding to the TO peak frequencies less than various recrystallization treatments (Fig. 4). They
512 cm ). approach the value of the single crystal silicon.

315
5. CONCLUSIONS

Several technological options for a low-cost

approach to multicrystalline silicop solar cell
production have been tested experimentally.
Recrystallization of initially deposited low-
quality silicon layer 'of thicknesses up to 50 ~m on
a low-cost substrate is emphasized as the main
technological problem to be solved efficiently and
economically. In evaluation of the thin film
deposition methods, the SSP method has been found
less energy consuming than the plasma spraying, but
some high deposition rate CVD methods (11) may be
more adaptable for a large volume production. In
recrystallization of relatively thicker deposited
layers, zone melting by focussed incoherent light,
combined with some crystal seeding or vertical-
constriction technique (12), seems to be an optimal
solution.
In a p-n junction formation, low energy (1
keV) unanalyzed ion implantation (6) and defect
annealing by rapid thermal processing in similar
optical furnaces as used for crystallization could
be the next, low cost and fast procedure. Then, the
solar cell processing may be completed by the
metallization and hydrogen passivation steps in
order to improve further the optoelectronic
properties of poly-Si solar cells.

Acknowledgements
The authors gratefully acknowledge Chronar
Corp., Princeton, NJ, for providing some
experimental facilities utilized in this work as
well as their colleagues at the Rudjer Boskovic
Institute for their help in performing various
experiments.

REFERENCE

(1) A.M. Barnett et al.: Proc. 8th E.C. PV Solar

Energy Conf., Florence, 1988, p. 149.

(2) T.F. Ciszek: J.Crystal Growth, 66, 655 (1984).

(3) F.V. Waldo Crystal-Growth, Properties and

Applications, Vol.5, Silicon, ed. J. Grabmaier
(Springer Verlag, Berlin, Heidelberg, New York,
1981), p. 147.

(4) C. Belouet: Proc. 5th E.C. PV Solar

Energy Conf.. Kavour i, 1983, (Reidel,
Dordrecht), p. 50.

(5) A. Eyer et al.:Proe. 7th E.C. PV Solar

Energy Conf., Sevilla, 1986, p. 714.

(6) N.B. Urli: Polysilicon Thin Films and

Interfaces, eds. T. Kamins, C.V. Thomson, and
B. Raicu, MRS Symp. Proc., San Francisco,
1990, Vol. 182

(7) D. Wruck and P. Gaworzewski: phys. stat.

solidi (a) 56, 557 (1979).

(8) H.S. Mavi et al. : J. Appl. Phys. 66,

5322 (1989) •

(9) Z. Iqbal and S. Veprek: J. Phys. C ll,

377 (1982) •

(10) M. peri!iin et al. : to be published

(11) A. Luque et aI.: Proc. 9th E.C. PV Solar

Energy Conf., Freiburg, 1989, p. 19.

(12) H.A. Atwater: MS Thesis, MIT, Cambridge, MA

(1983) .

316
lorn EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

POLYCRYSTALLINE PHOTOVOLTAIC SILICON INGOT PRODUCTION

I.A. Dorrity, B.J. Garrard, D.A. Hukin

Crystalox L1mited
1 Limborough Road
Wantage, Oxon
OX12 9AJ, U.K.

ABSTRACT. A production system for directional solidification of polycrystalline

PV silicon has been developed. The equipment currently produces 44 em square
cross section ingots, weighing up to 90 kg, with columnar grain growth and large
grain size. Solar cells fabricated from the material achieved average efficiences
of 13.~.

1. IN'l'llODU;l'ION An elevated platfonn surrounds the chamber and

allows ready operator access to the top of the
Recent teclmolodcal advances in _fer sawin& and system for assembly and lDBintenance of the
processin& have produced increasin&ly efficient furnace. The chamber lid and base plate may be
solar cells. This factor coupled with the independently raised and lowered by means of
continuin& growth of the terrestrial PV market hydraulic rams to give free access to the furnace
have resulted in a significant requirement for low interior. The furnace is fitted with a two stage
cost multicrystalline silicon in&ots. rotary and booster pump assembly which provides
rapid system evacuation and an ultimate VBCU1,IIIl of
In response to this demand, Crystalox has 0.001 mbar.
collaborated with Italsolar of Nettuno, Italy, to
produce an industrial scale Brid,iman. system, and The silicon charge is held in a 44 em square
associated process teclmolOiY for the COIIIIlercial section slip cast quartz crucible which is treated
production of hiih quality solar silicon in&ots. internally with a silicon nitride based surface
The system is fully automated to ensure process coating. This coatin& prevents wettin& of the
reproducibility and ease of operation, and crucible by the liquid silicon and subsequent
currently produces directionally solidified ingots stressing of the grown ingot durin& cooling.
weiihin& up to 90 kg.
The crucible is supported on a graphite base which
Special attention has been paid to ensure that the is mounted upon a water-cooled stainless steel
ruggedness of the design and the efficiency of pedestal. The pedestal can be rotated in the
operation are consistent with the demazds of a range of 1-10 rpn and slowly lowered in the
cost effective commercial production environment. furnace assembly. Hiih prec~swn growth
translation speeds in the range of 5 - 40 IDIl/hr
together with rapid positioning speeds are
2. &JJI"Ro1INl' DESIGN AND OPERATION provided using the Crystalox MSD4000 micro-stepper
drive unit, and 10,000 steps per revolution
The system is shown in Fig. 1. and consists of a stepper motors.
_ter-cooled stainless steel V8CU\.III chamber
containin& an induction heated graphite furnace. A 250 kW solid state induction power supply
operatin& at 8 - 9 kHz provides power to the tuned
output circuit of IDBtohing capacitors and work
coil via a step down transformer which limits the
voltage on the coil to 250 volts at maxi.mtn output
power.

The power supply is capable of operatilllt in either

hiih stability open loop power control, or in
closed loop temperature control usin&
theI1llOCOuples located wi thin the graphite
susceptor, providing +/- 0.1 0C control at 1550 oC
Both control modes are capable of bein& fully
programmed •

All controls, alams, and lOOI1i torin& functions are

displayed on the operators console, and the system
is capable of carryin& out the complete production
cycle under either manual or preprogranmed
automatic control usin& the on board progranmable
logic controller. All the necessary sensors and
operatilllt functions required for on line dats
logging and remote computer control are also
installed and interfaced.
Fig.1 Polycrystalline Silicon Production System

317
Hydraulic lowerina of the furnace chamber base A 15 x 15 em wire satm wafer cut perpendicular to
exposes the on.lOible which is then removed. by a the growth direction, with a thickness of 180 am
fork lift truck, and similarly replsced by a is shatm in Fig. 3 and provides further indication
preloeded cn.lOible. The exchaniIe of orucibles and of the large grain size obtained usina this
coumencement of another Ifrowth I'I.Ul 0Sl\ be executed. technique.
by one operator within period of 10-15 04..ti~Q'

The process cycle involves evacuation of the

system and preprojfI'8llllled. heatinlf to 900 ·C. ArJfon
jfaS is then admitted. to a pressure of 1 atmosphere
and a continuous flow maintained. as the
temperature is raised. to 1540 ·C usinlf 100 kW of
power. The susceptor temperature is held tmtil
completion of meltinlf of the charjfe, as clearly
indicated. by a recording thermocouple placed
i.DIoediately below the crucible base.

The t.eoJperature of the furnace is then reduced to

1460 ·C. and solidification initiated. by a
combination of Prolfraamed temperature reduction
and lowering of the crucible from the hot zone of
the furnace. After completion of crystallisation
the furnace temperature is ramped down to 1100 °C.
Power is then switched. off and helium gas
introduced to replace the argon and accelerate the
coolinlf of the furnace to about 400 °C. At this
temperature the crucible and ingot are removed and
a newly oharjfed. crucible introduced for the next
growth rtm.

Currently, the cycle time for the complete

process is 38 hours, givinlf a 1EXiDa.a annual
production capacity of approximately 11 tonnes.

3. MATBRIAL mlPBRTIES

The system produces 44 x 44 em cross section

inaots typically with a heiifht of 17 em and -ilfht Fig.3. 15 x 15 CD x 180 micron thick wire satm
75 kif. The inlfots display a hiifhly orientated. wafer cut perpendicular to the growth direction.
vertical jfrain structure over the whole voll.1lle
with individual crystallites showina cross Solar Cells were fabricated. uslnlf conventional
sectionalaI'eas up to 3-4 ant. Filf. 2 shows a coamercial cell processlnlf techniques on 2 ohm-Cm
vertical section of the inaot and illustrates the boron doped p-type wafers from 76 kif inaots.
excellent columar jfrain jfrOWth. Average conversion efficiencies of 13. 2X were
obtained. other parameters are sl.lllDarised. in
Table I.

Minority carrier diffusion lengths were measured

usinlf the surface photovoltage method and values
obtained. after diffusion ranged. from 129 to 184
microns. The 1DBXiJm.m whole area diffusion lenlfth
obtained from a 10 x 10 em wafer prior to
diffusion is 152 microns.

Table I TYPICAL CELL I-V CHARACTERISTICS

Cell Size 100 x 100 mm

Voe 0.592 V
Fig.2. Vertical cross section of 44 CD inlfot.
I.e 29.3 mA/ant
F'l'IR spectroscopic measurements on Boron doped
p-type inaots produced usina both scrap F.F. 0.76
Czochralski, and 'off-spec' electronic grade
feedstock, showed that carbon and oxyaen P•• s. 1.35 W
concentrations 3.5 and 5.0 x 101 ' respectively.
The low carbon and silicon carbide content make Bff. 13.2X
this material ideal for trouble free waferina by
wire saws.

Sixteen 10 x 10 CD blocks 0Sl\ be satm from each

inaot to produce 56 kif (75" yield) of material for
waferina. The low carbon and silicon carbide
content IIIBke the material particularly suitable
for wire sawlnlf, thus allowina more effective
inaot utilisation.

318
 Acknowledilements

The existing chamber and poIoler supply have been Grateful thanks are due to the following for their
designed to incorporate an enlarged coil/susceptor valuable assistance in the technical execution of
assembly which will aocoomodate 55 em square the work: described in this paper, tollether with
crucibles. A developDell.t prograume is currently the provision of characterisation services:
underway at Crystalox to implement this
modification thus significantly increasing the Mr L.S.V.Gardiner, Crystalox
productivity and Iliving an annual production Dr D.Marpdonna, ltalsolar
capehili ty of 17 tonnes. Dr F • Ferrazza , ltalsolar
Dr T •Bruton , BP Solar
Mr J.C.Vialert, Photowatt
Mr R. Brenneman, Solarex Inc.
Mr G.R.Smekens, ENE
Mr P • Lauers, ENE

319
 TOWARDS LOW COST MULTICRYSTALLlNE SILICON WAFERS
'FOR HIGH EFFICIENCY SOLAR CELLS

S. Martinuzzi, I. Perichaud, J. Gervais and D.' Sarti (*)

Laboratoire de Photoelectricite des Semiconducteurs (Case 231)

University of Marseille - F. 13397 Marseille Cedex 13

(*) Photowatt Int. S.A. - 6, rue de la Girafe - B.P. 6069 - 14002 Caen Cedex

ABSTRACT. Low cost techniques are used to improve drastically the electrical properties of
the base of multicrystalline silicon solar cells. Gettering of the samples by phosphorus
diffusion at 900°C for 4 hours increases markedly the minority carrier diffusion lengths,
which achieves 200 pm in the largest part of the sample area. Similar results are obtained
when the samples are gettered at 850°C and hydrogenated by annealing in gas flow at 400°C
for 30 mn. The external phosphorus gettering and the hydrogenation are ,found to be
complemental.

1. INTRODUCTION
2. EXPERIMENTAL
Multicrystalline silicon wafers contain all
the defects of the single crystals and also grain 200 pm thick wafers were cut out adjacently of
boundaries, twins and particularly dislocations. the same cast silicon ingots thanks to a wire saw.
In addition the concentration of impurities are In these wafers the density and distribution of
higher, and the interaction of impurities with defects are very similar due to the columnar
extended crystallographic defects is the main growth of the grains in multicrystalline material
source of recombination centres which limit (Polyx - Silso).
minority carrier diffusion lengths and lifetime in The wafers were submi tted to phosphorus
the material. diffusion at 850°C and at 900°C for 12 to 240 mn.
External gettering is a means of reducing 12 mn diffused samples are considered as standard
metallic impurity concentrations by localizing while longer diffused samples are called gettered
them in inacti ve regions of the devices. (It was verified that gettering is observed only
Phosphorus diffusion near the surfaces is a well when the diffusion duration is longer than 20 mn
known techniques of gettering which haves the /3/). Phosphorus diffusion were carried out in an
advantage to be used in the conventional cell open tube furnace using POC1 3 .
processing. The gettering sites could be the
interfaces between SiP precipitates and silicon Companion samples of 2 x 2 cm 2 were cut out of
matrix /1/, the heavy doped N+ region near the the wafers, and one N+ surface is removed by
surface /2/ or the dislocation network which is chemical etching. An aluminum ohmic grid
due to the different size of phosphorus and investigated contact is obtained by metal
silicon atoms. diffusion at 450°C for 30 mn (back surface).
In a preceding paper /3/ we have shown that in On the other surface (front surface) arrays of
multicrystalline silicon the optimal gettering mesa diodes were realized by classical
te~perature is in the range between 850 and 900°C, photolithography and chemical etching, at the same
however the influence of the gettering duration place in the matched samples. Fig.1 describes the
was not completely investigated. experimental sequences.
The recombination activity of crystallographic The improvements due to the gettering process
defects could be due also to the segregation of were evaluated by the increase of effective
oxygen and carbon dissolved in the grains /4,5/. electron diffusion length (Ln) determined from the
As oxygen and carbon cannot diffuse rapidly or
remain trapped within defects and precipitates, it spectral variation of the photocurrent in the near
could be also useful to passivate these defects by infrared and using for the spectral variation of
means of hydrogenation. The two treatments i. e. the absorption coefficient of the light /6/ :
gettering and passivation could be applied
separately or successively and their respective 86.074 2
influence like their eventual complementarity (------ - 78.150)
A
could be evaluated by means of minority carrier
diffusion length or lifetime measurements, using
matched samples. ( A in pm and a in cm- l ).
The aim of this paper is to demonstrate that,
in spite of the presence of a large density of Lifetime measurements were also done by
defects, multicrystalline silicon wafers could be reverse recovery storage time and by small signal
sufficie~tly improved in order that minority open circuit voltage decay.
carrier diffusion lengths achieve the thickness of Hydrogenations were carried out by annealing
the samples. These improvements result of the the samples, after the deposi tion of the ohmic
successive application of two low cost contact, at 400°C for 30 mn, in a gas flow.
techniques : gettering by phosphorus diffusion and The surface recombination velocity S was
hydrogenation by annealing in gas flow. Hydrogen evaluated before and after the treatments thanks
atoms penetrate in the wafers by the p type back to the spectral variation of the photocurrent in
surface, and in the base of the cells through a the near U.V. part of the spectrum.
aluminium backside grid.

320
 Fig.3. shows the evolution of effective
diffusion lengths in an other serie of matched
samples gettered at 900°C for 60, 120, and 240
I~ II tmirn P;:; 15 IOn - S fANOAIlO mn. The improvements are more marked than those
observed at 850°C, and in the samples diffused at
r----..,N' ~OO°C the value of Ln are frequently found to be
I ~ l I L~.JN' comparable to the sample thickness (i.e. 200 m).
As we have verified that in these samples the
I ~ I I ~ value of the surface recombination velocity S has

Ii.EI.L
tOI FF . P
not been decreased
(8!!!. 10 5 cm. s -1)
during the treatments
the increases of L
n
are
essential I) due to bulk improvements, revealed by
_---_N'
_____.IN+
the photocurrent variations in the near infrared.
The influence of 8 is well observed in the
measurements of lifetime T ,which does not

t
increase in the samples as diffusion length can
do.
N"/ However, like after gettering at 850° C, some
L::1 Au
regions of the sample are not drastically
improved, suggesting that in these regions, there

t are probably impurities which cannot be gettered,

or active defects or a combination of defects and
slow diffusers.
8IMS analysises indicate that in the N+ region
6fT! 0 Hyp of the investigate samples heterogeneous
accumulation of Fe, Ni and Cu are found, which are
similar in the well improved like in the poorly
Figure 1 : Experimental sequences of the investi- improved regions. Chemical etching of the samples
gation of phosphorus gettering and hydrogen reveal that in the well improve~ regions the
passivation. density of etch pits and that of grain boundaries
is very low, while grain and subgrain boundaries
The nature density and distribution of defects are present in the poorly enhanced areas.
were investigated after chemical etching of the As we have previously reported /3/, the
samples. phosphorus gettering modifies the etch pit density
of the wafers as shown by fig.4. This is
particularly true for dislocation lineages, which
3. RESULTS tend to disappear completely. We do not intend to
claim that the defects have physically disappeared
Fig.2 shows the evolution of three matched but rather that their chemical reacti vi ty to the
samples submitted to different phosphorus etching solution is drastically reduced after the
diffusion time at 850°C : 12 mn, 120 mn and 240 treatment.
mn. When the phosphorus gettered samples at 850°C
It is clear that the effective electron are hydrogenated, L increases again as shown by
diffusion lengths measured at the same place in fig.5 indicating thJt L values higher than 150~ m
the different samples increase neatly with could be obtained by t'he combination of the two
diffusion time. However the improvement mechanism techniques.
appears to be a continuous process, and some These two techniques appear to be additive or
regions of the sample are poorly improved even complemental as phosphorus gettering decrease the
after the longer treatment. concentration of metallic impurities while
hydrogen passivates the crystallographic defects,
probably by the saturation of 8i0 2 bonds /4/.

190

150

110 .

70 . 0

30 0

Figure 2 Influence of phosphorus gettering time at 850°C on diffusion lengths measured in mesa diodes.

321
 In!l'ml
200.
1/ 240

a 120
160.
• 60

.... IS ... n
120.

90.0-

40.0

2 6 10 14

Figure 3 - Influence of phosphorus gettering time at 900°C on diffusion lengths measured in

mesa diodes (as the sample thickness is about 200 ~m, the values of L in diodes 4 to 11
and in diodes 13 are arbitrarily limited to 200 ~m). n

These improvements are less marked in

restricted areas in which a high density of grain
and subgrain boundaries density are found,
suggesting that these regions contain impurities
which cannot be gettered and which have probably
formed clusters, precipitates or inclusions /7/
which the recombination centres cannot be
passivated by hydrogen.

4. CONCLUSION

Except in some restricted areas, it is

possible to increase drastically the minority
carrier diffusion length values in the largest
a part of the mul ticrystalline samples by means of
low cost techniques which are already used in the
conventional cell processing or which are
compatible with the cell fabrication technology.
The multicrystalline wafers become comparable
to CZ monocrystalline wafer and can support after
gettering, high temperature treatments.
Hydrogenation could be applied by the back surface
through the aluminum grid used to realize the
ohmic contact after the complete preparation of
the cell or during the end of the metallic contact
deposition when the temperature of the device
decreases and reaches 400°C.
It is needful to investigate the limiting
factors in the restricted areas where the minority
carrier diffusion lengths are poorly improved, by
b SIMS analysis of impurities and by the study of
the crystallographic defects.

This work was supported by CNRS-PIRSEM

Figure 4 - Evolution of etch pit density and
(France), by Agence Fran9aise pour la Maitrise de
distribution before (a) and after gettering by
l'Energie (France) and Commission of the European
phosphorus diffusion at 850·C for 240 mn. Communities (Joule Programme).

322
 Ln (pm) ~ DIFF. P I B50 C
145.
CP0 \0(!)f%)
115.

B5.0

55.0
G12 MN DE DIFF. P

0120 MN DE DIFF. P
25.0
C!)120 MN + HYD. 400·C 30 MN
2 6 12 14 18
DIODE NI

Figure 4 - Influence of phosphorus gettering and of hydrogenation of the minority carrier

diffusion lengths. Hydrogenation increases Ln even when Ln values are higher than 110 ~m.

5. REFERENCES

/1/ A. Ourmazd and W. Schroder, Appl. Phys. Lett.

45, 781 (1984).

/2/ J.S. Kang and D.K. Schroder, J. Appl. Phys.

65, 2974, (1989).

/3/ S. Martinuzzi et al., Proc. of 9 th European

Photovoltaic Solar Energy Conf. Freiburg -
RFA - 1989, pp.458.

/4/ L.L. Kazmerski in Polysilicon films and

interfaces MRS proc., 106, 1988, p.199.

/5/ S. Pizzini et al. Appl. Phys. Lett. 51, 676

(1987).

/6/ J. Gervais, private communication - 1990.

/7/ R.K. Brenneman et al. Proc. of 21st IEEE PV

Spec. Conf. Orlando, 1990 p.1290

323
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THERMAL ENVIRONMENT AND PROCESS OPTIMIZATION OF SOLAREX CAST

POLYCRYSTALLINE SILICON
Narayanan.S, Terry.J, Brenneman.R
Solarex Corporation
630,Solarex Court,
Frederick,MD,21701,USA.

ABSTRACT. There has been a steady improvement of the quality of

Solarex polycrystalline silicon since its commercialization. In
this paper a sUbstantial improvement in the distribution of high
efficiency cells from our casting stations is reported. This was
achieved by identifying and incorporating a critical thermal
regime during casting of the ingots.

the region and to a lesser degree for the

Introduction material higher up in the ingot. Stresses
generated during the solidification of
Cast polycrystalline silicon has emerged the ingot leads to the formation of a
as a low cost alternative to single high density of dislocation (up to lOE7)
crystal silicon as a substrate for solar near the bottom of the brick (7). The
cells. Nearly forty percent of stress -strain characteristics of the
crystalline silicon solar cells in 1989 material have been analyzed (8). The
were fabricated on cast polycrystalline critical yield stress of the material
wafers (1). It has been demonstrated that from the bottom of the brick is almost
the performance of the cast twice that of the material from the top
polycrystalline solar cell could almost of the brick. Also, the bottom region of
match that of single crystalline solar the ingot contains a higher percentage of
cells (2). Many companies have developed inclusions. Finally, there is a variation
processes to cast polycrystalline in the solar cell efficiency as a
silicon ingots in graphite or ceramic function of the height in the ingot as
molds. All these processes are based on well as in the yield of higher
directional solidification principle in efficiency cells from the bricks from
which the ingot solidifies from the station to station.
bottom to top with a certain degree of
controlled growth (3,4). The aim of the present work is to
decrease these variations and thereby
Solarex started the commercial production increase the average conversion
of cast polycrystalline ingots in 1982. efficiency of the solar cells fabricated
At that time approximately 40% of the from the ingot. The thermal regime during
ingot was waferable and the average the casting and its effects on the
efficiency of the solar cells fabricated material quality was investigated. A
on the material was about 9%. Since then series of experiments to identify the
substantial research and developmental major source for these variations was
efforts at Solarex were directed to conducted. Based on this work an
increase the waferable yield and to optimized process in all our production
improve the electronic quality of the casting stations was implemented.
wafers. Within five years ,the control of
the casting station was automated and
high purity ceramic molds were developed Experimental
at Solarex. As a result, the waferable
yield increased to 75% and the electronic During the first set of experiments ,
quality of the wafers improved (5). details of the casting run for each
Modifications to the insulation package station were recorded. The ingot from
further improved the electronic quality each station was cut into bricks. Bricks
of our material. Devices up to 12.5-13% from each casting station were then
could be routinely produced (6). In this wafered • Wafers from each brick were
paper, the improvements achieved by marked to identify their relative
optimizing certain other aspects of the position in the brick and the originating
thermal environment during the casting casting station. All these wafers were
are reported. coprocessed into solar cells and tested
Based on analysis of the solar cell test
results, an optimum thermal regime for
BaCkground ingot casting was identified as described
in the next section.
The details of the Solarex controlled
solidification process have already been The next step was to test the validity of
published (4). The silicon at the bottom the optimized thermal regime on a large
of the crucible solidifies rapidly. This sample wafers originating from many
produces small average grain size of the casting stations. During this trial,
material in this region. The premature corrective action was taken wherever
freezing of silicon significantly affects necessary to. achieve the optimized
the electronic quality of the material in thermal regime . Details of each casting

324
run were recorded. Some of the casting runs which met the thermal criterion. The
runs still did not meet the thermal optimization of the thermal regime has
regime criterion. Solar cells were significantly improved our capability to
fabricated after wafering all ~hese produce higher efficiency cells using the
ingots. The test results of the solar same cell process sequence.
cells are presented in the next section.
Conclusion
Results and Discussion.
There has been a steady improvement of
The parameters of the solar cells (from the Solarex casting process since it was
the first set of experiments) as a first commercialized. The optimization
function of position in the brick and the of the thermal regime during casting
casting details of the respective bricks resulted in the improvement of the
were analyzed to detect any correlation electronic quality of our material. The
using "Systat" statistical analysis thermal criterion has been successfully
software. The statistical analysis implemented in all our casting stations
(namely principal component analysis) of resulting in a 14% improvement in the
the casting temperatures and the cell yield of high efficiency cells. This
parameters revealed that the performance improvement has been achieved without any
of the solar cells were better if, during increase in the manufacturing cost.
casting run of the ingot, the crucible Further work is needed for a better
was within a critical temperature range understanding of the effects of the
prior to initiation of the thermal environment, and the role of the
solidification. The short circuit current defects ,impurities and stress on the
density, the open circuit voltage and the performance of the material.
efficiency of the solar cells as a
function of their relative position in Acknowledgement
the brick are presented in Figure 1 a-c
We can see that for the casting runs in The authors thank Dr. John Wohlgemuth
which the crucible was within the for his suggestions and other staff at
critical temperature range the solar cell Solarex for fabricating the solar cells.
parameters were better .The overall
electronic quality of the brick improved
, particularly for the material near the Reference
bottom of the brick. The relative 1. Private communication from Robert
improvement in the solar cell parameters Johnson, Strategies unlimited
and the distribution of the high
efficiency cells are shown in Table 1. 2. S. Narayanan, S.R. Wenham, M.A. Green
There is a modest increase in the solar ,"High Efficiency Polycrystalline Solar
cell parameters but a significant Cells" ,Proceedings of Fourth
improvement in the distribution of higher International Photovoltaic Science and
efficiency cells from the bricks. Also, Engineering Conference, Sydney, 1989,
it became clear that one of the major Page Ill.
sources of variations in the performance
of the materials from different casting 3.J. Wohlgemuth," Casting Polycrystalline
stations was the temperature of the Silicon for Photovoltaics", International
crucible. Corrective actions were Symposium Workshop on Silicon technology
initiated to ensure that each casting development and its role in sun belt
station meets the temperature criterion. countries, 1987, Islamabad, Page G1.
The improvement in the material quality 4. T.F. Ciszek, "Silicon for solar cells
could be mainly due to the decrease in " ,Crystal growth of electronic
the amount of silicon which prematurely materials, E.Kaldis ed, Elsevier Science
solidified. The stresses generated during Publishers, (1985) , page 185.
the solidification of the ingot are lower
due to the decreased solidification rate. 5.J. Wohlgemuth, R.K. Brenneman
Detailed analytical characterization of "Improvements in casting of
the material is underway and will be polycrystalline silicon for solar cells",
reported separately. Technical digest of the International
PVSEC-3 ,Tokyo, 1987, page 93.
During the analysis of the results of the
large scale trial, the test data of about 6.R.K. Brenneman, T.A. Tomlinson, "State
160,000 solar cells were classified under of the silicon casting art" ,Proceedings
two groups. The first group contained of the twentieth IEEE Photovoltaic
cells which originated from ingots specialist conference, 1988 , page 1395.
adhering to the critical temperature
criterion. The second group contained the 7.J. Wohlgemuth, S. Narayanan ,"Impact of
cells which belonged to ingots not defects on the performance of silicon
meeting the criterion. Only a small solar cells made on cast polycrystalline
fraction of the total casting runs did silicon" Work shop on " The role of point
not meet the criterion indicating that defects/defect complexes in silicon
with adequate preventive maintenance a device fabrication, 1990, Keystone,
very high degree of conformance can be Colorado.
achieved. The distribution of high
efficiency cells for materials from three 8.D.G. Ast ,Sandia Contractor Report,
casting stations and from all the casting SAND90-7029, 1990.
stations is presented in Figure 2 . There
was a 14% increase in the distribution of
high efficiency cells for all the casting

325
 NORMALIZED ISC VS. POsmON
ISC TABLE 1
l ~ r-------------------------~

t-----------------~""""-..,....._I PW- ..... """ IMPROVEMENT IN SOLAR CELL PARAMETERS

•• r--_ _7~"='_-"-'-···-···-····-···-····-·°.......j·· AND DISTRmUTION OF HIGH EFFICmNCY CELLS
... r--r==/7-"
..--
.....--
....--.--~
... ~ .... ,
Efficiency Current Voltage Distribution
Hi~h Eff. Cells
.'
.~ ~------------------------~ 1.6% 1% 0.3% 15%

POSmON IN BRICK

(a)
CASTING STATION PERFORMANCE
NORMALIZED VOC VS. POSITION
voc
lmr---------------__________-,

t--------------------=_~__l DtDMJTMUT UNn

~
.~ ....... .
.'
....... --
C····
-t-------~~S_~----------__l

.Mr-~~~~----------------~

.nr-------------------------~

rosmoN IN BRICK

Figure 2. Effect of optimized

thermal regime on the distribution
(b) of high efficiency cells.

NORMALIZED EFF VS. POSmON

u" ,----____________________-,
EFF

...........~

r.: . . .
~
/ .....
•Mr---~r<~~--------------~
. ~~----------------------~

rosmo IN BRIO;

(c)

Figure 1. Solar cell parameters as

a function of position in brick.

326
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A LOW-COST DIFFUSION PROCESS FOR THE FABRICATION OF SOLAR CELLS

J,A.S.Dias, N.V.Serran*; O.V.Avilez and A.M.Andrade+
LED/DEMIC/FEE/UNICAMP
Universidade Estadual de Campinas
*Hober S.A. Semicondutores
P.O.Box 6061 - 13081 Campinas, SP BRAZIL
+LME - Universidade de Sao Paulo
Caixa Postal 8174-01051-Sao Paulo-Brazil

ABSTRACT. Paper film dopants can be an interesting alternative to the

conventional diffusion sources used in the solar cells production. The
fabrication process using these paper film dopants (known as diffusant
pre-forms) is very simple. There is no need for gases in the process
and the diffusion furnaces can be loaded with thousands of wafers, so
a high productivity is achieved. The quality of the junctions obtained
with this process is not as good as with standard fabrication
procedures, but the diffusion cost is much lower when compared with
the other techniques.

1. INTRODUCTION. these steps must be performed twice, increasing the

In developing countries it is even more costs excessively. Also the number of wafers that
important to have products with a low price and a can be loaded in a long diffusion furnace cannot be
medium quality than to have high quality but expen much more than 200.
sive ones. Solar cells are a typical case where the
production cost of the device can be more 2.2 Planar Solid Sources
important than its efficiency. The processing costs These sources are very attractive for the pre-
for the fabrication of solar cells are still very -deposition steps of Integrated Circuits - very high
high, specially for high efficiency devices. This quality junctions are obtained with these dopants -
imposes a strong limit in the application of solar but they are not convenient for the fabrication of
cells in poor countries, even if they have a large solar cells. They are very expensive, extremely
amount of solar energy available during the whole sensitive to humidity (it is mandatory to have high
year, which is the case of Brazil. quality gases in the diffusion furnaces), and need
Paper film dopants have been widely used in special procedures for inserting/withdrawing of the
the production of rectifiers and power devices. The diffusion boat in the hot zones of the diffusion
main reason for their use is the simplicity, high tube or permanent damage can occur to the sources.
productivity and low cost associated with these pre In what concerns productivity, if the first and
-forms in the diffusion process. As far as we know~ last slot of the diffusion boat are o,ccupied by
these pre-forms have never been used for the dummy wafers (which is generally required) and there
fabrication of solar cells since they are normally are N planar solid sources in the boat, the ratio
used to make deep junctions. (number of wafers / number of sources) is given by
A process using diffusant pre-forms which is 2-(2/N), what means that for any number of slots in
much cheaper and easier than all other diffusion the diffusion boat the number of wafers is always
techniques is under development. In this paper we less than twice that of solid sources. This
will discuss the various characteristics of the restricts the maximum number of wafers in a batch
different techniques used to make solar cells and to about 150.
compare them with the pre-form diffusion process.
Some diffusion data obtained with the pre-forms are 2.3 Spin-On Dopants
also presented. Spin-on dopants are cheaper to use than planar
solid sources and do not present some of the
2. A COMPARISON BETWEEN PROCESS TECHNOLOGIES. problems found in the diffusion with gaseous
The conventional diffusion technologies used dopants. However, if the wafers are coinstacked in
for the fabrication of standard solar cells are the furnace to obtain a higher productivity, the
basically based on three types of diffusion sources: separation of the wafers after diffusion is a very
gaseous dopants, planar solid-sources and spin-on long and difficult process. The procedure of
dopants. These types of dopants present good sprinkling A1 powder between the wafers to minimize
performance but, when compared with diffusion pre- their adhesion (commonly used in the production of
-forms, have several drawbacks in what concerns power devices) cannot be applied due to the high
costs and productivity. A summary of the problems value of the diffusion coefficient of the aluminum,
commonly associated with these dopant sources is since this would cause the resulting junction to be
given below: very deep.
Furthermore, spinners with an automatic
2.1 Gaseous Dopants. dispenser, well controlled exhaust system and
Gaseous dopants produce good junctions but are precise acceleration/deceleration are needed if
expensive, difficult to handle, hazardous when not uniform striation-free films are to be obtained with
properly operated and require special equipment in a good repeatability. Depending on the spin-on
the diffusion furnaces. If the wafers do not have a dopant, a baking step is needed before the diffusion.
protection on the back side (typically made with If the wafers are not coinstacked, the number of
silicon oxide, what means an extra process step), wafers per batch is the same for the gaseous dopants
both sides of the wafer are 'doped during the pre- case.
-deposition step and a back side junction removal
step is needed. Furthermore, if one needs diffusions 3. THE DIFFUSION PROCESS WITH PRE-FORMS.
on both sides of the wafers (for example, N type in Diffusant pre-forms a re hi gh purity paper fil ms
the front side and P type on the back side) all impregnated with a doping impurity, like Boron or

327
Phosphorus. They are available in several After the diffusion, the coinstacked is dipped
diameters, from 1 to 8 inches, so that they match in a HF solution to separate the wafers, which then
exactly with a silicon wafer. I.t is possible to are cleaned to go to the metallization step. Next,
choose a pre-form with different concentrations and an electroless nickel plating is used to make the
minority carrier life-time. Therefore, des~gning a interconnection rails.
junction with a given surface concentratlOn and
also having control on the minority carrier life-
-time is simple and straightforward.
The diffusion process using the diffusant pre-
-forms is extremely simple. Using Nand P type pre- N TYPE DIFFUSION P TYPE DIFFUSION
-forms (they have differe~t colour~, to m~k~ it
easy to identify them) a c01nstack w1th: a s111con
wafer, N type pre-form, silicon wafer, P type pr~­
form silicon wafer, N type pre-form, ... , 1S
mounted in a coinstack style diffusion boat. In ps in (lIO
Fig.l it is shown a coinstack mounted as des~ribed.
The quartz discs are needed to keep the c01nstack
tight and assure that there will not be dislocations
in the wafers or pre~forms duri ng the burni ng of the
paper (if a pre-form is not well aligned with a Figure 2: Sheet res i sti vity map measu.red ina
wafer, some cross-diffusion can occur). 2 inches wafer doped with pre-forms.
The burning of the paper must be done at a.low
temperature (in the range of 600 OC) to obta1n a
very uniform deposition of the dopant. If the pre-
forms are burned very fast, the uniformity of the 4. CONCLUSIONS
deposition is poor. The burning of the pre-forms The di ffusi on process usi ng di ffusant pre-forms
can be made in a special furnace that is kept at a is a very cheap technique that can be used to the
low temperature. Next the diffusion tioat is fabrication of solarcells. The costs of the process
transferred to another furnace, where the drive-in are basically the costs of the silicon wafers, since
is done. For a high throughput process, the pre- the cost of the diffusion pre-form is about 1/25 of
forms can be burned in the mouth of the same the cost of the wafer, and there is no need for
diffusion furnace where the drive-in will be special gases in the whole diffusion. process. The
performed so only one tube is needed for the number of wafers that can be used in one batch is
complete process. After the end of the burning, about ten times those obtained with conventional
what takes about 15 minutes, the diffusion boat is diffusion techniques. These two factors result in
pushed slowly to the center of the tube. an extremely low-cost process, that can be
If one needs a very uniform deposition step, convenient to use if the efficiency of the solar
it would be convenient to use a very low content of cells are not the most important parameter to be
oxygen during the burning of the pre-forms. A taken into account.
mixture of N2 and 02 with 2-10% of oxygen generally
results in very uniform depositions. However, if a 5. ACKNOWLEDGEMENT .
very low-cost process is desired, the gases can be The authors woul d 1i ke to thank FINEP, CNPq· and
substituted by filtered air, which then is used INEPAR ELETROELETRONICA, which supported this work.
throughout the process. We would also wish to thank Prof. Anne Corinne
Using filtered air it is possible to o~ta~n a Gottberg for the helpful discussions and her
sheet resistivity uniformity of about 5% w1th1n a enthusiasm with the Solar Cell Program in our
wafer. The results of a sheet resistivity map test Laboratory - LED/UNICAMP.
performed in a 2 inch wafer doped with Phosphorus
P35% and B5% types pre-forms /1/ is shown in Fig.2. REFERENCES
This pre-deposition was performed in one step /1/ Filmtronics, INC., Diffusant Pre-forms Catalog.
(that is, using the same diffusion tube for the
burni ng and the diffus i on of the juncti on) and us i ng
filtered air. As seen, the uniformity of the
deposition is good.
Since the wafers are coinstacked horizontally,
the number of wafers in a batch can be as much as
2000, in a medium size diffusion furnace. No
special handling or storage procedures are needed
for the pre-forms, that have a shelf life of six
months.

Si

P TYPE N TYPE

Si "'''=.__ WAFER

Figure 1: Both P and N type diffusions are made

simultaneously with the pre-forms.

328
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

INFLUENCE OF HYDROGEN PLASMA IN THE PROPERTIES OF a-Si:K/a-SiC:H INTERFACES

M.C.A. FANTINI
Instituto de Fisica, Universidade de Sao Paulo
Caixa Postal 20516, CEP 01498, Sao Paulo,SP,Brasil
I. PEREYRA, M.P.CARRENO, A.M.ANDRADE
LABORATORIO DE MICROELETRONICA, ESCOLA POLITECNICA, UNIVERSIDADE DE SAO
PAULO
Caixa Postal 8174, CEP 01051, Sao Paulo,SP,Brasil

ABSTRACT. In this work we study the a-Si:/a-siC:H in interface multilayer

structures deposited by glow discharge technique. In order to improve the
quality of the interfaces we introduce an extra hydrogen plasma step between
the deposition of consecutive layers without interrupting the RF power. We
compare this method with the traditional one, just changing the gas mixture
without interrupting the discharge. To evaluate this method, identical
multilayer structures with and without the extra hydrogen plasma were
fabricated. The interfaces were characterized by small angle x-ray
diffraction using reflection geometry and by opto-electronic measurements.
The influence of the structure parameters on the experimental diffractograms
will be discussed and compared with theoretical results.

The existence of quantum size effects 200 R while the a-SiC:H layer was kept
in crystalline superlattices and constant at -50 R.· Another important
mul tilayer devices has been observed for parameter to consider is the residence time
many years, but only recently the existence (T r ), which can be controlled by the gas
of optical and electrical properties, flow rates and by the deposition pressure.
attributed to quantum size effects, in In order to obtain abrupt interfaces Tr
amorphous multilayers has been reported. should be much smaller than the deposition
Even more, lately several works have shown time (Td ) for the thinnest single layer. In
interesting results in solar cells our case we guarantee this condition with a
utili~ing multilayers as window material Tr of 10 s, which is an order of magnitude
due to their improved electrical and less than the smallest Td , 120 s for the
optical properties / 1,2/. Also it is well 25/50 R structure.
known that the role of the interface is
essential for the obtention of the desired
transport properties. So in this work the 2.- RESULTS AND DISCUSSION
properties of the a-Si:H/a-SiC:H interface
are analyzed by optical absorption, The optical gap, the parallel
spectral photoconductivity and small angle conductivity and the mobility lifetime
x-ray scattering (SAXS). product ().IT) of the mul tilayers were
measured and related with the a-Si:H layer
thickness (T S ·). The optical gap was
obtained from the absorption spectra and
1.- EXPERIMENTAL the ).IT product from the spectral
photoconductivity.
Two. set of multilayers of The first set of multilayers, grown
a-Si:H/a-sic:H were deposited onto 7059 without the intermediate hydrogen plasma,
Corning glass by plasma decomposition of did not show any special dependence of
appropriated gaseous mixtures of SiH 4 and these parameters with TSi. On the other
CH 4 . The samples were 20 layers hand, the second set, w~th the hydrogen
multi structures obtained by alternating plasma step, showed a well defined
the gaseous mixture without interrupting dependence of the optical gap, the parallel
the discharge. An extra hydrogen plasma conductivity and of the mobility lifetime
step between the succesively deposited product with the a-Si:H layer thickness, as
layers is introduced, in one of the sets, it is shown in Fig.l.
to etch the undesirable residual layer It can be observed that for an a-Si:H
grown at the change of the gas mixture. layer thickness less than 60 R the optical
The temperature and RF power of th~ gap increases and the).lT product decreases,
depositions were 250 °c and -25 mW/cm which is consistent with the existence of
respecti vely. The thickness of the discrete levels within the a-Si:H layer
individual layers was estimated from the /3/. Even more, the fact that the first set
deposition rate of thick films of a-Si:H of samples do not show this behavior could
and a-SiC:H (-13 R/min for both materials) be due to non-abrupt interfaces which, as
grown under identical conditions than it is shown below, is confirmed by the x-
those of the multilayers. One should ray analysis.
remember though, that the kinetics of the Small Angle X-ray Scattering (SAXS)
nucleation process depends on the substrate with reflection geometry and conventional
conditions and thus one should expect tube radiation was used to analyze the
different growth rates for the first few samples in order to obtain information on
layers. The a-Si:H layers in the structures the period of the produced structures as
were varied from approximately 25 R up to well as on the quality of the

329
 Table 2 shows the obtained
thicknesses after depositing ten pairs of
the a-Si:H/a-siC:H stack. The total
thickness measurements were made with a
profilometer.
Table 2. Thicknesses obtained after the
multilayer deposition.
Sample da-Si:H (R) da-SiC:H (R) d(R)
MHI 36 72 108
MH2 51 68 119
MH3 61 61 122
MH4 110 55 165
MH5 236 59 295

The SAXS measurements were performed

in a small angle camera (Rigaku), using
a reflection attachment. The slits were
chosen to high resolution alignment. A
Geigerflex (Rigaku) x-ray generator
operating at 1.2 kW, and Cu-Ka radiation
filtered with Ni, were used. The
30 90 150 210 measurements were taken in a step
scanning mode (steps of 0.01 0 in 29) for
TSi (lll 10 sec .. The angular variation stayed
between 0.2 0 and 4.0 0 in the 29 scale.
The obtained x-ray values for the
period of the superlattices, for
Fig.1 Dependence of optical gap and).lT those diffractograms where interference
product with the a-Si:H layer thickness maxima showed up, are I isted on Table 3.
(TSi) • The multilayers grown without the
hydrogen plasma did not present any sharp
interference patterns that could be used
interfaces. This technique is one of the for period determination.
most appropriate to characterize
amorphous superlattices, because with Table 3. X-ray mean period values.
the interference maxima one can calculate
the mean period of the multilayers and Sample
compare it with the reactor deposition
rates, serving as an external probe of MH3 155
the deposition technique. Besides that, by MH4 233 220
inspecting the overall x-ray intensity
features, like the peak broadening and comparison of the results from tables
the occurence of high order maxima, one 1 and 2 shows that the obtained
can evaluate the quality of the produced thicknesses are always larger than the
multilayers. The ability to produce high expected ones. In this way it can be
quality superlattices is connected to two concluded that, either part of the
important parameters: the homogeneity of deposition is due to resident gases inside
the individual layer thickness (or system the reactors, then increasing each
periodicity) and the abruptness of the layer thickness or the deposition rates
in Lerfaces {or interface roughness and for both layers at the beginning of the
mixing). Measured reflectivities are deposition is larger than the "mean" rate
compared to theoretical calculations that determined for a thick sample. On the
take into account those important other hand the difference between the
parameters and valuable informations on values obtained with the profilometer and
the studied systems are derived /4/. from the x-ray measurements, tables 2 and
The thickness of the individual layers 3, are around 25%. The most reliable
were estimated from the deposition rates resul ts are those obtained from the
which, as it was said before, were obtained SAXS measurements, when precise x-ray
by growing a thick sample of each one of optical constants are available.
the multilayer components and measuring The determination of the periodicity
their thickness in a profilometer. The of the studied multi layers uses Bragg's
obtained values, for both materials, were law corrected by refraction /4,5/ and
-13 R/min. Table 1 lists the expected x-ray optical constants of crystalline Si
values. and SiC, since better values for amorphous
materials are not available in the
Table 1. Thickness calculation based on the literature and may vary very much with
deposition rate. The "d" is the the sample growth condition.
superlattice period. The experimental diffracted
intensities for two representative samples
Sample da-Si:H (R) da-SiC:H (R) d(R) are shown on Figures 2a and 2b. Figure
2a shows the results for a sample grown
MH1 26 52 78 without the hydrogen plasma between
MH2 39 52 91 the succesive layers, while Figure 2b
MH3 52 52 104 shows the results for a sample with
MH4 104 52 156 hydrogen plasma, and clearly demonstrates
MH5 208 52 260 the effectiveness of the hydrogen etching

330
plasma step due to the existence of well is:
defined interference peaks.
(3)
o~~~~ __________________________-,
( al The rugosity and thickness variations are
included by assuming the presence of
inhomogeneities parallel (1) and
perpendicular (r) to the substrate, that
....: .. promote an exponential decrease in the
..~. ,:'" " reflectivity, through a factor given by:
.... :,,:.:-:::::.. .....:.:..... '" .•.; ....._.......;
",

p = exp (-Q a- ) 2 exp (_Q a- ) 2 (4)

ra IQ 1 II r
I

I 2
2 Theta Combining those effects one can derive a
single parameter (0) that contains those
contributions:
0r-____~----------------------,

i. M=I
""'. (b I
(5)
\~
I- The interface thickness is then equal to
20, and the measured reflectivity is
expressed by:

-5 + MH4 I IRK= (D/W) sin (e) IR exp[ (-21Ima-/d) 2] (6)

o 2
2 Theta
where m stands for the diffraction order.
The comparison between the perfect
superlattice result and the Gaussian one
Fig.2 Log of the normalized experimental shows a decrease in the intensity of the
diffracted intensity for (al Multilayer ____ ________________________--,
grown without the H2 plasma step and (b)
O~ -=~

Multilayer grown with the H2 plasma step.

The x-ray reflectivity of a N layer

stack, including the substrate and the air
interfaces as the boundary conditions, is
easily calculated for a given angle e,
using Fresnel's coefficients (Rn):

IR ~ (1)
I
o
abrupt
Qauss ian
The measured. reflectivity is given by:
-5 L-______________- L________- L____ ~

(D/W) sin (e) IR (2)

o 2
2Theta
where (D/W) sine considers the sample
irradiated area.
The theoretical calculated
reflectivity of an amorphous superlattice
with ten periods, identical to the sample Fig.3 Log of the reflectivity calculated
MH4, is depicted on Figure 3. This for an amorphous superlattice having 10
figure showns the calculations for a periods with abrupt interface and with a
perfect superlattice, with abrupt gaussian distribution of mixed materials.
interfaces, and for a multilayer system,
where mixing of the materials at the
interface, interface roughness and layer interference peaks due to the
thickness inhomogeneities are considered. imperfections. This kind of effect in the
In this model those contributions can not diffractogram is also expected when
be separated /6/. The calculations shown in roughness becomes the most important defect
Fig. 3 are for an estimated 20 R interface in those structures.
thickness. In this model, the imperfections Analogously, if one assumes that
of the structure can be introduced interface mixing is the most
considering a Gaussian distribution of the deleterious effect that may cause the
mixed material at the interface, such that decrease in the x-ray reflectivity power
the resul ting refractive index (n) as and peak broadening, it is possible to
function of the individual ones (nA and n B) model only its contribution, neglecting

331
roughness and thickness variation. Figure 0 ,..-___ """'_ _______________---,
4 shows the calculations obtained by
assuming a sine type mixing of
the a-Si:H/a-SiC:H at the interface (20 ~
thick), for a sample identical to the MH4.
0 r---__~----------------------_.

M=2

M=3
abrupt
nan- homogeneous d

-5 L -______________ ~ ______________ ~

obru pi o 2
s i ne 2 Th ela
Fig.5 Log of the reflectivity calculated
- 5 L---------------~-- ____________~ for an abrupt interface (perfect
o superlattice) and for a superlattice with
2 thickness variations of ± 10 H.
2 Theta

Then, the determination of the interface

Fig.4 Log of the reflectivity calculated thickness could be directly obtained.
for an abrupt interface and for a sine type Summarizing, it ,was demonstrated the
mixing of the materials. effectiveness of the hydrogen plasma step
in obtaining more abrupt interfaces. The
samples grown without the plasma step did
In this case, equation (3) is modified in not show interference effects in the
order to consider the proposed variation of diffractograms neither variations in the
n(z), such that: optical gap. This implies a mixing of the
interface greater that 20 H, according to
nA . nB the theoretical models. The reason for this
n(z)= --2--[1-sin(rrz/u)]+--2--[1+sin(rrz/u)] (7) however, can not be the residence time,
Tr , which as said before, is 10 times lower
than the smallest deposition time. So, it
where r is the interface mean thickness. must be related to the response time of the
Again the comparison is stablished with a mass flow controllers and valves, showing
perfect multilayer. The results indicate that a low residence time is not enough to
only a small shift in high order guarantee abrupt interfaces.
diffraction maxima, thus mixing of Eventhough the fitting of the
materials a:t the interface might be an experimental results with the theoretical
important parameter only for layers with models cannot be done due to the absence of
very different x-ray optical constants. precise optical constants, a comparison
The last case to consider is when the permits to evaluate which of the interface
variations inherent to the plasma chemistry parameters affects more significantly the
and kinectics cause an inhomogeneous diffractograms. In the case of a-Si:H/a-
distribution of layer thickness, SiC:H, materials which have similar
neglecting m~x~ng and roughness. Figure optical constants, the most important
5 depicts the comparison between the parameter to consider is the variation of
perfect superlattice and the MH4 the thickness of the individual layers.
stack calculations, sup~osing a thickness This hypothesis is reinforced by the
variation of ± 10 It for each experimental fact that the first deposited
component layer. The resul t shows peak layers are grown with different deposition
intensity decreasing and broadening. rates than the last ones.
It is important to point out that
the strict comparison between the 3.- REFERENCES
experimental results and the theoretical
ones cannot be done yet. The agreement (1) Y. Kuwano et al., J. Non-cryst. Sol.,
between them must be achieved with a more 97&98 (1987) 289.
precise knowledge of the x-ray optical (2) S • Tsuda et al • Proc. 18th IEEE
constants of the amorphous materials Photovol. Spec. Conf. 1295 (1985).
used in these studied superlattices. (3) B. Abeles and T. Tiedje, Phys. Rev
Measurements are in course in order to Lett., 51 (1983) 2003.
obtain those constants. Besides, one may (4) J.H. Underwood and T.W Barbee, AIP
need to combine the above described Conf. Proc. No.75, p.110. New York,
theoretical treatments in order to consider USA (1981).
a real system. This is usually done (5) M.C.A.Fantini, P.V.Santos, I.Pereyra
through fitting procedures, what and M.N.P. Carreno, 5th. Brazilian
reinforces the necessity of better School on semic. Physics, World
knowledge of the x-ray optical Scientific, Singapore (1991), in print
constants, to decrease the number of (6) P.V. santos,M. Hundhausen,L. Ley and
fitting parameters giving reliable results. C.Viczfan, J.Appl. Phys.,69, 178(1991)

332
 POSTER SESSION 2B

Amorphous Silicon and Related Materials II

10TH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THIN METALLIC INTERFACIAL LAYERS: A MEANS TO IMPROVE THE TIN-OXIDE/

BORON DOPED HYDROGENATED AMORPHOUS SILICON INTERFACE ?

P. Roca i Cabarrocas and U. Eicker*

Laboratoire de Physique des Interfaces et des Couches Minces (UPR
0258 - CNRS), Ecole Poly technique, 91128, Palaiseau, Cedex, France
* SOLEMS, 3, rue Leon Blum, 91124 Palaiseau, Cedex, France

ABSTRACT. Among the factors limiting the open circuit voltage of p-i-n solar cells,
the bending at the tin-oxide/p(a-Si:H) interface is one of the most important. To
reduce this band bending we evaporate thin metallic interfacial layers. As a
consequence we observe an increase of the open circuit voltage, provided that the
p(a-Si:H) layer is not too thick. Moreover, the interaction of a-Si:H with the metal
and the formation of a silicide, as well their effects on the solar cell performance
are investigated

1. INTRODUCTION factor FF and of J se , without any effect on

Vee are contrary to our expectations. We
The research on a-Si:H has led to solar point out that it is not sure whether
cells with efficiencies up to 11.7-% for platinum forms silicides with a:'SiC:H at
single junction devices and 13.3% for 200 °c and that the use of a thick p (a-
multijunction devices 111. Those high SiC: H) layer may screen the beneficial
efficiencies were mainly obtained through effects expected from the interfacial
an increase of the short-circuit current J se platinum layer. In our view, thin players
of the cells. An increase of their open have to be used if one wants to take
circuit voltage Vee is necessary to further benefit from the modified interface.
improve the cell efficiency. Now, it has In figure 1 we present the potential
been shown that for a thin p(a-Si:H) layer, profile at the Sn02/p (a-Si: H) interface
the open circuit voltage is largely limited determined from Kelvin probe measurements
by a downwards band bending, which develops and the change that one would obtain
at the tin-oxide/boron-doped a-Si:H through the introduction of a Cr or Pd
interface because of the low value of the interfacial layer, when assuming the
Sn02 work function 12,3,4/. As a following values of the work function for
consequence, a small reduction of the Sn02 the materials involved in the junction: 4>(p
substrate, usually thought to be a-Si:H) 4.85 /11/; 4>(Cr) 4.8 1121,
detrimental, increases the Sn02 work 4>(Pd) - 4.5 113/ and 4>(Sn02) - 3.7, deduced
function and helps to increase the from our Kelvin probe measurements. It is
efficiency of the solar cells 14/. In interesting to note that considering the
agreement with these experimental results, above values one would expect to obtain a
computer models on the effects of the high Voe by the use of a simple Schottky
height of the front contact barrier and of diode junction. In fact, the exper imental
the thickness of the p layer have shown values of Vee for Schottky diodes are of the
that the front bar·rier 4>be has to be order of 0.4 V, even though the Vee can be
increased (the downwards band-bending increased up to 0.7 V by the use of a Si02
reduced) to increase the efficiency of the interface /14/. Then, a thin p(a-Si:H)
solar cells 15,6/. layer is needed to pin the Fermi level
Thus our purpose was to improve the close to the a-Si:H valence band.
Sn02/p(a-Si:H) interface in order to Cr
increase the open circuit voltage without <I> - 4 .8 .V CB
any loss in J sc . To do so, we evaporated ---- -~---.~
thin metallic layers on the Sn02 substrate
prior to p-i-n solar cell deposition. We
have already shown 171 that good quality
Schottky barriers (barrier height - 1.1 eV
and diode quality factors 11 - 1) can be - 1 . 5 .V
obtained on Sn02/palladium/a-Si:H structures
when the palladium layer is only 20-A
thick. Moreover, the introduction of the
palladium layer did not result in any 0 .3 eV
appreciable decrease of the optical
transmission of the device in the 600-700
nm range, a fact which was explained by the = 0.7 eV
substitution of the metallic layer by a p(a-Si : B)
more transparent silicide compound. Indeed,
it is well established that a-Si:H reacts <I> = 3.7 <I> e 4 . 8 eV
with metals to form silicides 181, and
reports exist on the use of chromium Figure 1. Schematic diagram of the band
silicides as semitransparent electrodes structures at the SnOz/p(a-Si:H) interface
19/. However, the reported effects on and the expected reduction of band bending
ITO/Sn02/a-Pt-Silp(a-SiC:H)-i-n solar cells as a consequence of the introduction of a
/10/, i.e. the increase of the cell fill Cr interfacial layer.

335
 We here present a study on the effect 2 indicate that even when 20 A of metal are
of the thickness of both the metal (Cr, and used, the optical transmission of the
Pd) and the p (a-Si: H) layer on the solar devices is not reduced.
cell parameters.
90

••
Sn02/Cr/p-i-n
2. EXPERIMENTAL DETAILS 80 fl Sn02/Pd/p-i-n
0 Sn02/Pd
The samples used in this study Of' 0 0 Sn02/Cr
consisted of Sn02 substrates coated with 70 0
thin layers of palladium or chromium (10-60 s::
A) deposited by thermal evaporation. The 0 60
thickness of the evaporated metal was •.-1
III 0
monitored by a quartz crystal during the III 0
thermal evaporation of the metal. p-i-n •.-1 50
Ii! 0

• •
solar cells were deposited onto these
III
substrates in a hot-wall multiplasma- s:: 40
monochamber reactor described elsewhere

•
III
115/. The substrate temperature was set at 1-1

•
E-t 30
150°C and the total gas pressure at 40
mTorr for all the samples used in this
study. The nand p layers were respectively 20
grown from mixtures of 3 sccm of the dopant 0 10 20 5030 6040
gas (phosphine or trimethyl-boron, both
diluted at 2% in hydrogen in the gas Metal thickness
(A)
cylinder) in 12 sccm of silane at an rf Figure 2. The optical transmission of
power of 2 W. The i layer ( - 5000-A thick) Sn02/Metal substrates (open symbols) and of
was deposited from 15 sccm of pure silane Sn02/Metal/p-i-n devices (ful1 symbols) as
at 0.5 W. In some cases a mixture of a function of the thickness of the metal
methane in silane was used for the layer
deposition of the p(a-SiC:H) layer.
Aluminum dots were thermally evaporated to
provide ohmic contacts with the n layer. 3.2. Effects of the thickness of the p
Optical transmission measurements in layer
the 350 - 1100 nm range with the light In figure 3 we present the open circuit
incident on the glass side were performed voltage Vee as a function of the deposition
on the Sn02/metal substrates to determine time (rd S 1 A/s) for two series of p-i-n
the loss of the substrate optical solar cells deposited on Sn02 substrates
transmission due to the metallic layers as without any metallic layer. The dependence
well as on the Sn02/p-i-n devices in order of the Vee on the thickness of the player
to check if the formation of silicides is the one usually observed 141 and we
produces an increase of the device optical recall that the increase of Vee is
transmission. accompanied by a decrease in the short
circuit current of the solar cell. As Vee
3. RESULTS AND DISCUSSION saturates for tp 60 s, we chose a
deposition time of 40 s in the following
3.1 .. Effects of Pd and Cr on the optical studies of the effects of the thickness of
transmission the metal layer. Indeed, if the p layer is
Figure 2 shows the values of the so thick that the interface is screened out
optical transmission at 850 nm for both the (Vee independent of the thickness of the p
Sn02/Metal substrates (open symbols) and layer), then we do not expect to observe
the Sn02/Metal/p-i-n devices (full symbols) any effect of the modified interface on Vee.
used in this study. It can be seen that the
optical transmission of the substrates is
almost unreduced by the first 10 A of
800 ••
•
metal. However, for a 60-A thick metallic
layer, the optical transmission is reduced
from 83% to 55 % in the case of Cr and to
47% in the case of Pd. Taking into account
that the p-i-n itself transmits 44% of the 700
incident light, the whole structure should
transmit 24% for Cr and 22% for Pd at 850
nm. The measured values of the optical ~ u 600
transmission are 30% for Cr and 28% for Pd
(fig. 2). The relative increase of the >0
optical transmission is not significant
enough to justify a silicide formation. As
a matter of fact, it is well known that •
•
500
silicides do not form when impurities are
present at the interface. This is most
likely our case, with a thin metallic layer
evaporated on top of a rough Sn02 400~~--~----~--~--~~
substrate. We note that the annealing of
the Sn02/Metal/p-i-n devices did not
a 20 40 60
produce any change in their optical Deposition time (s)
transmission, which suggests that any
reaction between metal and a-Si:H did Figure 3. The open circuit voltage of
happen during the deposition of the p-i-n Sn02/-i-n solar cells as a function of the
solar cells. Moreover the results of figure deposition time of the p(a-Si:H) layer

336
3.3. Effects of the thickness of the Cr and Figure 5 shows the J se of the solar
Pd layers cells for which the Voe has been presented
In Figure 4 we present the Voe as a in figure 4. These results indicate that
function of the thickness of ~he metallic the introduction of a 10-A thick metal
layer for Cr and Pd. In both cases the layer does not produce any substantial
samples were annealed during 10 minutes at reduction of the short circuit current, in
150°C to improve the n/Al contact. We agreement with the optical transmission
observe the same behavior for both metals, measurements presented in figure 2. Then,
namely Voe increases by about 30 mV with from the values of the Voe' J sc , and FF
just a 10-A thick interfacial layer and presented above, it appears that the
saturates as the metal thickness is further introduction of a 10-A thick Pd layer
increased. From this we deduce that 10 A of slightly increases the Voe without any
metal are enough to modify the Sn02 work reduction of the other cell parameters.
function. However we note that 1) the Then 10 A of Pd is the optimal thickness of
increase of Voe is much smaller than what we the interfacial metallic layer to improve
expected from the schematic band diagram of the solar cell efficiency.
fig. 1 and 2) t!.voe is independent of the 14
metal we use. These results are similar to
those related to the open circuit voltage
of Schottky diodes /14/ and suggest that 13
• •
•
Cr
Pd

••
there is a pinning of the Fermi level which
does not allow a higher increase of Voe. 12
790 '"
a
.....
•
11
780 • !
• t:Jg 10
•
......
770 0 •
0 0 9
•
! u
760
0
8~~~~~~~~~~~-L~~~

o 10 20 30 40 50 60
~ Metal thickness (A)
750 4

740
• Figure 5. The short circuit current
deduced from the spectral response of the
p-i-n solar cells as a function of the
thickness of the metal layer
730 3.4. Use of a p(a-SiC:H) layer
0 10 20 30 40 50 60 In figure 6 we present the Voe of solar
Metal thickness (A) cells deposited with a p(a-SiC:H) window
layer. Filled dots refer to solar cells
Figure 4. The Voe of Sn02/p-i-n solar with a thin p (a-SiC: H) layer (td = 60 s +
cells as a function of the thickness of the 40 s buffer a-SiC:H) grown from a mixture
Cr(.) and Pd(O) interfacial layers of 60% methane in silane.Here the effects
of the Pd interfacial layer on the Voe are
When Cr was used to increase the Voe, comparable to those observed with a p(a-
and even when this layer was only 10-A Si:H) layer. However, when a thicker p(a-
thick, the fill factor FF of the solar SiC:H) layer (td = 80 s + 40 s buffer a-
cells was strongly reduced as indicated by SiC:H) grown from a mixture of 72% methane
the values of FF presented in table I. The in silane was used (open dots), then the
reduction of FF, directly related to a high interfacial metallic layer did not affect
series resistance of the devices, is the Voe at all.
attributed to the Sn02/Cr interface as this
effect was never observed in the case of
Pd. Moreover, it is well known from Kelvin
probe measurements /16/ that the nature of 840 0
0 o
the contacts between a-Si:H and Cr can be
completely modified if the air-exposed Cr
substrates are submitted to an argon plasma
before the deposition of a-Si:H. However,
the exposition of the Sn02/Cr substrates to ~ 800

an argon plasma before the depos it ion of

the p(a-Si:H) layer did not change at all "
~
•
the FF of the solar cells. 760

Metal ~A) 0 10 20 40 60
I 0
•
tp - 80 s + 40 s
tp ,. 60 s + 40 s
720L-~~~--~--~--~--L-~~~~
FF (Pd) 0.72 0.73 0.69 0.68 0.72 o 5 10 15 20
FF (Cr) 0.73 0.55 0.35 0.40 0.51 Palladium thickness (A)
Figure 6. The Voc of p-i-n solar cells
Table I. The fill factor FF of with a p (a-SiC:H) window layer as a
Sn02/Metal/p-i-n solar cells as a function function of the thickness of the Pd
of the thickness of the metal interfacial layer

337
 We interpret these results as a /8/ M.J. Thompson, N.M. Johnson, R.J.
consequence of the screening of the Nemanich and C.C. Tsai, Appl. Phys. Lett.,
interfacial layer by the thicker and 39, 274 (1981)
probably more defective p(a-SiC:H) layer. /9/ K. Seki, H. Yamamoto, A. Sasano, and T.
We note that in the case of the thick p(a- Tsukada, Appl. Phys. Lett. 44, 682 (1984)
SiC:H) layer we observed a small increase /10/ K. Hanaki, Z.Y. Xu, D. Kruangam, H.
of the J se of the solar cells, in agreement Okamoto and Y. Hamakawa, IEEE PVSC, Las
with a previous report /10/. Vegas (1985), p. 493
/11/ K. Winer and L. Ley, in Advances in
Disordered Semiconductors, Vol. 1,
4. SUMMARY AND CONCLUSIONS Amorphous Silicon and Related Materials,
edited by H. Fritzche, World Scientific
The results reported in this study have (1989), p. 365
shown that: /12/ A.J. Szadkowski, A. Kalnitsky, K.B.
1) The deposition of thin metallic layers Ma, and S. Zukotynski, J. Appl. Phys. 53,
between the Sn02 substrates and thin p(a- 557 (1982)
Si:H) or p(a-SiC:H) layers allows to /13/ CRC Handbook of Chemistry and Physics,
slightly increase the open circuit voltage 49th edition, edited by Robert C. Weast,
of the solar cells. The increase of Voe was Published by The Chemical Rubber CO.
found to be independent of the kind of /14/ A. Madan and M.P. Shaw, The Physics
metal and its thickness. and Applications of Amorphous
2) When the thickness of the metal layer is Semiconductors, Academic Press Inc. (1988),
less than 20 A, there is no reduction in p. 216
the short circuit current of the solar /15/ P. Roca i Cabarrocas, J.B. Chevrier,
cells. J. Huc, A. Lloret, J.Y. Parey, and J.P.M.
3) Even though Cr appeared as a better Schmitt, J. of Vacuum Science and Technol.
metal from the optical transmission point A, to be published
of view, the fill factor was strongly /16/ Measurements performed in our
reduced in solar cells with a chromium laboratory.
interfacial layer.
As a conclusion, thin interfacial Pd
layers allow to slightly increase the Voe
and the overall solar cell efficiency.
However, this effect is of the same order
of magnitude as the one obtained by just
submitting the Sn02 substrate to an argon
plasma prior to solar cell deposition. In
our view, further studies are necessary to
understand the mechanisms that control the
band bending at the Sn02/p(a-Si:H)
interface as well as the formation of
silicides in the Sn02/Metal/a-Si:H system.

ACKNOWLEDGEMENTS
This work is supported by PIRSEM/CNRS
and AFME contracts. We thank M.J. Surmont
and J.C. Dauny for their technical
assistance ..

REFERENCES

/1/ B.L. Stafford, W. Luft, B. von Roedern,

and W.L. Wallace, in Proc. of the 21st IEEE
Photovoltaic Specialists Conference,
Orlando, Florida, May 1990
/2/ F. Sanchez Sinencio and R. Williams, J.
Appl. Phys., 54, 2757 (1983)
/3/ K. Itoh, H. Matsumoto, T. Kobata, A.
Fujishima, o. Nakamura and K. Fukui, Appl.
Phys. Lett. 51, 1685 (1987)
/4/ P. Roca i Cabarrocas, in Amorphous
Silicon Technology, edited by A. Madan,
M. J. Thompson, P. C. Taylor, Y. Hamakawa,
and P.G. LeComber (Materials Research
Society, Pittsburg, PA, 1989), Vol. 149, p.
681
/5/ H. Tasaki, W. Y. Kim, M. Hallerdt, M.
Konagai, and K. Takahashi, J. Appl. Phys.
63, 550 (1988)
/6/ J .K. Arch, F .A. Rubinelli, J. Y. Hou,
and S. Fonash, Proc. of the 21st IEEE
Photovoltaic Specialists Conference,
Orlando, Florida, May 1990
/7/ P. Roca i Cabarrocas, S. Ramprashad ,
J.Z. Liu, V. Chu, A. Maruyama, and S.
Wagner, Proc. of the 21st IEEE Photovoltaic
Specialists Conference, Orlando, Florida,
May 1990

338
 IMPROVED THICKNESS UNIFORMITY AND OPTOELECTRONIC PROPERTIES OF a-Si:H THIN FILMS

H. Meiling, R.E.1. Schropp, W.G.J.H.M. van Sark, J. Bezemer, and W.F. van der Weg
Utrecht University, Debye Institute, Department of Atomic and Interface Physics,
P.O.Box 80.000, NL-3508 TA Utrecht, The Netherlands

ABSTRACT. We report on experiments with H2 dilution of Sill. during the growth

of a-Si:H. Both the optoelectronic properties and the lateral thickness uniformity of
10 cmxl0 cm glow-discharge a-Si:H films are investigated. A correlation is suggested be-
tween the enhancement of the uniformity and the decrease of the DC self bias voltage Vdc
upon dilution. With increasing H2 concentration, Vdc becomes more negative, thereby de-
creasing the lateral inhomogeneity of the electric field strength, and increasing the lateral
homogeneity of the amount of precursors near the substrate. This results in a thickness
uniformity within 1.5% standard deviation over the whole substrate area, obtained in a
20 cm diameter cylindrical reactor.
The optoelectronic properties of a-Si:H are strongly enhanced upon dilution of the source
gas. The photoresponse (fph/(fd increases from 105 when samples are deposited from
only SiH 4 to 107 at equal flow rates of SiH 4 and H2. The fraction of clustered hydrogen
atoms, as determined from infra-red absorption measurements, also strongly decreases
upon dilution. These changes in material properties are attributed to an increased surface
mobility (lower deposition rate) of the precursors, induced by a change in gas phase
reactions, and an increased Vdc'

1. INTRODUCTION fIl
fIl 1.8 - - typically only sm.
CD
Since high efficiency pt.i-n+ type solar cells and solar I:: ---- lluflOlf SiHu'He = 1/1
~
modules made of hydrogenated amorphous silicon, a-Si:H,
are increasingly being used for large area applications, the ....0 1.6 - - guflOlf Sm./Ha = 1/2

lateral thickness homogeneity over that area becomes essen- ......c:

tial. Especially in multijunction cells thickness uniformity "C 1.4
CD
over the entire panel is important, in order to preserve cur- .....
N
......
rent matching between the respective i-layers and to keep III 1.2
the light absorption in the tunnel junction laterally constant. E!
I-<
Little or no attention has been paid in literature to the lat- 0
eral thickness uniformity of the deposited films. We therefore Z 1
investigated the influence of various deposition parameters
[1). Results of experiments on uniformity with respect to H2 -5 0 5
dilution of S~H4 are presented in this paper, in Section 2. Distance from center (cm)
The influence of H2 dilution of SiH 4 on the structural,
optical and electrical properties of glow-discharge a-Si:H thin Figure 1: The normalized thickness of a-Si:H thin films, mea-
films has been a subject of investigation in several groups sured at different positions on a 10 cmx 10 em glass substmte, a8
[2-5] because it is known to improve material properties. In a function of gas composition.
Section 3 the influence of dilution on the structure and op-
toelectronic properties of a-Si:H will be discussed. In this figure the thickness of the films at different lo-
cations on the 10 cmx 10 cm glass substrates is plotted. All
thicknesses are normalized to the thickness in the center of
2. THICKNESS UNIFORMITY the films. The thickness is determined from optical reflection-
and transmission measurements with visible light. It is clearly
2.1 Introduction shown in this figure that the addition of H2 to SiH4 improves
In our multichamber system ASTER [6) we frequently uniformity: in the case of a gas flow ratio of SiH 4 to H2 of
observe large inhomogeneity in thickness when films are de- 0.5, the standard deviation in thickness is only 1.5%, while
posited with the use of SiH 4 only. In the 20 cm diameter in the case of deposition with only SiH 4 thickness differences
parallel plate deposition reactor, samples are mounted face between the center and the edge of the samples can amount
down on the 18 cm diameter grounded upper electrode. The to more than 80%.
lower, powered electrode has a diameter of 15 cm and is con- Thickness inhomogeneity must be due to lateral inho-
nected through a matching network to a 13.56 MHz power mogeneity in the creation of precursors that are responsible
generator (Alcatel ARF300). We usually add H2 to the sup- for the growth of a-Si:H. In order to find the origins for the
ply gas, which does not only improve film quality [5) but also changing lateral,distribution of precursors we have paid more
enhances uniformity, see Fig. 1. attention to the electric field strength distribution between

339
 ........
the glow-discharge electrodes. This was done specifically by >
looking in further detail at the way in which the RF-power .
........
~"'200 011'
is coupled to the plasma.
I
3011'
gj
2.2 Experimental procedure t>CI 2511'
In Fig. 2 a schematic representation is given of the match- ! 150
ing network and the capacitive coupling (through CD) of the -0
> 2011'
power supply Vi to the glow-discharge. Both tunable ca- III
1511'
a:I

-
pacitors CI and CL are used to influence the phase differ- :0 100
ence between the current and voltage in order to minimize
the reflected power, and to achieve perfect matching of the
glow-discharge impedance to the power supply. Cl and L2
Qj

U
III
•
-----.-
have relatively large values and serve for separation of AC 0 50
and DC voltages. The actual potential Vrf(t) on the powered 0 0.5 1
electrode is measured continuously with an oscilloscope. SiH,J(SiH4 +H2 )

Figure 3: Vele. the DC self bias voltage, as a function of the gas

composition for several input powers. Vde is defined as the DC
voltage of the powered electrode with respect to ground.

Vi ~---,--l I '----,--,-- , -¢ Vrf From this figure it can be seen that the DC self bias volt-
(scope) age decreases as the partial pressure of SiH. decreases (lower
SiH. concentration), at a constant input power. The addi-
load tion of H2 in the glow-discharge therefore results in a more
negative Y.!c and thus a higher Vrf , since these potentials are
proportional. This effect is more pronounced at higher input
powers than at lower input powers. The differences in Y.!c at
- - - - - - - - - _ _~_ _ _ _ J constant Pi amount to 30% when changing the gas mixture
from 90% SiH 4 to 90% H2.
Figure 2: Schematic representation of the coupling of the RF input
voltage VI to the glow-discharge system ('discharge') through a 2.4 Discussion
matching network. The input- and the output impedance of the The decrease of Vdc (towards a more negative value) with
network can be influenced with the tunable air capacitors CI and increasing H2 fraction implies a higher RF-amplitude Vrf and
CL of 1000 pF and 500 pF respectively. Cl is a blocking capacitor, an increase of the impedance of the plasma, see Fig. 2. This
L2 is a radio-fr«luency choke. decrease of Vdc was also reported by Kampas and Kushner
[10]. They attributed it to a decrease of the amount of neg-
When a periodic potential Vrf(t) =
Vrf sinwt is applied ative ions when the SiH 4-fraction decreases.
to a glow-discharge system with differently sized electrodes The fact that a more negative Y.!c can be responsible for
a DC self bias voltage will develop between the electrodes. a better uniformity of the films can be explained as follows,
This bias voltage, which is measured at the position labeled see Fig. 4.

1 f:
Vde (see Fig. 2), is negative with respect to ground, when the
powered electrode is smaller than the grounded electrode. Si-
multaneous measurement of Vde and Vrf shows that there is a
,.mple hold«
linear relation between these parameters. The proportional-
ity factor appears to be independent of the gas composition. substrate SSSSSSSSSl
Only the area ratio of both electrodes determines the slope c::::==> c:>
of the linear relation, in agreement with the model of Mei- B A
jer and Goedheer [7]. The influence of this bias voltage on
the material properties of the a-Si:H films has already been
studied by Aozasa et al. [8]. An important point is that the
potential difference between the substrate electrode and the
plasma, which is related to Vde , determines the energy of ions
which impinge on the surface of the growing film. Although
the density of the ions in the plasma is relatively low with Figure ~: Schematic representation of the upper electrode, with a
respect to neutral radicals [9], the' flux and energy of ions ar- sample mounted face down. A and B are used to denote different
riving at the surface have a great impact on growth kinetics. positions near the substmte. E,o. denotes the direction of the
electric field in the positive part of the RF voltage cycle.
2.3 Results of H2 dilution
In order to find out why H2 dilution of SiH 4 enhances In this figure a schematic representation is given of the
uniformity we used different gas flow ratios ranging from 10 electrode configuration in one of the deposition chambers
to 90% SiH 4 • The total gas flow amounted to 60 standard of ASTER. The sample is mounted to the upper electrode
cm3 /minute (sccm), the process pressure was 0.12 mbar, and and is held in place by metal clamps. The sample holder is
the interelectrode distance was 3.6 cm. The input power Pi grounded. Also the direction of the electric field during the
was varied from 10 to 40 W. Results of measurements of "de positive part of the RF-voltage cycle, B_, is indicated.
under these deposition conditions are given in Fig. 3. The When an insulating substrate like Corning glass is used,
parameter in this diagram is the input power. the electric field strength resulting from the DC self bias volt-
age in the center of the sample (position B) is lower compared

340
to the electric field strength at the edge of the sample, posi- 0.5 lOB
tion A. The reason is that at A there is a grounded surface,
while at B this is not the case. This means there is a lat- ..-.0 0.4
eral inhomogeneity in the DC electric field strength, and in 0
0 107
the production of ions and radicals. Therefore at position A ....s>
more growth precursors are created in the plasma than at B, +0 0.3 Q

resulting in a higher deposition rate at the edge of the sam-

r-
0 • lOB
"C
IT
'-.
....s>
........
ples than in the center of the samples. When H2 is added to
.........
0.2 Q
P-
the SiH 4-gas the self bias decreases (becomes more negative) 0
r-
by upto 30%, thereby decreasing the relative influence of the 0
.....
N 0.1 105
insulating substrate on the lateral homogeneity of the elec-
tric field strength. We suggest that this is the main reason
for the addition of H2 to result in more uniform deposition. 0 104-
With the use of a conducting substrate, such as glass 0 0.5 1
with a transparant conductive oxide on top of it, the unifor- SiH4/ (SiH4- + H2 )
mity is seen to be much better, as can be expected from the
better grounding of the substrate and therefore the better Figure 5: Material properties of a-Si:H deposited from different
lateral homogeneity of the electric field in this case. gas compositions. The left hand vertical axis denotes the amount
The improved uniformity is not attributed to changes of microstructure, obtained from IR-measurements. The right
in the gas phase chemistry upon H2 dilution, since these hand side vertical axis denotes the photoconductivity ratio r. The
changes will occur proportionally over the whole substrate solid lines are guides to the eye.
area. These changes do not alter lateral inhomogeneity in
the deposition rate. Changes in surface kinetics and plasma considerable fraction of the H-atoms is bound in the clustered
chemistry upon dilution will briefly be discussed in Section 3, phase.
in connection with results of measurements of the optoelec- In Fig. 5 also the photoresponse r, the ratio between
tronic properties of the films. the photoconductivity O"ph and the dark conductivity Ud, is
plotted as a function of the gas composition. Again it can
be seen that a small addition of H2 induces large changes in
3. STRUCTURE AND OPTOELECTRONIC PROPERTIES the value of this ratio. An increase of almost two orders of
magnitude in the photoresponse is obtained: from r = 105
when only SiH~ is used to r :::::J 107 when more than 30% H2
3.1 Sample preparation is added.
We will discuss some results of depositions with different
concentrations of H2 in the gas phase. The gas phase ratio 3.3 Discussion
SiH 4 /(SiH 4 + H2) was varied in the range of 0.3-1.0. For In earlier work we already reported on changes in material
all samples the input power was kept constant at 20 W, the properties due to H2 dilution [1,5,13). From the results pre-
process pressure amounted to 0.20 mbar, and the substrate sented in this paper we find that, in addition, the optoelec-
temperature was 200°C. trical quality of the material is improved, as can be deduced
from the change in the photoresponse r: the higher r, the
3.2 Results fewer electronic defects are present in the material.
From infra-red absorption measurements we obtain in- We suggest the following explanation for the observed
formation about the binding configuration of hydrogen atoms phenomena. Probably again the change in DC self bias volt-
to silicon atoms. The stretching mode in the IR-spectra, ab- age Vde upon dilution, mentioned before in Sect. 2, is respon-
sorption in the range of 2000 to 2100 cm-\ reveals informa- sible for the differences in material properties. A stronger
tion about the fraction of hydrogen that is not bound as iso- negative Vde implies that ions approaching the growth sur-
lated Si-H. A measure for this is the so called amount of mi- face obtain a higher energy, which could enhance structural
crostructure R, which is defined as the intensity of the clus- relaxation during growth if predominantly weak Si-Si bonds
tered H peak (absorption in the range of 2070-2100 cm- l ) are broken. In addition, the surface mobility of growth pre-
relative to the total absorption of the stretching mode: cursors may be enhanced by the increased energy flux to-
wards the sudace. This mobility increase may not only be
caused by the change in Vde, but can also be related to a
(1 ) change in the relative distribution of different types of pre-
cursors. The deposition rate decreases more than a factor
of two upon dilution. This suggests that the etching of the
sudace becomes more important when SiH~ is diluted with
Here /2000 denotes the integrated absorption band due to
isolated Si-H bonds at 2000 em-I. The integrated absorption H 2•
It is also seen that gas phase polymerization of Si-H2-
band at 2070 cm- 1 is indicated by /2070' This band may
chains, or dust formation, vanishes when 20% or more H2
include absorption by Si-H2 or Si-H3 bonds, (Si-H 2)n-chains,
is added to SiH 4 • Polymerization does appear to have a
and clustered H atoms on the surface of internal voids [11,12).
large influence on material properties, because most of the
In Fig. 5 R is plotted against the fraction of SiH 4 in
changes in material properties occur upon only moderate di-
the gas phase. It is clear that R decreases, from 0.35 to
lution. Addition of more H2 above 30% does not seem to
approximately 0.05, when H2 is added to SiH~ upto 30%.
influence material properties any further.
Beyond 30% H2 , R is independent of the amount of dilution.
When 30% or more H2 is added, the value of R is 0.05,
which means that the major part of the H-atoms is bound
as monohydrides, Si-H bonds. When no dilution is used a

341
4. CONCLUSIONS REFERENCES

In this study the influence of H2 dilution of SiH. on both [1) R. E. 1. Schropp, H. Meiling, W. G. J. H. M. van Sark,
optoelectronic properties and lateral thickness uniformity of J. Bezemer, and W. F. van der Weg, in Proceedings of
glow-discharge a-Si:H thin films is investigated. We suggest the Fifth International Photovoltaic Science and Engi-
a correlation between the enhancement of uniformity and the neering Conference, Kyoto, Japan, 1990, p. 379.
decrease (more negative) ofthe DC self bias voltage l'dc upon [2J P. Chaudhuri, S. Ray, and A. K. Barua, Thin Solid
dilution. When more H2 is added, l'dc becomes more nega- Films 113, 261 (1984).
tive. This decrease in l'dc is suggested to improve the lateral [3J P. E. Vanier, F. J. Kampas, R. R. Corderman, and
uniformity of the electric field, thereby decreasing the lateral G. Rajeswaran, J. Appl. Phys. 56, 1812 (1984).
differences in the creation of precursors near the substrate,
which results in more uniform films. [4J S. M. Lee, S. J. Jones, Y.-M. Li, W. A. Turner, and
The optoelectronic properties of a-Si:H improve upon W. Paul, Philos. Mag. B 60, 547 (1989).
H2 dilution. The improvement in film quality, both struc- [5) H. Meiling, W. Lenting, J. Bezemer, and W. F. van der
turally and electrically, is ascribed to enhanced surface mo- Weg, Philos. Mag. B 62, 19 (1990).
bility of precursors, due to increased ion bombardment or [6) C. A. M. Stap, H. Meiling, G. Landweer, J. Bezemer,
due to different gas phase reactions, leading to a lower de- and W. F. van der Weg, in Proceedings of the Ninth
position rate. E. C. Photo voltaic Solar Energy Conference, Freiburg,
In summary, it appears that a small addition of only F.R.G., 1989, edited by W. Palz, G. T. Wrixon, and
about 30% H2 to SiH. is enough to establish improvements in P. Helm (Kluwer Academic, Dordrecht, the Nether-
material properties, and to obtain 'device quality' material. lands, 1989), p. 74.
Further dilution does not enhance film quality, but it does
enhance film thickness uniformity, up to only 1.5% standard [7) P. M. Meijer and W. J. Goedheer, IEEE Trans. Plasma
deviation in thickness over a substrate area of 10 cmx 10 cm. Sci. (in press).
By choosing the proper deposition conditions, 'device qual- [8J M. Aozasa, R. G. Pyon, and K. Ando, Thin Solid Films
ity' a-Si:H can be obtained with a high lateral uniformity. 136, 263 (1986).
[9) K. Tanaka and A. Matsuda, Mater. Sci. Rep. 2, 139
(1987).
5. ACKNOWLEDGEMENT
[10] F. J. Kampas and M. J. Kushner, IEEE Trans. Plasma
This work was financially supported by the Netherlands Sci. PS-14, 173 (1986).
Technology Foundation (STW), and the Netherlands Agency [11J M. H. Brodsky, M. Cardona, and J. J. Cuomo, Phys.
for Energy and Environment (NOVEM). Rev. B 16, 3556 (1977).
[12] H. Wagner and W. Beyer, Solid State Commun. 48, 585
(1983).
[13J H. Meiling, M. J. van den Boogaard, R. E. 1. Schropp,
J. Bezemer, and W. F. van der Weg, MRS Symp. Proc.
192, 645 (1990).

342
 LOW TEMPERATURE SHORT TIME RESPONSE OF LIGHT INDUCED DEFECTS IN a-Sl:&

Jorge Kasaneva Reinoso. Jordi Andreu Batall6

Dept. of Physics. Dept. of Applied Physics
University of Antofagasta University of Barcelona
CasUla 170 Diagonal 645
Antofagasta Barcelona 08028
Chile. Spain
Telephone 227784 Fax 247786 Telephone 3307311 Fax 4110873

ABSTRACT. We have carried out both theoretical and experimental studies of the dinamics of
creation of light induced defects In a-Si:B, In a low temperature regime. We have
comprobated that for temperatures lower than -100 ·C, all the thin films, independent of
both dark activation energy E. and r, the power-index that relates photoconductivity with
flux of photons, are characterized by a thermal recovery nearly during the first 4 minutes
of 1/2 AHI light exposure. This increase in photoconductivity is linear with time
exposure, and the slope of this straightline is inversely temperature dependent. At the
end of nearly 10 minutes of illumination time, the photoconductivity decay due to light
exposure.

IHTRODOCTIOI.- In 1977 Staebler and Wronski (1) The conditions for light soaking and photo-
first-reported on light induced dark and photocon- conductivity measurements were as follows: the
ductivity changes in a-Sl:B, which were found to temperature range during light exposure was 173 to
be reversible by annealing at approximately 150 103 K, which was exactly controlled with a mean
.C. These changes: were found to be caused by the error of ~ 1% by a PID programmable temperature
creation of metastable defects within the bulk of controller. The light-soaking and photoconductivi-
the amorphous semiconductor. ty measurements were carried out in a vacuum of
Since the first report of these metastable 10- 0 Pa. The time of exposure was up to 3.5 hours
changes, numerous studies have been carried out to of white light of 50 mW/cm a •
understand both the origins and their effect on An IR filter between the halogen lamp and
material properties ,a, (a, '4'. the sample blocked the heat radiation from the
Experimental result. have shown that the lamp in order to ensure a constant temperature at
metastable defect. are, in fact, produced by elec- the sample. The white-light resistance signal, the
tron hole recombination ,e), although hole trapp- time of exposure and temperature were stored in a
ing alone may be sufficient to create defects with file produced by an automatic system of resistance
a much lower efficiency ,e). Low temperature de- versus temperature measurements '"0). After each
gradation studies of a-Si:B thin films are limited degradation experiment the, samples were automati-
to temperatures nearly -100 ·C ,7, ,a) because the cally heat treated for 1 h at 180 ·C in order to
validity of proposed models. Hany few studies are fully recover its originals conductivity and pho-
reported about low temperature degradation under - tocondUctivity.
100 ·C by both short and long time light exposure.
In this paper we analize the increase of RESULtS AND DISCOSSION.- Figures 1 and 2 show the
photoconductivity as a function of time due to sh- logarithmic representation of photoconductivity as
ort time illumination during the first 4 minutes a function of time of illumination with the tempe-
IR filtered 1/2 AMl white light exposure, and we rature as a parameter along all the time of the
develop a rate equation of light induced recovery experiment for sample 1 and 2 respectively. Figu-
of photoconductivity that is a function of time res 3 and 4 show the linear representation of
exposure and the temperature of the thin films. normalized photoconductivity as a function of time
during the first six minutes of illumination with
EXPERIMENTAL DETAILS.- The samples used in these the temperature as a parameter. We observe a li-
experiments are 80 intrinsic a-Si:B thin films near increase of normalized photoconductivity up
deposited on glass (Corning 7059) substrates in a to four minutes of light exposure and the slopes
capacitively supported radio frecuency glow-dis- of these straightlines changes inversely with
charge system from undiluted silane gas ,e,. As- temperature. This case may be considerer as a
deposited properties, thicknesses, electrical, light induced recovery of intrinsec defects in-
photoelectrical and degradation parameters are duced by recombination of carriers, phenomenon
given in '"0'. For the conductivity measurements that is prevalent during the first minutes of
nicrom electrodes with a spacing of 0.74 mm in a illumination.
coplanar structure were evaporated onto the film. In this regime, most of the recombination
inmediately after deposition in order to minimize traffic still takes place at the continous distri-
surface contamination. Table 1 summarizes the more bution of localized states in the movility gap,
important features of two representative samples and we may consider that in this short time the
used in this experiment: substrate temperature T. energy released by recombination in light created
during deposition, thickness a of the a-Si:B film, defects produce a decrease of total recombination
conductivity activation energy E. in the annealed processes which finally produce the recovery of
state A and the light-soaked state B, the dark the intrinsic mid-gap defects. We assume that the
conductivity ~RT at room temperature in the an- non-radiative energy released by recombination is
nealed state A and the light-soaked state B distributed more efficiently in the highly local-
and the power-index r at room temperature inthe ized states near the midgap. Although it may be
annealed state A and the light-soaked state B. argued that the direct e-h recombination releases

343
larger energy per unit event comparing with to From where D = K3-n(t)"-t
recombination through detects an thus more likely ReplaCing 5 in 4 we have
to break a bond, yet it one considers the dissipa-
ted energy "density, the situation is just oppos- r- Ka- n K3- n (t)a_t
O-

site. Since the recombination centers are more G

localized states (3·5 A), comparing to the band
tail states ot electron and hole which is larger and trom (1) no B Gr
than 10 A, the released energy density due to thus
recombination through defect is therefore more ,ACfph,(t:.)
B 6
than 5 to 10 times of that due to direct electron ~Ph(O)
hole recombination '1'). In a low temperature
regime the intrinsic defects remain "frozen", that may be written
allowing its recovery efficiently because the
surrounding lattice is not enough relaxed due to ll~"h''') fph''') - fph(O) II rot
low temperature. ~Ph(D) fph(D)
All the experiments of photoconductivity
versus flux of photons have showed that the power and thus
index r decrease when temperature is lowered, and
that r increases according to increases light- II 1 + rot 7
exposure of the a-Si:H, doing it each more time
near to monomolecular. Thus the recombination
process is dominated by deep localized centers. where temperature dependence of r is To/(To+T),
The short-time regime will end when the and To is a characteristic temperature that deter-
recovery rate of intrinsic or as deposited defects mines the shape of the distribution of states
recovered by recombination R of free carriers denSity in the film 'la). Finally,
equals light creation of defects also created by
recombination. For simplicity, we assume that the 1 +
free carriers concentrations are n(t)=p(t) and at
zero time
Where we observe the expected variations of nor-
malized photoconductiVity with temperature and
time of illumination.
Where G is the illuminating flux, R is the total
recombination, r is the power-law dependence of REI'IREIICES.
photoconductivity, no is n(t=O), and K1 is a sui- (1)
table constant. At time t, n(t)=no(t)+ An(t) and Staebler, D.L., Wronski, C.R. Appl. Phys.
the recombination rate at the light-induced de- Lett. 31, 292 (1977).
,a)
fects is equal to Ka-n(t)-D, where D is the den- Stutz.ann, M., Jackson, W.B., Tsai, J.J.
sity of light induced defects; hence, ~~!s. Rev. B 32, 23, (1985) and 34, 63, (1986).
- Guha, S., Huang, Y., Hudgens, S., Phya. Rev.
R = G 2 B 29, 5995, (1984).
,.) - Dersch, H., Stuke, J., Beichler, J., Appl.
If we replace nIt) in 2, Phys. Lett. 238, 456, (1980).
,a) _ Guha, S., Yang, J., Czubatyj, S., Hudgens,
G = Kdno+An(t»'/r - K2 -(n o +An(t»-D ~;; Hack, M., Appl. Phys. Lett. 42, 589, (1993).
- Nakamura, N., Watanabe, M., Nishikuni, M.,
G= Kl-n01/r-(1+dn(t)/no)1/r - K3 -n o -(1+An(t)/n o )-D Hishikawa, S., Tsuda, H., Nishiwaki, M., Ohnishi,
M., Kuwano, Y., J. Non-Cryst. Solids 59-60, 1139,
For n(t)/no«l; (l+An(t)/n o ),/r-1+(l/r)- An(t)/n o (1983).
::: - Eser, E. J. Appl. Phys. 59 (10), (1996).(2)
G - Kl-nol/r- (1 + (1/r)-An(t)/no ) - K3-nO-D - Hack, M., Guha, S., den Boer W., Phys. Rev B
33, 4, 2512 (1985).
G An(t) ,.) AndOjar, J.L., Canillas, A., BertrAn, E.,
G - G + -,.----n-o- - Ka-no- D 3 ~~~~nza, J.L. Vacuum Vol. 39 N 7/9, 795, (1989).
and - Rasaneva, J. Doctoral Thesis. Laboratory of
Physics of Thin Films, Dept. Applied PhYSics and
-
An(t) A~"h''') r- K3- D- nO 4
no ~.h(O) G Electronics, Faculty of Physics, University ot
For the creation rate ot defects we have trom
,.) Barcelona (1990).
'1') _ Wen-jer Tzeng., Hsiung-Kuang Tsai., Si-Chen
Lee., AlP Conference Proceeding 157 on "Stability
dD of Amorphous Silicon Alloys Materials and Devices-
5 ". Palo Alto Ca., 62, (1987).
dt ,la) _ Rose, A., "Conceps in Photoconductivity and
Allied Problems", Robert E. Krieger Publ. Co.Hun-
tington, N. Y., (1978).

344
Table 1.- Properties of representative saaples.
"
Saap. T.
nu •• ['C]
I
[~a]
E.A
reV]
E•
leV]
. _RT A
[SIca)
-RT-
. [S/ca)
rA ra

300 0. 33 0.58 1.16 1.2·10'6 I.B·I0-9 0.63 1.04

300 3.3 0.77 0.99 1.7·W· 4.7·W· o 1.031.04 ~J
t
:>
o
z
o
u * *
~2
*
~
a. * *
1( * )( )(

:i * *x )(

!
A A 6A 6
cr
~!!! *
!!!~~
o )(
+
z 1 + + + +
"" "
"
M

..................
~
g
• •
•• •
• • • .. * * •••
• * .... 't1t'WJt~
. ..
z *
8 ..""""-....." o + 30C'" TIt.4E (min)
~ 10-' II
o
:r
"
o -
- 20
50
C
C FIG. 3.- RECOVERY OF NOR..cAlIZED PHOTOCONDUCTMTY
a. *+ - - 80C
100 C
UNDER 1/2 AMI OF" LIGHT EXPOSURE AT VARIOUS
TEMPERA1uRES (SAMPLE 1).
M - 130 C
• - 1<10 C
• - 150 C
.. - 170 C "
10··'L-~~~~-L~~~~-L~~~~~WW

10·' 10 10' 10'

TIME (min)
FIG. 1.- PHOTOCONDUCTIVITY AS A FUNCTION OF
TIME OF" LICHT EXPOSURE WITH TEMPERATURE
AS A PARAMETER (SAMPLE 2).

Ii Ii Ii II iii
* * *
o 0 00 0
*
'x" 'x"
-¥'"
*.. *... ** *
++ +
* x
+ + +

~ "
x

.. .. """""
: ..*, • .....
)(:IIt

•••
:It

..........
JIi X X IIC:lCx X

"
.
:>
o ~
z
o
.. o + 30 C

o~ 10·'
,,- 0 C TIME (min)
o - 30 C
:r
a. *+ - - 100 70 C
C
FIG. 4.- RECOVERY OF NORMAliZED PHOTOCONDUCTIVITY
UNDER 1/2 AMI OF LICHT EXPOSURE AT VARIOUS
TEMPERA1uRES (SAMPLE 2).
" - 130 C
• - 140 C
• - 150 C
II - 170 C

10 10 • 10 •
TIME (min)

FIG. 2.- PHOTOCONDUCTIVITY AS A FUNCTION OF

TIME OF" LIGHT EXPOSURE WITH TEMPERATURE
AS A PARAMETER (SAMPLE 2).

345
lam EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

INVESTIGATION OF SPECTRAL MISMATCH IN

a-SiGe:H BASED TANDEM SOLAR CELLS BY OPTICAL MODELLING

M. Gorn and C. Curran

Phototronics Solartechnik GmbH, Hermann-Oberth-Str. 9, D-80ll Putzbrunn, Germany
A. Antoine-Labouret and J. P. M. Schmitt
Solems, 3 rue Leon-Blum, Z.I. Les Glaises, F-9ll24 Palaiseau Cedex, France

ABSTRACT: The issue of spectral mi smatch in two-termi na 1 stacked or tandem

solar cells is often presumed to be a drawback with respect to single bandgap
devices. A detailed analysis shows however, that at least for devices made of
amorphous materials, this effect is marginal, during most of the time during
the day, when mos t of the e1ectri c power is genera ted. The essence of thi s
work is the use of complete optical data of all individual layers in a-SiGe:H
based tandem solar cells.
1. INTRODUCTION
Every solar cell must be optimized with re-
spect to a certain solar spectrum, and devia- 4.5
tions from this spectrum should not cause an
overproportional reduction of power output. This 1.5
eventua 1 defi ci ency is called "spectral mi s- fT
match". Tandem cells are generally presumed to ";'
have a severe spectral mismatch problem with re- E0
spect to single cells as the results of calcula- 4.0
tions by Fan and Palm III and Chambouleyron 121 >
might suggest. They arrived at the conclusion n 1.0 ~
..,
that the benef\t of multiple-gap solar cells op- 0
timized for one spectrum is el iminated by the .....
current mismatch produced by different solar
~
spectra duri ng the day or throughout the year. 3.5
However, the drastically simplifying model they
used for the opti ca 1 properti es of the PV ma- 0.5
terials was surely pushing their conclusion to
the pessimistic side.
Here, we report on the first calculation of
spectra 1 current mi smatch effects on the basi s 3.0
of measured optical data in amorphous materials.
Stacked pi npi n a-Si:H based devi ces are of
interest due to thei r improved stabil ity under 2 3
light-soaking. Tandem cells using a-SiGe:H al- E leV]
loys with bandgaps of 1.6 to 1.5 eV, are repor-
ted to yield 13t efficiency 13/. If material
with a bandgap of 1 eV was available (e.g. Fig. 1: Optical data (solid lines) and
a-Ge:H), AM1.5G efficiencies of l8t could be (broken lines) for a-Si:H (curves no. 1) and
achieved, as results also from our calculations. a-SiGelH of bandgap 1.66, 1.56, 1.42, and 1.23
Using a straightforward optical algorithm eV (curves no. 2,3,4 and 5, respectively)
and a fi 11 factor model, the power output for
typical tandem or triple devices under different
spectra during the day is calculated. Gl ass (1 mm thi ck, n - 1.5) and TCO (200
nm, n a 1.9) are assumed to be non-absorbi ng.
2. OPTICAL DATA p+-layer data are taken from a typical film
a-SiC:H:B with 1.95 eV bandgap (thickness of 7
Opti ca 1 absorpti on, I)( , and refracti ve i n- nm kept constant here). No effects of light-
dex, 1\.. , have been determi ned in the wavelength trapping are included. Optical data for the
region of 400 to 1.100 nm for the layers backcontact (AI) are taken from a handbook 171.
p+-a-SiC:H:B (Eg - 1.95 eV), a-Si:H (1.78 eV), A more refined model would require measured pro-
a-SiGe:H (1.66, 1.56, 1.42, and 1.23 eV), by perti es of the backcontact and opti ca 1 trappi ng
combined transmission and ellipsometric measure- at all interfaces, as descri bed in the work by
ments. The latter are important for higher pho- J. Horris et al. 18/.
ton energies and lower bandgaps whenever the
one-oscillator model (Drude formula, see the 3. CALCULATIONS
standard literature 14/) is not valid. This can
be seen in fig. I, where for low photon energies A complete formalism accounting for all
a parabolic behavi our for n is observed. Except multiple specular reflexions and interferences
for a-Si:H, the n (E) curves for a-SiGe:H always in the multil ayer systems pi n, pi npi n, tandem
exhibit a maximum, the so-called resonance ener- and tri pIe cells on gl as slTCO with Al-backcon-
gy, see e.g. ref. 4. In this region and for tact, is used, based on the matrix method des-
hi gher photon energi es, the QC, and"", data are cribed in the literature 19/. In this way, the
taken from ellipsometry 15,6/. In fig. 1 the opti ca 1 absorpti on in the i-I ayers is obtained
Tauc-plots for ot are also shown. whi ch here is regarded to be the quantum

346
efficiency (certainly true for a good device or 4. INFWENCE OF VARYING SPECTRA
under reverse bias). Currents for a certain
spectrum refer to i ntegrati on from 400 to 1.100 Solar spectra vary due to various Influen-
nm. As seen in table I, single pin cells with ces in the atmosphere leading to seasonal, dai-
I-layer thicknesses 250, 370, and 470 nm are I y, wea ther- or po 11 uti on- induced changes. Dra-
used as a reference, yielding under AMl.5G (re- stic changes In the blue-to-red ratio occur In
commended by SERI) the same power as the three the course of the day at di fferent AM va lues,
current-matched s tacked pi npl n ce 11 sin the se- and the I nvestl gatlon of spectral ml smatch ef-
cond section of table I, respectively (assuming fects In tandem cells Is best done with these.
doubl e open-cl rcul t-vol tage and same fill fac- He used a representative set of measured
tor, as almost usually observed experimentally). daily spectra taken from the work by GI atfelter
Note that the current densities are slightly et a 1. 11 0/, plotted ,I n fl g. 2. The conti nuatl on
sma 11 er than ex perl menta lly observed due to the of these spectra from 1. 000 to 1.100 nm is a
absence of optical trapping In these calcula- guess but not unrealistic, If compared with me-
tions.
Hhlle for the stacked cells bottom layer Z.B.-_____~ln~t~en~s~l~t~y~[~W_a_-~Z~n._-~l~]____________;
thicknesses were chosen as 200, 300, and 400nm,
we restrl cted ourselves to 200 nm thl ck bottom AM! . l
i-layers In all cells containing SIGe, In order
to remain in a practical range for reasonable
photostability (results for larger thicknesses
turn out to be very sl mil ar as far as spectral
mismatch Is concerned).
The top cell I-layer thicknesses, in triple
cells also the middle cell, were adjusted such
that the currents for the AM 1. 5G spectrum (re-
commended by SERI) were equal to the bottom cell
current ins i de the devi ce. Thi s is ca 11 ed the
current-matched condition in table I. Under a
different spectrum, the currents differ from
each other. The resulting overall current in all
the two-terminal devices of table I is then al-
ways taken as the lowest current of all i-layers
in the devl ce. Thi sis the wors t case and can
only improve, Series connection effects are ad-
dressed further below in section 5. Wavelength [naJ
F1 g. 2: Spectra occurl ng in the course of the
day used for determl nation of mismatch effects
In multiple bandgap solar cells

current-matched unmatched condition

condition
j (Spectrum)1 j (AM1.5G) [1]
j [rnA/cm2]
thicknesses gap gap at AMI. I AM2.0 AM3.6 AM5.0 AM9.2
Il i2 13 E2 E3 AMI.5G (SERI) (13.30h) (08.59h) (07.45h) (07.08h) (06 . 45h)
[nm] [nm] [nm] [eV] reV] • 1001 • 1141 • 931 • 771 • 651 • 441

single 250 --- --- ---- ---- 11.0 111.2 84.6 62.1 46.5 22.0
pin
370
470
--- ---
--- ---
---- ----
---- ----
II. 7
12.3
111. 6
111.2
85.2
85.1
62.9
63.1
47.4
47.8
22.9
23.3

stacked 66 200 --- 1.78 ---- 5.5 109.4 84.9 59.9 43.3 18.9
76 300 --- I. 78 ---- 5.9 109.2 84.8 59.9 43.4 19.0
plnpln 84 400 --- 1.78 ---- 6.1 108.8 84.8 60.0 43.5 19. I
,._ - -

tandem
85
III
200
200
---
---
1.66 ----
1.56 ----
6.4
7.2
109.7
108.8
83.3
83.4
59.0
59.3
42.8
43.1
18.8
19.1
190
466
200
200
---
---
1.41 ----
1.22 ----
8.8
II. I
108.9
109.9
83.7
84.2
60.2
61. 7
44.2
46.2
20.0
21.9

triple 45 130 200 1. 78 1.66 4.3 109 . 4 83.0 57.8 41.4 17.6
55 210 200 1.78 1.56 4.8 108.7 83.1 58.0 41.7 17.8
I 78 550 200 1.78 1.41 4.9 109.5 83.3 58.6 42.3 18.3
.. -

triple 130 1120 200 1.66 1.22 7. 5 109.2 83.8 59.6 43.4 19.2
130 490 200 1.56 1.22 7.5 108 . 8 83.9 59.7 43.5 19.2
II 127 172 200 1.41 1.22 7.5 109.2 83.2 59.2 43.1 19.0

Table I: Devices of structure glass/TCO/piln(pi2n(pi~n»/Al with top iI-layer a-Si:H (Egap •

1.78 eV) and various bandgaps for subsequent cells. Thlcknesses such that currents match at AM1.5G
spectrum. For different spectra duri ng the day, normal i zed current outputs (with respect to matchi ng
condition, the current assumed to be fully limited by the restricting diode) are given.

347
teorological data. The current outputs of single As shown in fig. 3, the differences in nor-
pin, stacked pinpin, tandem, and triple devices malized current output between single pin and
under the different spectra are given in table I multiple cells almost disappear very quickly in
above. Sizeable differences between single and time, falling below 5"1. at AM 3. 6 (in our example
multiple cells only occur at extreme air mass 7:45 h).
values, e.g. AM 9.2. This corresponds in our ex- For the rest of the day, the di fferences
ample, after ref . 10, to an early hour in the are below the 3"1. level also at noon time (AM 1.1
morning, 06:45 h. There, the mismatch effects spectrum), see table· I or fig. 3. This means
can reach 20"1. (less current than single pin), that spectral mi smatch effects are unimportant,
although not all multiple cells perform equally for the multiple bandgap devices considered
bad, as seen in table I. The differences are not here, during most of the day when the energy ge-
very large. There is a small trend: Hith i n- neration is highest.
creasing overall device thickness performances Finally we address the issue why in these
at large air mass improve (only seen for single calculations spectral mismatch effects turn out
but not for multiple bandgap cells in table I). to be small when amorphous materi aI s are used.
This may explain the small trend seen in mUl- In refs. 1 and 2, where large effects are obtai-
tiple cells, that, for a given device structure, ned, the absorpti on edge cons is ts of an abrupt
wi th shri nk i ng bandgap in the bottom cell the cut-off at the Tauc-gap energy. Thi s 1eads to a
mismatch seems to decrease . However, this can be cut-off in the spectral response curve in the
due to the 1arger overall thi cknesses in these red which is not observed in solar cells made of
devices which are necessary for current matching amorphous materials. Their red reponse is deter-
at AM 1.5G. mi ned by the exponential decrease in 0(. at and
below the bandgap (Urbach-tail).
The non-abrupt spectral response tail at
;g higher wavelengths leads to overlap regions in
L 100 the spectral response of stacked pinpin and
'5 multiple bandgap cells. This can be seen in fig .
a. 4, where examples are plotted for a stacked pin-
So triple cell pin (with equal i-layer bandgaps of 1.78 eV) and
/1 (SiGe. 1.66 eV)
<: // AM2 an tandem (i 2-1 ayer bandgap 1.22 eV) cell,
~ /
/
both having i2-layer thicknesses of 200 nm .
a 50 /'1 Because of current-matching , the top-layer of
the second devi ce is much thi cker (466 nm with
~ / AM3.6
respect to 66 nm in the pinpin, see also table
N
1ii /1AM5 I). This leads to a much larger red response in
E , I
normalization: the top-cell of the tandem. Although the
15c: 1 current at AM 1.5 G (SEAl)
bandgaps are quite different, increased red
AM 9.2
response of the top-cell and some remaining
7 8 9 10 11 12 13 14 overlap in spectral response of top- and
time of the day [hI bottom-cell of the tandem keep the spectral
mismatch effect at a low level.
Fig. 3: Current output of single and multiple
cells in the course of the day (pinpin and tan-
dem cells fall between the curves shown, see
tabl e 1)

1.0
Q)
(/J
top cell
c ....--..... «tandem)
o
a.
\\ r,
/ . . ." \ / bottom cell

"
(/J
Q) \ / \ /\ (tandem)
L.. 0.5
~
\
"\
/' \
\./
\
\
......
L..
C,)
I \ bottom cell \
Q) / (pinpin) \
a.
(/J
\ /-

'-" \ ,
....
0.0
400 500 600 700 800 900 1000 1100
wavelength [nm]
Fig. 4: Calculated spectral response of pin (solid lines) and of a-SiGe:H 0.22 eV) tandem (broken
lines); thicknesses of i-layers such that currents match at AMI .5G (see text and table I).

348
5. POWER OUTPUT OF MISMATCHED SERIES CONNECTION A principal criticism against the above po-
wer output calculation of a stacked cell could
In case the currents of individual cells in be that there is a vertical tangent in the I-V
stacked devices are different, the worst case is characteristic at open circuit, being equivalent
to assume the resulting current to be the mini- to a zero series resistance. Such a series
mum of the individual currents (this is the case resistance would only slightly lower (if at all)
for the resul ts in the previ ous secti on, tab 1e current differences. However simplified this mo-
I). However, there will be an adaptation of the del may be, it shows that cell s with fi 11 fac-
currents put in series according to the indivi- tors of 601 when put into series with a current
dual I-V characteristics of the subcells. It may mismatch of the order of 51 cannot loose even 11
a1so occur that thei r fill factors are di ffe- in actual power output.
rent, but the resulting stacked cell fill factor Therefore the current mismatch effects cal~
will lie in-between. In fact, the combined cha- culated in the previous section are by far over-
racteristics are complex to calculate and we estimating the actual power loss. The fact that
will use a simplified model in o"der to illus- current mi smatch in stacked multi pIe j uncti ons
trate what can happen. For this purpose, a can be compensated by other effects, has aI so
simple parametrization of the I-V curve been demonstrated experimentally, see ref. 10.
V (I) = Vo (l_In/I~) 6. CONCLUSIONS
is chosen, from which a simple formula for the Using complete optical data for all the
fi 11 factor (FF) in terms of n fo 11 ows by cal- layers in a-Si:H and a-SiGe:H based solar cell
culating the maximum power point, yielding e.g. structures we obtained thickness combinations
FF s 0.58 for n = 5 or FF - 0.62 for n = 6, for current-ma tchi r.g in tandem and tri pIe devi-
etc., the analytical expression being ces for current-matching under the AM1.5G (SERI)
spectrum.
FF n (n + 1)-(n + 1)/n
= (2) Then the spectra were vari ed due to hi gher
air-mass values occuring during the .course of
By means of thi s parametri zati on .i tis the day. Current outputs of single pin and mul-
straightforward to calculate the maximum power tiple cells were compared. During most of the
output of a series connection PS* of diodes by day, when most of the energy is collected, the
addi ng the characteri sti cs of eq. 1. The result normalized differences are almost negligible
is not equal to the sum of the i ndi vi dua I maxi- (below 31). Therefore the benefit of stacked and
mum power outputs Pl* and P2* and we call multiple bandgap devices over single pin solar
ce 11 s, as regards s tabil ity and effi ci ency, is
~P/Ps=(Pl*+P2·-P;)/(Pl*+P2) (3) not lost because of spectral mismatch effects.
(1) J.C.C. Fan and B.J. Palm, Solar Cells, 10
the relative power loss in the series connection (1983) 81
due to the current mi smatch. It turns out that (2) I. Chambou 1eyron, So 1ar Ce 11 s , 12, ( 1984)
when Ii miti ng ourse 1ves to current di fferences 393
up to 301, voltage differences can be neglected (3) J. Young, R. Ross, R. Mohr, and J.P. Four-
and the power loss can be calculated in terms of nier, Materials Research Soc. Symp. Procee-
the relative current difference. This is plotted dings, vol. 40, 475 (1986)
in fig. 5. It is seen that the relative dif- (4) L. Ley, The Physi cs of Hydrogenated Amor-
ference in power output is always sma 11 er than phous Sil i con II, Spri nger VerI ag, Berli n,
the relati ve difference in currents, dependi ng 1984, p. 61
on n (or on the fill factor, being closely rela- (5) M. Heller and G. Bauer, priv. communication
ted to n). (a-SiGe:H layers and transmission data)
(6) M. Stutzmann, priv. communication (ellipso-
metric data)
IP -(~... f))1
(7) Handbook of Chemistry and Physics, CRC
R.+2 , -_ _ _ _ _ _ _ _ _ _ _ _ _ _----, Press, Inc., Boca Raton, Florida, 1984,
65th edition, E-364
(8) J. Morris, R.R. Arya, J.G. 0' Dowd, and S.
Hiedemann, J. Appl. Phys. 67, (1990) 1079
(9) H.A. Macleod, Thin-Film Optical Filters,
Adam Hilger Ltd, Bristol, 1986, 2nd edition
(10) T. Glatfelter, J. Burdick, J. Fournier, and
L. Boman, Proc. 19th IEEE PV Specialist's
Conf., New Orleans, 1987, p.194

.,
0.01 3 4 5 6 7 e 0.1 J
(i l- 12)j12

Fig. 5: Relation between the relative current

difference in a stacked cell pinpin and the re-
lative power loss in the whole device with re-
spect to the sum of individual powers. (The pa-
rameter n corresponds to a mathemati ca 1 repre-
sentation of the I-V curve in terms of In)

349
10TH EUROPEAN PHOTOYOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

CHARACTERIZATION OF a-Si:H BASED PIN CELLS BY SUBBANDGAP PHOTOCURRRENT SPECTROSCOPY

H. RUbel, H. Frammelsberger , R. Geyer, B. Scheppat, P. Lechner, M. Gorn , N. Kniffler

Phototronics Solartechnik GmbH (PST), O-SOII Putzbrunn, Germany

ABSTRACT. In a systematic study we measured the spectral distribution of the optical

absorption coefficient c(, (h \) ) within the subbandgap energy range. This was done by
constant photocurrent method (CPM) measurements on glow-discharge deposited a-Si based pin
cells. He evaluated typical CPM spectra for pin cells with thicknesses ranging from 0.4 f.m
to 10 pm. He discuss the correlation between the device characterizing quantity, the fill
factor (FF), and the number of defects which is related to the quantity ~ 0 (subbandgap
optical absorption constant). He show that the changes in the i-layer due to the creation
of metastable defects after light soaking or current injection - characterized by 01. D -
di rectI y corre I ate wi th the changes in so I ar ce 11 performance (FF). Further we try to
quantify 0<..0 in terms of a defect density of states (OOS)'

I. INTRODUCTION Hith an i-layer thickness of 0.4 }1m an AM

1.5 efficiency of S.st \$ achieved. For compari-
The distribution of the density of states son, some glass/TCO/nin/metal structures were
OOS in the bandgap of amorphous s11 icon is the prepared for space charge limited current (SCLC)
crucial parameter for the characterization of measurements.
this material . Since a-Si:H solar cells are ra-
ther complex thin fllm devices including doped 3. DEFECT CHARACTERIZATION
regions and he'terojunctions, the number of ex-
perimental techniques to study the intrinsic ma- The absorbance due to defects Ii es I n the
terial in the pin structure Is very limited. Va- energy range 0.7 eV < h ~ ( 1.4 eV. He use ~
I uab lei nformati on on the dens ity and the ener- at the photon energy h~ • 1. I eV ( fA (1 . 1
getic distribution of electronic states in the eV» and the defect related term 0<" 0 to
forbi dden gap can be gained from the spectra I characterize our pin cells (see Fig. I).
dependence of the absorption coefficient In Fig. 2 the values for r;I.. 0 . 1 eV) are
CA.- ~(h,») at subbandgap energies II, 2, 3/ . plotted as a function of the applied bias vol-
Further, it was shown that there exists a clear tage for pin cells with a thickness ranging from
correlat ion between photogenerated defects de- 0.64 pm to 9.5 fm. All pin cells show a fairly
termined by constant photocurrent method (CPM)
measurements and changes in the solar cell per-
formance (i.e. the fill factor FF) 14, 51.
In our investigation we used the CPM to
characterize our a-Si:H based pin cells . We stu- a - Si:H
died the influence of different bias voltage
modes (reverse and forward bias, short- circuit .....
c:
pin
case) for .cells of various thicknesses (0 .4 fm
<d <I0 JIm). Also we mod ifi ed the properti es of Q)
CPM
the cells by prolonged light exposure or forward .U
bias current injection in order to di scuss the
correlation between the device characterizing ::::
Q)
quantity FF and a defect indicating quantity de- o
duced from the CPM-absorption coefficient. u
2. EXPERIMENTAL c:
o
';::::'
In the CPM the photon fl ux is adjusted to C.
keep the photocurrent Iph constant when the 10-
photon energy E • hv is varied. Infrared light oC/)
enters the pin cell through the p-Iayer . The
signal from the cell (an ac photocurrent) is re- .c
corded using phase sensitive detection . The re- CO
lative CPM signals were calibrated by fitting
them to the absorption spectra determined by
transmission measurements on films deposited un-
der the same condi ti ons as the acti ve I ayer In
the cel I.
The samples were prepared by glow dis- energy E
charge-decompos i t i on of s i I ane on a TCO coated !..UJ..,..-1. :
glass substrate in a three-chamber load lock sy- The pri nci pa I feature of a CPM spectrum. The fi-
stem. The structure of the cells is: gure shows how to estimate ct.. O.
glass/TCO/p-a-SiC:H/i-a-Si :H/n-a-Si:H/AI.

350
 constant CPM signal for -3 V < Us < 0.5 V
(UB: bias voltage). This behaviour indicates
\ nearly complete coli ecti on of all photo-genera-
pin ted carri ers. For UB > 0.7 eV, there is an i n-
0.64
Ia at h v = 1.1 eV I crease in the 1.1 eV signal for all cells. This
increase is very strong especially for the thin-
ner pin cells (d - 0.64 ~m, 1.7 pm; the same was
observed for pin cells with d - 0.4 j-lm and 1.0
~m, not shown here) whereas in the case of the
o thicker cells it is more gradual (d - 3.4 p'm,
6.6 pm). For the cell with d - 9.5 pm no Mas
dependence exists any more.
We recently developed an empirical model to
explain this thickness dependent behaviour of
. / .0 pin cells when the bias voltage is varied 13/.
We have taken into account the asymmetri c occu-
.", pation of states throughout the i-layer. We con-
J..-_ _ _. _ .- .- .- 6.6 cluded that in a reverse bias mode, relevant
-1 ~'- ' transitions are restricted to an intermediate
i-layer region where the Fermi level lies near
to mi dgap 131. In the fo 11 owi ng, all CPM mea-
dC~ml: surements are made in the short ci rcuit mode,
because there is no change in the CPM signal for
.--7.I ' • • ,-
9.5
- 3V < U~ < + O.SV. Therefore, we are sure
that the slgnal is due to a real i-layer DOS.
In the following we want to establish a re-
Iati on between the solar cell performance and
-2 the defect quanitity ~ D. Therefore we i ntro-
- 3.0 0 0.5 1.0 1.5 duced different degradation treatments (using
1i ght or current i nj ecti on) for the same cell (d
= 0.4 pm). Then we evaluated the solar cell per-
formance Woc , I sc , FF, "1 ) and the CPM
f.ls......Z: spectrum. Fig. 3 shows the degradation behaviour
The absorpti on coeffi ci ent ~ at h» • I. I {V of the pin cell during light illumination (a):
as a functi on of bi as voltage Us for pi n cells AMI, 100 mW/cm 2 ) and current injection (b):
of different thicknesses 110 rnA). For both degradation methods it is

9
9
8 a -Si:H
;f 8 ;f 7
e..... e..... 6
pin
I=' 7 I=" b)
5
6 4
0.9 0.9
~
~

2:-
() 0.8 () 0.8
0 0
::> ::>
0.7 0.7
N N
E 14
0
":-6- 0 e:. A
E 15
0
....... 13 A --6 ....... 14
<X: <:
oS 12 oS 13
U ()
en 11 en 12
0.7 0.7

L.L.. u.. 0.6

L.L..
0.6
u.. 0.5
0.4
0.1 1 10 100 0 10 100
degradation time t [h] degradation time t [min]
.E.!.9.J:
Cell performance during degradation (a): by l1ght; b): by current injec-
tion) as a function of exposure time

351
 0.1
0.01 0.10 1 10 100 1000
degradation time t [h]
10.0 0.70
Q ........•. o
.. 0·····
b)
.... ·0)..'
'
' IJ.
.. 0' 0.65
!;l"
aD
........ :..: a- Si:H 0.60
..-,
E
.......
() 1.0 •.... pin
0.55 u..
u..
0 .'
·· .•.. /FF
~ ~
.r .....• 0.50
...... .
"
.................... ....
degradation current: 110 rnA 0.45

0.1 0.40
0.01 0.1 1 10
degradation time t [h]
~:
FF and <i.O as a function of degradation time t for light degradation (a)
and a current degradation cycle (b).

clearly indicated that the deterioration in ef- 4. CORRELATION OF r:J... 0 AND THE DENSITY OF GAP
fi c i ency is mainly due to the reducti on of the STATES (DOS)
fi 11 factor FF.
The quas I para 11 e 1 behavi our for the 1i ght Up to now we have been speaki ng about the
degradation and the current injection treatment, defect term 01.0 ' It would be desirable to
i . e. the increase of 01. 0 and the drop in FF quanti fy it in terms of the ~OS. A compari s i on
for prolonged degradation time t is demonstrated of CPM (~O) and space charge 1 imited current
in Fi g. 4 (same degrada ti on condi ti ons as in measurements (OOS obtained from SCLC) 161 on pin
Fi g. 3). and nin structures respectively, yields compa-
Figure 5 summarizes the unique relation rable results (Fig. 6) . The subgap absorption
between FF and the defect quantity d. 0 for term do. 0 as well as the DOS estimated by SCLC ,
different degradation treatments for the same grow with illumination time t following initial-
cell (the degradation conditions are specified ly, the well-known functional form (t l/3 ),
In the figure). The results clearly indicate predicted by the Stutzmann model 171 . The satu-
that the solar cell deterioration by 1ight or ration at relatively low illumination times may
current I nj ecti on is introduced by the induced be due to the temperature T - 60·C during illu-
changes (the creation of metastable dangling mination. If one chooses a proper constant, one
bonds) In the intrinisic (undoped) region. can sca 1e 0'.0 I n terms of the OOS.

352
 0.8 r---r--.,---r--.------.----r--.-----,
REFERENCES
light degradation:
.. AM 1.0 111 H. Klda, H. Yamagishl, T. Kamada, H. Okamoto
(100 rnW/crn·2 ) and Y. Hamakawa, Technical Digest of the 1st In-
current degradation: ternational Photovo1taic Science and Engineering
0.7 Conference (Kobe, Japan, 1984) 417
.110 rnA
o 55rnA 121 J. Ko~ka, M. Van~tek, O. Stfka, O. Dung
6 27rnA Trl ng, J. Stuch 1i'k and 1. THska, Proc. of the
8th European PV Solar Energy Conference (Flo-
LL rence, Italy, 1988) 724
LL 0.6
131 H. RUbel, R. Geyer, 8. Scheppat, P. Lechner,
M. Gorn and N. Knlffler, Proceedings of the 21st
IEEE Photovoltaic Specialists conference (Kissi-
0.5 a - Si:H mee, USA, 1990) 1389
pin 141 S.S. Hegedus and J.M. Cebulka, Proceedings
d = O.4llm of the 20th IEEE Photovoltalc Specialists
Conference (Las Vegas, USA, 1988)186
0.4 I.---L_-'-_-'-_-'-_-'--_'----'-_--' 151 H. Framme 1sberger, H. RUbel, P. Lechner, R.
o 2 6 8 Geyer and N. Kniffler, Appl. Phys. Lett., to be
published
161 H. Huber, diploma thesis, MUnchen 1990
~:
The fill factor FF as a functi on of the defect
related term ~ D for light and current degra- 171 M. Stutzmann, H.B. Jackson and C.C. Tsai,
dation. Phys. Rev. B 32 (1985) 23

.: SCLC 0: CPM
Glass TCO Glass TCO
......-.
..-
>Q)
I
~LI-M. W
~AI 10

......-.
(")
I
.,
n
\
I
\
n pin •
,.,.. - -
• ••
_---....0;..
T"""
I

E
E 10 17 dl = 0.4 11m 0 o
........
......
0 .;0......... •

.
(/)
o .;
........
0

.
0
0 .............
a - Si:H
t l13
,.,../ .
O ......... ................- - -
light soaking
......-
/ AM 1 [100 mW / cm 2]
......-
/
jOI6~{~L-~~~~__~LL~W-~-L~llUL-~~~~__~-W~
o 0.01 0.1 1 10 100 1000
illumination time t [h1
.Ei.9......-§:
0<. D from CPM and the DOS from SCLC measurements as a function of ill u-
mlnatlon time t.

5. CONCLUSIONS
He have shown that the changes in the I n-
trlnsic layer of a pin cell due to the creation
of unstable defects (o'..D) can be directly
correlated with the changes in solar cell per-
formance (FF). Finally we were able to correlate
~D with the measured DOS.

ACKNOWLEDGEMENT
This work was supported by the Bundesminister
fUr Forschung und Technolgie

353
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

STABILITY OPTIMIZATION OF a-Si:" BASED SOLAR CELLS

P. Lechner, H. RUbel and N. Kniffler

Phototronics Solartechnik GmbH, Hermann-Oberth-Str. 9, D-8011 Putzbrunn

ABSTRACT: The stability of pin solar cells is essentially dependent on both thickness and quality
of the intrinsic material. Intrinsic a-Si:H was prepared at various deposition rates and substrate
temperatures as bulk films and, with the same set of conditions, in pin devices. Photoconductivity
and defect density (from Constant Photocurrent Measurements) are correlated with the material
structure evaluated from IR transmission measurements. Material having a low fraction of structural
inhomogeneities shows the highest photoconductivity and the lowest defect density before and after
light soaking. Howeve~ the photoconductivity ratio taken before and after light soaking is unsuited
to assess material stability. The stability of pin cells is significantly improved with higher sub-
strate temperatures and lower growth rates, which also lead to a more homogeneous i-layer material.
Thus properties of intrinsic films can be directly correlated with the stability of pin devices.

1. INTRODUCTION
The films with a thickness between 1 and
Although intensive examinations of light 2 pm were provided with coplanar Al electrodes
induced phenomena of a-Si:H have been done over for dark and photoconductivi ty measurements and
the past years, a-Si solar cells still reveal an constant photocurrent measurements (CPM). The
efficiency loss of typically 15 to 30'L before films on c-Si substrates were used for FTIR
stabilization. Many efforts have been made to transmission measurements. cH was estimated
identify the decisive parameters for the deposi- from the absorption mode at 640 cm- l . The
tion of stable material and solar cells. The absorption at 2090 cm- l (12090), usually
subject of this paper is to assess a correlation identified with both the stretchlng modes of
between the light-induced degradation of pin SiH2 and of SiH due to inner surfaces, was
solar cells and intrinsic films. Therefore, pin evaluated by deconvolution of the two overlap-
devi ces were produced ina mul ti chamber system ping absorption modes at 2000 and 2090 cm- l .
where the deposition conditions for p and n lay- Solar cells, with an i-Layer thickness
ers were kept constant, whil e the i-I ayer was between 0.2 and 0.6 pm were grown on smooth TCO,
grown at various conditions yielding different in order to reduce the effects of field non
material quality. In parallel to this series uniformities due to surface roughness.
i-layers were deposited on Corning glass under The I-V curves were measured with a solar simu-
the same set of conditions. In order to optimize lator calibrated to the global AM 1.5 spectrum.
the i -layer qua 1ity, the vari ati on of two depo- The fi 11 factors FF were around 0.7 except for
sition parameters was studied: Growth rate and those cells deposited at very high Ts.
substra te temperature Ts' Both control the A11 1i ght soak; ng experi ments were perfor-
hydrogen content cH of films, which seems to med under xenon 1amps with ali ght i ntens i ty of
be one of the most crucial material parameters, 100 mW/cm 2 , that led to a steady state tempe-
d£termi ni ng the suscepti bi 1i ty to 1i ght-i nduced rature of 55·C. The cells were kept at open
degradation (1). circuit during degradation. The light from the
Besides the i-layer quality the second xenon source is absorbed nonuniformly especially
equally important feature for the design of in thick films. However, as the optical gap of
stable solar cells is the i-layer thickness (2). all the i-layers shown here varies only between
The improved stability of stacked cells is based 1.70 and 1.76 eV, the non uni form creati on of
on thi s thi ckness dependence (3). In order to defects can be considered to be similar in dif-
assess qualitative differences of the i-layers ferent films, and hence the effects of light
in pin cells, it is important to use the same soaking are thought to be comparable.
i-layer thickness in these cells.
3. RESULTS AND DISCUSSION
2. EXPERIMENTAL PROCEDURE
The stability of a pin solar cell is
The intrinsic films and the pin solar cells strongly dependent on its i-layer thickness.
were deposited by rf glow di scharge in a three- Fi g. 1 shows the i niti a1 effi ci ency '10 conti n-
chamber system with load lock. Since separate uously increasing from 7'L to 8'L with i-layer
chambers were used for intrinsic and doped thickness, due to higher short circuit currents.
layers, a comparable contamination level for the However, after 500 h 1i ght exposure thi ck cells
bulk i-material in both films and pin cells is revea 1 the lowest effi ci ency, whereas cells
expected. In the fi rst seri es Ts was vari ed in between 0.2 and 0.4 pm thickness show a maximum
the i-chamber from ISO· to 300·C, while the in the efficiency, in this case about 6.5'L. This
deposition rate was kept constant at 2 'Als. The value can be further improved by means of opti-
second series was grown at Ts = 200·C with cal light trapping. The ratio 11,/1\0 decreases
silane pressures between 0.1 and 0.3 torr. This here from 0.88 for 0.2 pm to 0.72 for 0.6 pm
resulted ina variation of the growth rate from thick devices. This dependence clearly shows
1.5 to 6 "t,/s on Corning glass and c-Si substra- that a comparati ve asses sment of the i ntri ns i c
tes, and a somewhat higher rate up to 7.5 'Als on material quality can be made only in cells with
TCO substrates. the same i-layer thickness.

354
 to The dark- and photoconductivity, GO and
GPh, of films deposited at rates between 1.5
and 6 AI s , and at Ts between 150· and 300· C
9 .9 are shown in Fig. 2 and 3, respectively. With
increasing rate, GPh and GO show a downward
8 .8 trend for the ' as deposited' state (state A).
After 100 h light soaking (state B) both Gf>h
0
.7 ~
........ a - Si:H
~
~

6 .6 10 parameter:

T5\ _ - - - - -
5 11 (500 h) .5 ......-
/--/ _ _ 2- _ _0 - -
4
0.0 0.2 0.4 0.6 0.8 1.0
.4 0-/ \ 0
I - layer thickness l rate

-i -
[~m] 5
Fi g. 1: Effi ci ency before and after 500h 1i ght
soaking, and normalized efficiency of
pi n ce 11 s as a functi on of the i-I ayer
thickness. The lines are a guide for
the observers eye
·3r-~~~-'~~~~~-,
O~----~----~-- __L - J
0.0 0.1 0.2 0.3
:o=:~-
-4
12090 / (1 2000 + 12090)
-5
1oohAM1 Fig. 4: Hydrogen content cH versus the struc-
~
a.... tural parameter 12090 / (12000 +
E lZ090) 12000 and lZ090 are the
u lR absorption modes at 2000 and 2090
~ -7 cm- l , respectively
tl
Cl
.2 -8

~u_
and GO are reduced by one order of magnitude.
-9 This is the classical observation of the
Staebler-Wronski Effect (SHE). The interesting
-10
feature here is that the ratio ~PhCA)/GPh(B)
is independent of the growth rate. In FI g. 3,
• 100 h AM 1 GPh significantly decreases 'for substrate
-11 '----'................--'---'--'--'--'---'--'---1 temperatures below ZOO· C. Again, the reduction
o 5 10 of GPh due to light soaking turns out to be
deposition rate [~ I 51 one order of magnitude, independent of Ts . The
Fig. 2: Photo- and darkconductivities of in- ratio G'Ph/Gb in state A is above 105 at
trinsic films versus the deposition rate low and about 103 at high Ts. After light
soaking this ratio is even improved for films
~r-~--~---r--~--~ deposited at high Ts. He thus conclude that
neither the ratio GPh(A)/GPh(B) nor the
ratio ~h/GO before and after degradation is
as deposited an unambi guous fi gure of meri t for the i-I ayer
stability.
-5 Fi g.4 shows the hydrogen content cH as a
function of the ratio I2090ICIZOOO+IZO~0)
which is a measure for the fraction of materlal
inhomogeneities. As the growth rate is increa-
sed, cH remains almost unchanged at 7t
(TS·ZOO·C), whereas the incorporation of
hydrogen is more and more shifted from the
SiH configuration to the void rich configura-
tion. This indicates that at higher rates (here
4-6 'l.ls) the structure of a-Si:H becomes more
-9 i nhomogenous . By lowed ng Ts from 300 to ZOO·C
cH increases from 5t to 7t without a major
-10
change in the void fraction. However, as Ts
approaches 150·C, both a higher cH (up to 9t)
and a significant structural change are indica-
-11 ted. It is generally agreed that the SHE depends
essentially on the amount and bonding configura-
150 200 250 300 tion of the incorporated hydrogen. Therefore, we
deposition temperature T5 rCJ plotted in Fig. 5 the G"Ph values from Figs. 2
and 3 as a function of the structural parameter
Fi g. 3: Photo- and darkconducti viti es of i n- I2090ICI2000+12090)' The photoconductivity
trinsic films versus the substrate tem- which for a-SI:H is a measure for the mobility-
perature Ts lifetime product of the electrons is clearly

355
reduced with increasing material inhomogenei- After discussin-g the light induced changes
ties. Device quality a-Si:H containing only a of intrinsic material in terms of dark- and
small fraction of hydrogen in the SiH2 confi- photoconductivity, defect density, hydrogen
guration reveals the highest ~Ph values in content and its bondi ng confi gurati on, we will
states A and B. However, the ratio describe below the stability of single junction
~Ph(A)/uPh(B) is the same as for rather in- cells, that were deposited with the same set of
homogeneous material. conditions as the individual i-layers. In
The defect density NO which was estimated Fig. 7, the efficiency after 500 h light
by CPM is shown in Fig. 6. Here a clear correla- soak i ng. normal i zed to the i ni ti a 1 effi ci ency,
tion can be established at least for the light is plotted as a functi on of the i-I ayer depos i-
soaked samples. Although for state A the data tion rate for two series of cells, 0.2 and
show more scatter, the same trend can be obser- 0.4 pm thick, respectively. The trend to higher
ved: Void rich material probably has a higher degradation at high growth rates is obvious. The
initial NO, but definitivly has a higher NO fact that rather thin cells reveal the same
after degradation, as compared to device quality trend as thicker cells, although at a higher
stability level, indicates that bulk material
.3r-----r-----r-----.-, properti es play the domi nant role in the degra-
dation. It is important to mention that the
initial performance in contrast to GPh(A) was
a - Si:H not a function of the growth rate. This can be
o~

-_0--
explained by the fact that in an undegraded cell
o -- _ as deposited
the dri ft 1ength EJI't (where E is the e 1ectri c
·4 0 -~o /0 field in the i-layer) is much larger than the
i-layer thickness. Even for inhomogenous mate-
o ri a 1 where p'r, deduced from the uPh values, is
•
.c;
a.. about a factor of two lower than in device-grade
t> . ..... 100 h AM 1 material, this still holds. However, in state B,
-5 --.. --.. __ I Gph and hence pT decrease by about one order
--.. .
Ol
.Q . ..... . of magni tude. Now the dri ft 1ength is on the
-..;
order of the i-layer thickness, and any varia-
~-- tion of IJ'r significantly influences the carrier
transport leading to the observed stability
·6 L-----~----~----~~ differences in cells.
0.0 0.1 0.2 0.3 The variation of the substrate temperature
turned out to be more criti ca 1 than the vari a-
12090 1(1 2000 + 12090 ) tion of the growth rate. It is unavoidabl e that
both the TCO/p and the p/i interfaces are i n-
fluenced at high Ts. As a result, Vac and,
Fig. 5: Photoconductivity versus to some extent, the FF significantly deterio-
(I2000 + 12090) rate. The initial efficiency is too low for a
decent stabi lHy comparison. As a consequence,
Ts is restricted to the regime from 150· to
220· C. From Fi g. 8 it is evi dent that the nor-
a - Si:H malized efficiency after 1000 h light soaking is
significantly improved from 0.74 to 0.88, as
100 hAM 1 )--

--
\ --'~
--- -< 1.0 500 h AM 1
••••
• o

0/
o ~
Q)
0.9 0-0

o~~ o .!::!
CO
E 0.8
o
-0
~m
c

~o
~

as deposited 0
oS
f=' 0.7
0.0 0.1 0.2 0.3
12090 I (1 2000 + 12090 ) 0.6
Fig. 6: Defect density of the intrinsic films
estimated by CPM, versus o 5 10
12090 / (12000 + 12090)
deposition rate [A / s1
material. Thi s resul t i s fully consi stent with
the behaviour of GPh as the carri er 1ifetime Fig. 7: Normalized efficiency of pin cells
is controlled by the defect density. after 500h light soaking as a function
It is important to menti on that Park et a 1 . of the i-layer deposition rate. The
(1) found a correlation between the saturated i-layer thickness is 0.2 and 0.4 JIm.
defect density and cH' In our case such a respectively
correlation can be only established when both
cH and the s tructura 1 i nhomogenei ti es increase
(T s variation). However structural changes at
constant cH (rate variation) can also entail
changes in GPh and No.

356
 6.REFERENCES
(1) H.R. Park, J.Z. Liu, P. Roca i Cabarocas,
1.0 1000 h AM 1 A. Maruyama, M. Isomura and S. Wagner,
Materials Research Society Symposium
Proceedings, Vol. 192, 1990, p. 751
as 0.9
(2) T. Tsuda, N. Nakamura, K. Watanabe,
.!::!
til T. Takahama, H. Nishiwaki, M. Ohnishi and
E 0.8 Y. Kuwano, Solar Cells, 9, 1983, p. 3
o
.s
I=" 0.7 ~
(3) P. Lechner, B. Scheppat, R. Geyer,
H. RUbe I , M. Gorn and N. Kni ffl er,
Materials Research Society Symposium
0.6 Proceedings, Vol. 149, 1989, p. 583

150 200 250

deposition temperature Ts rC]

Fig. 8: Normalized efficiency of pin cells

after 1000h light-induced degradation
versus the i -layer deposit; on tempera-
ture. The i-layer thickness is 0.3 ~m

Ts is increased from 150· to 200· C. It is

unclear, whether a further enhancement of Ts
would lead to more stable cells. This has to be
investigated in the future. At least for the low
Ts regime a clear correlation between cell
stabil ity and G'Ph(B) has been establ i shed,
similar as for the variation of the growth rates.
4. CONCLUSIONS
Intrinsic a-Si:H films prepared at a low
deposition rate and high substrate temperature
exhi bit a low fracti on of s tructura 1 i nhomoge-
neity that is correlated with high photoconduc-
tivity and low defect density. Even for high
depos iti on rates and low subs trate temperatures
that lead to void rich material, the ratio of
photoconductivities before and after light
soaki ng does not change and it is thus unsuited
for stability assessment.
As the stability of pin cells is strongly
dependent on thei r i-I ayer thi cknes s, opti mi za-
tion of the intrinsic material has to be done in
devices of the same thickness. Cells with
i-layers grown at higher rates (5 ~/s) show a
s i gnifi cantly 1arger degradati on than those
deposited at 2 1>./s. The variation of the sub-
strate temperature yields sufficiently large
initial efficiencies only in the lower tempera-
ture regime to permit a meaningful comparison.
The stability of the cells significantly impro-
ves with increasing substrate temperature, while
the corresponding films indicate a struc- tural
change to more homogeneous material. Thus
intrinsic film properties are reflected to a
hi gh degree in the stabil i ty behavi our of pi n
devices.

5. ACKNOWLEDGEMENTS
The authors like to thank A. Scholz from Univer-
sitat Kaiserslautern for the IR measurements.
Helpful discussions with H. Schade are grateful-
ly acknowledged.

357
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

ACCELERATED STABILITY TESTING OF AMORPHOUS MODULES

Ursula Eicker, Solems, 3, rue Leon Blum, F-91124, Palaiseau Cedex France. Tel. (33) (1)
60.13.34.40. Fax. (33) (1) 60.13.37.43.

ABSTRACT. An effective process optimisation of end of life efficiencies imperatively

requires accelerated degradation methods, reducing the delay of stability results from
typically 1000 h to 1-2 hours. We have used forward bias current injection at intensities
between 100-400 rnA/cm' to reach saturation levels within 0.5 hours. The saturation time
varies inversely proportional to the square of the generation rate and the fmal defect
density is lower at high temperature, i.e. high annealing rates. The accelerated aging,
however, creates more unstable defects with lower activation energies than those generated
by conventional AM 1.5 light soaking, which slowly anneal even at room temperature.
However, the energy levels of such defects are comparable to light generated defects.

1. INTRODUCTION 3. DEGRADATION KINETICS

The degradation kinetics of amorphous PIN modules The first aim of our study was the determination of
depend not only on the carrier densities in the i layer, i.e. for the current density necessary to reach end of life values in
a given illumination on absorption in the p layer, aging at V 00 the shortest delay. The current density was varied from 100
or load conditions, recombination statistics etc., but also on rnA/cm' as shown in Fig.2. It has to be noted that at 200 to
the quality of the interfaces and the susceptibility of the 400 rnA/cm', the current density in the interconnection scribing
materials to degradation. Stabilised end of life values under lines of 100 /-lm width reaches already 20 A/cm', above which
AM 1.5 illumination are only abtained after typically 1000 the interconnection is easily damaged.
hours. lbis becomes a serious obstruction to a fast feed back
deposition process optimisation. Fig.l shows the degradation
kinetics of two technologies of 30 x 30 em' modules for
which a judgement of their stability can only be made after
more than 1000 hours of light exposure. Accelerated testing In!luence of injection inLensiLy
70
using forward bias current injection has therefore been
developped to reduce the delay of results to some hours. · 100 mA/em2
65 • 150 rnA/ omll
• 200 rnA/omll
70
o ~2~~nology o 250 mA/orn2
x 300 rnA / orn2
65 ~~Bhnology • iOO rnA/orn2
55

50 ~------~--------+--- ____- +_ _~~~

0. 1 10 100 1000
55 o 0 Time (h)

50~----------~------~-----+----~
Fig. 2 : Kinetics of the Filljactor degradation as a
0. 1 10 100 1000 10000 function of current intensity (100-400 mAjcm').
Time (h) The arrows indicate the saturation time
calculated from t"",=k JjG2 with k=5xJO" em'·.
Fig. J : Fill factor degradation under a Na lamp for two
module technologies.

2. EXPERIMENTAL DETAILS
The parameter represented is the Fill-factor of the
The modules used in this study consisted of 10 0. SnO, whole module measured under high illumination which has
substrates with a pin structure of 2000-3000 A thickness and ?ee~ s?own to correlate directly with the defect density in the
a standard metallisation. The pin cells were series connected mtrinslC layer (1). At 400 rnA/cm', saturation is reached within
by integrated scribing lines to give 24 individual cells on a 30 30 minutes. We can compare this value with the results
x 30 cm' module or 6 cells on a small 5 x 5 cm2 module. The obtained by the Ptinceton group using high intensity laser
modules were deposited in a monochamber plasmabox reactor illumination to degrade intrinsic samples (2). Using a
at deposition temperatures of 200° C, a total pressure of generation rate of 3 x 1022 em's" they obtained saturation of
loo/-lbar and deposition rates of ~ 1.3 Als. The dopant gas for defect densities I!fter one hour while a generation rate of 5x
the carbonated p layer was trimethylboron. 10'0 em" s" required ~ 8000 hours. The final defect state

358
 density was comparable in both cases (10" cm-')_ The
saturation time was inversely proportional to the square of the
generation rate : c~ : susceptibility to degradation
G : generation rate
tat = k ltG' v0 : attempt to escape frequency

with the constant k at .. 2 x 10'" cm"_ The carrier density in In order to have saturation values for high injection intensities
our modules of 2500 A thickness is 10" cm-' S-1 for a current similar to those for low intensities, either the activation energy
density of 400 mA/cm'. Using the constant k determined by has to become lower, i.e. the defects generated are more
Wagner et ai, the calculated saturation time is 12 min., unstable, or the temperature has to be higher. The lower
corresponding rather well with our first measurement point at activation energy for high level current injection is directly
30 min. The degradation kinetics as a function of the current confirmed by comparing room temperature relaxation kinetics
injection level, i.e. the carrier density, confirms furthermore the of a population of 32 pin modules which have been aged
ltG' dependence of the saturation time (as indicated by the either by conventional 0.7 sun Na lamp illumination or 200
arrows in Fig. 2). mA/cm' current injection, respectively. Whereas the end of life
values are comparable (table I), the modules stabilized at high
A comparison of modules intensity current injection relax at room temperature while the
aged under a Na lamp (generating about 70 % of light soaked modules remain stable.
the photocurrent generated by the sun) and with 200 mA/cm'
current injection is shown in Fig. 3.

70
o IIlumlnoUon No Type of aging Yee Icc FF Rse n
la mp (mY) (mA/cm2) (%) (OIuncm2) (%)
65 Current injection
at 200 mA/cm2 Initial values 800 12.6 62 14 6.2

2 hours current injection 775 11_9 54 17.5 5.0

~60 at 200 mA/cm2
U.
U. 650 hours lamp Na 778 12.2 55 16.7 5.1

55 ---------'------- AFTER RELAXATION (1 month)

Current degradation 789 12.7 57 15.2 5.7

50~------~--------~------~------~ Ught soaking 775 11.8 55 17.8 5.0

0.1 )0 100 1000
Time (h)

Fill-factor degradation by Na lamp illumination

and current injection for the same type of
modules_ The module temperature during light
soaking is about 5C1'C versus 6C1'C during Table 1 : Statistics of 32 PIN modules (30 x 30 cm') aged
current aging. by Na lamp or current injection_ After aging all
modules were stored at room temperature in the
Whereas 1000 hours light-soaking did not yet give a dark for one month and remeasured under
saturation of the Fill-factors, 2-4 hours current injection clearly AMI.5 illumination.
saturated the defect density. 2 hours injection at 200 mA/cm'
is used as our standard degradation to generate comparative On the other hand, aging similar modules with forward bias
statistics for a. larger amount of modules (see table 1). currents at the same intensity as the photocurrent generated by
the Na lamp led to stable end of life values. These
experiments indicate thus lower activation energies for high
current injection intensities. CPM measurements on PIN cells
aged both with low intensity light and high level current
injection showed however, that the defect shoulder is centered
SATURATION VALVES
around the same energy level in the optical gap for both
methods (1). We suggest that the defects generated by current
H one includes a thermal annealing term in the defect
generation model of Stutzmann, Jackson and Tsai (3), defect injection are the same as those generated by low level
density saturation values can be obtained, which depend on the illumination, but that the stabilisation of such dangling bond
temperature during degradation as well as on the activation pairs by hydrogen switching is less efficient in the short
energy for annealing. Recent calculations have shown (4) that timescale of the fast aging.
in the high intensity regime (carrier densities > 2x 10" cm-' A further critical parameter for both the defect saturation value
S-I) the defect saturation value N.., can be expressed and the saturation time is the temperature during degradation.
aIUIlytically as : High temperatures enhance the annealing term and should lead
to lower defect saturation values in a shorter time.

359
We varied L\e module temperature for a given current injection CONCLUSION
level (200 mA/cm2) by changing the cooling temperature of
the substrate holder for the low temperature experiments (0- High intensity forward bias current injection can be
70°C) or by placing the modules into a conventional oven (> used to obtain amorphous module end of life efficiencies
70°C). The temperature was measured by a thermocouple on within a delay of some hours. The time to reach saturation is
the module metal surface. shown to vary inversely proportional to the square of the
injection intensity. The activation energy for annealing at high
intensities is lower than for low intensity light soaking or
70 current injection, leading to a slow room temperature
relaxation. End of life stabilised values and the kinetics of the
65 o 200 mA / cm2 al O'C
cell degradation depend on the temperature of the module. In
.. 200 mA/em2 at "O'C
conclusion, accelerated aging allows a very efficient feedback
,. 200 mA/rm2 al 70'C
for process optimisation, yet cannot be used to obtain aged
X 200 mA / cm2 III IOO'C
modules stable in time.

- II=~---
REFERENCES

45 1. W. Frammeisberger, H. Rubel, P. Lechner, N. Kniffler,

Appl. Phys. Lett. (to be published).
40~----------------------~----------~
O I
.
10
Time (h)
100 2. H.R. Park, J.Z. Uu, S. Wagner, Appl. Phys. Lett. 55
(25), 2658 (1989).

3. M. Stutzmann, W.B. Jackson, C.C. Tsai - Phys. Rev.

13, 32, 23 (1985).

Fig. 4 : Influence of module temperature during 4. M. Gom, B. Scheppat, P. Lechner - Denver Stability
degradation. The module temperature varied Meeting, 1991).
according to the thermal conductivity of the
substrate holders.

For times larger than 10 hours, the defect saturation value

becomes lower at temperatures above 70°C, i.e. the module
fill-factor increases (Fig. 4). On the other hand, the annealing
term does not seem to interfere at short time scales, where the
degradation is actually faster at high temperatures. This effect
could be attributed to an improved energy transfer between
electron-hole recombination and local phonon generation
resulting in a temperature activated Staebler-Wronski
susceptibility C.... More detailed experiments are required to
investigate these initial kinetics.

360
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &--12 APRIL 1991 LISBON, PORTUGAL

LIGHT TRAPPING DUE TO TCO ROUGHNESS IN a-Si:H SOLAR CELLS

MEASURED BY SPATIALLY-RESOLVED PHOTOTHERMAL DEFLECTION

F . Leblanc*S, J. Perrin*, P. OllivierS, A. Antoine-Labouret$

$SOLEMS S.A.
Z.I. Les Glaises, 3 rue Leon Blum
F - 91124 - PALAISEAU Cedex, FRANCE

*Laboratoire de Physique des Interfaces et des Couches Minces

UPR A0258 du CNRS, Ecole Poly technique
F - 91128 - PALAISEAU Cedex, FRANCE

ABSTRACT: Because of the roughness of the front transparent

conducting electrode in a-Si:H p-i-n solar cells, light
scattering occurs in the different layers. Refractive index
differences produce light trapping in the structure.
Photothermal Deflection (PD) is used to analyze the contribution
of trapped and absorbed light due to total internal reflection
at the glass/air interface. Each irradiated spot gives birth to
a diffuse ring, with a radius of approximatively twice the glass
substrate thickness. The radial distribution of absorbed light
is derived by spatially resolved PD from a spot irradiated by a
Kr laser beam at 647 nm. Moreover the overall "glass/air"
trapping efficiency is measured by a spatially averaged PD
method from 350 to 800 nm. This trapping efficiency ranges
typically from 2% at 350 nm'to 12% at 800 nm.

1. INTRODUCTION the TCO film, so that reflected light is

scattered. Because of the high refractive index of
Over the past few years, it has been clearly a-Si:H (n) ~ 3.7 at 600 nm) compared to the TCO one
shown that a-Si:H p-i-n solar cells deposited on (n2 ~ 1.9), scattered reflected beams above a
appropriate textured Transparent Conducting Oxides critical angle Bc,2) ~ Arcsin (n2/n) will be trapped
(TCO) exhibit strongly enhanced conversion in a-Si:H by total internal reflection. M::>reover,
efficiencies. Rough interfaces reduce reflection a fraction of the scattered beams transmitted in
coefficients and favour light scattering. Part of the TCO will be again trapped by total internal
the scattered light is confined in the different reflection at the TCO/glass interface above a
layers due to the refractive index differences. critical angle 1\0,12 ~ Arcsin (ndn2) ~ 52 0 • This
Two types of trapping can be distinguished (see corresponds to a scattering angle in a-Si:H of 9c 13
Fig .1): i) "local" trapping of incident light Arcsin (nl/n) 24 0 • Eventually only the
scattered and absorbed in the cell structure over scattered reflected beams between 0 and 24 0 in a-
a lateral distance of the order of the TCO and a- Si:H will be transmitted into the glass.
Si: H film thicknesses ; ii) "non-local" trapping
due to total internal reflection at the glass/air
interface.
Rough TCO scattering properties are usually
investigated by measuring the haze ratio /1/ or
the angular resolved reflectance. Shade and Smith Air
/2/ have compared TCO scattering properties to the no .1
Mie theory. With simplifying assurrptions, M::>rris
et al /3/ have also developped a rrodel on
absorption 'enhancement for long wavelength (>600
nm) in actual solar cells. Gil ..
In a previous paper /4/, we have presented a "1 •1.5.2
photothennal method, based on a calibration by
Joule heating, to measure the optical absorbance TCO
of rough as well as SIOClOth TCO films. In the n t _ 1.8
present paper, we analyze the non-local trapping .oSl:H
n.) ~ 3.7
by two photothennal methods. Both of which are m.t.1
based on the sensitivity of PD to the power
locally dissipated along the probe beam path. Fig 1: Schematic description of light trapping in
First we analyze the spatial distribution of a solar cell deposited on textured TCO
absorbed light due to non-local trapping for two
sarrples with different glass thickness and TCO
roughness. Secondly we analyze the spectral 2.2. Non-local trapping is effective at any
dependence of the spatially averaged non-local wavelp.ngth. When using a textured front-electrode,
trapping The spectral dependence of the non-local part of the incident light is back-scattered at
light trapping is investigated by varying the the TCO/a-Si:H interface. In addition, for red
wavelength of the incident light. light, another part comes from scattered
reflection by the metal which is not locally
trapped (see §2 .1). Assuming that the refractive
2. PHENOMENOLOGICAL DESCRIPTION OF LIGHT TRAPPING index of float glass is constant over the visible
spectrum and equals to nl = 1. 52, all the light
2.1. Local trapping is effective only for red back-scattered in the glass at angles greater than
light, which is weakly absorbed by a-Si:H, and the critical angle 9c.Ol = Arcsin nO/nl = 42 0 , can
m:)stly reflected by the rear metallic electrode. not escape from the structure. This alOClunt of
This latter interface reproduces the texture of energy is totally reflected at the glass/air

361
 interface and eventually partly absorbed in the irradiated spot. The radial distribution of
structure, at a minimum distance of 2tan (9c ,01) xdg, absorbance A(r) is correlated to S(y) as follows:
where dg is the glass thickness. For a radial
rM
distribution Rsf (9) for the diffu&'! reflectance SCY)=2f A(r)rdr (4)
(where 9 refers to angles in the glass), Rs being Yr2_y2
the total diffuse reflectance and y
where rM is the limiting radius defined by
i~ f (9) sin (9) d9 = 1 ( 1) the sample dimension (rM = 10 mm) at which

f
A(rM) = 0
the radial distribution of absorbance for non- The inverse transform, called Abel's transform 151
local trapping can be expressed as : allows the determination of A(r) :
3

~ ~~
Rs f (9) Rg (9) A (e) cos (e)
A(P) (2)
8 7t A(r) = - rH dy (5)
so that the total absorbance for non-local -V y 2 _r2
trapping AnlT is given by
r

AnlT = iP~ (P) 27tp dp (3)

Moreover, by integrating the energy deposited in
the ring and the energy deposited in the
irradiated area (from the measured curve S (y) or
where p = r/dg = 2tan (9), and A-I is the maximum from the Abel transform), one can evaluate the
radius defined by the dimensions of the sample. In long distance trapping efficiency:
Eq. (2), Rg (9) is the glass/air reflectance (= 4 % fr·S(y) dy
for 9 < ec ,Ol and 100 % for e > 9c ,01), and A(e) the r*
=
J
absorbance in the cell structure for a beam angle llnlT (6)
of incidence e. In this model, second order r
< S (y) dy
contributions of rescattered light are neglected. o
Assuming a Lambertian diffusion law f(9) where r* corresponds to the radius of the area of
2cos (9), and A(e) = 1, the variation of A(p) local absorption : r* = 0.5 nm. The plateau of
normalized to Rs /4x (so that the value for 9 = S (y) for r < 2tan (90,01) has to be extrapolated down
o is equal to Rg) has been corrputed according to to r* to corrpute the integral at the numerator of
Eq. (2) and is shown on Fig.2. Eq. (6)

0.4 y
a
Ol
0.3 He-Ne
!:!
...
<C
probe
laser
'tI 0.2
-- ---- . ., --
G.I
N

l'l modulated
E
... 0.1 Kr laser
0
Z beam

0 b modulated
0 2 4 6 8 1 0
He-Ne monochromatic
r/ dg light
probe
Fig.2 : Radial distribution of the absorbance for laser
non-local trapping, corrputed for a Lambertian
diffusion law and normalized to Rs /4x

3. EXPERIMENTAL

Non-local light trapping can be easily

observed by irradiating a cell with a laser beam :
a diffuse ring clearly appears around the
irradiated spot. Two experiments have been
performed to analyze the energy deposited in the
diffuse ring: Spatially Resolved Photothermal
Deflection (SRPD) and Spatially Averaged
Photothermal Deflection (SAPD).

3.1 Principle of SRPD

In SRPD, a modulated collimated Kr laser beam
at 647 nm is focused on the sample surface. By
shifting the focusing lens vertically without
modifying the probe beam positioning, it is
possible to vary the distance between the
irradiated spot and the probe laser (see Fig.3a). Fig.3 : a) Principle of SRPD ; b) Principle of
Thus one can plot the energy Sty) dissipated along SAPD. The circles indicate the effect of spatial
the probe beam path versus the distance y from the averaging from individual spots at the surface

362
3.2 Principle of SAPD 10 0
In SAPD, the light source consists of a lkW
Xe arc larrp dispersed by a IOClnochrorrator. The a
outgoing IOClnochrorratic beam is ho=genized by
placing on the exit-slit a highly diffusive 10' 1
surface and then collimated, in order to get a >:
parallel pump beam covering the whole sarrple area. en
A 1 rom wide slit, longer than the sarrple, is
placed in front of the sample and aligned parallel
«i 10' 2
I:
to the probe laser. When using the slit, the IJl
energy sent by each spot out of the probe path, is
'iii
not taken into account (see Fig.3b): the PD signal c
D- 10. 3
Sl does not include the additional absorbance due
to non-local trapping. When reIroving the slit,
contributions from every irradiated spot, even at
few millimeters from the probe, are taken into
account : the PD signal S2 corresponds to the 10' 4
global absorbance. The non-local light trapping 0 2 4 6 8 1 0
efficiency is then deduced directly Y (mm)
(7) 0.005

Changing the wavelength of the incident light

gives the spectral dependence of the phenomenon. 0.004
>:
3.3 Sarrples
en 0.003
In both experiments, the chopping frequency «i
I:
is set to 13 Hz, and the deflection ·medium IJl
consists of ethanol. However the sarrple could not 'iii 0.002
be simply immersed in a quartz cell filled with c
the deflection medium. This would have strongly D-
affected the non-local trapping phenomenon by 0.001
changing the glass/air interface into a
glass/ethanol interface. Therefore a special
sarrple mounting has been built in which the sarrple
0.000
is used as the front wall of the PD cell, the
0 2 4 6 8 1 0
glass-side being exposed to the air and the film-
side immersed in ethanol. The probe laser beam is Y (mm)
passing behind the metallic back contact of the a-
Si:H film. Fig.4 : Normalized measurements of SRPD a) log
Two types of float-glass substrates with scale ; b) linear scale
textured Sn02 have been studied. There
characteristics are presented in Table 1. They 2,0
were covered by a single a-Si:H p-i-n structure #2
and then metallized.

Sarrple #1 #2 1,5
ot
glass thickness dg 1.1 mm 2mm ,...
0

~ 1,0
TCO thickness 700 nm 1200 nm
...
~

TCO texture height 170 nm 300 nm :or

0,5
TCO sheet resistance 6.S (l/o 9 WO
TCO absorbance AMl.S 6.9 % S.34 %
0,0 4-~~L.,--~--r"':::;:::=:;;::::::;i="';"-~-r-~"""':'-I
o 2 4 6 8 1 0
Table 1 (mm)

Fig.S : Radial distributions of absorbance derived

4. RESULTS AND DISCUSSION by Abel transform of Fig. 4b. The minimum peaks
before the thresholds at r = 2tan (8e • 01) xdg are due
4.1 SRPD measurements to numerical effects.
After normalization to the central absorption
peak, the measured absorption levels S (y) range
from SxlO-3 to 10-4 at the limits of the ring (see law. The good agreerrent between experimental A(r)
Fig.4a and 4b). The Abel transforms (Eq. (S)) curves and the theoretical curve of Fig. 2
indicates that the Lambertian diffusion law is a
corresponding to the radial distributions A(r) of
fair approximation if the angular dependance of
absorbed light are shown on Fig. S. A steep
increase of the absorption is clearly observed at the absorbance A(8) is not taken into account.
a distance y = 2 mm for sarrple #1 and y = 3.6 mm
for sarrple #2, exactly corresponding to 4.2 SAPD measurerrents
2tan(6e ,Ol)Xdg = 1.8xdg for each sample. At 8e,Ol' the The spectral dependence of the non-local
"glass/air" reflection coefficient abruptly trapping efficiency llnlT is shown in Fig. 6. The fact
reaches 100 %. For 8 > 8e ,Ol' the amount of that the results obtained at 647 nm by integrating
scattered reflected light decreases as cos 4 (8) the SRPD data fall on the curve of SAPD
according to Eq. (3) for a Lambertian diffusion measurerrents is an indication of the self-

363
consistency of the experimental approach. Three
observations can be made : i) as expected, 1lnlT is A 0.7 1 47~T'I (10)
Tl lT = uTln1TX 0.3 x 0.5 = . 'IniT
higher for the IOClre textured sample #2, ii) for
short wavelengths which cannot be reflected by the Thus for sample #2, the local trapping
rear metal contact due to the strong absorbance of efficiency at 800nm for which <l.1lnlT (A) = 9%
a-Si :H, 1lnlT is only due to diffuse reflectance at reaches 42% of the diffuse reflectance at
the TeO/ a-Si : H interface and a=unts to 2 to 3% the a-Si: H/metal interface. Moreover the
with little wavelength dependence, iii) in the red overall diffuse reflectance Rs in the glass
region above 600 run the increase of 1lnlT is IOClstly is roughly 21lniT= 24%.
due to the additional contribution of diffuse
reflection at the a-Si: H/metal interface. However 5. CONCLUSION
at 800 run where a-Si:H becomes transparent IOClSt
absorbance occurs in metal. Photothermal Deflection Spectroscopy has been
used to analyze the phenomenon of light trapping
14~--------------------------------' in a-Si:H p-i-n cell on textured TCO. Non-local
trapping due to total internal reflection at the
- 12
glass/air interface has been spatially resolved
~
10 and the corresponding trapping efficiency TlnlT has
>-
u
c been measured. Moreover it has been shown that TlnlT
8
Q) is 50% of the diffuse reflectance in the glass i.e

-~
Q)

C)
6

4
equal to the diffuse reflectance in the air.
Eventually the local trapping efficiency TllT in the
red region has been quantified with respect to
.E TlnlT
a. 2
a. mirror
...
as
I-
0 ACKNOWLEDGEMENTS
We thank E. Cornil of SOLEMS f'Jr providing
-2
900
samples and for expertise in TeOs. We also thank
300 400 500 600 700 800
J. Huc for technical help and fruitful discussion
Lambda (nm) about the PDS set-up.
Fig.6:Spectral dependance of absorbance due to
non-local light trapping measured by SAPD. The REFERENCES
points at 647 run have been obtained by integration /1/ R. G. Gordon, J. Proscia, F. B. Ellis and A.
of the SRPD curves of Fig. 4 . The lower curve E. Delahoy, Solar Energy Mater. ~ (1989) 263-
corresponds to SAPD measurements on a flat mirror 281
to indicate the typical precision and accuracy of /2/ H. Shade and Z. E. Smith, Appl. Opt. 2i (19)
the method. (1985) 3221-3226
/3/ J. Morris, R. R. Arya, J. G. O'Dowd and S.
4.3 Discussion Wiedeman, J. Appl. Phys. fil. (2) (1990) 1079-
The value of R", i. e. the total diffuse 1087
reflectance at a given wavelength, can be derived /4/ F. Leblanc, J. Perrin, J. M. Siefert, J.
from the value of 1lnlT . Indeed, llnlT is the ratio Schmitt and C.. Godet, Proceedings of the 5th
of the total absorbance for non-local trapping AnlT International Photovoltaic Science and
given by Eq. (3) to the absorbance at normal Engineering Conference, Kyoto, 1990, pp. 253-
incidence A(9=O). If one assumes A(a) constant 256
/5/ D. R. Mook, IEEE Transactions on Acoustics,
A(a) = A(e=O), then using Eqs. (2) and (3) Speech and Signal Processing, ~ (4)

f
Tl n 1 T = R s

o
PM
f (a)
3
Rg (a) cos (e)
-----'--'----"-'-----'------'---'--
8 1t
21t P dp (8)
(1983) 979-985

The integral in the left hand side of

Eq. (8) has been evaluated for a Lambertian
law and is equal to 0.46 for PM = 10. Thus
Rs is roughly twice the value of llnlT . This
means that non-local trapping represents 50% of
the diffuse reflectance.
In order to evaluate the local trapping
efficiency TllT in the red region, we can argue that
the observed increase of the non-local trapping
efficiency
~Tlnlr<A) = TlnlT (A >600run) - TlnlT (350-600run) (9)
corresponds to 50% of the fraction of thE: diffuse
reflectance at the a-Si:H/metal interface which is
transmitted into the glass via the a-Si:H/TeO
interface. This fraction corresponds to angles
smaller than 9c,13 = Arcsin (n1/n3) = 24° in a-Si:H
(see §2.1). If a Lambertian diffusion law is
assumed for the a-Si :H/metal interface, the
fraction of transmitted in the glass, normalized
to the total diffuse metal reflectance, can be
evaluated by a formula similar to Eq. (8), and a
rough estimation gives 30%. Eventually

364
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE S--12 APRll.. 1991 LISBON. PORTUGAL

PHOTOINDUCED EFFECTS IN HYDROGENATED AMORPHOUS SILICON

E. Fortunato, R. Martins, M. Vieira and L. Guimaraes.

Faculdade de Ciencias e Tecnologia da Universidade Nova de Lisboa
Quinta da Torre, 2825 Monte da Caparica, Portugal.

ABSTRACT: Photoinduced effects observed in ITO/a-Si: H/ITO

structures are studied through the analysis of the J-V curves,
in order to determine the role of interface in the structure
performances. At the dark, the J-V curves show an ohmic
behaviour, whilst under illumination, they are rectifying-like,
exhibing a photoinduced potential, caused by the difference in
hole and electron mobilities. As this photodiffusion effect is
controlled by the surface recombination velocity and by the
carriers diffusion length, an effective diffusion .length is
also determined.

1. INTRODUCTION
-
-
light _
In amorphous silicon solar cells, the
diffusion length (L) is an important
parameter closely related to the density of
states [N (EF) 1 of the intrinsic layer and
x
responsible for current limitations due to d
electrons and holes recombination. Since the
last years, a several techniques have been
developed to determine L, like Surface
Photovoltage (SPV) (1,2); Steady State
Photocarrier Grating Technique (SSPG) (3)
and Flying Spot Technique (FST) (4).
Besides the diffusion length, the Figure 1: Schematic diagram for discussion "o f the
surface recombination velocity (s) is also Dember effect
important, since it rules the solar cells
interface behaviour.
In this paper, we present results Now, taking into account that the
conc~rning a photoinduced effect surface-recombination velocity and that the
(photocurrent, JD and photovoltage, VD), current density for electrons (In ) is equal
observed in ITO/a-Si: H/ITO ohmic and opposite to the hole current density
structures. The photovoltage detected, is (Jp ) , we have:
called photodiffusion or Dember effect and
allow us to determine an effective diffusion In = leis M 1.
length (L' = L + s't'), where 't' is an
effective lifetime for holes and electrons. where ~n ~ ~p (high illumination level) is
Here, L' ~epresents the competition between the excess of generated carriers .
the average length that the carriers can The solution of the diffusion equation:
diffuse freely in the bulk and the
corresponding distance before being captured
by the interface states. 2.

2. FUNDAMENTALS is given by (6):

When strongly absorbed radiation falls

upon a semiconductor surface, the electron
~p(x) = =
GnL
D
3.
and hole pairs generated at one side of the
sample diffuse towards the opposite side. cosh [ (x-d) /Lj- (L/Lsd) sinh [ (x-d) /Ll
Since the diffusion coefficients of both (l+If/LsdLsO) sinh (d/L) + (L/LsO+L/Lsd) cosh (d/L)
carriers are different, a photovoltage is
generated in the direction of the
illumination. This effect is called where L = ~ D is the ambipolar diffusion
photodiffusion or Dember effect (5). coefficient;
D
Therefore, if we consider an intrinsic LsO s' at the illuminated
amorphous semiconductor with thickness d, surface (x = 0) and Lsd is the corresponding
irradiated by white light and transparent quantity on the opposite surface (x = d) .
ohmic contacts on opposite sides, and if the Since the current density is given by:
electron mobility OJ. n ) is larger than the
hole mobility (l1p), a VD potential is
observed, as shown in the figure 1.
J nx + J px = crE - lei (Dp - Dn)~ 4.

365
the corresponding Dember voltage, is
2e-i
obtained when J nx + J px = 0, giving:

I e I (Dn - Do)
Vo [~p (0) - ~p (d) 1 5.
(1

For simplifying case of d »

the L;
(l - 1 < d;
Ls 0 Ls d and Iln » IIp,
~p(O) - ~p(d) is obtained from equation 3, 0.5 1.0
leading to: V (V)

6.

where CPT is 2kT/q and (l the absorption

-2e-i
coefficient.
From the above equation, is expected an
hyperbolic behaviour of Vo as a function of
L' . Figure 2a): J-V characteristics in the dark for
It is also possible to determine the sample ai 104
Dember short circuit current density (Jo),
assuming that:
2
9E ~p(O) - ~p(d)
7.
dx d

In these conditions, we have:

-1. 0

8.

where Go is the photon flux density N

EO
(photons/cm2 s) and ~ the quantum efficiency. u
"-
~ -0.1 0.3
-6
3. SAMPLE PREPARATION I"J

The amorphous silicon specimens under -8

analysis were grown in a TCDDC (Two
Consecutive Decomposition and Deposition
Chamber) system. The details of the
preparation method are found elsewhere Figure 2b): J-V characteristic under illumination
(7,8). The a-Si:H films were sandwiched for sample ai104 . The insert shows a magnification
between ITO/ITO (Indium Tin Oxide) contacts, of the interested region of the J-V curve, were the
prepared by reactive thermal evaporation (9) Oernber effect is clear noted
with a thickness around 1000 A and with 85%
of transmittance, in the visible range.
The films thickness varied between 1 and The obtained L' values (through
4 11m and were measured by a Sloan Dektak equation 6) reflect the influence of Land s
apparatus. on film properties. In Table I we summarize
the VD, JD and L' values, for samples
prepared at different deposition conditions.
4. J-V CHARACTERISTICS We also show the corresponding N(EF)
inferred through Space Charge Limited
The J-V characteristics were measured by Current (8).
a Semiconductor Parameter Analyser 4145 HP,
at room temperature, in the dark and under
illumination with a ELH lamp (1900 W/m2) .
In figure 2 a) and b), we show the Table I
recorded J-V curves in the dark [a) 1 and
under illumination [b) 1, for a particular
sample. In the dark, we observe an ohmic Vo Jo L' N(EF)
behaviour of J-V whilst under illumination, Sample (mV) (A/cm2 ) (run) (eV-1cm-3)
the curve shape is no longer ohmic and the
detected photovoltage is in the direction of
the incident light. The observed J-V curve (li102 200 1 . 8e-4 130 5X10 16
shape, under illumination, is attributed to
(li103 250 3.2e-4 104 3x10 16
the role of the surface/interface states.
The rectifying behaviour is explained (li104 140 1. 5e-3 186 8X10 16
by the appearance of a dipolar like-layer (lil06 100 4.8e-5 260 4x10 16
due to carriers that are captured and charge
the surface interface states. Thus, a n+p- (li117 90 7.7e-5 280 2X10 16
like junction is formed in the direction of (li118 70 1.7e-5 380 8XI0 15
the incident light, whenever s is
significantly.

366
 In figure 3 we plot the dependence of VD Thus, high quality devices are ascribed
on inferred L', for the different specimens with structures showing s < 10 3 cms- 1 and L'
under analysis. Such data are well fitted by larger than 200 nm.
a hyperbolic function, as ~redicted by
equation 6. From this figure, we observe
that VD and L' have reverse behaviours, 6. FUTURE PROSPECTS
showing that if s -+ 00 VD -+ 0, whilst, when
s -+ 0 no rectifying behaviour is observed As demonstrated in this paper, the
and L' = L. For intermediate values of s, Dember effect can be a powerful tool to
its influence on film properties is infer the diffusion length and the surface
determined by N(EF) as will be discussed in recombination velocity.
the next section. So if we take into account that VD is a
function of a-I (see equation 6) by changing
300 , . . - - - - - - - - - - - - - - - , the light wavelength, such plot gives a
straight line which slope and the intercept
at VD = 0 are related with Land st'. This
work is now under progress.
o
:> 200
ACKNOWLEDGEMENTS

The authors wish to express their thanks

100
to N. Carvalho for preparing the specimens.
The financial support by the Commission
of the European Communities (project
contract nQ JOUR-0077-C (TT)/OASIp project)
and Portuguese national R&D programme: Junta
0L--~-...1---'----'--~----'
Nacional de InvestigacAo Cientifica under
100 200 300 400 Eureka project nO 366 and Instituto Nacional
de InvestigacAo Cientifica/Centro de Fisica
L' (nm) Molecular das Universidades de Lisboa is
gratefully acknowledged.
Figure 3: Plot of Vo as a function of L'. The
fitting follows a hyperbolic function, as predicted
by equation 6 REFERENCES
(1) A. R. More, Appl. Phy. Letters 40, 403
5. DISCUSSION AND CONCLUSIONS (1982) .
(2 ) A. R. More, J .Appl. Phy. 56, 2796
The figure 4 presents the data of N (EF) (1984) .
versus L' for the set of samples indicated (3) D. Ritter, E. Zeldov, and K. Weiser,
in Table I. Appl Phy. Letters 49, 79 (1986).
(4) R. Martins, M. Vieira, E. Fortunato, I.
Ferreira, F. SoareS and L. GuimarAes,
10 17
M
I • presented at the MRS Spring Meeting, S.
Francisco (1990).
g
• (5) H. Dember, Physik. Z. 32,554-856
....1
••
()
(1931) .
:> ( 6) K. Seeger, Semiconductor Physics - An
~ Introduction, 4th Edition, Springer
10 16 Verlag (1989).
~
~
• (7 ) R. Martins, L. GuimarAes, E. Fortunato,
A. Macarico, M. Santos and I. Ferreira,
:z 8 th ECPSE Conf. Florence, 653-660
(1988) .
(8) E. Fortunato, A. Amaral, L. Rodrigues,
N. Carvalho, L. Guimaraes and R.
10 15 Martins, Technical Digest of the
0 50 100 150 200
( 9) C. Carvalho, O. Figueiredo, E.
L' (nm)
Fortunato, R. Martins, A. Macarico and
Figure 4: Dependence of N(EF) on L'
L. GuimarAes, 8 th ECPSE Conf., 801-805
(1988) .

There, we observe as a normal trend that

L' is high for samples having a low N(EF) .
Nevertheless, for L' values lower than
100 nm, the unexpected behaviour shown, can
be attributed to the increasing role of the
interface states. From these data, together
with those ones referred in figures 2 and 3,
we conclude that the Dember effect is not
only sensitive to the surface recombination
velocity, but also to the diffusion length
of the carriers. So, for s < 10 3 cms- 1 (as
can be estimated by JD) L' is mainly
controlled by L, being VD an easy way to
infer the diffusion length of the carriers.

367
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

INFLUENCE OF THE RF POWER DENSITY ON THE ELECTRICAL PROPERTIES OF

GLOW-DISCHARGE AMORPHOUS SILICON

A.Amaral, L.Rodrigues*, L.Guimaraes*, R.Martins*

Instituto Superior Tecnico da Universidade Tecnica de Lisboa
Av.Rovisco pais, 1000 Lisboa, Portugal.
*Faculdade de Ciencias e Tecnologia da Universidade Nova de Lisboa
Quinta da Torre, 2825 Monte da Caparica, Portugal.

D.Mencaraglia, Z.Djebbour, J.P.Kleider, C.Longeaud.

Laboratoire de Genie Electrique de Paris (URA D0127-CNRS)
Universites Paris VI et Paris XI, Ecole Superieure d'Electricite
91192 Gif-sur-Yvette Cedex, France.

ABSTRACT. The optimization of the devices efficiency and their

costs is an important request to promote the application of
amorphous silicon thin films devices on industry. In this
perspective, it is interesting to try to increase the growth
rate of the intrinsic layer, without degrading the electro-
optical properties of the materials. This possibility was
studied varying the RF power density in an in-home glow
discharge system since this is an easy controlled production
parameter. Being the materials properties a consequence of
each specific reactor, we want to determine the highest RF
power density, in our case, leading us to the production a
good quality material. The photoconductivity of the samples
was measured and informations on the deep gap states
distribution were inferre<;\ from space charge capacitance
measurements.

1. INTRODUCTION our a-Si:H layers (ranging between 1 and 3

microns) obtained this way have been found
Aiming a good application of amorphous in good agreement with the values obtained
silicon for industrial purposes, either in from direct measurement using a Sloan
sO,lar cells or in the microelectronics Dektak surface profiler.
field, it is economically interesting to The capacitance measurements C(V,T,ro)
increase the growth rate (r) of the undoped were performed using a 5208 PAR lock-in
layer, without degrading the material amplifier, in the temperature range
properties. 310-420 K, using a set of frequencies from
We have studied this possibility in 15 Hz to 7000 Hz. The amplitude of the
order to, optimize the deposition RF power modulation ac signal was around 25 mv rms.
density (dp), which is an easy controlled
production parameter. To study the The DOS at the Fermi level N (EF) is
evolution of the electro-optical properties determined from the dependence of the
of the material with the RF power density capacitance on temperature and frequency at
of the plasma deposition, we used two types zero or reverse dc bias using a procedure
of experiments: i) capacitance versus bias, proposed by Lang et al (1). A plot of
temperature and frequency C(V,T,ro) from C2(dC/dT)-1 versus T is performed at
different frequencies. In a given range of
which we determine the density of states temperature, this plot, theoretically, is
(DOS) near the Fermi level and the DOS expected to obey the following relation:
integral between midgap and Fermi level
(I *) and ii) photoconductivity
measurements. C;:2 (dC/dT) -1 = (eS/LD) (T-To) (1)

With e the dielectric constant of the

2. EXPERIMENTAL material, S the cross-section of the device
and LD the Debye length associated to the
Undoped a-Si:H layers were deposited localized states at Fermi level, which is
over lTD coated pyrex glass in a RF glow given by:
discharge PECVD system. The ITO (Indium-
Tin-Oxide), 1000 A thick, was produced by
reactive thermal evaporation. The (2)
deposition conditions of the a-Si:H films
were the following: pressure = 100 mtorr To(ro) is a turn-on temperature above which
and substrate temperature = 200°C. The the localized states at the Fermi level
electrode geometry of the PECVD system was begin to follow the ac modulation at the
a grid one and the RF power density was
varied from 1.6 mW cm- 3 to 25 mW cm- 3 • The angular frequency, ro. In the samples
Schottky contacts were achieved by the investigated in this paper, we effectively
deposition on the a-Si:H film of an observed a well defined linear region in
electron-beam evaporated Pt layer. The the C2 (dC/dT) -1 versus T plot, from which we
sample thicknesses were deduced from the could easily derive the value of N(EF). An
geometrical capacitance of the Schottky example of such a plot is given in
diodes measured at room temperature with Figure 1.
high frequency signal (50 KHz), assuming
that the value of the a-Si: H dielectric
constant is the same as that of crystalline
silicon (e = 10- 10 F m- 1). The thickness of

368
 10-3 (roughly one order of magnitude) of the DOS
I'e;
from EF to Ei *. This increase in the deep
.... t> gap states density may be related to the
'e: stress in the films produced by a more
~'"
'0"- ..... .. ... ..
energetic bombardment during the film
deposition with the highest
densities. This could be origin of the
power
... ...
10- 4
.. peeling of the films when power density is
further increased (above 25 mW cm- 3 ) •
TABLE II-Comparison of the calculated Ie with
measured values of r*

Sample dp N(Ep) Ec-Er I" Ie

10-5L-____~______~______~____~ (roW cm-3 ) (cm- 3 eV-1 ) (eV) (cm- 3 ) (cm- 3 )

o 10 20
dp (mW cm -3) ClilOS 1.6 1.8e 16 0.64 4.2e 15 3.2e 15
aillS 14.5 3.6e 16 0.15 1.1e 16 2.3e 15
Figure 4: Photoconductivity (Gph) versus the power ail19 24.9 2.2e 16 0.11 3. ge15 1.0e 15
density
When analyzing these data we observe
that all the samples present a good 4. CONCLUSION
photoconductivity. (Cfph) around 10- 4 Q- 1 cm- 1 ,
when dp varies from 1 to 15 mW cm- 3 • The The present data show that the DOS
samples produced with higher power around the Fermi level as well as the
densities could not be studied due to photoconductivity values, at room
peeling off. The fact of these parameters, temperature, practically remains .unchanged
when the power density increases up to
N(Ep) and Cfph remain unchanged in this range 25 mW cm- 3 pointing out the possibility to
of power densities seems to point out that obtain, at greater power densities, a good
there is no relevant modification in the quality material. Unfortunately a film
density of states in the neighbourhood peeling occurs to power densities values
of Ep. greater than around 20 mW cm- 3 precluding
However, analyzing the data of the use of higher power densities to
Figure 5, we see a significant shift in the achieve faster deposition rates than around
position of the Fermi level inside the 2 A S-l in a classical standard glow
mobility gap. Indeed, the activation discharge system. This peeling is probably
energy, increases by 0.25 eV when dp related to internal stress in the material
increases from 1.5 to 25 mW cm- 3 • occurring at high power densities. A
solution to eliminate this stress is to
promote the dilution of the silane in
0.8r-----------------~------_;
~ helium as referred by some authors in this
Q) conference (5).

ACKNOWLEDGEMENTS

• The authors would like to express

0.6 their thanks to N. Carvalho for the help
given during samples preparation and
E .Caristan for the deposition of the Pt
• rectifier contacts.
0.5L---~--~--------~------~ This work was supported by a
o 10 20 30 Cooperation R&D programme between Portugal
(INIC-Instituto Nacional de Investigac:;:ao
dp (mW cm -3) Cientifica) and the French Embassy, and by
the Comission of the European Communities
Figure 5: Variation of the activation energy CEC ( project contract nO JOUR-0077-C
versus the power density (TT)/OASIS project).

This suggests that the density of

states N (E) should undergo some REFERENCES
modifications. This is confirmed by the
evolution of N(E) between EF and Ei* when (1) D.V.Lang, J.D.Cohen and J.P.Harbison,
inferred from the comparison of N(EF) and Phys.Rev.B25, 5285 (1982).
I* in the same samples. Indeed, if we (2) J.P.Kleider, D.Mencaraglia and
assume a constant density of states between Z.Djebbour, Journal of Non-Crystalline
E F and Ei * w·e can calculate the Solids 114, 432-434 (1989).
corresponding integrated DOS Ic. The (3) M.Tanelian, H.Fritzsche, C.C. Tsai and
calculated values of Ic are compared with E.Symbalisty, Appl. Phys. Lett. 33,
the measured values of integrated DOS I* in 353 (1978).
Table II. (4) I.Solomon, Fundamental Physics of
For the smallest power density, we can Amorphous Silicon, Spring-Verlag, 33
see that the assumption of a constant DOS (1981) .
between EF and Ei * is .well satisfied. On the (5) P.Roca i Cabarrocas, R.Vanderhaghen,
contrary, the comparison of the results Y.Bouizen, M.L.Theye, D.Mencaraglia, Z.
obtained for the two highest power Djebbour, J.D.Sib, J.P.Kleider,
densities suggests a significant increase C. Longeaud, to be presented in this
conference.

369
 1. 5e-7
2 between the intrinsic level (Ei·) close to
~ slope 2.05 nF midgap and the Fermi level, is directly
';' N(E F ) derived from the slope of this linear
~

E-< region, which is independent of the

"0
...... measurement frequency in a given range.
U The photoconductivity measurements
"0 were performed on coplanar geometry with
'"U ohmic aluminium contacts, using a sun ligth
1. Oe-7 simulator (= 1000W/m 2 ) with the samples
placed in a vacuum system dynamically
pumped down to 2.10- 6 torr. The a-Si:H films
were grown over alkaly free glasses
(AF 45), preventing interface problems
(3,4) .
The Tauc optical gap, obtained from
direct optical transmission measurements,
in the visible range, was around 1.63 eV
5.0e-8 for all the investigated samples.

3. RESULTS AND DISCUSSION

Figure 3 shows that the growth rate
varies within a factor 3 for the
investigated dp range: the growth rate
remains small, always below 1.6 A.S-1. The
N(EF) values is nearly constant as shown
320 340 360 380 400 420
Table I.
T(K)
2.5
Figure 1: Typical plot of C2(dC/dT)-1 curves versus .,
temperature. The slope of the linear region give ......
...:
us the DOS around the Fermi level

Note that the region of linear

. 2.0

variations is shifted towards higher •

temperatures with increasing frequencie.s, 1.5
while the slope remains unchanged, as •
expected from relation (1). From the • •
dep~ndence of To on the measurement angular 1.0 ••• •
frequency (ro = '\)oexp (-~E/kTo», we can
determine the value of the activation •
energy of the material, ~E = Ec-EF. We show 0.5
in Figure 2 the results obtained from the 0 10 20 30
curves of Fiqure 1. dp (mW cm -3 )
Figure 3: Growth rate versus the power density
N '
;I;
8
TABLE I-Geometrical and electrical characteristics
of.the investigated samples at room temperature

Sample t dp r N(EF) Ec-EF IJph

(,.,.1 (mW cm-3 ) (A/s) (cm-Je~l) (eV) (O-lern- I )

ailOS 2.23 1.6 1.0 1.8e 16 0.64 1. 7e-4

5
ailOe 1. 43 2.4 0.7 1.De ' • 0.60 1. 7e- 4
ail06 1. 49 3.2 1.0 1.1e ' • 0.58 7.1e-S
4L-~L-~~~~~-L~-L~~~~
ail16 1.44 4.8 0.9 2.7e 16 0.63 1.1e-4
3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 uil15 1. 94 6.4 1.2 1.5e 16 0.54 1. ge- 4
1000/To (K -1) ailO? 1. 33 8.0 1.3 4.4e 16 0.68 1.2e- 4
ai117 1.30 9.6 1.0 1. 6e ' • 0.72 1. 8e-4
Figure 2: Example of a Arrhenius plot of the 1. 5e- 4
ail18 3.27 14.5 1.1 3.6e ' • 0.75
measurement frequencies versus 1000/To. To is the
value when the linear region, from the curves of ai119 2.96 24.9 1.6 2.2e l • 0.77 --
C2(dC/dT)-1, intercepts the temperature axis.

The capacitance dependence on the

applied reverse bias, C (V, ro), is plotted in
tpe form of [dIes/C) /dVj-2 versus the This point is in agreement with the
applied bias (V) for each measurement photoconductivity results measured for the
frequency. Such a plot has been shown to be same samples as shown in Figure 4.
linear (2) in a given range of the applied
bias. The integral of DOS:
EF
I* = J N (E) dE (3)
Ei'

370
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Temperature dependence of the performances of a-Si:H solar cells

Luigi Schirone, Raffaella Vincenzoni

Dipartimento di Elettronica, Universita "La Sapienza", Via Eudossiana 18,00184 Roma

ABSTRACT. The temperature dependence of the performances of a-Si:H

pin solar cells is investigated by the analysis of Current-Voltage (I-V)
characteristics under standard AM1.5 illumination, of a cell exposed to a
prolonged irradiation and of the same cell after one hour annealing at 140
°C . The experimental results are related to the variations of parameters 11.
and 't by a model describing in a simplified way the operation of the
device.

1. INTRODUCTION variations of other parameters, being Rs the

most effective factor.
A temperature variation is a simple
experimental method to achieve a 2. EXPERIMENTAL RESULTS
modification of the physical parameters
affecting the operation of a semiconductor 2.1 OPEN CIRCUIT VOLTAGE
device. In this work we investigate the In fig. ! the experimentally detected open
temperature dependence of the performances circuit voltage is reported as a function of T,
of a-Si:H pin solar cells by the analysis of for the sample in condition a) and b).
Current-Voltage (I-V) characteristics under
standard AM1.5 illumination. The 0,82 -r"I::------------....,
experimental results are related to the a Voe 1. soaked
0,80 • Voe armealed
variations of physical parameters 11. and 't by
a model qescribing in a simplified way the 0,78
operation of the device. The comparison of
the achieved experimental results to the ;;; 0,76
behaviour predicted by simple relations, '-"
~
even if. approximated, allow us to get a >o 0,74
certain physical insight of the weight of
different parameters on the determination of 0,72
the global performances of these devices.
A characterization was perfom1ed of
the cell in different operating conditions: 290 300 310 320 330 340 350 360
a) after a prolonged exposition to T (OK)
AM1.5 irradiation, at room
temperature
b) after one hour annealing at 140°C. Fig.!: The experimentally detected open
The open circuit voltage Voc, the short circuit voltage is reported as a function of
circuit current Iscand the series resistance temperature, for the sample in condition a)
Rs are taken in account. The modifications light soaked and b) annealed. A linear
of the efficiency and of the fill factor of the behaviour is observed.
solar cell result by a combination of the

371
 A linear dependence on the T is effective lifetime and by the relations
reported, and a straight line is reported for developed for the crystalline semi-
better comparison. Moreover we can conductors.
observe that the exposure to the light causes
a temperature-independent reduction in open 16~---------------------,

circuit voltage.
The observed behaviour can be
predicted by simple considerations based on
general equations describing current flow in
semiconductors.
The open circuit voltage Voc is
defined by the condition J = 0
After a substitution of this condition
in the expression of the current in a 6 annealed
semiconductor, we achieve the following theoretical
expression for the electric field: 13 ~~~~~~~~~~~~

290 300 310 320 330 340 350 360

(x) = -DnV'n + DpV'p
Eoc ~n + JlpP T (OK)
Fig.2: The experimentally detected short
Jln Vn +Jlp Vp = VT Ke (x·)
__ '(JT-
Y' '---'----'---'----'-- circuit current is reported as a fUllction of
Jlnn + JlpP temperature, for the sample in condition a)
The voltage present between light soaked and b) annealed. A behaviour

f
the electrodes can be calculated as: like Jsc (T) = Jsco TO.5 is observed.

V"" =Vbi - VT K.,dL An expression for the current

flowing in the solar cell as a function of
where Vbi e Ke(x) are weakly physical parameters can be found solving
affected by temperature variations. the continuity equation in the -i- region.
Following one of the methods adopting the
2.2 SHORT CIRCUIT CURRENT regional approximation (see for example
In fig. 2 the short circuit current [3]), it is possible to express the carrier
exhibited by the sample is shown at densities nand pas:

~ + N2e~ +f Gn(1..)e-a(x). dA
different temperatures. The experimental
trend can be fairly well approximated by the ~
n (x) N, e-
continuous curve reported in the picture and
expressed as:
J~ (T) = Jsc (300) ( 360 )0.5. p (x) ~ P, e- r';, + P2 e {p +f Gp(1..}e- a(x), dA
As a starting point of our analysis of
this experimental behaviour, we observe
that, in the considered interval of The derivatives of these functions
temperature, the variations of both of Jl and can be expressed in the form:
't can be described by a TO.S law. This n' (x) = ~ fn (x) ,p' (x) = ~ fp (x)
behaviour is described for extended states
mobility in the work by Spear [1]. We being Ln = YDn 'tn , Lp = YDp 'tp
achieved the carriers' lifetimes and their The achieved expressions can be
dependence on temperature by a numerical

+ +n
adopted to write the total current as:
analysis of the recombination rate based on J t = q E (Jlnn + Jlpp) +
the dynamics proposed in [2]. Moreover,
we observed that 't is not strongly affected q (VTP·5 [ ~ fn (x) fp (X)]
by the amount of carriers' excess: this
If we suppose that the functions
feature allows us to describe the behaviour n(x), p(x), fn(x) and fp(x) are weakly
of the amorphous semiconductor by an
affected by the temperature, we can attribute

372
 all the modifications occurring in J sc to Il, 't The dependence of Rs on
and VT. Taking in account the proper tenwe:ature cannot be explained only by the
dependences of these parameters on the vanatlon of these parameters: RTCO is
temperature, this equation gives a function characterized by a value as large as a few
Jsc (T) = Jsco J'l.5 Ohms and is not affected by T while the
contribut~o~ ?y. the exponential 'part of the
as experimentally detected. charactenstlc IS mcreased by an increment of
T. vye propose to interpret the
2.3 SERIES RESISTANCE expenmentally detected reduction of Rs as a
The series resistance of a solar cell is contribution by the resistivity under
defmed as: illuminati(:>ll of the intrinsic layer, that is
exponentIally reduced when T is
Rs=I~~lv=voc increased.
The experimentally detected behaviour for 18 .---------------------~
Rs at different temperatures is shown in fig.
3: a l<l!ger value and a stronger dependence 16 A
on T ~s .shown by the cell in light soaked
COndItIOn than in annealed 14
condition. ---S
fI.)

..c: 12
14 . - - - - - - - - - - - -..... 0
'-'
fI.) 10
e&:: a:Si-H cell
13 8 intrinsic layer
..-.. theoretical
5 12 6

-
..c: 0 20 40 60 80 100 120 140 160 180
0 x soaked
•
11 t (min)
f I.)' annealed
e&:: Fig.4: The series resistance Rs is shown in
10 ~o~p~rison wit~ the resistivity of an
mtnnSlC film~ at di~fe~ent times of exposition
to AM1.5 lrradiatlOn. The same time
290 300 310 320 330 340 350 360 dependence is observed.
T (OK)
Fig.3: The experimentally detected series A confirmation of the weight of the
resistance is reported as a function of last contribution can be found by
t~mperature, for the sample in condition a)
ex aminating the behaviour of the solar cell
lIght soaked and b) annealed. A decreasing during the prolonged exposition to an
behaviour is observed in both of the curves ~Ml,5 irradiatio~. In addition, in fig. 4 the
with a stronger dependence on T shown i~ tIme d~pe~de~ce .IS shown of the resistivity
a) condition. of an mtnnslC fllm, deposited under the
~ame conditions of the -i- layer of the cell. It
This behaviour can be described IS possible to observe that both of the curves
modeling the operation of the cell by a can be approximated by a time dependence
t
current generator Jsc in series with a diode R = Ro + Rl
e-Ii"
and a Rs' resistance. The value for Rs is the
result of a number of contributions: the one where B = 100'.
given by the exponential characteristic A further confirmation of the validity
typical of the diode is increased by the of this model can be taken by the
contribution given by the resistance of observation th~t Rs. is more strongly
contacts, particularly by the resistance dependent on T m a lIght soaked cell: this
be~aviour can be associated to a bigger
associated to the interface between the a-Si
and the TeO film operating as a superior weIght of the contribution of the series
electrode. resistance, with respect to the annealed state.

373
 ACKNOLEDGEMENTS REFERENCES
We wish to acknowledge prof. F.P. [1] W.E. Spear, in "Amorphous silicon
Califano for continuous encouragement and and related materials", pp. 721-765,
prof. F. Palma and G.P. de Cesare for 1988 WSPC
providing the samples. [2] M. Hack, S. Guha and Shur, Phys.
Rev. B, .N , 12, (1984)
[3] F. lITera, F. Palma, to be published
in Solid State Electronics

374
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON, PORTUGAL

ELECTRONIC TRANSPORT PROPERTIES OF HIGH DEPOSITION RATE a-51: H nATERIAL

T. Mohammed-Brahim,A. Rahal,N. Ababou,N.Beldi,M.Aoucher

Laboratoire des Couches Minces et Semiconducteurs-U.S.T.H.B.
BP 32 EI-Alia Bob-Ezzouor, Alger,Algeria

D.Mencaragl i a,C.Longeaud,J.P .KI ei der,O.Glodt.Z.D j ebbour

Laboratoire de Gimie Electrique de Paris (URA 00127 CNRS)
Ecole Superieure d'Electricite-Universites Paris VI et XI-91190 Gif/Vvette France

ABSTRACT. Electrical transport properties of undoped a-Si : H films, prepared at high rate
(10 Als) by dc magnetron sputteri ng in an argon I hydrogen di scharge, are studied usi ng
various characterization techniques: Capacitance - Temperature - Frequency (C-T-w),
Capacitonce- Bias (C-V), Pulsed Thermally Stimulated Current ( PTSC), Time of Flight
(TOF).
A density of states (DOS) at Fermi level Ef of 4-5.EI6 eV-l cm-3, a DOS integral between
intrinsic Fermi I evel and Ef of 3-4.E 16 cm-3, a conduct ion band tail charateri st ic
temperature Tc of 280 K obtained on first a-Si : H samples produced, indicate that the
material presents relatively good electrical properties. Such an interesting deposition
method, faster than currently used techniques, may be lead to further developments.-

1. INTRODUCTION 2. EXPERIMENTAL DETAILS

Industrially, the film deposition rate is an 2.1 Samples preparation

important parameter in the use of Hydrogenated Undoped a-Si : H films are prepared by high
amorphous silicon for solar panels production. rote de magnetron sputtering. The apparatus
High electronic Quality a-Si : H ftIms are cur- used is described in detail elsewhere (4). An
rently produced by glow discharge decompOSi- undoped Sl1lcon target (99.99995 purity) is
tion of Silane. However the deposition rates, sputtered in dc discharge by an argon/hydrogen
in this process, are relatively low (aroundlA/s). mixture. Both gases are 5N purity. Prior to sput-
This is why some effort has been done in order teri ng the chamber was backed to about 100 C
to increase this deposition rate. Usually it is and degased to a pressure less than 4E -4 Pa.
difficult to accomplish both high deposition The deposition conditions where chosen in order
rate and high Quality film. But the use of a very to give samples with good electrical properties.
high -frequency glow discharge deposition tech- Some of these conditions are:
ni que (1) or He di I uti oni n RF glow di scharge (2) Td = 260 C P = 95 W
has showed the possibility to obtain good P (Ar) = 0.6 Pa P(H2) = 0.08 Pa
Quality films. Films, 1.5 J.l-m thick, have been deposited on
Good Quality films presented here are obtained three kinds of substrate : fused quartz plate,
by dc reactive sputtering of a Silicon target. crystalline silicon wafer and carbon substrate.
This is another promising way (3). Its advantage The deposition rete has been calculated from
is the ability to uncouple the two source mate- the ratio of the measured film thickness to
rials, Silicon and Hydrogen, and to optimize deposition time. The hydrogen content has been
independently the role that each plays in the measured by ERD and the optical constants have
film growing and the resulting electronic been determined from transmission measure-
properties. This type of control is not possible ments in the 0.4-2.4 J.l-m range.
for glow diSCharge decomposition of Silane The results show a deposition rate of 10 A/s, on
where the source materials are initialy bonded optical gap of 1.9 eV and a hydrogen content
in the Silane. of 18 %.
The electroniC properties presented in this work Schottky diodes were processed on these
deal principally with the deep gap states materials to perform the electrical measure-
characteristics. They were determined from dc ments made in a sandwich configuration. These
current - voltage, capacitance versus reverse samples are Pt/a-Si : H/Cr staggered structures
bias, temperature and frequency C (V,T,w ), deposited on a glass substrate. The semi-
Pulsed thermally stimulated current (PTSC) and transparent platinium film ( 200 A thick) was
Time-of-Flight (TOF) measurements on Schottky electron-beam evaporated and then annealed at
diodes. 160 C to form the Schottky contact.

375
2.2 Electrical and optical characterization: The PTSC experiment is a new method which has
Prior to the characterization of the gap been developed in order to obtain the DOS
stotus, the conducticity and photoconductivity profile above the Fermi level, up to around 0.25
of the f11 ms at room temperature are measured eV below the conduction band edge Ec. It is
in a gap configuration. The photoconductivity based on the conventional TSC experiment, but
is measured under 10 mW/cm2 white light. repetitive short voltage pulses are applied to
For the electrical characterization of the the sample whl1e heating, and the resulting
gap states, four different types of experi- transient currents are monitored. Compared to
ments have been performed on the same set of the conventional TSC, the experimental condi-
Schottky diodes. These are dc current-voltage tions are closer to those assumed in the
measurements, capacitance versus reverse analysis: in particular, during the pulses the
bias, temperature and frequency C(V,T,w), sample is submitted to a high electric field and
Pulsed Thermal1y Simulated Current (PTSC) retrapping of the emitted charges is reduced.
and Time of Flight (TOF). Details of this technique have been published
Four different types of experiments have elsewhere (7).
been performed on the same set of Schottky The TOF technique is another kind of
diodes. These are dc current-voltage measure- experiment used to characterize the transport
ments, capaCitance versus reverse bias, tem- properties in amorphous semiconductors. In this
peratue and frequency C(II,T,w), Pulsed Ther- experiment performed on Schottky diodes,
mally Stimulated Current (PTSC) and Time of photocarriers are created at the Schottky
FI i ght TO F). electrode by a laser pulse and the transient
From the dc current-voltage I(V), we have photocurrent corresponding to the drift of
derived the classical electrical parameters of electrons (which are collected at the back
the structure, i.e. the saturation current den- electrode by applying a bias pulse prior to the
sity Js and the bamer height Cl>b of the laser pulse) is analysed. In addition to the
Ptia-Si:H contact. Js is derived in the usual determination of the drift mobilitY,it is
way from the exponential region for direct possible to derive a DOS profile from this
biases (V < 0.4 volt). To get easily an order of experiment. Indeed, from the pre-transit part of
magnitude of Cl>b, we assume the current to be the transient photocurrent (i.e. before the
thermioniC emission limited and deduce <l>b arrival of most of the photocreated electrons at
from the classical formula: the back electrode)we can deduce the shape of
Js = AT E2 exp (Q<l>b/kT) the CBT ( 8 ), whl1e the shape of the DOS
taking for Js the extrapolated value as indica- corresponding to deeper states in the gap can.be
ted above and using A= 120A cm-2 K-2, the obtained from the analysis of the post trans1t
value of the Richardson constant in crystalHne port of the photocurrent (9).
silicon (note that the value of Cl>b is not very Optical determination of the valence
sensitive tothe precise and unknown value of band-tail Urbach parameter Eo and the defect
A). density Nd between near 1 eV ond near 1.5 ell
From the capaci tance measurements we from the conduction band edge, Is performed by
can obtain some informations on the deep gap Photothermal Deflection Spectroscopy (PDS)
states characteristiCS, namely the DOS near measurements.
the Fermi'level N(Ef) and the integrated DOS 1*
between midgap and the Fermi level. The DOS
near the Fermi level N(Ef) has been derived 3. RESULTS AND DISCUSSION
from capacitance versus temperature and fre-
Quency . measurements C (T,Cal) for a fixed A dark conductivity of 3.E-8/0hm/cm and a
reverse bias of the Schottky structure, using a photo-dark conductivity ratio of 2.E3, are
procedure described elsewhere by Lang and obtained. The activation energy of the dark
al.(5). In this procedure, N(Ef) is directly conductivity is 0.65 eV.
related to the slope of a C2/(dC/dT) versus From the 1(11) characteristiCS, an example of
T plot, theore- lically expected to be linear which is presented in Figure 1, we have derived
deri ',led from the C(T ,Cal) measurements. The the following typical values of the electrical
DOS integral 1* is deduced from a new parameters of the Schottky structure:
treatment of the C-V measurements on Js = 4.8E-10 A cm-2
amorphous semiconductors, recently proposed <l>b = 0.95 eV.
by some of the authors (6). This integral is The treatments of the capacitance measure-
determined from the slope of the ments are presented In Figures 2 and 3 respec-
(d( 1/C)/dV)E -2 versus V plot, 'Nhich has tively. As expected from the theories cited in
been shown to be linear when the diode is the preceding section, we observe well defined
reverse bi ased. linear behaviors from which we can derive the

376
DOS 1*. The aven~ge values of N(Ef) and 1* 6re The DOS profile obtained with the PTSC
equal to 5. E16 cm-3.eY-l 6nd 4.E 16 cm-3 technique is shown in Fig. 4. We found that the
respectively. charactenstics temperl!lture Tc of the deepest
states in the conduction bond t6il (CBT)
(corresp'onding of energies around 0.25 eY
below Ec ) is about 260 K. The DOS profile at
deeper energies is found without significant
structure down to 0.5 e'l below Ec. However, a
little bump centered at 0.S6 eV below Ec
le·06 o reverse seems to point out, with a DOS maximum value
equal to 1E17cm-3.e'l-1. This is also sustained
by the DOS value at the Fermi level found from
1e·07 capacitance measurements (S.E 16cm-3.eV-l at
0.65 eV below Ec}.
1e·OS
The yalue of the electron drift mobility
deduced from TOF at room temperature is
le·09 0.23 cm2 V-I s-1 . the DOS profi I e obtai ned
from the combination of pre- and post-transit
1e.1 0 1..cL.:u.LJlJ...L.UlJ...L.U....u..J.....u..J.-Li..Ll."-'-'--LLLLLllLJ measurements is also shown in Fig.4. It can be
o 0.20 0.40 0.60 O.SO 1.00 seen that the gener61 shape of the DOS is in
good agreement with that obtained from PTSC;
BIAS (V) in particular, the value of Tc is confirmed to be
around 280K.
Figure 1 : Typical current denlity J versul The PDS measurements show a valence band-
the applled bial V tail Urbach parameter Eo of 64 meV and a
defect density Nd of 7.E16 cm-3.
The results are summarized in Table I.
-1 -3
F (Hz) N(Ef) (eV em )

a ... 45570 4.88 10 16

b .... 18230 4.22 10 16

-3
c .... 7292 4.99 10 16 F ( Hz ) I·e em )

d - 2917 5.34 10 16 a ..... 187 3.59 • 10 16

b ..... 467 4.52 • 10 16

c .... 1167 4.34 • 10 16

v
[ C 2/ C dC / dV ) ) 2 ( a .• u )

.= J E
e

.
I geE) dE
,
(IJ
E•

• • • •
•
TEMPERATURE (K)
POLARIZATION (VJ
310 350 390
-3 -2.5 -2
Figure 2 : Treatment 01 the capacitance Figure 3 : Treatment of the capacitance
C(T,W) versus the temperature C(V.w) venus the apJ)l1ed bias
for the four lndlcaied lor the three lnd lcated
frequencies to derive N(Et). frequencies to derive I.

377
 a-Si:H Deposition Barrier height Saturation DOS at Ef I" Nd Tc Eo
prepared rate at 300 K current at 300 K N(Ef)
by eMs) (eV) (A/cm2) cm-3 eV-l (cm-3) (cm-3) (K) (meV)
Standard glow discharge 1 1.05 1.9E-l1 1 E15 2 E15 2 E15 200 52
(best materiall
dc magnetron sputtering 10 0.95 4.8E-l0 5 E16 4 E16 7 E16 280 64
(this world
Table I : comparison of the electronic characteristics of a-Si:H samples

ACKNOWLEDGMENT

The authors acknowledge Dr. l. Chahed of

Laboratoire d'OptiQue des Solides (Pierre and

-
P,e-TOF Marie Curie University Paris) for the PDS
Tc=280K measurements.

\i !
REFERENCES
C-T-w
1016 eni'eV·' IOPtical (1 )A.Shah, E.SauvinN;Wyrsch,H;Curtins,B;Leutz,
D.S.Shen,V.Chu,S.Wagner,H.Schede and H.W.A.
I
: absorption
Post-TOF I E ov =64 me V Chao
C-v-w II I (T.=740K) Proceedings of the 20 th IEEE Photovoltaic
13 '016Cni31 I I
PTSC Specialists Conference, Las Vegas (1968)
I I I (2)P.Roca i Cabarrocas,R.venderhaghen . M.LTheye
0.2 0.4 1.6 1.8 T V.Boui zem,D.Mencaragl i 8,Z.D j ebbour,J.Si b,
Eg J.P.Kleider and C.Longeaud
(Ee-E) (ev) This conference
Figure 4 : General Shape of the DOS. (3)M.Pinarbasi,N.MaleY,A.Meyers and CJ.R.Abelson
Thin Solid Films 171,217 (1989)
4. COMPARISON WITH BEST GLOW DISCHARGE (4)N.Ababou,N.Bel di ,M.Aoucher emd T.MohClmmed-
AND CONCLUS ION BrBhim
submitted to Thin Solid Films
We have .also reported in table 1 the typical
(5)D.V.Lang,J.D.Cohen and J.P.Harbison
values of parameters we currently measure on
Phys.Rev. B 25,5265 (1962)
the best standard a-Si:H material produced by
(6)J.P .Kl ei der,D.Mencaragl i a and Z.D jebbour
capacitively coupled diode RF glow discharge.
J.Non-Cryst. Sol1ds 114,432 (1989)
We can see that the electronic quality of the
(7)C.L ongeaud,J.P .Kl e1 der,D.Mencaragl i a,
material is beUer for the glow discharge one.
T.Mohammed-Brahlm and A.Rahal
However, in this case the deposition rate is
MRS Spring Meeting San Fr8ncisco (1990)
less than 1 A Is (10). This can be considered as
(6)C.Longeaud and R.Vanderhaghen
a drawback from an industrial point of view,
Phil.Mag. B 61 (2),277 (1990)
especially for solar panels production when
(9)G.F .Seynhaeve,R.P.Barcl aY,G.J.Adri aenssens
high throughput is needed or for applications
8nd J.M.MarshCl11 Phys.Rev. B39, 10196 (1989)
such that radiation detectors or xerography
(1 O)P.Rocca i Cabarrocas,B.EQuer,J.Huc,A.Lloret
where thick layers are needed. This is why
and J.P.M.Schmitt
some effort has been done in order to increase
7 th Photovoltaic Solar Energy Conference
this deposition rate. Recent results show
Sevi 11 a Spai n ( I 986)
that one way to obtain low defect denSity
P.Rocca i Cab8rrocas Thesis P8ris VII
a-Si:H films at high deposition rates is to use
university (1968)
He dilution in RF glow discharge ( 2 ). The
method presented here using magnetron ring is
another promising way. The electronic
transport properties obtained on first films
produced and reported in this paper stay in the
range of good qU81ity a-Si:H.
Horever ,optimization of the deposition
parameters , especially those controlling the
hydrogen content in the films prepared by de
magnetron sputtering, is still in progress.

378
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A-Si:H AMBIPOLAR DIFFUSION LENGTH AND EFFECTIVE LIFETIME MEASURED BY

FLYING SPOT TECHNIQUE (FST)

M. Vieira, R. Martins, E. Fortunato and L. Guimaraes.

Faculdade de Ciencias e Tecnologia da Universidade Nova de Lisboa
Quinta da Torre, 2825 Monte da Caparica, Portugal.

ABSTRACT. The determination of the ambipolar diffusion

length, L*, and the effective lifetime, t*, in p/i and a-Si:H
Schottky barriers (ITO/p/a-Si:H/Al-Si; Cr/a-Si:H/Cr/Ag) have
been determined by Flying Spot Technique, FST. This technique
consists in the transient analysis of the
photocurrent/photopotential induced by a laser beam that moves
perpendicularly to the structure with a constant motion ratio,
at different velocities. The charges induced during the beam
motion, are analysed and correlated with the continuity
equation, to obtain the photocurrent/photovoltage distribution
profiles. So, taking into account the competition between the
diffusion/drift velocities of the excess carriers and the
velocity of the flying spot, it is possible to solve the
transport equations and to compute separately L* and t*, from
the asymmetrical distribution responses.

1. INTRODUCTION 2.2.Lateral photoeffect

When a thin transparent metal
The behaviour of minority carriers is of semiconductor or p/ i structure is
considerabl~ interest for the illuminated with monochromatic localized
characterization of the optoelectronic light, perpendicular to the plane of the
quality of the materials. So, the analysis interface, a lateral photovoltage is
of the diffusion length values in a-Si:H developed in addition to the transverse (5),
films (and their alloys) and its correlation depending on the position of the light spot.
width the density of states, DOS, is Considering such structures and a light
important when we want to use such materials beam which injects carriers at the point A
in device applications, such as solar cells. (Fig. 1). From the equilibrium condition
Commonly, diffusion lengths in a-Si:H have
been evaluated from Surface Photo Voltage,
SPV, (1,2) and from Steady State Photocarrier light ~
Grating Technique, SSPG, (3). All these
techniques are based on the analysis of the
steady photovoltage and photocurrent. In
this paper we report a simple new technique y
for the determination of the ambipolar
diffusion length, FST (Flying Spot
d e:e x
Technique), where instead of a steady light eie
beam, a beam with a constant motion ratio is
used to generate the carriers. So, taking
e:e a-S:H

into account the competition between the

diffusion/drift velocities of the additional
generated carriers and the velocity of its
generation, it is possible to solve the
transport equations and to compute
separately L* and t* from the asymmetrical
distribution of the photocurrent/
photovoltage induced during the spot Figure 1: Charge movement and potential
movement. for the lateral photoeffect
(before injection) the junction goes over a
2. PRINCIPLES OF THE FLYING SPOT TECHNIQUE new steady state where most of the electrons
stay at the illuminated point, while the
2.1 Introduction holes are collected by the metal. This is
The effect of illuminating a accompanied by a shift in the Fermi level,
semiconductor junction is, as known, the constituting the transverse photovoltage
appearance of a photovoltage effect between indicated as VA.
the two sides. If the illumination is As the conductivity of the metal (p+) is
nonuniform, an additiorial effect arises much larger than that of the semiconductor,
resulting in the development of a it may be thought as an equipotential region
photovoltage parallel to the junction and so, the holes will instantaneously
(lateral), in addition to the transverse redistribute themselves uniformly over the
photovoltage. region.

379
 At any other point B (Fig.l) a deviation
from equilibrium will appear, resulting in a M
Deplection Bulk
lateral field (El)' The potential V(y) e Semiconductor
cl t
~
characterizes the effect of t.his lateral / ' J dif
light a
field, and it depends on the position of the yL
1
light spot.
0 w d x
2.3 Transverse photoeffect
When a junction is illuminated through a Figure 2: Depletion layer photoeffect
very thin transparent metal or semiconductor
contact a photovoltage (Vph) is developed
between the surface and the bulk due to the 3. CARRIER TRANSPORT OUT OF EQUILIBRI17M
diffusion of the carriers and its subsequent (NONUNIFORM GENERATION)
charge separation in the surface field.
The carriers induced during the spot
movement depend on its velocity. So,
assuming that at any time, t, the sample is
where Jsat is the saturation density spli t into two zones: the dark and the
illuminated one (Fig.3), and that the light
current; VT = n KT and n the diode quality spot is moving with a constant velocity, c,
q
factor. (in the yy direction), the excess hole
The total photocurrent density through concentration depends on time through
the depletion layer will consist of two y' = y-ct. In these conditions we have:
contributions (Fig.2):

2. 7.

one, Jdl, due to the carriers generated and the transport equation takes the form:
inside the depletion layer and another,
Jdiff, due to carriers generated inside the
bulk and diffusing into the junction. c
~=D*~
iJy iJy2 8'
Recombination and thermal generation of
carriers, inside the junction, will be

,
neglected (6).
The generation function g(x,t) (density Light C=I.l*E apl
I
of carriers generated per unit time) is g(x,O) y'
split into a time dependent function .(t) J x'
and a space dependent function g(x) = ae-ax, -1 -0/2 0 +0/2 +1

[xr
thus: I
,I
g(x,t) = .(t)ae- ax 3.
[ [ [
Dark Illuminated Dark
where a is the absorption coefficient; .(t)
describes the light spot movement as a
function of time and is like a total Figure 3: Nonuniform generation
"effective" incident photon flux impinging
the sample at the instant t. Jdl is found to Based on this, the spot movement can be
be equal: thought as an applied electric field of
strength c/I.l*, competing the "drift"
w velocity of the carriers (in the yy
Jdl(t)x=O = q .(t) f g(x)dx 4. direction) with its diffusion velocity
o (L~/'t*) .
As boundary conditions we assume that
The hole current in the bulk is obtained the hole concentration and the current must
by solving the ordinary field-free diffusion be zero for y'=±~ and y'=O respectively, and
equation with recombination (7), suitably
modified to take into account the nonuniform the same at y'=iJ/2 (iJ is the spot width):
generation (8): QQ QQ
(Dp iJ) y=0/2 (dark) = (D PiJ ) y=0/2 (ill)
.(t)aL*
Jdiff (t) x=w -q e-aw 5.
l+aL* p(iJ/2)dark = p(iJ/2) ill 9.

If we do not take into account the Jp(Y'=O) = 0 Jp(Y'= ± ~) = 0

fraction of the majority carriers
photogenerated in the depletion region that The solution of Eq. 8 is given by:
can diffuse to the interface and reduce the
photocurrent (7,9),the total current density Dark region [g(x)=O]:
is given by:
l-t (y'-iJ/2»0 and 2-t (y'-iJ/2)<O
Jtot(t)x=o _q .(t) (1- e-aw ) 6.
l+aL* P(Y)1,2 = Gillt*Bl,2 exp(kl,2 y) 10.

380
Illuminated region [g(x)~O]: From equation 14 and as Q (y) = Q (x)
(charge conservation principle) we deduce:
ly'-dl2l<O x
y = (l/d)x and t = f [l/(dc)] dx
p (y) = Gill't*. o
Now, for I (l/d)x-ctl > d/2, we got:
11.
P(t)1,2=Gill't*Bl,2 exp[kl,2(d/l)ct] 18.
and:
with:

Gill = GO
o
fa e- 1lX dx; Whilst, fori (l/d)x-ctl < d/2, we have:

A = 1/[exp(kld/2)-(k1/k2)exp(k2d/2)]; p(t) Gill't*{A[exp(k1(d/l)ct)+

+(kl\k2)exp(k2(d/l)ct] } 19.
B1 = 2 A B = 2 (kl/k2) A; and:

~
_1_
kl - [2~* + (~) 2 +
2D* L*2
Since both type of carriers contribute
to the transport mechanism, instead Dn, Dp,

~
_1_
k2 = - [2~* - (~) 2 + 12. Ln, Lp we use the quantities D*, L* as
2D* L*2
effective diffusion constant and ambipolar
diffusion length, respectively (11).
GO = Iph (l-R) 11 depends on photon flux According to Eq.5 and the above
intensity, Iph' and it is corrected by the relations, since the total photocurrent is
quantum efficiency of the material, 11, and proportional to ~(t), and so, to:
by reflection losses, R.
Taking into account the above relations,
an expression for the time dependence on the
generation ratio can be derived equating the
average charge "collected" at x=O during the
time t, as shown below. we expect to obtain an asymmetrical signal,
for the photovoltage as a function of time,
4. METHOD OF ANALYSIS with a peak dependent on the transit time of
the light spot. By doing:
From equation 7 and introducing the
characteristic time:
y
it is possible to compute separately L* and
t=f~ 13. 't*, from the asymmetrical response. So, we
o
determine 't*, L* and D* from the inverse of
it is possible to correlate the x,y, the slopes (Ll and L2) :
coordinates of a generated charge, with the
position of the spot at the instant of
't* = (LI-L2)/C ; L* = ~LIL2; L* = ~D*'t*
generation.
The charge collected (equal to fJdx) is 5. EXPERIMENTAL DETAILS
a measure of the average distance moved by
the carriers (10). So, at x=O, in the yy The films under analysis have been
direction, the collected charge, due to the produced by plasma decomposition of silane
carriers generated at x,y is given by: (PECVD) (12) in the structure configuration
referred previously. The DOS was inferred by
y SCLC as described elsewhere (13), in the
Q(y) J Qo exp(-y/c't*)dy 14. same type of samples.
a
That gives:
6. EXPERIMENTAL SET UP
~ if y<c't* 15.
Q(y)x-o ~ 1 For FST measurements, the experimental
set up is shown in the fig.4.
Qo is the total generated charge at The sample is excited by a moving HeNe
(x, y); d is the film thickness; and 1 the laser beam spot 0.. = 632.5 nm with an
contact width. enclosed path to avoid stray light),
In the xx direction: controlled by a motor drive. The signal from
the sample is detected by a LeCroy
x oscilloscope and the reference signal is
Q(x) fQoexp(-x/L*) dx 16. given by a silicon photodetector which is
o also used to calibrate the angular frequency
of the light motion and the space traveled
That gives: by the spot (12).

Qo x
Q(x)y-O'" d if x<L* 17.

381
 ,
8. CONCLUSIONS
YeNe-LASER
,s~Z'~
+~-te
~ o W" The recorded data show that the FST
c==3-- - - ------~~~ measurements are quite suitable to determine
separately the effective diffusion length
I
and the effective carrier lifetime.

r--"'------1 n rnm
o 00
Limitations of this technique are related
with film thickness (uniform absorption
through all film) and the relaxation time,
A!V
AMPLIFIER 'tr (as 'tr decreases and approaches the value
of 't*, the results are meaningless) .
We have also observed that L* and 't* are
influenced by the DOS, which determines the
lifetime of the carriers through
recombination.

Figure 4: Sketch of the FST technique ACKNOWLEDGEMENTS

The authors wish to express their thanks

7. RESULT AND DISCUSSION to F. Soares, N. Carvalho, J. Figueiredo, G.
Lavareda and I. Baia for the help given
In figure 5, we show the distribution during films production and
curves obtained by FST measurements at characterization.
different spot velocities. It can be seen This work was supported by Portuguese
that curve shapes and peaks are affected by national R&D programme: Junta Nacional de
the spot velocity, in agreement with the Investiga~ao Cientifica under Eureka project
final solution (Eq. 5, 18, 19). nO 366 and Instituto Nacional de
Investiga~ao Cientifica/Centro de Fisica
1.2
Molecular das Universidades de Lisboa under
project B and by the CEC project contract
~ 1.0 nO JOUR-0077-C (TT)/OASIS project.
..; 0.8
REFERENCE:
0.6
(1) A.R. More, Appl Phy Letters 40,403
0.4 (1982) .
(2) A.R. More, J.Appl Phy 56,2796 (1984).
(3) D.Ritter, E. Zeldov, and K. Weiser,
Appl Phy Letters 49,79 (1986).
(4) D.Ritter,K. Weiser and E. Zeldov,
J.Appl Phy 62,4563 (1987) (Acad.
o 100 200 300 400 500
Press, New York, 1980).
t (IlS) (5) J. Torkel Wallmark, Proc. of IRE 45,
474, (1957) .
Figure 5: Curves for the FST measurements
(6) Wolfgang W. Gartner, PhY. Review, .Vol
at different spot velocities, as 116, N1, 84, (1959)
indicated (7) A.R. More, J.Appl Phy 54,222 (1983).
(8) R. Martins, M. Vieira, to be published
In Fig.6 we show the dependence of DOS (9) Joseph Reichman Appl Phy Letters 38,4
on L*. Data show that the DOS decrease as L* (1981) .
increases, leading to a high quality a-Si:H (10) K. Hecht, z. Physik 77,235 (1932).
films. For the same set of samples under (11) J. van den Heuve1, PhD thesis, Delft
analysis· 't* varied between 3X10- 8 s to University of Technology, the
8.8x10- 9 s, in agreement with the behaviour Netherlands (1989).
shown in Fig.6 for L*. (12) R. Martins, M. Vieira, E. Fortunato,
I. Ferreira, F. Soares and L.
Guimaraes, presented at the MRS Spring
Meeting, S. Francisco (1990).
10 18 (13) E.Fortunato, to be presented at this
M
conference.
I

. E
U

>
•
~
- r.. 10 I
~
Z

10 '
100 200 300 400
L* (nm)

Figure 6: Dependence of DOS (inferred by

SCLC) on ambipolar diffusion
length.

382
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Thermal Stability of a-Si:H/a-SiC:H Superlattices Studied by

Hydrogen Effusion, X-ray Diffraction, IR Spectroscopy and
Photoluminescence

T. Fischer, T. Muschik, P. Hanesch, J. Kolodzey, and R. Schwarz

TU Miinchen, Physik-Department E16, W-8046 Garching, FRG
G. Neff, H. Schneider, and M. Stanger
TU Miinchen, Physik-Department, Kristall-Labor, W-8046 Garching, FRG

Abstract
To study the thermal stability of amorphous a-Si:H/a-Si1_"C,,:H-multilayers hydrogen evo-
lution experiments were done at 410, 525, 600 and 750°C. The Si-H and Si-H2 bondings,
as observed by fourier transform infrared spectroscopy (FTIR), show a stronger temperature
decrease compared with the C-H-bonding which corresponds to the higher bond strength of C-H
(4.3 eV) compared to Si-H (3.3 eV). In Photoluminescence (PL) measurements it is found that
for multilayer films with small silicon sublayer thicknesses the decrease of PL efficiency with an-
nealing is less pronounced as compared to samples with larger wells. Structural stability of the
multilayers is demonstrated by X-ray diffraction (XRD). The stability against crystallization
up to 750°C is shown by the absence of the [111] XRD peak.

Introduction Heating steps have been performed in an UHV appara-

tus with a small quartz tube (diameter = 2 em, length =
The wide bandgap of amorphous SiC alloys has made them
15 em) which can be heated by a movable oven. A qua-
an attractive material for practical applications, especially
drupole mass spectrometer allows the detection of partial
for window layers in solar cells. They can be used eit-
pressures.
her as single laye.r;s or as alternating a-Si:H/a-Sh_"C,,:H-
multilayers [1]. For both cases, the study of such multilay- Table 1.
ers is of primary importance, since it offers the opportunity
sample no. d Si [nm] d sic [nm] d [nm] Egap leV]
to measure physical properties from an accumulated signal
M177 30 30 1200 1.75
of large number of interfaces. Information about stability
M183 10 10 1200 1.77
at higher temperatures is needed for some fields of applica-
M184 1 1 1200 1.85
tions and it will give hints for long term stability at lower
M198 1 1 40 1.85
temperatures.
S213 bulk 960 1.72
To evaluate the changes at an interface, a well defined
picture of the original state is needed. For the interface
sharpness XRD, transmission electron microscopy (TEM) Hydrogen Bonding
and the observation of the plasma parameters are very use-
ful tools [2]. With rising temperature interdiffusion bet- After the different annealing steps the H bonding modes
ween the multilayers occurs, which can be monitored by were measured by FTIR with a Bomem spectrometer (mo-
the time dependence of the Bragg peak signal at a given del DA 3.02). The results can be summarized as follows:
constant temperature [3]. Annealing for approximately 15 min. at 410°C and at 525
Since hydrogen plays a fundamental role for the pro- °C reduces both the Si-H and Si-H2 absorption to about
perties of amorphous silicon alloys and multilayers, we stu- 30% and less than 20%, respectively (see Fig. 1), whe-
died the chemical bonding and thermally stimulated ou- reas the C-H bonding at 980 cm- 1 remained constant after
tdiffusion of H, measured by FTIR spectroscopy and H the two temperature steps. After heating at 750°C for
effusion. Some consequences of the defect creation during 15 minutes almost no signal could be obtained except for
annealing of the multilayers are then discussed in connec- the C-H absorption, which remained at about 30% of the
tion with PL experiments. Finally the structural stabi- original value. This behaviour may directly be attributed
lity is investigated by XRD to find out whether there is to the higher bond strength of C-H (4.3 eV) compared to
microcrystallinity after annealing. Si-H (3.3 eV).

Preparation and Annealing Hydrogen Diffusion

The multilayers were deposited by RF-glow discharge in a The hydrogen effusion curves taken during isochronal an-
UHV reactor having computer-controlled fast gas switching nealing typically show a single peak around 550°C in ho-
as described elsewhere [4]. For this study we use multilay- mogeneous a-Si:H films and a low- and high-temperature
ers from a deposition series with a carbon content of Xc peak at about 450 and 700°C, respectively, in silicon-
= 0.38. The bandgap of the a-SiC:H layers was approxi- carbon alloys or a-Si:H/a-SiC:H multilayers [4,5]. Here we
mately 2.25 eV with barrier and well widths ranging from measure the hydrogen diffusion coefficient D by monitoring
10 to 300 A (see Table 1). From photothermal deflection the time dependence of the effusion rate rH at temperatu-
spectroscopy (PDS) a single interface defect density of less res slightly below and above the low-temperture peak and
than 1011 cm- 2 was found [4]. compare our results with literature.

383
 where Wo is an attempt frequency and a is below 1. In Fig.
3 we show that the measured data clearly follows the power
M183 law for D(t) with a power of 0.6 ± 0.05.

10-7~------------------------~
:i
IU83
.9 Mt77
til
0 M198
6
./J
./J
.~

E
(/)

6
L.
CH2;J-
I- slrelch

si-Cf-I:J-rock/Wag

1000 2000 3000

WavE>numbE> r (em-I)

Fig. 1. FTIR spectra of the multilayer sample M183 5 10 15

illustrating the stability of the C-H bond. t [min]
Fig. 2. Hydrogen effusion as a function of annealing
A number of measurements were reported showing that
D has an activated behaviour. The prefactor Do and the
time at T A = 410°C for three different samples. For the thin
sample M198 an exponential fit to eq. (1) can be performed
activation energy Eo were, for example: 0.0117 cm 2 /s and
(straight line).
1.53 eV [6J, 0.01 cm 2 /s and 1.5 eV [7J, and 0.00166 cm 2 /s
and 1.45 eV [8J, respectively. Assuming hydrogen diffusion 10-6~------------------------~
through the film and not surface reactions to be the rate-
determining step we follow the analysis given by Shewmon
for the calculation of D [9J. For low values of D and for
thick samples rH follows a square root time dependence.
At high enough temperatures and/or for very thin samples
the effusion rate can be approximated by:
2Po 1l"2
rH = -;p:D exp(- 4cPDt) (1)
where Po is the accumulated total pressure after complete
effusion and d is the film thickness.
Figs. 2 and 3 show the effusion rates for low and high
temperature, respectively. At 410°C, taking a value of D
from the literature, the diffusion length L = VDt after 5 D = 3.2 X 10- 11 em',s
mb is about 40 nm. So for sample M198 in Fig. 2 we can
use eq. (1) to determine D = 5xl0- 14 cm 2 /s (see straight
line). At times larger than 5 min the diffusion coefficient 10 15
seems to change to smaller values indicating dispersive dif- t [min]
fusion as described below. The other samples should show
a square root law. However, the response of our apparatus Fig. 3. High temperature effusion of a 1.2 J-Lm thick
is too slow to correctly show this time dependence. sample with a fit to the dispersive diffusion behaviour given
At 600°C also the thick samples should follow an expo- in eq. (2) (dotted and dashed lines). The straight line is a fit
nential time dependence. Again, using eq. (1) we determi- to eq. (1) at 5 min.
ned the diffusion coeffient D = 3.2xlO- 11 cm 2 /s for sample
M183 as shown by the straight line in Fig. 3. Our values A comment should be made concerning the low-
are by a factor 8 and 2 below the values given in ref. [7J at temperature results. Beyer attributed the low-temperture
410 and 600°C, respectively. Compared to ref. [8J we find effusion peak of a-SiC:H alloy films to desorption of hydro-
the same results within the precision of the experiment. gen from internal surfaces and (fast) out diffusion through
It is interesting to interpret the obvious change in D structured material [l1J. This could lead to a much lower
with time by comparison with the recently described di- activation energy of the 1.5 eV often reported and here
spersive diffusion which leads to a stretched exponential found for our samples. We have measured earlier that an
behaviour of hydrogen diffusion in a-Si:H [lOJ. This effect additional amount of hydrogen of about 1014 atoms /cm2
is believed to arise from a continuous distribution of energy is present at multilayer interfaces [2J. This is an indication
barriers for hydrogen motion and leads to: that a larger amount of voids is present there. However,
we presume that hydrogen from those locations has effused
D(t) = Do(woW" (2) already before the effusion curve in Fig. 2 has reached the
exponential decay region after t = 3 min.

384
 Photoluminescence This double peak structure of M177 is more clearily vi-
sible after the 525 DC anneal, at which a drastic decrease of
Photoluminescence (PL) measurements can yield informa- 1/PL,S; occurs, whereas 1/PL,S;C is nearly unchanged compa-
tion about the density of dangling bonds and the width of red to T A =410 DC (i.e. the SiC-layers become the dominant
the band tails. Heat treatment of hydrogenated amorphous source of PL from the multilayer) (see Fig. 4). This beha-
silicon films leads to outdiffusion of hydrogen and thus to viour is a direct consequence of the fact that the excitation
an increase in the number of dangling bonds. These form energy is larger than the bandgap of the a-SiC layers by
states located near the middle of the energy gap which about 0.15 eV. A small amount of light is absorbed in these
are known to act as non-radiative recombination centers. layers, resulting in a PL spectrum peaking at approxima-
Therefore, with increasing dangling bond density the PL tely the same energy as in the case of a-Si:H. In untreated
efficiency drops [12J. samples the contribution of PL from the barrier layers can
After each annealing step PL spectra were taken in or- be neglected, because in addition to the small absorption
der to investigate the influence of the multilayered struc- coefficient it is known that a-SiC alloys exhibit a smaller
ture on the stability of the mentioned properties. The sam- PL efficiency compared to pure amorphous silicon [13J.
ples were excited by ~ 50 mW /cm 2 of laser light with an After the last annealing step at 650 DC there is no de-
energy of 2.41 eV. The effect of dangling bond creation by tectable silicon-PL and 1/PL,S;C has dropped by a factor of
annealing is observed as a decrease in PL efficiency 1/PL in 2 compared to TA = 525 DC.
all of our samples. There is, however, the clear tendency The results obtained for sample no. M184 with d s; =
that for multilayer samples with a small silicon sublayer 10 A are different to those of M177 with d s; = 300 A in
thickness d s; the reduction of 1/PL is slowed down compa- that there is no PL signal from the SiC-layers. For both
red to films with larger ds;. This result is displayed in samples the thicknesses of the SiC-layers and of the Si-
Fig. 4, where the PL spectra after a 410 DC anneal and a layers are equal. Thus, in sample M184 with 10 A thick
525 DC anneal are shown for two samples with d s; = 300 barriers the photogenerated carriers in the a-SiC:H enter
A (no. M177) and d s; = 10 A (no. M184), respectively. the Si-layers by diffusion. This phenomenon is also known
After annealing at 410 DC the emission energy of the from silicon germanium multilayers [14J.
silicon PL of sample M177 is significantly lowered compa- The most important difference in the behaviour of the
red to the initial state, suggesting a decrease in bandgap, two samples is that the decrease of 1/PL in M184 is much less
which might be due to out diffusion of hydrogen from the pronounced in going through the annealing steps compared
silicon sublayers. The increase in the number of Si dangling to M177. E.g. the ratio of the PL efficiency for M184 when
bonds is also indicated by the concommitant reduction of heated to 525 DC compared to the as deposited state is ~
the silicon PL efficiency 1/PL,S; by a factor of about 7. At 0.6, whereas for M177 this value is ~ 1.6xlO- 3 • For M184
this annealing step a small PL signal is observed at 1.3 eV the PL efficiency is large even after annealing at 650 DC.
which can be attributed to the SiC-layers. The results from other multilayer samples with varying d s ;
confirm this tendency.
One possible explanation for the above observations
is that due to the stronger C-H bonds compared to Si-H
100 bonds hydrogen is driven out of the Si-layers at ~ 500 DC,
whereas the SiC-layers are stable up to at least ~ 650 DC.
80
Therefore, the barrier layers can act as a hydrogen source
after 410 °c anneal for the Si-wells. Samples with a small d s ; can be supplied
60
with hydrogen more easily than films with large d s; so that
after 525 °c anneal
40 the Si-film quality is preserved at high temperatures and
.....
VI the reduction of PL efficiency is less pronounced.
c 20
::J
0 Structural Description by XRD
a;
..=. Our X-Ray diffraction measurements aimed at two questi-
....
>. ons: the stability of the multilayers against thermal inter-
diffusion and the stability against crystallization. All mea-
VI
cQ) 100 M177 Id si =300 Al surements have been performed on a diffractometer with
..... a primary monochromator using the Cu K" line. To an-
c 80 after 410°C anneal
I-i swer the first question, the XRD patterns shown in Fig. 5
after 525 °c anneal
I
....J 60 yield positions and amplitudes of the Bragg peaks as well
a... as those of the so-called Fresnel interferences caused by re-
40 flection at the air/sample and sample/substrate interface.
The results for sample M198 (Fig. 5) show that there is
20 a shift of the Bragg peak, which can be explained by a
shrinkage of the layers above ~ 400DC caused by hydro-
0 gen effusion [5]. Another feature of the XRD patterns is
the drop in intensity of the first order Bragg peak after
annealing at 525 DC, which is indicative of a beginning in-
0.5 to 1.5 2.0 2.5 terdifusion of the multilayers. Similar results have been
Photon Energy reV) obtained recently giving an interdiffusion coefficient D of
4x10- 2D cm 2/s at ~ 500 DC [3].
Fig. 4. PL spectra of two samples with ds;=300 A and The check for microcrystallinity was done by observing
ds;=10 A after annealing at the indicated temperatures. the [111]-peak of crystalline silicon. Fig. 6 compares the

385
 104
Conclusion
In conclusion we have determined some aspects of the
,.-, 103 M198 (dSi = 10 A.) thermal stability of a-Si:H/a-SiC:H superlattices. The
III
P- amorphous structure is maintained even above 750°C
c.>
......... whereas interdiffusion of carbon and silicon already sets
....... 10
2
...... in at ::::: 500°C. High stability or even a slight increase in
~
the number of C-H bonds with annealing was observed.
s:l
....enbD 101 The changes in PL intensity and PL peak energy suggest a
higher thermal stability in multilayers than in bulk a-Si:H.
0
10°
~
'-'
Acknowledgements: This work was supported by the
BMFT under contract no. 0328853A8. The FTIR mea-
bD
surements have been performed by F. Miiller and R. Os-
.....0 10°
termeir. A helpful discussion with W. Beyer is gratefully
acknowledged.
10°
-5 -4 -3 -2 - 1 o References

[IJ S. Tsuda, PVSEC-5, Kyoto, Japan (1990), p. 21

Fig. 5. Grazing-incidence X-ray diffraction pattern of the
multilayer sample M198 at different temperature steps. [2J R. Schwarz, T. Fischer, P. Hanesch, T. Muschik
and J. Kolodzey, E-MRS, Symp. D, Strasbourg
(1990) -
peaks of a heated a-Si:H sample with that of the sample
M198. Clearly, there is no significant indication for crystal- [3J J. Kolodzey, P. Hanesch, T. Fischer R. Schwarz,
lization even after the 750°C anneal. The stability against G. Zorn and H. Gobel, ibid.
crystallization could be due to the fact that for silicon car-
bon alloys with x ::::: 0.3 the crystallization temperature is [4J N. Bernhard, M. Kirsch, R. Eigenschenk, M.
~ 900°C [15J and thus is higher than the temperatures Bollu, C. Wetzel, F. Muller andR. Schwarz,
used here. Similar results were found in a-Si:HJa-SiN:H Mat. Res. Soc. Symp. Proc. 192, 237 (1990)
multilayers [16J.
[5J G. Zorn, J. Kolodzey, H. Gobel, T. Fischer,
P. Hanesch and R. Schwarz, EPDIC-l, Munich
(1991 )
[6J D.E. Carlson and C.W. Magee, Appl. Phys Lett.
33,81 (1978)

[7J W. Beyer and H. Wagner, J. Appl. Phys. 53,

8745 (1982)
[8J S.F. Chou, R. Schwarz, Y. Okada, D. Slobo-
din, and S. Wagner, Mat. Res. Soc. Symp. Proc.
95,165 (1987)
[9J P.G. Shewmon, in Diffusion in Solids, McGraw-
Hill (1963)
[10J J. Kakalios, R.A. Street and W.B. Jackson,
Phys. Rev. Lett. 59, 1037 (1987)
[11J W. Beyer J. Non-Cryst. Solids 97 & 98, 1027
-29 . 5 -29.0 - 28 . 5 -28 . 0 -27 .5
(1987)
20 [0]
(12J R.A. Street, J.C. Knights and D.K. Biegelsen,
Fig. 6. Comparison of the [l11J peak of crystalline silicon
Phys. Rev. B 180, 1880 (1978)
in the bulk a-5i:H layer 5213 with that of the multilayer sample
M198 after a 15 minutes anneal at 750°C. [13J W. Siebert, R. Carius, W. Fuhs and K. Jahn,
Phys. Stat Sol. (b) 140, 311 (1987)

(14J T. Tiedje, C.R. Wronski, P. Persans and B. Abe-

les, J. Non-Cryst. Solids 77 & 78, 1031 (1985)

[15J A. Morimoto, T. Kataoka, M. Kumeda and T.

Shimizu, Phil. Mag. B 50, 517 (1984)
[16J S. Miyazaki, Y. Ihara and M. Hirose, J. Non-
Cryst. Solids 97 & 98, 887 (1987)

386
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON, PORTUGAL

LIGHT INDUCED DEFECT INFLUENCE ON a-Si FIELD TEST

PERFORMANCE

W. B. Berry and M. J. Hahn

Department of Electrical Engineering
University of Notre Dame, Notre Dame, IN 46556

Abstract: The Solar Energy Research Institute has had amorphous silicon modules
under field performance and evaluation test for since 1985. This paper addresses the
evaluation of that performance in light of the Staebler-Wronski effect (1). This six years
of data obtained by the SERI test program is analyzed in this paper for these early
modules of the mid-eighties. The performance of these modules show strong
dependence on light induced defect generation as described by Faughan, et ai, for
cell laboratory performance. The model description is linear with log(time). Since the
modules are tested to a real outdoor environment the performance shows a significant
annual annealing cycle. Because of the relatively strong annealing effects, more than
two years of field test data are required to adquately predict the degradation trends.
The performance of these earlier modules is compared to module sets of more recent
manufacture. The later modules show better stabilization following a stronger initial
degradation. Analytical models are used to evaluate performance to date and to
project 30-year performance. Many modules project reasonable percentage
pe rformance.

1. INTRODUCTION preceeded it to field test. Standardization of marketed

modules has made great strides but is still in the
The Solar Energy Research Institute has been conceptual stage, and, therefore, modules placed under
conducting a test program to evaluate the field test often have some of the same physical
performance of amorphous silicon modules since these characteristics.
units have begun to be marketed. The modules, which In a test program it is useful to have more than
have been placed under test, represent a sampling of one element under test. In this program sets of modules
the developing technologies and are representative of of different manufacture or manufactured by different
those which can be purchased in the open market. As corporations have been selected. To distinguish
such these modules are expected to present a between the sets of modules being tel>ted, the module
representative experience similar to that which a sets are labeled according to the period in which they
consumer might experience. One goal of this program is were procured and placed under test. In this paper, two
to establish and maintain a life history data bank for periods of procurement are included: those placed
representative sets of modules for as many of the under test in years 85-7 and those so placed in years
developing technologies as can be accommodated. 88-9. There are module subsets in each procurement
These technologies include the different materials which period which correspond to subclassification such as
make up the basic cell structure which are in turn being size and/or manufacturer. The source of manufacture is
studied for thei-r advantages and disadvantages, the not identified since the purpose is to establish a
sub-technologies for each material which continue to comparative history of the modules technological
emerge in the development and exploration phases of development and the manufacturers are considered to
learning how the materials might be better exploited, be reasonably competitive. As the historical
and the escalation in size necessary to meet the end developmental patterns emerge it is expected that the
objectives of bulk power generation. manufacturers will continually leap-frog their competition
The objectives of this program include: while maintaining competitive effectiveness.
This paper compares the degradation of the
- analysis and evaluation of amorphous silicon module earlier module sets to those of more current technology.
field test data to determine the progress in The comparison is made for the full period of test and
performance and stability of emerging evaluation and for the period following that when the
technologies, initial light induced defect (LID) generation ,i.e.,
- modelling the ferformance data to predict productive Staebler-Wronski(SW) effect [1), has eased and the
module life, modules appear to have begun to stabilize. The
- comparison of module performance with the long term comparison which neglects the early SW effects has
objectives of the photovoltaics program, and been labeled the post initial decay and the prefix, pi, is
- determine the modes of degradation and the used to indicate the module response for that data
underlying physical mechanisms and processes. period.

Amorphous silicon is only one material under 2. LIGHT INDUCED DEGRADATION

development for photovoltaics which has reached that
state of technological development where it makes The degradation model used in the analysis
sense to have such a record on field performance. With below assumes the log(time) dependence [3]. This
this data bank, each new sub-technology development model was been shown to fit laboratory data reasonably
can be compared to the general history of those which well and has been used in earlier papers to analyze

387
field test fill factor data[3.4]. The field test data is not as elimination of a minimum test period. This observation
readily analyzed since the modules exit in a hostile is in consonance the current recommendation that the
environment which includes the four seasons during ratings for amorphous silicon modules be those
which amorphous silicon undergoes different rates of following a 600 hour light degradation period.
annealing. The seasons' temperature characteristics
are reasonably cyclic and no attempt is made here to 3. FIELD TEST DATA
quantify the seasonal differences.
The basic model was formulated for the fill factor Field test data is presented for two modules, one
as each from the 85-7 and 88-9 module sets. Each period
FF = FFo - B log(tlto), group of modules has twenty six modules under test.
There are two subgroups for the 85-7 period and five
where t is measured in months, to is the reference time subgroups for the 88-9 period. The two modules which
period, FF is the fill factor and FFo is the equivalent fill have been selected to represent the respective 85-7 and
factor at the end of the first time period, i.e., one month. 88-9 periods of procurement are labeled A3 and NEA3.
The same model form will be used in the analysis of the The data for these modules are presented in TABLE I
efficiency. It is recognized that that is not precisely and Figures 1 and 2.
accurate, but the fill factor is part of the analytical
expression for the effiency. 3.1 Tabular data
11 = FF (Voc Isc I <l> ), where <l> is the The data presented in TABLE I is typical of that for
an average module for each deployment period. The
insolation, Voc is the open circuit voltage, and Isc is the model letters (ad) and (pi) refer respectively to the
short circuit current. The efficiency, as analyzed below, portion of the data included in the analysis. (ad)
is presented as percent of initial eficiency. indicates that all data obtained from initial deployment
has been included. (pi) indicates that the initial loss in
performance or degradation data have been excluded.
EFF = 11/110 The exclusion periods for '85-7 and '88-9 modules sets
are the same for each module in the deployment period
The efficiency and the fill factor would be proportional if sets but are different for each deployment period.
the (Voclsc) product varied in proportion to the Efficiency is presented in normalized form. That
insolation, which in essence means the Voc remains is, the measured efficiency at month mo divided by the
constant and Isc is proportional to the insolation first month efficiency. The fill factor data is not
intensity. These are, for example, influenced by shunt normalized since it is presented in the equivalent of a
and series resistance respectively. percentage.
In this degradation study, the change in fill factor The EFFo and FFo values are the first month
with time under test is compared for the total period from following deployment intercepts (at '0'). These values
initial placement under test and for the sub-period under are back projected values which come from the
test· following the initfal decay. For the modules regression analysis. It is possible to pin the analysis to
procured in the 85-7 period, the initial eight months were the initial value, but that approach is less representative
removed from the data to determine post initial (pi) of the data for least square error analysis. In addition,
degradation performance. For those modules procured no attempt has been made to normalize the data with
in the 88-9 period, the first two months were removed. respect to that intercept value.
The difference in the choice of. removal period was The EFF30 and FF30 values are intercepts
determined by estimating when the strong initial decay projected to the 30-year end of life. The abscissa value
transferred to a more moderate decay. The initial decay at 30 years is 2.56.
for the newer modules was so rapid that it enabled the 'B' is the models module degradation coefficient.
It determines the rate by which the module degrades
upon exposure to the environment. In this case the
TABLE I: Log(t/to) coefficients for Module Efficiency model is determined specifically for the early response
and Fill Factor. Module A3 and Module NEA3 are 85-7 to light induced defect generation and does not account
and 88-9 modules respectively. for a saturation limit on that generation. If the
coefficients are expressed in percent of initial back
MODULE A3 projected value and the loss goal limit is ten percent
~ over the 3D-year deployment life, then the maximum
~Q[tD Eft EFEQ B EEElQ allowable value for 'B' is 3.9. This 'B' value is
ad 99.2 17.8 53.5 dependent on the initial value if the above proceedure is
pi 93.4 14.1 57.4 not followed.

3.2 Graphical presentation

Eill Eas;;tQr FFo B EElQ The graphical presentation of the data given in
ad 56.8 6.0 41.5 TABLE 1is shown in Figures 1 and 2. Figures 1
pi 53.9 4.1 43.4 represent '85-7 Module A3, and Figures 2 re[resent 88-
9 Module NEA3. The least mean square error fit to the
MODULE NEA3 data is presented for the 3D-year life with the endpoints
at the first and thirtieth months. Note that the sybols at
M.Q.d.el the the endpoints for the efficiency figures do not
t::jQrm Eff EFFQ B EFFJ.Q represent data. The open square data points represent
ad 88.5 10.1 59.7 those removed for the (pi) analysis.
pi 73.3 -1.0 76.3
4. DISCUSSION
Fill FgctQr FFQ B FFaQ
Analysis of the data has been carried out for the
ad 54.3 5.3 40.6 amorphous silicon module sets. The individual module
pi 49.4 0.9 47.1 data are similar to that shown in TABLE I and Figures 1
and 2. These individual module data have been

388
 combined with that for all modules in the sets to obtain a more rapid decrease. The former is an indication of
representative average. These averaged data values saturation in the generation of light induced defects and
are presented in TABLE II. the latter is an indication of the advent of another
100
t_----,-------.--------, degradation mode.There are indications of both of these
D
CAl
phenomena in the figures.
~ 90 • piAl The fill factor data for module A3 show the peaks
and valleys expected in the variation of seasonal
response. Saturation appears to be beginning to take
~ 80 place when all the data (ad) is considered, but not for
the (pi) data alone. The efficiency data seem to be
~ 70 trending the other way as if resistive effects were
becoming important to degradation. However, the data
~ 60 1a. EFFICIENCY
are, in general, well represented by the log(t)
characteristic. Note that, for both efficiency and fill factor,
that the (pi) result is an improvement over the (ad) result,
50 ~----~-----~--------~ i.e., the slopes and therefore the final operating
o 1
Log(mo)
2 3 performance is better.
60.-----r-----?-____T -_ _--~_ _--~
The 88-9 Module NEA3 performance is very
different. The initial rapid decrease in performance is
57.5 followed by a period of approxomately two years of near
55 stable performance. The seasonal variations are
25 52.5 apparent, but the continued decrease in average
performance is absent for this module. It is this
!:) 50
response which makes it appear that the amorphous
it 47.5
silicon technology has begun to solve part of the light
;j 45 induced defect problem. The processing techniques do
u. 42.5 not address the issue directly by dealing with' a change
40 1h . FILL FACTOR in the basic mechanism, but structure the cell design to
37.5 avoid or delay its effect.
0 .5 1.5 2 2.5 4.1 Average Module Response
Log(mo) The above discussion deals with a sampling of
Figure 1a and 1b. Efficiency and fill factor data for two period modules. These. modules are examples of
85-7 Module A3. The straight line fits to the data are the better modules and are taken from sets of modules
least square error regression amnalysis which placed under test to obtain a representative
include(ad) and exclude (pi) the initial decay. The preformance experience. TABLES II and III summarize
endpoints are not data. the average data for the module sets efficiency and fill
100 factor for modules currently under test. There are
considerable differences in the modules design based
on the earlier experiences gained from observations of
~ 90
o NEA·3
• pINEA·3 deployed modules. "Currently marketed" modules

;
always lag behind in developing technologies. This is
~ 80 clearly the case with amorphous silicon.

70 TABLE II: AVERAGE LOG(T) EFFICIENCY

COEFFICIENTS
60
2a. EFFICIENCY
50 Eff2.., a Eff3...!L piEffQ, pia piEff3Jl
0 2 3
Iog(mo) MOOULE SEI 8~-7 A
103.8 23.5 44.8 110.1 27.6 41.8
60
o NEA·3 MOQU!..E SET 8~-Z B
• piNEA·3
!5 55 97.0 24.6 34.4 105.5 30.4 31.6
!:)
it 50 MOOULE SEI 88-& NEA
;1 86.2 9.9 60.9 74.5 -0.7 76.5
u. 45

2b. FILL FACTOR MQQULE SET 88-& NEC

40
87.2 14.2 50.7 78.3 6.1 62.3
o 5 1.5 2 2.5
MOC!ULE SEI 88-& SEA
Ioglmol 83.3 10.0 57.5 69.5 -2.5 76.0
Figure 2a and 2b. Efficiency and fill factor data for
88-9 Module NEA3. The straight line fits to the data are MQQU!..E SET 88-9 SEa
!east square error regression amnalysis which 82.5 16.4
Include(ad) and exclude (pi) the initial decay. 40.2 68.8 3.9 58.8

. In general there are two trends which might apear MODULE SEI 88-9 sec
Inthe I?g(t) graph~. hese are a leveling off of the 84.4 12.9 51.3 73.2 -2.5 76.0
successive data POints and the transition to a region of
------------------------
389
TABLE III: AVERAGE LOG(T) FILL FACTOR to three months of exposure to the environment.
COEFFICIENTS The '88-9 stabilization following the rapid initial
decay shows improved performance in some cases. On
average these modules have degradation coefficients
f..E.Q. B FF3.2. RiFFQ., RiB RiFFaJl
which project the desired less than 10% loss of
MQOULE SEI 85-7 A performance for the desired 30-year lifetime.
58.0 8.5 36.3 61.2 10.5 34.0 '85-7 module fill factors show a strong log(t)
dependence which is indicative of Staebler-Wronshi
MQDULE SET 85-7 B phenomena.
49.4 7.4 30.4 52.1 9.7 27.4 End of 30-year life fill factors for the '88-9
modules are projected to be in the range 40<FF<55.
These values are respectable when compared to the
MQOULE SEI 88-a NEA '85-7 modules which have maximum values in the mid
54 .. 9 5.8 40.0 49.2 0.7 47.3 40's

MQDULE SET 88-a NE~ 6. ACKNOWLEDGEMENT

53.4 4.9 41.1 50.9 2.5 44.4
This work has been supported in part by the U.S.
Department of Energy under Contract No. DE-A C02-
MQOULE SEI 88-9 SEA 83CHl0093 to the Solar Energy Research Institute, subcontract
54.9 5.1 41.6 48.7 -0.6 49.9 XX-9-19024. SoeciaI thanks to Laxmi Mrig for providing the
data on the deployed modules.
MQDULE SET 88-a SEB 7. REFERENCES
57.1 7.5 37.9 50.6 1.7 48.8
1. Staebler, D. L. and Wronski, C. R., "Optically Induced
MQOULE SEI 88-a SE~ Conductivity Changes in Discharge Produced Hydrogenated
Amorphous Silicon", J Appl Phys, 51, 3262 (1980).
57.5 6.0 42.2 53.7 2.6 47.2
2. Smith, Z. E., Wagner, S. and Faughnan, B. W., "Carrier
The reader is invited to determine the normalized Lifetime Model for the Optixcal Degradation of Amorphous
average coefficients. The procedure is as follows. Silicon Solar Cells", Appl Phys Lett, 46, 1078 (1985).
Divide the three coefficients in the set, (ad) or (pi), by the 3. Berry, W. B., Hahn, M. J. and Mrig, L. c., Amorphous
'0' coefficient. The limit on the B coefficient is 3.9 for ten Silicon Technology-1990, Mat Res Soc, P. C. Taylor, et al
percent 30-year life degradation. Eds., 192, 51 (1990).
The variance in average coefficient values is, in
general, not great. Obviously there are exceptions. 4. Hahn, M. J., Berry, W. B. and Mrig L. c., "Comparative
Recall that the '0' values are statistical projections. For Shon Term-Long Term Field Test Performance and Stability of
the 85-7 Modules the (ad) values are, on average, less Tandem and Single Junction a-Si Modules", Proc 21st IEEE-
than the (pi) values. This is simply a strong indication PVSC, vol II, 1057-61, 1990.
that some of the modules have experienced other
modes .01 degradation and are beginning to fail. The
later data more strongly show that increase in
degradation. This is also born out by the lower (pi) '30'
value. projections. The 85-7 B values are greater than
the ten percent life degradation goal in all cases.
For the 88-9 Modules, the '0' values for the (ad)
data are generally higher than those lor the (pi) data,
which is indicative of increasing stabilization in module
performance. Further and in concert with the forgoing,
the '30 year' projections for the (ad) data are generally
lower than tpose for the (pi) data. The B coefficientsfor
the (ad) data are still to high. Those for the (pi) data are
generally in an acceptable range, i.er., close to a
normalized 3.9 value. Note that some of the average
values are negative which indicates considerable
recovery following the strong initial degradation. It is
expected that one or two additional years data will show
a leveling of these values and perhaps the anticipated
downward trend.
The general trend in the projected performance is
positive and shows real improvement. The real values
for the efficiencies are still too low, however, and require
continued ingenuity in the technological developments
to come.

5. CONCLUSIONS

It is early to draw definitive conclusions on the

eventual performance of the later technology modules.
These modules show improved stability following strong
initial degradation over the earlier modules.
The initial degradation of the '88-9 modules is
accelerated when compared to the '85-7 modules.
However, the performance appears to stabilize after two

390
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

EFFECT OF UNUSUAL EXPERIMENTAL PARAMETERS IN R.F. PLASMA

DEPOSITION OF AMORPHOUS'SILICON BASED MATERIALS

G. Bruno, P. Capezzuto, G. Cicala and P. Manodoro

Centro di Studio per la Chimica dei Plasmi.
Dipartimento di Chimica-Universita di Bari
Via G. Amendola, 173- 70126. Bari (Italy)

ABSTRACT.The effect of frequency of the r.t. field (13, 27. 75 MHz) and of plasma modulation
(l0-100 Hz) on plasma cOIrq:)osition and material properties is investigated in the plasma-
enhanced chemical vapor deposition (PECVD) of amorphous silicon and silicon-germanium
alloys. The growth rate of a-'Si:H from SiH4-H2 plasmas is found to increase with frequency,
as well as the electron temperature in the plasma region near the grounded deposition
electrode. Modulated wave (MW) plasmas of SiF4-H2 and SiF4-GeH4-H2 mixtures are shown to
produce a-si:H,F and a-SLGe:H,F films, respectively, with higher homogeneity (in both
thickness and composition) than that observed in continuous wave (CW) discharges.

1. INTRODUCTION Some results on material char'acterization

are also reported and preliminarily discussed.
Although the plasma-enhanced chemical vapor
deposition (PECVD) of amorphous silicon films 2. EXPERIMENT
(a-Si:H, a-Si:H,F) and related materials (a-
Si:Ge, a-Si:C, a-Si:N) /1/ is already used in 2.1 Reactor configuration and experimental
the industry for solar cells and optoelectronic conditions.
devices, considerable research efforts have been The experiments reported here have been
focused, in the last few years, on some aspects carried out in two apparatus having different
which are still considered unsolved. configuration (hereafter reported as reactor
In particular, the identity of the growth type-A and -B), but equal reactor geometry and
precursors, the plasma-surface interaction, and equipped with the same diagnostic techniques.
the microscopic parameters affecting the film The type-A apparatus includes cwo UHV chambers
quality are problems still debated in the (reactor and load lock chambers) whose base
scientific community. vacuum is in the 10- 8 torr middle range, while
On the other hand, a wider technological type-B apparatus consists of a single chamber
development of PECVD technique i.n the amorphous with a base pressure of'10- 6 torr.
silicon field requires further important goals The reactor has a typical parallel-plate
to be reached: (a) the increase of the film geometry. The diameter of the electrodes is 10
deposition rate, for low cost production; (b) the em , and the electrode gap is 3 cm and 5 cm for
improyement of surface homogeneity, for the type-A and type-B reactors, respecti.vely. The
scaling-up process to large area devices; (cHhe r.f. power is fed to the lower electrode, while
achievement of the topmost material quality, for the upper electrode, holding the substrate
high efficiency devices. (Corning glass-7059, silicon and molybdenum) and
:TO give answers to the above open questions· equipped with a temperature controlled heater,
and eo achieve the technological goals, two is grounded.
different approaches are normally used. First, The study of the frequency effect in the
the optimizq.tion of the process and material is deposition of a-Si:H from SiH4-H2 mixture has
carried· out by acting on the external been carried out in the apparatus type-A.' Three
macroscopic plasma parameters (r.f. power, different r. f. generators working at 13, 27 and
deposition temperature, gas pressure, feed 75 MHz have been utilized; the r.f. power has
composition, and reactor geometry). Second, the been varied between 5 and 50 watt at each
role played by internal plasma microscopic frequency.
para-meters (electron density, electron The plasma modulation effec~ has been
temperature, density of neutral and ionic investigated, in the apparatus type-B, for the
species, etc.) on the deposition process and deposition of a-Si:H,F and a-Si,Ge:H,F from
material quality is investigated. SiF4-H2 and SiF4-GeH4-H2 systems, respectively.
Recently, some workers have addressed their The r.f. power is supplied through a linear
attention to the study of novel growth methods amplifier by a 27 MHz generator which can also
in which unusual parameters, such as frequency produce a square wave modulated signal in the
of r.f. field 12-4/, plasma confinement 15/, audio frequency range at different periods (10-
light irra..diation /6/ and plasma modulation /7- 1000 Hz) and duty cycles (15-75%). Growth
9/ are involved. Among chese, the frequency of conditions for a-Si:H, a-Si:H,F and a-Si,Ge:H,F
the applied field in the VHF band range (30-150 films are summarized in Tab. I
MHz) and the square wave modulation of the
plasma in the audiofrequency range (1-1000 Hz) Table I-Discharge conditions for the deposition
see!!', to be parameters strongly affecting the of a-Si:H,a-Si:H,F and a-Si,Ge:H,F films.
surface.kinetics and the fluid dynamics of the
process, as well as the film properties. material a-Si:H a-Si:H,F a-si,Ge:H,F
In this paper, the effect of frequency and
plasma modulation on the deposition of ~~orphous gas mixture SiH4:H2 SiF4:H2 SiF4:GeH4:H2
silicon and silicon-germanium alloys is gas composition 1:9 10:1 10: (0.005-0.1):1
analyzed. Results on plasma diagnostics and flow rate(sccm) 20 11 11
material characterization are reported. In the pressure (torr) 0.3 0.3 0.3
first part, the effect of the frequency of the dep. terrq:). ('{:)a 350 300 300
r.f. field (13-75 MHz) has been studied in the r.f.power (watt) 5-50 30 30
deposition of a-Si:H films from SiH4:H2 frequency (MHz) 13,27,75 27 27
mixtures. Data on the growth rate and on some modulation:
internal plasma parameters are discussed. In the period (ms) 10-100 10-100
second part, the effect of the plasma modulation dllCY cycle (%) 15-· 75 15-75
(lO-100Hz) has been studied in SiF4-H2 and SiF4-
GeH4-H2 systems for the deposition of a-Si:H,F (a) nominal value.
and a-Si.Ge:H,F films, respectively. To this aim
ti.me resolved optical emJssion spectroscopy 2.2 plasma diagnostic techniques.
(TROES) has been utilized and the obtained data The diagnosis of the plasma process and in
allow to argue on the origin of emitting particular of the plasma phase, which is a key
species. on the formati.on kinetics of the active point of our approach, has been performed by
species, and on the improved surface means of four diagnostic techniques:
homogeneity.

391
Time Resolved Optical Emission Spectroscopy reported diagnostic techniques.In the following,
(TROES), for the analysis of the emitting the obtained data are separately presented and
species present in the plasma and to measure discussed.
their time evolution during the modulation. The
light emitted from the discharge' is monitored 3.1 Effect of the applied frequency in SiH4-H2
through a sapphire window by a gatable system.
intensified array detector (PAR-OMA III) Figure 1 shows the trends of deposition
attached to a Jarrel-Ash triple-grating rate, ro, at three_ different frequency (13,
monochromator. The emission lines of F, Ge, and 27,75 MHz) YS. tm applied r.f. power.
H atoms, and the bands of SiF3, SiF2 SiF, and
GeH have been detected. Each emission intensity,
followed as a function of the experimental
:3
conditions, has been taken as a measure of the
relative density of the correspondent species in
.....
Q
II
the excited state. The plasma region examined
with OES is close to the deposition electrode. ....CII
....
-
Mass spectrometry (MS). A quadrupolar mass
spectrometer (QMS-311, Balzers) has been Q'
2
utilized to control the gas feeding mixture and
to detect the end gas products. In particular, ~

..
the analysis of the mass spectra has provided c:
the values of the total depletion of the feeding .~
0
gases.
'ft
Langmuir Electrical Probes (LEP) , utilized to
fulfill the need of identifying the microscopic ,.•
parameters of the discharge lelectron &
concentration and temperature), which are the
real 'fingerprints' of a plasma reactor. LEP 0
measurements have been performed in single (SEP) 0 10 20 }O 40 50
and double floating probe (DFP) configuration. r .f .power, (watts)
The analysis of I-V curves provides the.values
of electron density (ne) and temperature (kTe) Figure I:-Deposition rate of a-Si:H films
and roughly the electron energy distribution vs r.f. power at three different r.f.
function (eedf). frequencies.
Laser Interferometry ILl), which allows 'the 'in
situ' determination of the deposition rate, rD, The observed increase of ro with frequency
during the growth and consists of He-Ne 1mW confirms the results obtained by Curtins et
laser and a photodiode detector. The ro values al./21 in plasma fed with pure SiH4, In
are calculated from the duration (4t) of the addition, the trends of rDwith increasing r.f.
oscillations which occur as a result of optical power can result from two different
interference between the rays reflected at the phenomena: (a) the parallel increase of ne 114/,
outer surface and those reflected at the which explains the almost linear trend of rD in
interface with the substrate. the low power region, and (b) the strong
depletion of SiH4, which accounts for the
2.3 Material characterization. observed decrease of ro at higher frequencies in
The diagnostic techniques utilized for the the high power region.
material characterization have been addressed to As for the frequency effect, the discussioh must
the evaluation of the chemical composition of take into account the important role of'the
the samples by IR spectroscopy and X-ray discharge anatomy; i.e. the spatial profiles, in
microanalysis, to the optical gap rneasure by the electrode gap, of the internal microscopic
optical transmission spectroscopy (OTS) and to parameters such as electron temperature', kTe,
the evaluation of electrical properties. and electron density, ne, the shape of the
The lR data have been taken by a Bornern FTIR electron energy distribution function, the
spectrometer Michelson 100, immediately upon floating and plasma potential, the reduced
removal of the samples from the reactor. Films electrical field. Figure 2, for instance, shows
on c-S1 substrate have been used for the typical ne and kTe profiles measured by DFP
evaluation of typical bands involving S1-H, Ge-
R, S:-F bonds and for the rough evaluation of technique in SiH4~H2 plasmas at 13.56 MHz.
the H content 'in the films /10/. Films deposited
on molibdenum have been utilized in X-ray
microanalysis (EDX-SEM, Stereoscan 360 !pi
Cambridge) for the evaluation of both Ge/Si
composi tion and film thickness /11/. ;;; 8r fie Up
I 1 60

~ .f r'
The O";'S data, for photon energies in th'? range
1.4-3.0 eV at room temperature, have been sne:lIn .....
obtained by means 'J! 3. Car'y 219 spectrometer.
..
~

~4r /
120 ;
The films used for these measurements were
deposited onto C'~'rning glass substrates. The E
absDrbance ,:Ul.ves were analyzed using the methc)d
reported i:ly Freeman and Paul 1121. 'l'he optical ~
80 ...
gap values Eq have been obtained from the linear
extrapolation of Tauc's plot /13/. ,.
'0
The electrical characterization has Deen
performed, in coplanar Ag-contact configura~ ,.,
cion, to measure dark conductivity (OD) and
pho:o-conductivity 10PH), also at dIfferent ~ 2r 40 ~

illuminations (OpH IY). Measurements of the

conducti-lity as a function of the temperatu-re : of l.._ _ _ _ _ _ _' - - _........--=---'-_--'-_--'
have been done for the evaluation of the
activation energy of the conductivity (Eo).
o
anooe
10 20 }O
caCflO<Je
DIstance, (mm)
3. RESULTS AND DISCUSSION
Figure 2:Electron density (n e ) and
The frequency and plasma modulation effects temperature (kTe) spatial profiles measured
on internal parameters of the plasma phase, on in SiH4:H2 plasma at 13.56 MHz. Dashed line
the deposition rate and on the material is the reduced field E/P.
properties have been studied with the above

392
In the ,s ame figure, the estimated reduced the data reported in the inset of fig,4, Here
electrical field (E/p) profile is alse reported trie Ge content, XGe. in a-Sil-xGe>::H ,F alloys,
115/. As it can be seen, a very complex anatomy deposite1 under cw- and Mt-v--d isch arges, is
is present in these sys tE:lTIS, in t~1a.t. a rna rked plotted as a functi on of the axial position
variat.ion of n e , kTe and E/p occurs, going feom

r----
along t.h~ flow
,------ di.r~ct.ion.
._---
l h .
el ..?ctrode ::~urfa-::es to tl~e cent€~r of th.e r,lasma.
£ ~-'--l 80 .,'iVI47 i
d
10 ~ / \
\
i
)20 0,6
1
; 00 ~,<-n--;;v4P=1 ']

\ iI ....
f I
I t ~40 ~ ' ,_ • I
I
!,"3. . .. ~ ...... i

" or
I
j
.... 8r-r
\fil.' 16 ., CW44'- -,
I ......
pi) , 11
-.-------
r'

...... 6~
.
1;;;----+'' .
~
.iF />-~~.\'.____
' \
,! " • ..:
0,4 ~
I
o! i

..
~ 10 20 30 40 50 60 70
12
~
I
fI
/!/;i ,. ' n·...'
( /
\\ 'B~~
. / )l
"
0'>
...
f AXiOI pos1t1on, (mm)

"" 4~ ~\,/.-/ ... : ~

= -
a,!
d ~'---'
kTe
.,,~, 8
.... 0.2 r
. o a-SI'H,F
~___ ~.. ---==--~_ ;-~-A

I
"')~
-'::::="-V-'~-- \.
I

14 ••
[ 1'0
~
o • a-51 ,Ge:>1,F

\0
~ ,

I , 0 0

- "
I I
0 1 I 0'~~~~1------~~~~~10~2r-~--~~~c~w~~
0 10 20 30
anode c3thoae ModulatIon period, ems)
Distance, (mm)
r-igure 4:Troend of un iformity index (UI) in
Figure 3: Electron temperat,ure U:Te) and· modulated plasma deposi.tion ::i.S a tUDe t ,i,)n of
~lectron density (ne) profiles ITleasured in the modulation period. The inset shows the'
At' plasma at 13(./0), 27tAJ'A'j and 75(.,0 ) G02 content (x GE=) in a-·;::;i,Ge:H,F films
~'1Hz (p=C.3 t':,:!:"r , ~\]= 10 !t/att, <PAr'::' 10 seem,. d f2'p:)s i t8d und-2l' HW and cw ,:ondi t iC:1S :,L$:.
a:·:i.al position a.long the flow direction.
Preli minari ly, this t~~lpe of s t.udy has been d''JnE~
on Ar plasma at three fr~quency values in che As for the efft~ct on cl'iEmical, opt i cal and
Cype-A reactor. For the sake of limited spa (:e~ ~lec'.::.rical oroc(,~r t. ies, Drelimina'c'l l(::sul ts on a-
in fig. 3 only trle data of kTe and Ee, rr.easul"ed Si:H,F' sample'z are summarized in Tab.II. The
by single Langmuir proLe, dre Ytf'Jortec1 . It can data of two representative ~'~-samples obtained
be noted that the 13 MHz plasma is characterized at the same period (100 msec ) and at different
by a large asymmetrical dishornogeneity in both duty cycles (DC.z..IW-24=15%, DCMW-25 cc75%) are
kTe 'and ne values, which is also associated \d.th compared to those measured fc,r a typical CW-
a vf2ry thic}: sheath at the powerf2d electrode, At sample. The increase of 11ydrogen content and of
higher frequency kTe and n8 profiles are the photo-to-dark cOIJductivity ratio are I,orthy
smoother C!nd correspondingly the cathodic sheatll to be underlined.
becomes very thin. The fact that the plasma
becom es more eneraetic near the q rounded Tab. II-Deposition rate and electrical, optical
depo£itio:1 e lectr ()de is the link -wi th rD and chemical properties of a-Si:H,F samples
increasing with frequency. deposited under continuous-wave (ew) and
As fQr the material characterization, films modulated-wave (MW) plasma conditions.
grown at similar deposition rate and different
frequencies have been compared, in order to sam;21" 32-':W 24-HW :S-M'il
isolate the frequency effect from those due to Pr,~'perties
surface kinecics. Preliminar results have showr, rD, (A/ sec) 1.5 0.8 1.4
no d iff erence among them. Typical data for
°D, (Q em) -1 1.lxlO- 9 6.1xIO- 1O 1.4z10- 9
sarnples depositec. at 0.9 A/sec ar e as follows:
sD=lO-12 (fiCIr)-I, OPH/OD(AM1)=5'10 6 , y- °PH, cm)-l 1.2:<1 0 -. 5
( fi 3.4x10-· 5 6.;;::10- 4
factor =1) . 92, Eg=l. 83 eV, £0=0.92, Cii=18%. OPH/OD 1,lxl0 4 5.5xlO 4 4,5xJ.05
y 0.79 0.81 0.89
3.2 Effect of plasrr.a mo6ulaticn in SiF4-H2 and Eo, (eV) 0.8 0.86 0.84
SiF4-GeH4-H2 systems. Eg, (eVi 1. 76 1. 8':: 1. 80
Th~ aill, of this investigation is to decrease <;;li~'.1.L 8 13
the dishomogeneity of the film thickness and
then of ·the deposition rate along the gas flow Figure 5 shows a typical tridimensional i)E~:
direction. In fact,a common aspect in the spectnml in the 200-500nm spectral region where
material deposition from these systems is the the emissioll bands of SiF*(443nmi SiF2*
strong depletion of H2 and GeH4 with respect to (395nm) and SiF3* (2 40mn ) are present.
SiF4; the direct consequence is that the
material is frequently f ound to be non-uniform
in thickness and composition. The on and off
discharge switching during the plasma modulation
(MW) strongly decreases the concentration
gradients of the species responsible for the
arowth, with reSDect to the continuous-wave
discharge (CW). '
preliminary results on film thickness and c'n
material composition measured along the gas flow
direction show a positive effect of the plasma
modulation; the data are reported in fig. 4. The
lower values of uniformity index, ur ( UI=(rmax-
rmin) !2 rav; suffixes max., min. and av. refer to
Wa"elenGth, (p hl. l)
maximum, minimum and average values of the
deposition rate), are indicative of a much Figure 5: OES tridimensional spectrum
higher homogeneity going from CW- to MW obta ined fTom SiF4-H:2 plasma (modulation
discharges.Also the compositional homogeneity is p e riod 10 msec, duty cycle 50%)
strongly improved as it can be easly read from

393
The acquisition of each spectrum occurs every H) also in the ground state and on their
200 ~sec. From a rough inspection of the figure formation kinetics;
the following are the points for a worthy -the surface homogeneity in a-si:H,F and a-
discussion: (a) the non-zero values of the Si,Ge:H,F film deposition is significantly
emission intensity of SiFx* sp~cies at the improved in modulated plasmas with respect to
discharge ignition point (t=O) are certainly continuous plasmas.
due to the fact that SiFx radicals survive in The effects of these unusual parameters on the
the afterglow region, as the origin of the structural, chemical, optical and electronic
properties of a-Si:H, a-Si:H,F and a-Si,Ge:H,F
SiFx* is from a direct electron impact samples are presently under study and will be
excitation of the SiFx/16/ reported later.

SiFx + e -+ SiFx* + e REFERENCES

(b) a stationary value for all SiFx* species is (1) A. Matsuda, Pure & Appl. Chern. QQ(5) , 733
reached at increasing time values going from (1986) .
SiF3 to SiF2 and SiF species. Hence, the rate (2) H. Curtins, N. Wyrsch, M. Favre, V. Shah,
constants of the process Plasma Chern. & plasma Proc.2(3), 267 (1987).
(3)A. Matsuda, T. Kaya, H. Tanaka and K. Tanaka,
Si F4 + e -+SiFx + (4-x)F+ e (x=3,2,l) Jpn. J. Appl. Phys. 21, L567 (1984).
(4) S. Oda, J. Noda and M. Matsumura, Mat. Res.
accordingly decrease with decreasing x value. Soc. syrup. Proc. 118, 117 (1988).
The above discussed points are better evidenced (5) T. Hamasaki, M. Ueda, A. Chayahara, M.
in Fig. 6, where the time evolution of ISiF~ is Hirose and Y. Osaka, Appl. Phys. Lett.
reported at duty cycles ranging from 15% to 44 (11), 1049 (1984).
75%, for a 10 msec constant modulation period. (6) M. Vieira, R. Martins, A. Masarico, 1. Baia,
F. Soares and L. Guimaraes, Mat. Res. Soc.
Symp. Proc. 118, 113 (1988).
(7) K. Ebihara and S. Maeda, J. Appl. Phys. 22,
5000 2482 (1985).
(8) L.J. Overzet and J.T. verdeyen, Appl. Phys.

-
..... 5000 Lett. ~, 695 (1986) .
ft (9) G. Cicala, D.L. Flamm, D.E. Ibbotson and
J.A. Mucha. in Plasma surface Interactions
; 4000 and Proc. Mat., edited by 0 Auciello, A.
oy
Gras-Marti, J.A. Valles-Abaca and D.L.
"': 3000 Flamm, NATO. ASI-Series, vol .. 176 (Kluwer
..... Acad. PUb. 1990) , p.l?1 .
10... (10) G. Bruno, P. Capezzuto, G. Cicala, P.
iii 200C
..... Manodoro and V. Tassielli, IEEE Transactions
on Plasma Science ~(6), 934 (1990).
1000 (11) H. Yakowitz and D.E. Newbury, in Proe. 9th
Ann. SEM Symp. Chicago: IITRI, 1976, vol. I,
p.1S1.
(12) E. C. Freeman and W; Paul, Phys. Rev. B
2 4 5 10 ~(2), 716 (1979).
T1me, (ms) {l3) J. Tauc, R. Grigorovichi and A. Van:cu,
Phys. Status Solidi 15, 627 (1966).
(14) G. Cicala, G. Bruno, P. Capezzuto, P.
Figure 6:Time resolved optical emission Manodoro and V. Pische 9th Intern. symp.on
intensities of SiF* in SiF4-GeH4-H2 plasma Plasma Chem.edited by R. d'Agostinovol.
as obtained with modulation period of 10msec III, 1405 (1989).
at five different duty cycles. (15) Y. Ohmori M. Shimozuma and H. Tagashira J.
Phys. D: Appl. Phys. 19, 1029 (1986).
F"urtherly, the increase of ISiF' values at t=O (16) G. Bruno, P. CaDezzuto and G. Cicala J.
for higLer duty cycles is due to the increasing Appl. Phys. 69(9) ,-I 11991).
am~unt of SiF radicals survived in the
afterglow re~ion. From the kinetic p,)int of
vie~l( a plot. of ISiF*(t.:::-:i)) values as a function
of the afterctlo'w duraticin t.im-= ",vill ali.o\o} us to
evaluate th~ SiF lifetime. Also, ttle analysis
of the ISiF~ profile during the time on of the
discharqe will aive information on the SiF
fornation kinetics. This kine~ic approach is
beeing developed. and. extend'-2d tel other emi tting
spe.~ies .

CONCLUDING .REMAR.KS

The preliminary results on the effect of

freqlJency and plasma modulation in the plasma
enhanced chemical v"apor d~position of amorphous
silicon and silicon germanium alloys indicate
that:
- the increase of ehe electron temDerature in
the plasma region near the grounded-· electrode,
going from 13 MHz co 75 MHz activated
discharges, is able co ezplain the increase of
about 5 times in the deposition rate of a-Si:H
films from both pure SiH.1 and SiH4-H2 systems;
-the time resolved optical emission
spectroscopy (TROES). applied to modulated
p2.asIHas, is a p:)\-.ierful tool fJr <:haract'2rizing
the plasma phase. TROES provides us with
semiquantitative ir~fc;rmati_on3bout the density
)f elr.itting species U3iF3, SiF'2, SiFt Geh, Ge,

394
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

INTRINSIC LIMITATIONS OF SINGLE p-i-n a-Si SOLAR CELLS

Jose M. Ruiz

Instituto de Energia Solar. E.T.S.I.Telecomunicaci6n

Universidad Politecnica de Madrid.
Ciudad Universitaria. E-28040 Madrid. Spain

ABSTRACT. The behaviour of single p-i-n structure amorphous silicon solar

cells is analysed by means of a new transport model which, unlike common
models, takes into account the contribution to electric current from band-tail
carriers. First order solutions undoubtedly show the need for including tail
transport, if the experimental upper efficiency of a-Si solar cells, around
10%, is to be accounted for with realistic parameters. Approximate analytical
solutions, aimed for sound physical interpretation, are obtained in this paper
for both shortcircuit and opencircuit conditions. Band-tail transport is shown
to impose unavoidable (intrinsic) limitations to the opencircuit voltage
rather than to the shortcircuit current. Those limitations being the ultimate
reason for the low practical efficiencies of a-Si solar cells.

1. INTRODUCTION the current generation nor recombination. This is

an idealized (optimistic) case but representative
After several years of intensive and worldwide of technological upper limits.
experience with a-Si solar cells, an efficiency
saturation point, near 10% for single p-i-n
structures, seems to have been reached. The fast 2. DISCUSSION ON CONVENTIONAL MODELS
efficiency growing from the initial few per cent
has been the result of improvements in the material Leaving aside qualitative or semi-empirical
quality, mainly through the decrease of the gap DOS models used in the first stages of a-Si solar cell
(density of states). However, the 10% efficiency development, many different quantitative models
practical limit is far below the predictions of have been presented in the literature (references
some theoretical models and one demands why the /2/-/8/ constitute a partial list, far from being
growing tendency nearly stopped in the more recent exhaustive). In spite of their different approaches
years, in spite of the increased technological all of them rely on the same basic principles, that
efforts. Actually, slight improvements over the 10% is: the coupled set of Poisson and current
have been obtained, but at the price of increasing continuity differential equations, but using the
the complexity (homo- or hetero-structure tandems) same free carrier appro~imation as in the
and with the additional benefit of improving the crystalline case. We will later discuss on this
stability. In spite of this, two principal crucial point.
questions remain from the above panorama: Are there Differences among models mainly stand on their
some intrinsic reasons for the low efficiency objectives, solving methods or set of parameters
limit, specific of amorphous materials, not taken used. Specific a-Si DOS distributions were not used
into account by conventional models? And, if so, in preliminary numerical /2/ or quasi-analytical
do the intrinsic limitations mainly affect the /3/ models. Exponential-tail shaped DOS was
shortcircuit current, the opencircuit voltage or introduced in /4/ and has been used for
both? case-to-case dark current (and opencircuit voltage)
Answers to the above questions have been numerical analysis /5/ or to derive analytical
anticipated in a previous paper /1/, shortly approximations /6/. More recent numerical models
expressed, respectively, as: 'yes, because of /7/,/8/ (extensions of that in ref. /5/ with quite
band-tail carrier transport' and 'limitations different parameter values) tend to offer a more
mainly affect the opencircuit voltage'. These complete (and complicated) picture, with multiple
answers have been based on the development of a recombination paths, through both the exponential
transport model, specific for p-i-n a-Si solar tails and D-states, thus involving a large number
cells, which includes the usually neglected of parameters.
contribution of band-tail carriers to the electric This growing complexity of a-Si models
current. reflects the apparent difficulty to characterize
It is the main object of the present paper to the material by a small number of averaged or
deepen the understanding of the band-tail integrated transport parameters, because the many
conduction model (i.e. a model in which conduction uncertainties (and controversies) that still remain
is dominated by localized band-tail states as in a-Si knowledge.
opposed to standard models in which conduction is The basic transport equations in the free
dominated by free - or extended state - carriers). carrier approximation are used without question by
A breef discussion of common literature models will all models. This approximation comes, in fact, from
allow us to develop a parallel formalism for both a parallelism between the concepts of mobility gap
the standard and band-tail models. Then, in order in amorphous and energy gap in crystalline
to obtain simplified close form analytical materials. It implies that highly photoexcited
solutions, with a minimum number of internal carries relax to near and above the mobility edges
parameters, and to compare them, a p-i-n structure in the very short relaxation time, and then very
wi th transparent emi t ters wi 11 be considered. By slowly by free carrier transport and recombination
transparent emitters we mean here p and n regions through gap states in the much longer lifetime. The
whose contribution is limited to the establishment populations of extended band states (free carrier
of the built-in voltage but neither contribute to densities) are linked to their respective

395
quasi-Fermi levels, but the populations of The preexponential factors being effective
localized gap states (both deep and on the tails) densities of states (per unit volume) and Eet , E t
are conjointly linked to both quasi-Fermi levels. the characteristic energies of the exponentially
Moreover, carriers trapped in these states are decreasing band tail DOS (about 27 meV and 43 meV,
assumed not to contribute to current because of the respectively in a-Silo It is to be noted that
low hopping mobility. expressions (2) take this form because the
However, a different picture is perhaps more characteristic energies are higher than room
plausible in amorphous materials . We think that temperature thermal energy kT. Otherwise, this
photoexci ted carriers initially relax to the thermal energy should replace E or E (or both)
mobility edges and also down to band-tail states in (2) and the approximations rt~ f+lv~ill no lon-
within the short relaxation time, and then slowly ger hold. This is, for example, what happens in a
by conjoint free and trapped carrier transport and similar but "weaker" crystalline case: that of
recombination through deep states. That is to say heavily doped (degenerate) semiconductors.
that band tail carriers strongly interact with
nearest band edge states and very weakly (because
energetically unfavorable) with the opposite band. 3. FREE CARRIER VERSUS BAND TAIL MODELING
The population of both extended band and band-tail
states are mutually linked to the corresponding Modeling band-tail limited transport is
quasi-Fermi level and is only the population of performed with the same basic formalism (Poisson
deep states which is linked to both quasi-Fermi and continuity equations) as in the free carrier
levels. In support of this hipothesis we argue the approximation. Because free and band-tail carriers
results of time-of-fligth experiments /9/,/10/ are assumed to conjointly contribute to drift and
which directly provide both the drift (combined) diffusion currents and participate in generation
mobility and lifetime, rather than merely the free and recombination processes through other deep
carrier ones. These are obtained, a fortiori, by states, only current expressions are to be
the model assumption of equal 11. < products for redefined in terms of the global (added) carrier
drift and free quantities. densities (n, p, for electrons and holes by the
Moreover, we think that, in spite of the low, following) as well as the corresponding effective
but not zero, hopping mobility band tail carriers transport parameters (11, D and T with e or h
do contribute to current transport, especially if subindex). For the one-dimensional case:
minority carriers are dealt with. In support of
this hypothesis we note that experiments dealing J qll n/;+qD (dn/dx) qll [nt;+V (dn/dx)] (3)
with drift mobilities are usually performed in e e e e et
conditions where transport is dominated by majority
carriers. In these circumstances the free to J qllhP/;-qDh(dp/dx) qllh[pg-Vvt(dp/dx)} (4)
h
localized majority carrier density ratio is not so
low as to compensate the high free to hopping with:
mobility ratio and hopping conduction may be llefnf+llelnl llefnf+llelnl
unimportant (perhaps for electrons, not probably 11e = D (5)
n +n q(n /kT+n /E )
for ·holes). However, transport in diode or p-i-n f 1 f 1 ct

structures is dominated by drift-diffusion and

recombination of minority carriers and, for them, IlhfPf +Ilhl P I Ilhf P f+ llhl P I
the free to localized carrier density ratio may be Ilh D (6)
Pf+P 1 h q(p /kT+p /E )
much lower ' than the hopping to free mobility ratio, r 1 vt

thus making hopping conduction increasingly D n +n E

important so as to dominate transport. V =
e f 1 ct
et q(n /kT+n /E )
(7)
Illustration of the precedent paragraph is 11e q
f 1 et
given. in figure 1 where calculated equilibrium
carrier densities are represented as functions of D Pf +Pl E
V =
h vt
the Fermi level position. In the general non- vt q(p /kT+p /E ) '" (8)
Ilh q
f 1 vt
equilibrium case they can be expressed in condensed
forms in terms of the quasi-Fermi energies/1/: Equations (7), (8) rule as modified Einstein
relations for the global carrier densities. They
E -E E -E are obtained from equilibrium conditions (null
n =N exp-~ p =N exp-~ (1) drift+diffusion current for every group of
f ef . kT f vf kT
carriers). The approximations in their final
E -E E -E members hold as long as n1»n. and PI»P f , which is
n =N exp-~ '" n P =N exp-~ (2) usually the case. Note also tnat these results are
I el E f+l I vI ' E
et vt independent on the existence of hopping conduction .
Expressions for current densities (3),(4) are
102 ' formally identical to those in the free carrier
- - free • localized model, provided both V t and Vvt are replaced with
- - free
". 10' j/ the thermal voltage kT7q, often denoted as V. The
'E parallelism is even higher (especially for §olving
~ purposes) if the effective mobilities are constants
>-
f- or quasi-constants . For this, little more than non
(i; lO' zero localized state mobilities is required, as
Z
w deduced from equations (5) and (6).
0 lOa
c: Concerning recombination processes, the cross-
w relation between both basic models is established
c:c: 10 6
through the assumptions:
l3 10 3 11 e T e = 11 ef < ef (9)

with the experimental values /10/ (N being the

integrated density of deep states): S

Il< N 2.5.10 8 (V.cm)-l Ile ~ 0.5 cmz/V.s

e e s
Figure 1. Free and localized carrier densities as PO)
-3 Z
functions of the Fermi level position.
IlTN
h h s
Il ~3.10 cm/V.s
h

396
4. FIRST ORDER SOLUTIONS (n' (x) and p' (x». Having in mind that the electric
field is high and we are especially interested on
The case to be solved is that of a one- the shapes of n' for x»O and of p' for x«W (the
dimensional p-i-n structure with transparent p and equil ibrium crossing point is near the middle of
n regions for both generation and recombination, the i-region and, in any case, sufficiently far
that is, limited to the effects on the i-region. In away from the 0 and W I Imi t s) , n' and p' are
actual devices, recombination in the emitters or conveniently expressed as:
interfaces is known to be important and even the
dominant term at opencircuit (also so obtained, n' (x) ~ J (x)/q/l € (15)
e e
with both models, in our previous paper /1/).
Modeling them adds no special difficulties but does Introducing the initial guess for J ,J (from
add a number of parameters without altering the (13) with R~O) into (15) and recalcul~tin~ (13)
fundamental conclusions searched in this work. with this result (as a sum of monochromatic
Other simplifications required for our first components) one obtains, for the a-dependent
order analytical solutions are the following: quantum efficiency:
Illumination through either the n or the p side
without light confinement (i.e. without back or
(16)
mul tiple internal reflections devised for photo-
response enhancement). Constant electric field in
with (x' denoting W-x ):
the i-region (neglecting its narrow, horn-like, e e
increasing profile near the interfaces tends to
-ax' -ax' -aW
overestimate recombination there, but the slightly
1 = x'-
l-e e
1 =
e e - e
- x e
-aW (17)
lower than the average in the thicker middle zone e a h a e
has the opposi te effect). High internal quantum
efficiency of the i-region, i.e. low recombination for illumination through the p side, and:
at shortcircuit (this is merely a convenient
-ax -aW -ax
condi tion to express first order deviations from
1 =
e - e
- x'e -aW 1 = x -
l-e e (18)
the maximum in a compact manner). a h a
Wi th the x-axis pointing in the n.,i.,p sense
(in which the net photocurrent flows, and the for n-side illumination.
electric field is positive) and using 0 and W to The position of the crossing point is slightly
denote the limits of the i-region, a first spectrum dependent, both on its shape and intensity
integration of (3) and (4), with constant e, gives: (see figure 2 for illustration on a particular
monochromatic case). But the crossing point to be
JXJe(U) e(x-u) considered in eqs. (16)-(18) is the one
ex
n(x) n(O)exp-V- + ------ exp------- du (11) corresponding to the whole spectrum (the
et q/leVet Vet equivalence with experiments is with bias-light
0
methods). Actually it can be approximated to equal
e(W-X) +fJh(U) e(u-x) the equilibrium one.
p(x) p(W)exp--V-- exp----V--- du (12)
vt q/lhVvt vt
x

nCO) and peW) represent also the majority

carrier densities in the nand p emitters. For the
band-tail model they equal the respective ionized
donor (N+) and acceptor (N-) impurity densities.
In the fPee carrier model, ~quations (11), (12) also
apply' (with V t and Vvt replaced by kT/q and n/O),
p (W) obtainea from eqs. (1), (2».
f On the other hand, continuity equations, in
their integral form in terms of the generation (G)
and recombination (R) rates, are expressed (for the
transparent emitter case) as:

Je(x) = qIw(G-R)dV (13)

x 'igure 2. Carrier denSity profiles in the i region.
the total current density being: Shortcircuit profiles correspond to
illumination from the n-side with a
J = J e (x) + J (x)
h
= J e (0) = J h (W) (14) monochromatic radiation.

To obtain first order solutions from (11)-(14)

one proceeds as the first step of a more precise From the above expressions, very similar
iterative scheme. It is as follows: An initial results are deduced with the free carrier and band
guess for currents, deduced from (13), (14) is tail models (because the equal /l.T products and the
incorporated in (11), (12) to obtain carrier small significance of their different x ). We thus
profiles. Then, the crossing point x for equal n conclude that band-tail conduction doesenot intro-
and p is obtained. It separates mifiority carrier duce intrinsic limitations on the shortcircuit
zones and allows to perform the recombination current.
integrals in (13) in terms of the over-equilibrium Other interesting features are: The so-called
excess minority carrier densities in each zone, collection lengths (or IlT€ products) should not be
divided by the corresponding lifetime. This added to give an overall collection length. Rather,
is the end of the first order process. the quantum efficiency is limited by the collection
length of those minority carriers in the zone aside
4.1. Shortcircuit. illumination (especially for highly absorbed light:
Under this condition (V=O), g equals its aW»l, the most contributor to total photocurrent).
equilibrium value: ~ /W. The built-in voltage ~ is Because IlT for holes is much lower than for
obtained as E -(E _E b ) -(E -E ) , using eq. (2).b The electrons, the experimental finding of best
first terms iR (lI),f1~) e~uar gquilibrium profiles efficiencies with p-side illumination is fully
and the second ones represent the excess profiles confirmed by ~he model (see figure 3).

397
 5. CONCLUSIONS

Band-tail transport has been shown to have a

0.8
limiting effect on the opencircuit voltage and,
thus, on the efficiency of a-Si solar cells.
0 .6 However, no specific limitations arising from
band-tail transport have been deduced for the
0.4 shortcircuit current .
Two ways seem possible for some (differential)
increments over the highest actual efficiencies
0 .2
with appropriate technology refinements: One is to
reduce light losses (due to reflection, escaping or
o L-L....,O.1..4-'--O""'.-=-S--'----:"01-::.6---'---:0.1..7-'--0=".-=-8--'---' inactive absorption) to increase photocurrent. The
WAvELENGTH (mlcromsl other is to reduce deep states density to increase
lifetime and (somewhat) the opencircuit voltage.
A third way, the most effective if practical,
Figure 3. Internal quantum efficiencies of the is only theoretical: To reduce band-tail densities.
i-region for p and n side illuminations. Because they are intrinsic to the long range
disordered structure of a-Si, this way implies
increasing the "crystallinity" of the material.

4.2. Opencircuit . 6. REFERENCES

In this case the electric field is much lower
than in equilibrium: e = (~-V l/W. Local current III J . M. Ruiz. "Opencircuit voltage limitations of
densities are low, chaRge s~gnOCsomewhere in the a-5i solar cells: A simple model". Tech. Dig.
middle zone and oppose mutually, adding together to 5th Int. PVSEC, Kyoto, 335-338 (1990).
a net zero value. Thus, first terms in (11),(12) 121 G.A. Swartz . "Computer model o( amorphous
are much higher than equilibrium ones (especially silicon solar cell". J. Appl. Phys., 53,
as minority carriers are concerned), and second 712-719 (1982).
terms are (to the first order) negligible. 131 H. Okamoto et al. "Variable minority carrier
According with (13), (14) integrated generation transport for amorphous silicon solar cells".
and recombination must equal at opencircuit. Or, in Solar Cells, 8, 317-336 (1983).
current terms, photocurrent must equal dark diode 141 M. Hack and M. Shur. "Computer simulation of
current. Searching for the electric field bringing amorphous silicon based alloy p-i-n solar
this condition, one obtains: cells" . IEEE Elect. Dev. Let., EDL-4, 140-143
(1983).
151 R. J. Schwartz et al. "Numerical modelling of
V ~ (V +V Jln{qwln(O)p(w)(Vct+V vt )} (19) p-i-n hydrogenated thin film silicon solar
OC b ct vt J (~-V ) T T cells". Proc. 17th IEEE PSC, 369-373 (1984).
L b oc e h
161 K. Misiakos and F.A. Lindholm . "Analytical and
equivalent-circuit models based on numerical
'Equation (19), although somewhat implicit, solutions for amorphous silicon pliln solar
clearly points out the band-tail effects with cells". Proc. 19th IEEE PSC, 243-248 (1987).
respect to free carrier models: ~b is equally 171 J.W. Park et al. "A self-consistent nume~ical
obtained in both models for any given structure but model of thin film silicon hidrogen-alloy
the substracting term is higher in the band-tail solar cells" . Proc . 20th IEEE PSC, 55-60
model because both the logarithm argument and (1987) .
prefactor are higher. Thus, tail transport implies 181 J. L. Gray. "A computer model for the
a stronger reduction of V with respect to ~ . In simulation of thin-film silicon-hydrogen alloy
many cases free carrier mgaels predict V ~alues solar cells". IEEE Trans. Elect. Dev., 36,
close to ~ and the difference with the gg,nd-tail 906-912 (1989) .
.model may ~e as high as 250 mV as illustrated in 191 R. A. Street. "Trapping parameters of dangling
figure 4 for a·typical case (see 111 for details) . bonds in hydrogenated amorphous silicon".
Note also the (logarithmic) dependence of V Appl. Phys. Let., 41, 1060-1062 (1982) .
on lifetimes but its independence on mobilities. oc 110/ R.A. Street et al . "Effects of doping on
transport and deep trapping in hydrogenated
amorphous silicon". Appl. Phys. Let., 43,
1()3 672-674 (1983).
....E with band tail transport
C)
w ithout band tail transport
::(
J L (12 rnA/ em"
)-
l- 10- 3
iii
z /'

//
UJ
0 10- 8
I- ~ Voc
Z
W
a: 10- 9 -250 mV )
a:
::l
(>
10"20
0 .2 0 .4 0 .6 0 .8 1.0 1.2
V (Volts)

Figure 4. Dark current curves obtained with the

band-tail and free carrier models,
showing the Voc shortage because band
tail transport.

398
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON, PORTUGAL

OUTDOOR EVALUATION OF AMORPHOUS SILICON SOLAR CELL MODULES

J.e. JOSHI AND P.I. IONAR

Photovoltaic Laboratory, Centre for Energy Studies,

Indian Institute of Technology Delhi
Hauz Khas New Delhi-110016
INDIA

ABSTRACT: In this study, the field performance of eight different amorphous modules (manufactured ~
~ and ~ manufacturers) has been carried out over a period of three years in Indian climatic
condition. The degradation of these modules hod been evaluated through measuring the modules
efficiency (1t), fill factor open circuit voltage (V ) and short circuit current (Isc) at operating
temperature and solar insolation periodically. °After three year outdoor exposure, the module
efficiency has been degraded from 16% to SO% for different modules. It has been also ob~erved,that
the rated value of these modules are always higher under similar condition (1000 W/m , 2S0C) of
field. The temperature effects on the out put power and V indicates that dp/dT/P ; 0.35 to
0.4S/oC and dV /dT;2.5 mV/oC to 2.7mV/oC. These modules have algg been tested in diffuse radiation
environment (c~8udy day in field). This investigation show, that the power output and efficiency of
these modules except one are linearly dependent of the diffuse radiation as that of the same as the
predicted at global radiation.

1. INTRODUCTION using precision radiometer (epply pyronometer,

thermister). In order to mlnlmlze the variation
Amorphous solar cell modules of various kind in temeprature and solar radiation the data has
are available commercially in market. They have been recorded during clear and sun shine day with
the solar cells of different size and various wind speed of about 1m/sec. The data acquasition
design of encapsulations. The solar cells are system consists of a mUltiplexer, which select the
arranged in series or series parallel combination modules after fixed interval (10 min) of time and
in modules. The basic considerations in the record the maximum power, V , I and radiation.
design of these modules are low cost, high The operating temperature igcrec~rded after every
efficiency and stability. The amorphOUS solar thirty minutes.
cell modules show great promise and modules of
these solar cells with modules efficiency in the The I-V curve has been plotted using maximum
range of 5-S% are available commercially, but the power point tracker or variable load.
detail field study have not been done in Indian
environment. Many investigators (1,2) have been The effect of temperature on efficiency has
conducting field aging studies, accelerated been deter~ine at constant isolation (variation
testing program and application experiment using <SO Watt/m). The cell temperature measured with
broad variety of first generation amorphous respect V . the degradation study of these a-Si
silicon (a-Si) modules. It has been shown [3] modules c~fried out exposing to field condition
that the temperature has similar effect on the for more than three years. The performance data
efficiency in a-si as observed for crystalline Si- has been recorded every month for five to ten
modules. The extensive research (4) has been clear days. Beside of this modules also studied
going on to understand the mechanism such as light for diffuse radiation in cloudy days. Electrical
induced metasta'ble changes and the origin of light parameters have been measured.
induced defect.
The typical field parameters measured of
In the present work field performance of different modules are given in Table-I.
eight a-Si solar cell modules manufactured by
various, manufacturer have been studied and 3. RESULT AND DISCUSSION
compared in Indian environmental condition. The
outdoor testing research is important to identify The typical performance of these a-Si solar
the problem related with modules performance, cell modules at initial stage are shown in table-1
which will further require the laboratory mounted on a fixed latitude (2S.63 degree). It
investigation to modify the thin film flat plate has been found that the under actual operating
modules reliability effect [5]. condition A,C and El modules would be slightly
inferior than the other modulus. The rated values
of efficiencies of these modules are always higher
2. EXPERIMENTAL DETAIL than the 2 observed values under similar condition
1000 W/m, 2S0°C, and lm/s). The efficiencies of
A experimental set up for long term stability 4.4, 5.2, 4.3, 5.1, 5.4, 5.6, 3.44, 5.96 have been
evaluation of a-Si modules is shown schematically observed instead of 5, 5.6, 5, 5.5, 5.S, 6.1, 4,
in Fig.l. The modules of different manufacturers 6.5 for these different type A,B,C,D l ,D 2 ,D3 ,E l and
from A,B,C,D,E have been mounted on the fixed or E2 modules respectively. The power loss ln these
tracking structure. All the modules mounted on modules can be due to ohmic, mismatch, shadow and
the fixed structure (latitude tilt), the surface clear index. The loss in power is about S to 15%
of modules is kept toward to the sun while taking which is much higher than usual ohimic and
instantaneous data. Each modules is connected mismatch losses. The main causes is due to
data cquisition system to provide the individual scatter radition, which is contribute about 15%
performance data. The solar radiation and ambient even on the clear days but do not generate the
and module glass temperture have been measured proportional electric power.

399
Table-I: Typical Field Parameters of Amorphous Solar Cell Models
------------------------------------------------------------------------------------------------------------
Modules A B C D E
------------------------ ------------
D1 D2 D3 E1 E2
------------------------------------------------------------------------------------------------------------
Dimension (mxm) 0.33xO.33 O. 594x1. 204 0.315xO.315 0.338x1.02 0.336xO.36 0.336xO.36 0.31xO.31 0.32xO.32

Cell size (cm) 1. 19xO. 29 4.1x4.1 1.0x30 0.9x99 0.9x32 0.9x32 6x27.7 6x28.2

No. of solar 25 420 29 30 30 30 5 5

cells

Active area of 2 0.0855 0.672 0.087 0.297 0.0947 0.0942 0.083 0.084
solar cells (m )
Packing factor (%) 78.5 93.9 90.0 85 78.3 78.3 85.0 83.0

Electrical Parameters

Insolation Watt/m 2 1000 827 876 883 952 912 929 976

Pmax (Watt) 4.8 29.9 3.7 13.3 4.7 4.9 2.4 4.9

I sc (A) 0.465 2.28 0.264 1.12 0.334 0.349 1.38 1.85

V (V) 19.2 20.8 22.6 21.9 22.8 23.2 3.6 3.8
oc
Fill Factor 0.53 0.60 0.59· 0.54 0.61 0.61 0.53 0.68

Module efficiency 4.4 5.2 4.3 5.1 5.42 5.62 3.44 5.96
(%)
Module temperature 48.0 41.2 41.0 46.6 48.2 50.5 46.0 42.2
(,OC)

Ambient temperature 28.0 31.2 34.0 31.6 35.5 34.5 34.5 31.0
(oG)

dVoc/dT (mV/oC) 2.5 2.61 2.68 2.51 2.50 2.50

dP/dT/p%°C 0.44 0.35 0.37 0.35 0.48 0.48 0.35 0.34
------------------------------------------------------------ -----------------------------------------~-----

The effect of temperatu're on power and V found 80.5, 48, 35, 33, 27, 26, 23 and 17
are shown in Table-I. The result indicate thg£ percentage for modules E2 ,E 1 ,A,D,c,B,D 2 and D3
dP/dT/P for modules are lies in the range from respectively. The degraaat10n occurs due to
0.34% to 0.48%/G, where V decreases linearly ohimic contact, mismatch, clearness index and
with increasing temepratu~~ with slope of light induced effect. The dominant parameter in
approximately between -2.5 mV/oC to -2.7mW/oC per the a-Si material degradation of modules is the
cell. decrease of the fill factor due to a reduced
carrier collection length.
The several types of a a-Si solar cell
modules consists of various dimension are given in The loss in power is 17 to 80%, which is much
Table-I. The long term field performane of the higher than usual ohimic, mismatch and scatter
modules greatly depends on the insolation and cell radiation. The scattering radiation generally
temperature. In the outdoor measurement contribute (3) about 15% in the single crystal
meteorological condition such as illumination even on the clear day, but do not generate
intensity and temeprature of the cell vary day by proportional electric power. In order to find out
day. Therefore the measured out put performance the effect of diffuse radiation on a-Si solar
of the modules is 20rmalized for an standard cells, modules, have been studied on cloudy day.
condition (1000 Watt/m ,28°C). Fig.2 (a,b,c) show The various elctrical parameters have been
the variation of P ,efficeincy and fill factor measured with the variation of diffuse radiation
as a function ofmafiumber of years elapsed. The on the earth surface as shown in Fig.3(a,b,c).
average initial conversion efficiency of these Figures exhibit that out put power linearly
modules lies in the range between five to six increases with increase of diffuse rad~ation at
percent. The figures exhibits that initial higher intensity more than 130 Watt/m and non
degradation occurs abruptly within forty five linear behaviour show at lower intensity. Where as
days, and after that there is a smooth decrease in in the case of efficiency and fill factor, the
out put of electrical parameters except three efficiency slightly increases with increase of
modules. These three modules show another drastic diffuse raditions at higher intensity more than
change in electrical characterisitc after nine 130 Watt/m and than rapidly decreases at lower
months interval. The percentage of degradation intensity. For fill factor no change observed wit2
has been ealuated lies in the range from 16% to increase diffuse radiation more than 130 Watt/m
81%. The aging effect also obtained for P , and at lower inteisty drastic decrease has been
fill factor and efficiency under the same pe~~d observed for fill factor.
sf times. The maximum aging values have been

400
4. CONCLUSION
In this paper we have reported the performance of
single junction a-SiH solar cell mpdules in sun
light and diffuse radiation. These studies
indicate that the performance of flat plate
modules are simialr except three modules. The El MODULES
effect of temeprature on modules power and voltage - - E2
is slightly dffer from each other. The stability l- OUT PUT POWER STAN DARD __ 02
result indicates that in field condition a-Si:H
I-
« CONDITION 11000 WATT Im?28C I -*-- D)
modules show 15% reduction of efficiency in 3= GO A
approximately 45 days. The rated value of
efficiency is not achieved for any of these
...::> ~ :;.= 5,
Q. .~
modules in field condition. The diffuse radiation S )0
effect is more prominent on electrical o --..-.--.........->-
characteristics of a-Si solar 2ell modules at a:
higher intensity beyond 130 Watt/m . 'g 20
o
Q.

REFERENCES 10

1. D.L. Morel, J . P. Rumberg, R.R. Gay and C.F.

Gay, Field Performance of thin film silicon
modules, Proc . of 8th binnial conress, ISES 1987 1988 1989 1990 1991
Perth p.1517, Aug.1983. YEARS
2. Laxmi Mrig, Stove Bummel, Dave Wacddington
and and Richard De Blasio, Outdoor stablity Fig.2(a) Degradation of the power with ~oar years
performance of amorphous silicon modules, exposure of modules.
solar cell 24, 381-386, 1988.

3. Mukash Chander, J.C. Joshi and K.L. Chopra,

Field performance of flat solar cell modules,
Proceeding 8th European photovoltaic Solar
Energy Conference Congress Centre Florence
Italy May, 1988.

--
MDDUL ES
4. H. Sakai and Y. Uchide, Current Progress
toward reducing light-induced degradation in UJ
8 EFfiCIE CY STANDARD -.a..
a-Si:H film and solar cells solar cells 22, <!)
C OITIO N 11000 WATT Im~]8 'CI
275-285, 1987. S
Z
UJ

5. P.K. Konar and J.C. Joshi, Study of long run ~ 6

!oJ
degradation of flat-plate photovoltaic Ilo
modules. Proceeding 21st IEEE photovoltaic ,.:
u
specilists conference Orlando Kissimmea Z
!oJ
Florida May 21-25, p.1080-1084, 1990.
it....
!oJ
2

1987 19BB 1989 1990

YEARS

Fig.2(b) Performance of efficiency with four years

of exposure of modules.
-- ---. I
- - - - -- - , I ~

----- ~' ;~

; ~ , .::t

~. ~ ~r-~ i
~.i;·;
L.:.:,;~ -~
--tII ~ ;::-;-1
I
~ -' ,• •- --'

'------ -- - --- - - - -- - -- - - - - -- - .. .~
Fig.1 Block diagram of outdoor test system,

401
 15

D2 MODULES
D3

--
E2
Fil l FI,e fOR STANOAFIO A MODULES
....4- 8
( O" UI PO - ~- B -"- C
--0- C 01
0.8 -6-· D, A
- E 10 -.- 02
Q:
03
e
u
0.6
0::
uJ
El
Ll!
...J
...J 0.4 ~
G::
'3
n.
'3 5
02 o

o
1=98=6~--~19~8~7------19L8-8------19L8-9------19L9-0----
YEARS

Fig.2(c) Variation of fill factor with elapsed

of years 200 400 600 000 1000
DIFFUSE INTENSITY ( WATT 1M2)

Fig.3eD.) Varictio4 of ?o~.;er lrlit:l diffuse

rc::.dieticn.

MODULE:; -0- C
- 0- 8
6 01
A
D2
La
5 D3
EI
0.8
4
;:::...
i>- ~ ~ MOOULES
u
ffF 0,

!I
Z -0-
uJ - --- -tr- 02
Q 0.4

:=:: 1:
LL
LL
uJ 2

~f~,___~____.~____~__~___
0.2

OL-__
OL-__ ~ ____ ~ ____L -_ _ ~_ _ _ _...J-__
o 200 400 600 800 1000
o 200 400 600 800 1000 DifFUSE INfENSITY (WAT l/1m1 )
DIFFUSE INTENSITY (WolI/m 2)
Fig.3(c) Change of fill factor versus
Fig.3(b) Variation of efficiency with diffuse diffuse radiation.
intensity.

402
HITH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONPERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

ANALYSIS OF PERFORMANCE EVOLUTION OF AMORPHOUS SILICON MODULES

BY EXPERIMENTATION IN INDOOR AND OUTDOOR CONDITIONS

P. RAGOT*, A. CHENEVAS-PADLE**, H.S. COSTA*, D. DESMETTRE*

Commissariat a l'Energie Atomique
* C.E.N. Cadarache DER/SCC/LVT 13108 St Paul Lez Durance Cedex (France)
** C.E.N. Grenoble LETI/OPT BP 85 17, ave des Martyrs 38041 Grenoble Cedex
(France)
E. ROSSI, H. OSSENBRINK, R. VAN STEENWINKEL
Commission of the European Communities
Joint Research Centre-Ispra establishment
European Solar Test Institute 1-21020 Ispra (VA) (Italy)

ABSTRACT

As part of the continuing photovoltaic module evaluation effort, this paper

presents the results obtained over the last eleven months with amorphous silicon
modules. The experiments included both field tests at the Cadarache facility and
laboratory tests at Grenoble and at the Ispra Centres.

1. INTRODUCTION array 5 : American manufacturer D (tandem

junction) ,
The ASILEC programme "Analysis of the array 6: American manufacturer E (p-i-n
behaviour and ageing of photovoltaic a-Si:H junction) .
modules coupled to a load under natural solar
Each array (fig. 1) composed of 36 modules
exposure", managed by the Commissariat a are divided into 2 sub-arrays :
I' Energie Atomique and the Joint Research Centre sub-array 1 modules n° 1 to 8 for
of Ispra (CEC) , as partner and financed by
laboratory tests (t-O, t-3 months,
CEC/DG 12, aims at measuring, identifying and
t=l year),
analysing the quantitative phenomena likely to
sub-array 2 modules n° 9 to 36 which
involve a decrease of performance with time, of
remain continuously in sunlight exposure
a-Si:H generators under natural environmental and
operating conditions. during 2 years.
We present the evolution test results of 216
Each sub-array 2 is composed of 7 strings
commercially available a-Si:H modules of six
connected in parallel to 4 modules connected in
different types made by American and French
series.
man\lfacturers.
The modules of manufacturer A are prototypes The facility (fig. 2) consists of modules
and are not commercialized for the moment.
exposed to sunlight at a fixed 45° tilt angle
(latitude from Cadarache), with switching
2. PROCEDURE
prOVl.Sl.ons allowing a sequence of tests to be
conducted on a various series, parallel and diode
A standard procedure prescribed by the
configuration.
different partners defined complementary
investigations as follows : The data acquisition and processing includes
2 HPIOOO micro-computers, a HP3497 control
·CEN Cadarache: Performance measurements
command unit and a meteorological station.
under natural solar exposure : A filtered cell with the same spectral
module ageing with irradiation and response as amorphous silicon is used for
temperature,
calibration purposes.
daily power output in correlation with The electronic loads are driven by the
meteorological data, micro-computer.
annual power output of a parallel-series
A total of 15 voltage points are used to
module grid connected to a fixed load. plot an individual I(V) curve.
CEN Grenoble : Fundamental analysis :
About 20 minutes are required to scan all
analysis of p-i-n layers through
216 modules.
Secondary Ion Mass Spectroscopy (SIMS)
analysis of defects through the Infrared 3.2 C.E.N. Grenoble (LETI)
Stimulated Current method (IRSC).
JRC Ispra: In laboratory electrical This laboratory has perfected Secondary Ion
measurements : Mass system under ultra-vacuum for the
I(V) curve characteristics, quantitative analysis of different components in
absolute spectral responses, p-i-n layers.
temperature coefficients (current, A new experiment based on the Infrared
voltage, fill factor, maximal power) . Stimulated Current method [1] should allow the
localized state density to be measured in the
3. DESCRIPTION OF THE TESTING FACILITIES bandgap associated with the photo-degradation.

3.1 C.E.N. Cadarache 3.3 J.R.C. Ispra

The 216 modules are divided into 6 arrays : The experimental systems used by Ispra are :
array 1 : French manufacturer A (p-i-n the multi-intensity-simulator (2] for I(V)
junction) , curves,
array 2 : American manufacturer B (p-i-n the spectral response facility [3] for
junction) , spectral responsivity measurement over large
array 3 : French manufacturer C (p-i-n areas,
junction) ,
the climatic chamber associated with a
array 4 : French manufacturer C (p-i-n
simulato~ for the temperature coefficient
junction) ,
measurement.

403
 spectral response for 4 manufacturers. We observe
different response values versus the wavelength
range :
Best responses for modules A and C in short
wavelength (UV, blue range), to a lesser degree
for module E and a weak response for module B.
In the visible range, the maximum responses
(3 A/W) are between 550 and 600 nm for modules A,
C and E, 575 and 625 nm for module B with a
maximum of 3.5 A/W.

ASR (AjW)
~.;-----------------------------~~~~~
---
~--------,

,,-- __ .~ B

!'_~ 'y-- c
0.3
/ "~ y-E
/.-..7 / \:--/ \ ___ A
/1 '. /.
/ ,/ .~\ \

j
0.2
i..' \\ \
Fig. 1 - Scheme of connecting array
. 1 \\ .
V.;,/." \\ \.
0.1 " \

:/ I \\ \
.: i ~ \
o ~ •
300.0 400,0 500,0 600.0 700.0 800.0 900,0

I(nm)

r-;:;aJ~
Fig. 3 - Absolute spectral response for 4 modules
I (Initial measurement t-O)
I In the long wavelength (red, infrared range)
I module B has a better response than the others.
Figure 4 shows the evolution of absolute
I spectral responses for 2 manufacturer C modules
after 3 months of sunlight exposure in different
load conditions :
4a in short circuit,
4b in zener diode load (V-14V).
The evolution is weaker for the module in a
short circuit condition. This effect has been
shown in the past [4,5) . These authors proved
C5 C6 that this depends on the electric field in the i
•••••• layer. For the strong electric field (short
• • • circuit, reverse bias), the decrease of the short
circuit current is less than in the other load
Outdoor les~ng facility
conditions (positive bias, open circuit).
Fig. 2 - Block diagram of monitoring and data ASR (AjW)
acquisition and processing system
... ,------------ -----------,
Al to A6 Swi tching boxes
Cl to C6 Modules arrays
CEl to CE6 Electronic load
11, 12 Printers
M Meteorological station
MI, M2 Micro-computers
Tl, T2 Terminals
TC Plotter
UCC Control command unit
Power cables
Measurement cables
• Temperature sensors
4. RESULTS

4.1 In laboratory electrical measurements

(J . R. C.Ispra)

4.1.1 Absolute spectral response 900

~(MI)

This experiment is carried out for a-Si:H p-i-n

modules at the beginning and after 3 months of 4a - In short circuit
sunlight exposure. Figure 3 shows the absolute

404
 ASR(AfW) ASR(AfW)
n4 ~------------------------------------, OAr----------------------------------,
1 - 0 (Jsc • 12.01 ""'.an·2)

G.J

02
\,

01
\
~(nm)

5b - Module E
4b - In zener load
Fig. 5 - Evolution of absolute spectral response
Fig. 4 - Evolution of absolute spectral response for 2 modules in zener load
for modules C
4.1.2 Temperature coefficients
Figure 5 shows the evolution of absolute
spectral response for 2 modules of Table 1 indicates the temperature
manufacturers A and E after 3 months of sunlight coefficients of the main characteristics (Isc,
exposure with the same load condition (VL-14V). Voc, FF and Pmax) before and after 3 months of
For module E, we observe a weak evolution in sunlight exposure .
the short wavelength (300-550nm) comparated to a We notice different values of coefficients
strong evolution for module A, independently of following the type of modules and a great
wavelength with a decreasing on short circuit variation for manufacturer C. However, the
current density influence of these variations is of secondary
module E : 2 X order on the characteristic extrapolation to the
mpdule A : 11 X standard test conditions (1000 W.m- 2 , 25°C).
In this case [6]
this phenomena depends on the thickness of
the undoped layer (i) cells with thin (less Temperacure coefficients
than 300 nm) layers are relatively stable, Current Volta,_ Fill factor Max Pover
adding small amounts (1 ppm) of boron to the (1M.. ·C-l.clI- 2) (IIV."C· 1 ,e.- 2) (Z ·C· 1) (Z 'C· 1 )
undoped layer generally improves cell
stability.
•
(p-i-D) .0063 -2.14 .168 .. 041

C
ASR(AfW) (p-i-n) ,0029 .0072 -2.76 -2.64 . 140 .053 - .097 .. 20
n4r-------------------------------------, E
(p-i-n) ,0060 . 005S -2.57 ·3.05 ·.29 •. 33

A
(p-i-n) .0051 .0083 -2.24 -2.57 .139 .116 -0.85 - . 110

(p-i-n/p-i-n) .0046 .0051 ·5 . 0 ·5.14 . 106 .161 - . 09.5 - .030

TABLE I - TEMPERATURE COEFFICIENTS OF CURRENT,

VOLTAGE, FILL FACTOR AND MAXIHALE POWER
1 : t - 0, 2 : t - 3 months of sunlight exposure

4.2 Performance measurements in outdoor

conditions (C.E.N. Cadarache)

4.2.1 Initial measurements (t-O)

The electrical characteristics are indicated

in Table 2. We notice that modules E are the best
with a high fill factor (61.4 X) due to a low
series resistance and a high shunt resistance.
Modules B have a higher current density
(14.23 mA.cm- 2 ) but a weaker fill factor
(50.4 X). For module D (p-i-n!p-i-n), the current
5a . Module A density and the voltage values are relative to
tandem junction technology.

405
 Vo. the network (mismatch),
MAllUFACTtlRER J.e cell Ita ltah FF Eff series diodes,
(mA.cm- 2) (V) (Olml.cDl 2 ) (ObJll.cm2 ) (X) (X)
cables.
B
(p·t·n) 14.23 .765 7.69 308 50.40 5.48 Following the array type :
the electrical losses in the series diodes
C1
(p·l·n) n.39 .803 12.60 860 52.50 4.81 and module interconnection cables are
comprised between O. S to 1.1 X according to
C2
(p·l·n) n.53 .791 13.65 860 52.40 4.78
the array type,
the network mismatch losses depend primarily
E
5.87
on the dispersion of the module rating and
(p·l·n) n.52 .828 5.73 1100 61.40
the array configuration. It was comprised
A between 0 to 2.5 X.
(p.l.o) 10.17 .806 7.49 535 51.10 4.18
The total losses are relatively weak and are
D less than 2.6 X from set up maximal power.
(p·l·n/p·l·n) 5.58 1.680 76.70 2200 55.70 5.23

TABLE 2 - INITIAL CHARACTERISTICS OF AMORPHOUS Maxima Losses (W) Total

IlANUFACTIJRER Power Losses
SILICON MODULES TO STANDARD TEST CONDITIONS Mismatch Series Cables
(1000 W.m- 2 , 2S'C» (W) diode (X)
(1st day of sunlight exposure 19 March 1990)
B
(average values for 28 modules by manufacturer) (p·i·n) 115.45 0.94 0.50 0.13 1.36

Cl
4.2.2 Evolution of electrical (p·i·o) 296.73 0.65 1.43 1.04 1.05
characteristics
C2
(p·i·n) 287.26 0.88 1.43 1.00 1.15
Table 3 indicates the ageing indicators
representative of the time-evolution for each E
electrical characteristic, defined by the (p·i·o) 133.80 2.75 0.51 0.20. 2.59
following relation (where A-Isc, Voc, Pmax, FF, A
Rs and Rsh) : (p·l·n) 80.53 0.41 0.34 0.08 1.03

D
A module X (t) (average for (p·i·n/p·i·o) 447.05 0 2.43 2.41 1.08
R(A)
A module X (t-O) 28 modules)
TABLE 4 - PERFORMANCES OF AMORPHOUS SILICON
After a first decreasing of the electrical ARRAYS (28 MODULES BY MANUFACTURER TO STANDARD
characteristics during the first three months, we TEST CONDITIONS (1000 W.m-2, 2S'C)
observed two stabilization steps in relation to (t - 40 days of sunlight exposure)
two average ambient temperature values (26'C in
summer, -S'C in winter). 4.3 Fundamental analysis (C.E.N. Grenoble)
We observed a decreasing for Pmax from 8 to
27 X (summer) and from 18 to 36 X (winter). The preliminary tests of characterisation
The evolution of modules D (p-i-n/p·i-n) is carried out on 3 a-Si:H p-i-n modules :
slower than the modules with a p-i-n junction. the ionic analysis (Secondary Ionic Mass
For these the series resistance increased, but Spectroscopy) allowed to display the
not for the tandem junction modules. presence of all interesting components (Si,
C, N, AI, P, B, 0, H, Sb) as well as element
RUse) R(Voc) R(PtnlK) R(FF> R(b) R(Rsh) combinations (OH, SiH, A1203).
MANUFACTURER
S W S W S W S W S W S W
Thanks to a ultra-vacuum chamber, elements
such as 0 and H have been analysed up to levels
8
<p-j-n) 0.97 0.91 0.96 0.92 0.81 0.70 0.87 0.83 1.47 1.45 0.52 0.48
10.16 and 10.19 cm- 3 respectively.
C(1)
(p-i-n) 0.94 0.86 0.97 0.96 0.82 0.73 0.91 0.88 1.23 1.25 0.59 0.57 4. CONCLUSION
e(2)
(p-i-n) 0.91 0.83 0.96 0.95 0.73 0.64 0.83 0.81 1.54 1.77 0.47 0.46 The collaborations between C.E.N. Cadarache,
C.E.N. Grenoble and J.R.C. Ispra allowed to
E
(p-;-n) 1.0 0.92 0.96 0.94 0.86 0.74 0.90 0.85 1.36 1.58 0.46 0.40 perform the first complementary investigations
necessary for the entire study of phenomena which
A
(p·l·n) 0.95 0.85 0.95 0.92 0.83 0.67 0.92 0.86 1.60 1.83 0.71 0.59 occur in the process of the performances
D
evolution of amorphous silicon modules.
(p- i -n/p- i on) 1.03 0.95 0.97 0.95 0.92 0.82 0.93 0.91 0.86 0.87 0.74 0.71 During the first year of sunlight exposure
the electrical characteristics variation of
TABLE 3 - RELATIVE CHARACTERISTIC PARAMETER 216 modules were quantified (S manufacturers) in
VARIATIONS VERSUS SUNLIGHT EXPOSURE TIME OF relation to the evolution of the ambient
154 DAYS (S) SUMMER (MAX DAILY AVERAGE temperature.
TEMPERATURE - 26'C) The spectral responsitivity measurement
317 DAYS (W) WINTER (MIN DAILY AVERAGE showed different evolution of response versus the
TEMPERATURE - -S'C) load condition and for some modules versus the
(average values fur 28 modules by manufacturer wavelength.
The experimentation in laboratory and in
4.2.3 Performance of the 6 arrays outdoor conditions will be continued until the
end of 1991.
Based OIl measurements of individual modules,
The next run of tests for fundamental
strings and string clusters on the I(V)
studies (SIMS and IRSC) should provide
performance characteristics at irradiance levels complementary data for the analysis of the
from 850 to 1000 W.m- 2 and extrapolation from the different factors which cause performance
Standard Test Conditions (1000 W.m- 2 , 2S'C), we
evolution in amorphous silicon modules.
present the performance of 6 arrays (maximale
power) and the different losses (Table 4) in :

406
ACKNOWLEDGMENTS

The authors are grateful to the Commission

of the European Communities, Directorate General
12 for providing financial support.
H. S. COSTA thanks the Education Minis try of
BRAZIL/CAPES for their grant.

REFERENCES

[1] L. WU, D. HAN, J.I. PANKOVE

Excess optical absorption in a-Si:H after
illumination with bandgap.
Journal of Non-Crystalline Solids 114 (1989)
p. 606 - North Holland, Amsterdam
[2] H.A. OSSENBRINK
The multi-intensity-simulator of the Joint
Research Centre.
7th Eur. Photov. Solar Energy Conf. (1986)
p. 319.
[3] R. VAN STEENWINKEL
Measurements of spectral responsivities of
cells and modules.
7th Eur. Photov. Solar Energy Conf. (1986)
p. 325
[4] J.L. LATHROP, C.W. DAVIS, E.L. ROYAL
Accelerated stress testing of commercial
a-Si:H solar cells.
7th Eur. Photov. Solar Energy Conf. (1986)
p. 293.
[ 5] K. TAKAHASHI, M. KONAGAI
Amorphous Silicon Solar Cells
North Oxford Academic Publishers Ltd. (1986)
p. 210
[6] E.S. SABISKY
Status of research on the light-induced
effect in a-Si materials and a-Si solar
cells.
Journal of Non-crystalline Solids 87 (1986)
p. 53 - North-Holland, Amsterdam.

407
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ANALYSIS OF DIFFERENT COMMERCIAL TECHNOLOGIES a-Si SUBMODULES:

CHARACTERIZATION AND SHORT-TIME OUTDOOR DEDRADATION TEST

F. Chenlo and N. Vela

CIEMAT-IER Avda. Complutense, 22 28040 MADRID (SPAIN)
Tfno. (34)1-3466743 FAX 3466005

ABSTRACT. External optical and electrical characteristics and short-term outdoor degra
dation behaviour for serveral a-Si~ corresponding to 4 different commercial -
technologies together with 65 a-Si modules corresponding to other 4 different commer--
cial technologies were measured and analyzed. The test carried out were: Illumination
JV characteristics at different irradiance levels and temperatures; dark JV characteris
tics; spectral reflectance; spectral respons 2 and short-term degradation test consis-:
ting in an outdoor exposure during 150 Kwh/m of integrated radiation. The main results,
and conclusion were:
The efficiency measurements at different irradiance levels showed a bad submodule de-
sign due to problems such as high resistivity in the metallization grid, TCO or semi-
conductor material. Maximum efficiency values of undegradated samples from 2.2 to 7.8%
for submodules and from 5.4 to 7.0% for modules were measured.
Spectral reflectance measurements show a relatively low repeatability concerning the -
deposition process of submodules (control of p, i, nand TCO film thickness), which al
so has influence in the spectral response measurements.
In the outdoor degradation test different parameter evolutions were observerd among the
same technology submodules and the different ones, with efficiency losses from 3.6 to
21.3%. During this test some samples presented anomalous degradation due to corrosion -
and edge problems; these samples 'were rejected.
In general, the submodules presented worse characteristics than the corresponding equi-
valent modules due to problems like connections, encapsulation and the cutting process.

INTRODUCTION The test carried out with the submodules were:

A) Undegradated submodules. Both illumination and
The successful use of a-Si as alternative .mate- dark J-V characteristics, spectral response and
rial of c-Si for photovoltaic applications is based spectral reflectance.
in a whole characterization of their electrical cha B) Outdoor degradation test. Outdoor e~posure du-
racteristics and functional behaviour including as ring a time equivalent to 150 Kwh/m of energy.
pects like light-induced degradation due to Staebler: C) Degradated submodules. Both dark and illumina-
Wronski effect (1), thermally-induced degradation - tion J-V characteristics at different irradian-
due to high operation temperature (2) and corro-- ce levels (from 0.1 to 0.9 suns) and differents
sian effect due to e.g. humidity penetration (3). temperatures (28 and 45°C).
So, it is neccesary to perform both short and -
long term and real-time field tests (4) to assure a RESULTS
clear knowledge of their proper parameters and func
tional behaviour for their use in PV system wich - From the initial measurements of the J-V illumi
sufficient fiability. nation characteristics of the submodules noticea--
The purpose of this work is to study and charac- ble differences in performance were observed among
terize of the external optical and electrical para- both the different technologies and the samples of
meters for ~everal a-Si submodules corresponding to the same technology.
different commercial current technologies and to es Table II show the initial mean and maximum va-
tablish a comparison of the results before and af-- lues of the electrical characteristcis of the sub-
ter light exposure. modules. The dispersion values were in most cases
very high.
SAMPLE COLLECTION AND TEST PROCEDURE
Fig. 1 and 2 show the initial spectral reflec-
tance of the submodules SAl and SSl at the view
The samples were selected from different commer
of the graphs a noticeable scattering in the posi-
cial technologies trying to obtain samples as simi
tion of maxima and minima is observed in measure-
lar as possible to the modules and at the same ti:
ments; This means that in general there is a rela-
me an optimum size to minimize measure problems. -
tively low repeatability concerning the deposition
The submodules consist in several serially connec-
process of the submodules. A comparison between --
ted cells with the purpose to study the problems
before and after outdoor exposure, in all cases a
due to this association type. Table I shows the phy
decrease is observed, wich we attribute to chang-
sical characteristics of the samples. -
ges in the glass surface because there are not --
TABLE I. Physical characteristics of the samples. changes in the positio~ of maxima and minima.
After the 150 Kwh/m outdoor exposure test a ge
TYPE .NA OF
SAMPLES
TOTAL AREA
(em x em) ( ..
N2 OF CELLS
)
ACTI~
(cm )
AREA neral degradation of the characteristics were mea-
sured. Table II shows the percentages of degrada:
SAl 20 2.0x4.0 4 1.2 tion of each parameter in reference to the initial
SSl 20 5.0x6.5 7 3.8 values. (Fig. 5.)
SS2 20 1.5x5.0 5 1.2 The most important decreases correspond to the
SC1 6 7.7x7.7 7 7.5 Jsc and efficiency values. .

( .. ) This column correspond to the number of

cells serially connected.

408
 TABLE II: Initial electri~al parameters and theirs percentages of degradation
after 150 Kwh/m outdoor exposure test.

TYPE ISC 2 llrsc Voc llVoc F.F. llF.F. T) liT)

(mA/cm ) (%) (mV/cell) (%) (%) (%) (%) (%)
SAl MEAN 1;3.0 2.4 850 3.4 66.8 16.2 7.4 21.3
MAX. 13.2 1.9 861 4.2 68.7 14.6 7·8 19.7
SSl MEAN 8.3 17.6 839 2.2 25.0 1.7 3.6
MAX. 8.6 17.1 863 3.0 29.6 3.4 2.2 21.9
SS2 MEAN 13.3 8.3 833 1.5 40.5 7.6 4.5 14.2
MAX. 13.6 7.2 859 2.9 39.7 5.8 4.6 14.9
SCI MEAN 13.3 6.1 713 1.0 40.6 5.7 3.9 9.1
MAX. 13.7 3.0 687 - 2.3 40.5 5.4 3.8

In general the evolution of the parameters pre- MODULE PERFORMANCE EVALUATION

sent a exponential decrease with a sh~rply decre~
se during the first period (0-50 Kwh/m ) while much 65 a-Si modules manufactured by 4 different
more ~lowly one during the second period (50-150 technologies (5 types in total) were procured and
Kwh/m ). tested as soon after receipt.
Fig. 3 plots the efficiency v.s. irradiance le- The cuantities were: type MA1=10 modules;
vels for one a-Si submodule of each technology, - type MS1=20 modules, type MC1=15 modules; type
from this plot a bad design of the submodules can MC2=10 modules; type MV1=10 modules.
be inferred. Table III shows the active area and the initial
Fig. 4 shows the spectral response of one sample electrical characteristics of the modules including
of each technology. Anomalous results in these mea the medium, maximun values measured and theirs pe~
surements can be attributted at problems such as - centage of deviation.
serious association cell effect. The submodules -- All the measurements were performed in ~utdoor con
consist of different number of cells connected in ditions and corrected at STC (1000 W/m2 and 25°C)~
series. The short-circuit condition for the array
does not mean that every individual cell is also in
short-circuit, it means that differences in perfor-
mance may cause a situation in wich some cells
could have voltage close to the open circuit while
some others ones are operating at highly voltage.

TABLE III: Active area and initial electrical characteristics of the modules
including medium and maximum values and their percentage of devia
tion.

TYPE Isc (mA/cm 2 ) Voc (V/cell) F.F. (%) T)(%)

MAl MEAN 11. 3 851 69.6 6.7
MAX. 11. 7 864 71.8 7.0
0(%) 3.3 1.5 3.4 8.8
MS1 MEAN 12.6 834 62.1 6.5
MAX. 12.7 869 66.4 6.9
0(%) 2.5 3.7 8 16.7

MC1 MEAN 11.4 843 55.5 5.1

MAX. 11.4 857 58.4 5.4
0(%) 2.5 1.8 5.1 8.1

MC2 MEAN 12.7 797 58.7 5.8

MAX. 13.0 826 61.8 6.2
0(%) 3.9 8.9 10.22 15.5

MV1 MEAN 10.8 936 64.0 5.8

MAX. 11.8 862 68.1 6.4
6(%) 6.4 4.1 4.6 15.3
CONCLUSIONS

In general a relatively low repeatability con- Jsc 2.4 to 17.6 %

cerning the deposition process of the submodules is Voc 1.0 to 4.2 %
observed from spectral reflectance measurements. F.F. 3.4 to 16.2 %
The plots of efficiency v.s. irradiance level T) 3.6 to 21.3 %
indicate a bad design of same submodu1es.
The submodules do not seem sufficiently represen
Maximum inital va~ues obtained were: tatives of the commercial modules due to problems -
Jsc(mA/cm) Voc(mV/cell) F.F(%) T) (%) like cutting encapsulation and contact cell problems
Submodules 13.7 863 68.2 7.8 The incoherent results of spectral response re-
Modules 13.0 869 71.8 7.0 vealed problems such as non-fullf'ilment the short-
For the four measured technologies the range circuit condition for each cell of the submodule.
parameter degradation percentages were:

409
ACKNOWLEDGEMENTS REFERENCES

This research has been carried out in the fra- (1) Applied Physics Letters Vol. 31, n. 4, (1977)
me work of the Community Research JOULE Programme
(2) F. Willing, M. Bennett and J. Newton. 19th
contract number JOUR-0077-C throught which part of
IEEE PVSEC New Orleans (1987)
its funding has been provided. This work has also
been partially supported by EURATOM contrac number (3) G.R. Mon, L. Wen and R.G. Ross. 19th IEEE
3353-87-12 ED ISP E. PVSEC New Orleans (1987)
(4) R.G. Ross JR. 19th IEEE PVSEC New Orleans (1987)

8r--
Ii
.5 E
C
.r.

§ ...
.c
§~... tID
.,
>
C
.-"
CO
:. »co
:.

§ §

8....

Fig. 1. Initial spectral reflectance of the SAl Fig. 2. Initial spectral reflectance of the SSI
submodules. submodules.

~(%) 10 ,------------------------------, SA (~t ,------------------,

mill
O.H

O.l

I . IS

t.O u lit

( x)
D UI f $$1 • SSl • 5(1 o $$1 • XI • SSZ • 5.11
x= Irradiance/lOOOW
Wavelenght (nm)

Fig. 3. Efficiency v.s. irradiance level of one Fig. 4. Spectral response of one sample of each
sample of each submodule type. submodule type.

410
 VodmV/cel)
14,------------------------------------, m
~

I~

141
III
110
II
III
!iii

II 7911
7111
m
II 7111
N
74t
7lI
710
711
701
110
iI4
10 101 110 141 114 It
K.... h/m 2
D $II • SSI • ss.! • XI D SA l • SSI • SSl • XI
(a) (b)

FF(%)
71 ~----------------------------------,

41

JO

10 100 lit 140 I~ 10 100

10 lit ItO IIG
Kwh/m Kwh/m2
D S.II t SSI • ss.! • XI
D SAl • SSI .SSl • XI
(e) (d)

Fig. 5. (a) Jse, (b) Voe, (c) FF and (d) efficiency

degradation during outdoor exposure for the
fou r types of submodules.
The abcissa represents integrated radiation.

411
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EVALUATION OF TWO THEORETICAL MODELS IN SIMULATING THE PERFORMANCE

OF AMORPHOUS - SILICON SOLAR CELLS

C. Protogeropoulos, B.J.Brinkworth, R.H.Marshall, B.M.Cross

University of Wales College of Cardiff
Department of Mechanical Engineering & Energy Studies
Solar Energy Unit
Newport Road, PO Box 917, Cardiff CF2 1XH, Wales, U.K.

ABSTRACT. Six amorphous-silicon solar panels have been 2 tested and

evaluated. They are rated at lOW peak power each, _at 1kW/m insolation.
After analyzing the experimental data, we conclude that the one exponential
term model, (either including the series resistance in the equivalent
circuit or not), describes satisfactorily the performance of the solar
cells under several irradiance levels. Other results, such as solar panel
maximum efficiency, fill factor and performance characteristic are
available.

A. INTRODUCTION B.2 Amorphous-silicon Models

According to [1], the series resistance should
The amorphous-silicon panels were tested be taken into consideration in order to .explain the
indoors, at the Department's Solar Simulator, and low fill factors measured for amorphous-silicon
outdoors under realistic and very stable weather solar cells. Also, Wilson et al. [2], report that a
conditions. The panels are part of a small - scaled version of (1) with a single exponential term would
hybrid energy system, whose other components are a fit the I-V curve of an allorphous-silicon cell. This
wind turbine generator, rated at 50W peak power at reflects the shape of the amorphous-silicon I-V
10m/sec and a lead-acid battery of 90Ah measured curves, which do not have the characteristic "knee"
nominal capacity. The tests were carried out for of the crystalline silicon solar cells. Again from
each one of the panels and for the whole array, [2], the series resistance is considerably reduced,
connected in parallel. The performance of the while the shunt paths across the diode are low,
amorphous-silicon panels was examined by obtaining relative to the values for crystalline cells.
the I-V curve under a wide range of global Finally, the solar panel consists of 30 solar cells
insolations. Two theoretical lIodels were tested in connected in series. Applying these remarks to
order to establish a useful tool in predicting the equation (1), the following models for the solar
photovoltaic panel/array output. The models can then panel are established :
be used in sizing allorphous-silicon photovoltaic
energy systems. V
(4)
-~

B. PHOTOVOLTAIC MODELING or
B.1 General Theory
The performance of a photovoltaic cell is I = Iph - 10 ' [ex p ( ~~~~~;T ) - 1] - V::~R- (5)
usually expressed by its I-V curve which is produced
for several insolation levels. The general with the number of cells in a panel Ns = 30.
mathematical expression for the illuminated I-V Equation (4) is the 5 parameter model and is
curve for a single solar cell is given by the explici t as the cell current I does not appear in
following two - exponential transcendental equation, the right hand side. Equation (5) is a 6 parameter
Van Overstraeten [1] : implicit model. The parameters under identification
are PI and P2 from (2), P3 is the reverse saturation
1=1
ph
-I
01
[exp (- -Y+I'R.)]
. - y - -1 -1 [exp (Y+I.R.)
T}1 T
02 .......,-:v- -1]- -V+I·R.
2 T
R - - (1)
ah
current 10, P4 is the diode quality factor ~, P5 is
the shunt resistance Roh and finally for (5) only,
P6 is the series resistance R-. It should be
considered though that the reverse saturation
with, I the cell current, V the cell voltage, Iph current 10 is strongly dependent on the cell
the photogenerated current, 101 and 102 the reverse temperature variations (which occur at different
saturation currents, ~1 and ~2 the two diode quality
or ideality factors, VT the thermal voltage, Rs the insolation levels) :
equivalent series resistance and Rsh the equivalent 10 • Io( Tc )
shunt resistance. The photogene rated current and the The following expression from [3], has been
thermal voltage are given by the following
incorporated in the model
expressions
I ph = fl{G) = (P1'G + P2) (2) 10 = Kdev'~i 2( Tc ) (6)
with, Kdev a device-specific constant, independent
k'Tc of cell temperature and insolation and ~i is the
and, VT = q
(3) intrinsic carrier concentration, expressed by eq.(7)
for silicon
where PI and P2 are 1I0del parameters, Tc is the
absolute cell t*mperature, G is the global ~i2( Tc ) = 1.54 X 10 33 • Tc 3 '10-60BO/Tc (7)
irradiance in kW/m , k is Boltzmann's constant and q
is the electron charge. Equation (1) can be applied The equivalent circuit of equations (4) and (5)
to monocrystalline, polycrystalline or is presented schematically in the Figure 1 :
amorphous-silicon solar cells.

412
 6 PARAMETER MODEL
PARAMETER VALUE STANDARD ERROR
PI [Am 2 /kW] 9.5322E-I 7.061E-3
Figure 1 : Equivalent Panel Circuit
P2 [Al -4.6685E-2 1.321E-3
C. EXPERIMENTAL PROCEDURE 1.4801E-4 @ l80w/m~
7.882E-5
P3 = 10 8.9851E-4 @ 374W/m 2
6.215E-5
C.1 Description and Instrumentation [A] 1.8710E-3 @ 563W/II Z
9.354E-5
Panel tests were first carried out indoors, in 3.037lE-3 @ 725W/m
2.087E-4
the Department's Solfr Simulator. The maximum 3.6844E-3 @ 870W/m z
1. 708E-4
irradiance was 730W/m. All the panels showed a
quite uniform performance as their short circuit P4 = n 4.5824 1.011E-3
currents, open circuit voltages and I-V curves were P5=Rsh rOl 2.4832E+2 3.528E-2
close to each other for several constant irradiance
levels. Outdoors, the same panels were tested at the P6=R. [0] 4.3 1.167E-2
hybrid energy system site under steady global TABLE 2. 6 Parameter Model Results / Outdoor Tests
insolation levels. The panels were inclined 45 ,
orientated to the south. The outdoors tests were
carried out in July 1990, while the photovoltaic SOLAR INS2LATION STANDARD DEVIATION
array has been installed in April 1990, therefore a MODEL r Willi 1 OF FITTED CURVE
significant degradation of the panels had meanwhile
occurred. As expected, a decrease in the total 180 4.642E-3
performance of the panels in comparison to the 374 5.433E-3
results of the indoors tests was noticed. 5 563 4.54IE-3
The instrumentation included a Kipp &: Zonnen 725 2.197E-3
eMIl pyranometer with time constant 5sec, an'ambient 870 1. 38'3E-2
temperature sensor and a data logger. Each I-V curve 180 6.80lE-3
was achieved by varying a variable resistor which 374 3.459E-3
was connected to the panel/array. The time for each 6 563 4.252E-3
I-V curve was about 2min during a period of fairly 725 8.403E-3
constant global insolation and steady ambient 870 6. 744E-3
temperature.
TABLE 3. Standard Deviation of Fitted Curves
C.2 Cell Temperature Estimation
According to [41, the ce 11 temperature can be In Graphs 1 and 2 in the appendix, are plotted
approximated from the formula : the measured against the predicted I-V curves using
Tc = Tub + kc'G (8) the calculated parameters froll the 2 models.
where, TC and Tamb are the cell and ambient The device-specific constant Kdev was
temperature respectively, kc is the solar cell calculated, it~ value being 1.2487E-24 A/K 3 and
temperature coefficient and G is the solar 1. 7611E-25 A/K for the 5 and 6 par8.lleter model
insolation. The value oJ kc typically ranges between respectively.
2.0E-2 and 4.0E-2 in C/(W/.. ). It was considered
that kc = 2.3E-2 would be suitable for the
calculations at all insolation levels. E. FURTHER RESULTS
E.I Panel Maximum Efficiency
D. PARAMETER IDENTIFICATION The efficiency of a solar cell is given by
For the parameter identification, the non
linear routine Amoeba was 'used. The routine is based Pout I'V
TJ = - - - x 100 ~ x 100 (9)
on the downhill simplex method, requiring only C Pin
function evaluations, not derivatives, [5]. The
routine converges rather rapidly to the values of For ~otal panel area (including the framework),
the parameters but it is sensi ti ve to the initial A = 0.3m, the installed amorphous-silicon panel
values of them. The estimation of the starting efficiency is calculated by (9). In Graph 3, the
values of the parameters was based on the numerical maximum measured installed panel efficiencies are
analysis of [6]. Other statistical features that plotted for different insolation levels.
have been incorporated into the routine are the
determination of the standard deviation of the data E.2 Maximum Power Points
set and the standard errors of the parameters. The maximull power points for several measured
In the Tables 1, 2 and 3 are summarized the I-V curves at different insolations are plotted in
results of five fitted I-V curves Graph 4. For these points it is :
5 PARAMETER MODEL Pmax = I max Vmax (10)

PARAMETER VALUE STANDARD ERROR E.3 Panel Fill Factor

By definition the Fill Factor FF is the ratio
PI [Am 2 /kW] 9.5322E-I 3.255E-2 Pmax I Vmax
max
P2 [Al -4.6685E-2 1.244E-3 FF = (11)
I BC Voe I VOC
1.2454E-3 @ 180W/m~ 9.070E-4 Be
P3 = 10 5.6419E-3 @ 374W/m 2 1. 551E-3 In Graph 5, the installed panel Fill Factor
[A] 1.1457E-2 @ 563W/m 2 1. 553E-3 curve is presented for several insolation levels.
1.8042E-2 @ 725W/m 2 8.419E-4
2.2519E-2 @ 870W/m 5.476E-3 E.4 Connection with a 12V Lead-Acid Battery
P4 = TJ 6.7725 7.788E-3 The battery charging current for an array o~
six panels has been measured for up to 1000W/m
P5-Rsh rOJ 2.9754E+2 2.005E-2 global irradiance, as is shown in Graph 6. Taking
TABLE 1. 5 Parameter Model Results I Outdoor Tests into account the losses due to wiring and battery

413
internal resistance, the input power to the battery APPENDIX
can be calculated from
Pbal = Ppv - Ipv' r - Ipv'Rbat,ln{Ipv) (12)
N
with, r the equivalent wiring resistance and Rbat,in N

the battery internal resistance, which is a function

of the operating current. o
N

F. DEGRADATION OF AMORPHOUS-SILICON SOLAR CELLS

During almost one year of operation on site, we
have experienced a significant degradation of the
amorphous-silicon panels. In February 1991, a second
photovoltaic array was installed. It consists of 6
amorphous-silicon panels, connected in parallel. U1
From the first data retrieved and in comparison with :::J
o
the measurements of the first photovoltaic array, it :r:
has been estimated that the panel efficiency has a.
0::
decreased approximately 4% a month. o
::::;;
There have been a few published results so far «
on the degradation of the amorphous-silicon panels.
Work carried out by [7] for instance, shows that the
degradation progresses rapidly during the first 30
to 40 days of exposure and tends to stabilize
afterwards but following an almost linear drop of
efficiency as time passes.
The degradation phenomenon is currently under
study in order to establish a valid model to predict
the power of amorphous-silicon panels over time. ...:

G. CONCLUSIONS
Two models for simulating the performance of
amorphous-silicon modules have been presented. The
predicted I-V curves are in good agreement with
those measured under several insolation levels. It
appears that the major part of the difference is due 0Cl ": ~ ~ v ~ "'! o
o o 000 000 o
to the approximation of the cell temperature (sJedw\t) IN::J~~n:) l::JN\td
determination. The diode quality factors in both
models are rather high as well as the series
resistance .of the 6 parameter model. As a result the
panel efficiency and fill factor are low and the N
N
overall panel performance is poor.
It is concluded that the models are
satisfactory for system simulation purposes, though o
N
a rapId degradation of performance has been noticed
and that remains to be modeled.
---'
w
H. ACKNOWLEDGMENT z
«
CL
Mr. Trevor Thomas from the Department of
Physics of the same university, has helped with his
constructive remarks on the theory of
amorphous-srlicon solar cells.
REFERENCES
U1
[1]. R.J VAN OVERSTRAETEN, R.P. MERTENS, "Physics, :::J
o
Technology and Use of Photovoltaics", Adlllll Hilger :r:
CL
Ltd, 1986, ISBN 0-85274-487-0 a::
[2]. J.WILSON, J.McGILL, P.ROBINSON, "Doped o
::::;;
Amorphous Silicon Schottky-Barrier Solar Cells". «
Photovoltaics Workshop, Nov. 1977, National Physical
Laboratory, Teddington, Middlesex
[3]. L.GOLDSTEIN, G.CASE, "PVSS-A Photovoltaic
System Simulation Program", Solar Energy, Vol.21
[4]. MATTHEW BURESCH, "Photovoltaic Energy Systems,
Design and Installation ", McGraw Hill, 1983,
ISBN 0-07-008952-3
[5]. W.PRESS, B.FLANNERY, S. TEUKOLSKY, W. VETTERLING,
"Numerical Recipes", ClIIIIbridge Univ. Press, 1989,
ISBN 0-521-38330-7
[6]. D.AGCHBAYAR, N.ENEBISH, R.MARSHALL, B. CROSS,
"Numerical Analysis of Current-Voltage and Spectral
Response Characteristics of Solar Cells", 1st World
Renewable Energy Congress, Reading, U.K. 1990
[7]. R.ZILLES, E.LORENZO, I~egradation of I-V Curve
Characteristics of Amorphous Silicon Modules Exposed 0Cl N o
to Sunlight ",1st World Renewable Energy Congress,
~
o o
~ ~
0,0
~
0
~
0
~
0 o o
(sJadwv) IN::J~~nJ 13NVd
Reading, U.K., 1990

414
 0.9 -, 3.0
G RAP H 4.
AMORPHOUS - SILICON PANEL G RAP H G.
~
0.8 ~ A7nfW/m?\
I - V Characteristics &
Maximum Output Power Points
~
For Severol Insolation Levels
~0.7 ~
~ 725("/m2) >- 2.5
u
'"
EO.6 z
w
:'> u
L;:
!z05 563(W/m2) "-
W
W
a:: c;p.O
a:: z
::J 0.4
u «
0..
...J AMORPHOUS - SILICON PANEL
w ' 374(W/m2)
z 0.3
« Maximum Measured Panel Efficiency
0..
276("/m2) 1.5 ] / At Severol Insolation Levels
0.2
180(W/m2)
i
:::<

01 ]
~ 82:w:m'l
0.0 iii j I Ii i I I iliiiil~i~iiil 1.0
0 2 4- 6 8 10 12 14 16 18 20 22 0 100 200 300 4-00 500 600 700 800 900 1000
PANEL VOLTAGE (Volts) GLOBAL INSOLATION (Watts/m2)

.".
v.

3.5 J 0.6
G RAP H 6.

AMORPHOUS - SILICON ARRAY J G RAP H 5.

~ 3.0 (6 PANELS IN PARALLEL) / 0.5
0.
E Charging Current
« With Battery Connection ~
::- 2.5 vs
z Global Insolation ~0.4
w U
a:: «
a:: LL
3 2.0 AMORPHOUS - SILICON PANEL
~ 0.31 Panel Fill Factor For Several
' " Insolation Levels
'"z~ 15
....J
« .;- w
I ....
u ~ 0.2
D...
~ 1.0
w "
1=
« ~y 0.1
m 0.5
"
':. ,
;/.,
0.0 I I I iii I I I I iii I I I I iii i I I Iii ill i I I I i I t [ i i 0.0
I i' 0 100 300 4-00 500 600 700
0 100 200 300 400 500 600 700 800 900 1000 200 800 900 1000
GLOBAL INSOLATION (Watts/m2) GLOBAL INSOLATION (Watts/m2)
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A NEW PCA-SiN:H/A-Si:H)-I(A-Si:H)-N(A-SiC:H/A-Si:H)

SUPERLA TTICE SOLAR CELL

Argyrlos C. Varonldes, Electronics Engineering Department,
UnIversIty of Scranton, Scranton, Pa, 18510.
Allen Rothwal f, Electrical and Computer Engineering Department,
Drexel unIversIty. PhIladelphIa, Pa, 19104.

1, Introduction
Quantum phenomena have been observed in
amorphous superlattices [1 J. The amorphous
superlattices composed of two amorphOus
semiconductors such as a-Si:H/a-SiC:H, have
ABSTRACT recently been investigated [2]. One of the
properties of superlattices of the type mentioned
above, is the existence of minibands in the
conduction and the valence bands. Quantization of
The objective of this work is (1) the carrier motion leads to discrete ener'gy eigen-
evaluation and control of the Fermi level position solutions within the quantum wells. Each solution
in an amorphous hydrogenated silicon p-i-n Solar possesses a number of available states thus
Cell through its carbon (or nitrogen) alloy n-and p- forming minibands.
contact regions. The Fermi leve Is in both p- and n-
The Fermi level exists in the gap and usually
regions of a
many kT's away from the band edges. I t is uSeful
then to be able to predict the position of the Fermi
P(a-Si:H/a-SiO.8NO.2:Hl- N(a-Si:H/a-
energy in the gap of a compositional amorphous
SiO.5CO.5:Hl
silicon-based superlattice. On one hand, it is well
established that the Fermi energy is pinned in the
superlattlce solar cell are computed. (2)
gap and on the other, there is hope to be able to
prediction of open-circuit voltage values for this
novel device. The p- and n- regions are replaced by shift it towards a band edge (closer to Ec for an n-
superlattices, in which the Fermi levels in the type device, and closer to Ev for a p-type device),
wide gap barrier regions are, on an absolute scale, in order to Improve the electrical properties of
closer to the valence band or conduct ion band devices. I t must be pointed out, that degeneracy is
edges of the low band gap material (a-Si:H). In very unlikely, since high doping levels do not tend
order for the p- and n- superlattices to yield a to change the activation energy as drastically as
larger built-in ,voltage, the denSity of states in the in the case of the crystalline semiconductors. The
doped wide band gap material, must be much latter are clean materials with no dangling bonds
grpater, and the Fermi level of the wider band gap whatsoever. In an amorphous semiconductor
layer has to be closer to the corresponding band instead, the impurities introduce more and more
edge than that of the undoped low gap quantum defects and, as a result, the Fermi level gets
well. Hence, the wide gap material must be heavily pinned well below the edges. In a-Si:H for
doped and have effective tails wider than that of instance, the distance of the Fermi level EF from
the undoped low band gap material. As a result, the the conduction band edge is on the order of 0.2 eV
final Fermi level will lie between the for an n-type sample [3], while for a p-type sample
corresponding bulk Fermi levels of the layers of EF is 0.4-0.6 eV away from the valence band. It
the superlattice. was proposed in [4], that a superlattlce device will
For the n- region superlattice, the cause a shift of EF towards the corresponding
barrier must be in the conduction band; hence, we edges, namely conduction and valence bands,
propose a-5il-xCx:H and· a-5i:H as the thereby improving the upper limit on the open
constituents. For the p- region, the barrier must be circuit voltage [4J.
in the valence band, and we propose a-5i l-yNy:H The potential efficiency gains in amorphous
and a-5i:H. In this way, the Fermi levels will shift silicon solar cells is primarily based on
towards the corresponding band edges,
improvements in VOG' It is also known that the
contributing to improvements in Voc and
efficiency. maximum limit of open circuit voltage of a single
junction cell Is the built-in voltage itself. Work on
crystall ine and amorphous superlattlces provided
the possibility of designing devices in such a way
that the Fermi level can move closer to the

416
corresponding edges, in spite of the existence of
defects and impurities. With the layers thin p-reglon
enough, so that quantum effects are poss.ible,
sufficiently high doping of the wide band gap As the bulk Fermi level EF..., moves aw,ay fr()1'Y)
material will cause a shift in the Fermi level. This the midgap of semiconductor 2, the fmal EF
effect combined with the best one can get for the compromises in the range AEF = EF l-EF2' and it
bulk a-Si:H layers (n-type a-S1:H Ec-Ef = O.2eV),
wi 11 force the Fermi energy to move to a remains to be seen if it can get closer to the
"compromised" position between the highest silicon valence band edge or not. The silicon
nitrides seem to be good candidates for p-region
position of EF in the wide band gap material, and
wide gap semiconductors [13,14,16]'
the lowest position in the narrow band gap
material (i.e., EF will be between EF2 and EF I)' Integrating one obtains:

Ep~~~v2J
At the same time, one could not use just the
wide band gap material as a contact layer by
itself, since the value Ec-EF or EF -E v Is greater
d2] [Nv2]
[ d N exp--
[AEp]
ex p[ -
n p x
than that of the lower band gap material, and thus 1 ,,1 kT Epl - Evl
would have a very low conductivity, and also a exp kT·
1
barrier at the interfaces for carrier collection
would exist. Therefore, a superlattlce based design
offers the possibility of carrier tunneling through
the wide band gap material (for thin layers) and
hence a larger built-in voltage obtained without
losses In conductivity.

2. Gap states diStribution In the amorphous

sem Iconductors.

The crystalline semiconductors are clean

unless they are doped with impurities. What The picture is the following:
characterizes them is the fact that there are no there exists a superlattlce structure, where the
defect states in their energy gap. The DenSity of higher band gap material is doped with donor
States (DOS) of a 2 Dimensional Electron Gas Impurities [e.g. PI. At the same time, the lower gap
(2DEG) Is constant. In the case of a superlattice, material Is kept intentionally undoped or lightly n-
the DOS spectrl..llll becomes a staircase [5). In doped, whi Ie both layers are very thin in order to
amorphous materials the neutrality equatlon (ND, ensure: (i) quantization of the energy of the
In thermal eqUIlibrium, needs to be modified. In carriers in the wells (11) negligible band bending
addition to the "staircase" behavior of the DOS, due to space charge potentials [11,12) and (Iii)
another distribution of states exists In the gaps of adequate conductivity. Analogous situation' holds
both semicohductors. for the p-type region of the p-i-n structure.
Charge neutrality will determine the final Fermi
3. Neutrality condition. level position at equilibrium. The p+-reglon
includes p+ -type wide band gap material
The neutrality equation must be modified (amorphous si I icon nltride alloy) and a p-doped
appropriately, to Include the existence of the amorphous silicon layer. The reason for selecting a
trapped' carriers. Since the total number of
nitride alloy of amorphous $Ilicon is that one
electrons Is conserved. At T=O, electrons will
needs the barrier to be in the valence band, since
leave the material with the higher Fermi energy, In thiS case, EF will be shifted to lower Qosltlons
material2 (wide gap semiconductor) and they wl11
towards the valence bands. One would need a
fill states in material 1(narrow gap
downward shift of EF In the p-type region and an
semiCOnductor), In order to satisfy the
requirement for constant Fermi leve Is. CONCLUSIONS

This work discusses the Fermi level position

In the gap of amorphous' silicon and alloy
compositional superlattices. The relation of final
EF position is then connected to the open circuit
voltage, efficiency gain, and conductivity.

417
 UPDIHCIS [14)S.L. Were mi R.O. ClleDc, • Properties of
a- Si:Hla-SiNx:H SUpemUices v1tb. dopants
[1] R.R. Arye., A. Ca1alano, mI. J. O'Dow, • added In .. SiNx:H subJa)ll8Ill,· in AmoJpllo1l3
!ffec1 of SUperldte Doped ta,ell on 1118 sWeen TechroJocy, Mat Res. Soc. PlOC.,
PexfomIm::e of .. Si:H P-I-N Soler Cells,· No 118, pp. 170-179, P.G. Hemakaftedr.
Mat Res. Soc. s,Dlp.,Pxoc. Vol. 70, p. 517, [15IT. Yoslllm1Im, et al, • Influm:e of .. SiNx:H
1986. compositiln, on tl8DSfer-dopinc mi elec1ron
(2) M. Yosbimot) et al, "Cmier ttuIspOrt In .. ttapplDg effec1s In e.- SiNx:H1a-S1:H
Si:HIa-Sil-xCx:H 1Il1ra- 'tbin m1lltila)ll8Il1,· supeIia1ti:es,' PhIlosophbl Meeazlne B,
Philos. Mec.B60, No 1, pp. 89-99, 1989. 1987, VoL 55, No 3, pp. 409-416.
[3) A. Madan am M. P. Shav, Tho Phm of [16 )Von Roedem .t al, • Pho1Oemisstm studies
AllX!mbous Semj;tmlDl3 Acaiemic on in situ prepered h)'drocena1ed emDIPIlo1l3
~,p. 116,1988. s111con PIlms,' pbl))s. Mee. B40, p.433,
(4) A. Roth'Rl1am A. YeroUlu,' SuperlaUice 1979.
CoD1llCt La)ll8Ill for Hich Open Cilcuit Yol1Bce
a-stH Solar Cells," 21st IE!! PVSC, (17)LeComber, P.G. Madan, A. Speer, J. Non
OIIaMo PloIlda, pp. 1555-1559, May CO'St SOl1ds zo.. Z39. (1972).
1990. [18)McMehon 8t al, J. Non CIJ3t. Solids, ~
[5) C. Weisbuch In,· Pumammtal Properties of 375, (1984).
m-Y Semicolld.ucm Tw-Dimellsional [19 IS. S. Heged1l3 • TIle Open Cilc1Jit Vol1Bce of
Q1Iantlze4 Sttuctmes: Tho Basis for OptIcal AmoIpIlo1l3 Slli:on Pol-N Solar Cells", 201b
am Electtoni: Device Appli:atilm, • PVSC, pl02, (1988).
SemIooJducms am SeIDjnM!II:! 24 [20IT. Yoshida at al, "Open Circuit Vol1tga of p-
Acal.emic Press, WIlIar4son-Beer eds., pp 1- I-N a-Si Based Solar Cells", 20th PVSC,
133,1987. p335, (1988).
[6) T. Tiedje, C.R. WloIlski, P. PerseDS, B.
Abeles, "Recent Bxperimental Res1ll1l on ..
Si:H/a-Ge:H Superla1ti:e Structuxes", J. of
NoD-CIyst. SOl1ds, 77&.78, pp. 1031-104),
(1985).
[7) K. W. Boer, Sumy of Se~r
EJm;tL p. 432, 1990.
[8) J. Terse!! In "1lle 1beoI)' of lle1eroj1l.nc1ml
bam liDeup3,· He!elOj!mcmn Bam
Discon!inyities . P. Ce.puso and G.
M811aIitlDdo eels., North-HolIaDd Ph"ics
Publlsbm, pp. 3-57, 1987.
(9) P. l!Y8ICelis1i et al, • Pho~ studies
of a-SizCl-x:H I .. Si:H mI. a-Sizel-
x:HIh)1drogena"" e:omplxI1llI Silicon
Inroj1lIlCtiom,' Appl. Ph". Lett. ~ (8),
pp. 764-766, 1985.

[10)A. Varonides, • SembJDducm Superlatdces

vtal an appUcallon t) High l!!flcle:ocy
AmoJpOO\13 Silicon( AJkly) Bued Solar
Cells", Drexel Unive13ity, PlIiladelphia, Pa,
1990.
[11)1. Chen, • 'Dloore1ical enaly,IU of space
charge doping in emDIPoous semicoDduc1Or
superla1Uces,· Phys. Rev. B, ~ pp. 879-
889,1985.
(12)G .H. Dohler, • TIleoxy of AmoIPho1l3
structUIes (SUpemtti:u),' J. of Non-Cly.lt
Solids, 77&.78, pp 1041-1050, 1985.
[l3)M. Hirose am S. Miyazaki, "L\Ullirie3ceIM:e
mI. Transport in a-S1:Hla-Sl1_xNx:H
Q1l8JItUJn Well Stnx:tUIeS, • J. Non-CI}'S1aIl1De
Sotids,66,p. 327, 1984.

418
 a-Si(0_5)C(0.5) : H

O.295eV _
H+H+t-H-H T
l.75eV

2 .2OoV
..·S! : H

e·S! : H
.·SI(0.8)H(0 2)-H I e·S! :H

"Tho eample .. p+n saweell ==. TIle p. Olea io eomprt.ed of a-S!:H aM .·Si(0.8)N(0.2):H
leye,,: TIle n. ma m:l1ld....el of .-Si:HIa-Si(0.5)C(0.5):H leye".
Table : , TIle Ol,eneneIgle' ore '00""
for each case.

Openeireuitvollage valut.Voc,PenniJe...,l posi1ion Efin the gap oftbt P-IYPt region,

(a-SiN:H I a-Stli) 0llI! dark pe!peMiculorOlll! pora1lelcoMUC1ivitle. ".I.' "II' dl=d2=25A.,
Noyl=NoY2=IO AI9 emA·3 tV A·I,al=2.7,a2=4.5 , m" ~ 1.0 m ,l!f1-l!vl=O.95 tV,t.Ev=0210
eV ,t.I!f =Efl- I!f2 relative", Evl ,Egi = 1.75eV ,Eg2 -1.96 eV ,S= IOA6cml.ec,J(sc) = 18
mAlem'2,Ne= 10 AI8cm A_3 . a,300IQ em. Voc(nonsuperlattiee) -O.93V.

Ef2 -EY2 (eV) E12·Evl (eV) Ef-E!1 (eV) Ef-Evl (tV) t.Ef(eV) Voc(V) ~(Slcm) a (Slem) !;Voc(V)
/I

0.56 0 .35 0.438 0.512 0 .60 0.818 5.08:<10A-12 1.40xl~·7 -{).I12

0.46 025 0.525 0.425 0.70 0.905 1.6II~-10 4.56xl~-6 -{).02S

0.36 0.15 0 .612 0 .338 0.80 0.992 4 .94xl~-7 l.48xl~-4 0 .062

0.31 0.10 0 .656 0294 0.85 1.036 2 .087xl~-6 8.62xIOA-4 0.106

No1lt:
!;Voe= Vac(superlattice) - Voc(nansupeIIlattice)

Tablet
openeiIl:1!itvoltoge val... aM Penni Ievel' po.i1ionin tbt gap oftbt p-region (a-SIN:HI a-Sl:H).
dl -d2 =25A.,NOYI =N0v2-IO AI9cmA·3 eVA-I,e<J -2.7,<:<2 - 4.5,m*= 1.Dm ,Efl-E ..I =0.4oV ,tBr
=0.210 eV ,AEl'= Efl-Ef2 rela1i...,,,, Evl , EgI =1.75 eV,Eg2 =1.96 eV,S-IO"6<mIsee,
J(.e)-18mAlem'2, Ne= 1~18emA-3. ".-300IQ em . V0C(b1llk) -0.93V.

I!f2 -l!Y2 (eV) E12 -1l..1 (eV) /;Ef (eV) Ef -1! ..I(tV) ~.<S/em) a (S/cm) Voc(V) !;Voc(Y)
/I

0.$ 0.340 0.060 0 .341 2.373xIOA-9 6.576xl~-5 0 .989 0.059

0.51 0.300 0.100 0 .300 I 223110A-8 3.39OXID-4 1.030 0.100

0.46 0250 0.150 0250 0.904xIOA-7 2.SOSxl~-3 1.000 0.150

Table '3

Penni Ie...,l Et,activatDn t .. rgyEa,aM cond,.,tivitie.(a ," ),ofthe n-rype 'OJ>erlettice (o.-S!C:H la-S:H).
41 = 25A.,!;l!c = 0.45oV , e<J = I D4 [18],<:<2 - 4.00, Ee I-Efl = O.4eV , Ee2 -E12 _ O.60V ,Noel = Nac2 =
1~22cmA-3 eVA-I, /;Ef =02SOaV ,!;l!o -SOmev (= energylevel in !be quanlum von ,Ea = (l!c I-Ef) +!;l!o)
d2=boniervilth,dr=peliodofsuperlattice. cr. '10' /sz.' ....

d2(A) dr(A) Ecl-Et(eY) Ea(eV) ".J.(3Icm) ",jSICm)

so 75 0233 0.313 0 .324xlQA·9 3.652110A-4

4J 65 0.239 0.319 1.1S4xlQA-8 3.315xIOA-4
30 $ 0245 0.325 1.11211~·5 2.525xIOA_4
20 45 O~ 0.335 3 .949%1~-5 3.082110A-4
15 4J 0.262 0.342 0.536II~-4 2.1 47110A-4

419
IOTHEUROPEANPHOTOVOLTAICSOLARENERGYCONFERENCE 8--12 APRIL 1991 LISBON. PORTUGAL

PHOTOELECTRIC PROPERTIES OF a-Si:H

WITH LOW GAP-STATES DEPOSITED BY VUV PHOTO-CVD

T.Shirafuji, M.Yoshimoto, T.Fuyuki, and H.Matsunami

Department of Electrical Engineering, Kyoto University,

Yoshida-honmachi, Sakyo, Kyoto, 606, Japan

Abstract: Hydrogenated amorphous silicon films are deposited by a direct vacuum ultraviolet
photochemical vapor deposition method. The light source is a microwave-excited Xe resonance
lamp (wavelength=147nm). Undoped films are deposited using Si 2H6 at substrate temperatures
T,ub of 200 - 400·C. Photo- and dark conductivity ratio of the films is as high as 105 under
AM1(100mW cm- 2 ) illumination for all the Tsub. The film deposited at 300·C shows low
photoconductivity degradation of 10% after 3-hour illumination and a low defect density of
2.2 x 10 15 cm- 3 . Doped films are deposited by the addition of PH 3 for n-type and B2H6 for p-type
at T,ub of 2S0·C. Dark conductivity is controlled from 10- 11 to 10-3 S cm- l . The phosphorus
doped films in particular show much higher photoconductivity than the dark conductivity.
This is explained by the increase in rJJ1.T by the shift of Fermi level to the conduction-band side.
Besides, relatively low defect density is suggested to bring about the additional increase in rJJ1.T.

l.Introduction elsewhere[4]' A window made of MgF 2 was used between the

reaction chamber and the lamp. The window was blown with
Photochemical vapor deposition (photo-CVD) is a promis- N2 at a flow rate of 2S0cc mi'n- l to prevent the deposition
ing method for the preparation of hydrogenated amorphous of the film on it.
silicon (a-Si:H) films, since the bombardments of high-energy Undoped films were deposited using 100% Si2H6 with a
electrons and ions on the growing film surface can be avoided. flow rate of 2.Scc min-l. The substrate temperature was
In addition, the photo-CVD method offers the selectivity varied in the range of 200 - 400·C. The total pressure of the
of reactions with particular threshold energies in compari- reaction chamber was 2.0Torr.
son ~ith a glow discharge (GD) deposition method in which Doped films were deposited by the addition of controlled
source-gas molecules are dissociated by electrons with a con- amount of PH 3 and B2H6 to Si 2H6. Both of the dopant
tinuous and widespread energy distribution. gases are diluted with H2 • The flow rate of PH 3 and B2H6
A low-pressure mercury lamp (18Snm; 6.70eV)[1] or an was varied in the range of O.OOOS - O.lsccm. The substrate
Arl<' excimer laser (19Snm; 6.36eV) [2] is used as a light source temperature was kept at 2S0·C. The total pressure of the
and Si 2H6 as a source gas in conventional photo-CVD meth- reaction chamber was 2.0 - 2.5Torr.
ods. The energies of photons emitted from these light sources The typical deposition rate and the thickness of the films
are lower than threshold energies to excite Si 2H6 to low-lying for both the undoped and doped films were 0.6nm min- l and
excited states[3]. As far as we utilize the same photon energy, 300nm, respectively.
the reaction path will not be changed, and the properties of
the films will not either. Thus, an alternative light source
which emits photons with different energy is required.
We developed a novel direct photo-CVD method using vac-
uum ultraviolet (VUV) light (147nm; 8.43eV) emitted from 10-4
a Xe resonance lamp[4]. The energy of the photons emit-
ted from this lamp is higher than energies to excite Si 2H6 _0 _ _ 13
0-
to some low-lying excited states. Geometrical and electronic 'i 0
structure of the films deposited by the photo-CVD method 8.., 10-6
were repoited before[5]. The hydrogen concentration of the ~ PHOTO (AMI 100mW cm-')
films is less than lOat.%, and reaches Sat.% for substrate ><
temperature(T,ub) of 300·C. The hydrogen bonding configu- ....
Eo<

ration is mainly Si-H. In spite of the low hydrogen concen- ~ 10-8

Eo<
tration, the defect density of the films is less than 1016cm -3 U
;:J
and reaches 2.2 x 10 l5 cm- 3 for Tsub of 300·C. Cl
In this work, we investigate the photoelectric properties Z DARK
0 10- 10
of undoped and doped a-Si:H films deposited by the photo- U
- e - - e _ _e
C.VD method. e-
e

10-I'~~~~~~~~-L~~~~~~~-w
2.Experimental 200 300 400

Films discussed in this report were deposited by a direct SUBSTRATE TEMPERATURE (·C)
photo-CVD method using VUV light (147nm) emitted from Figure 1: Photo- and dark conductivities of undoped
a microwave-excited Xc resonance lamp. A schematic dia- films. The P.hotoconductivities were measured under
gram of the deposition chamber and the lamp was shown in AM1(100mW cm- 2 ) illumination.

420
 -0--0--0 0-

-
UP2/U/,O0 _______0 , I
o
bQ.
0.8
Upl/UpO .,...., 10" \ 0

\0
I 0
N o E

]1]
~
bQ. 0.6 ;..
o ...
Eo<
<J')
Z
~

... ~
~

I U
~ 0
10<
0.2 ~
~

1015
200 300 400 200 300 400
UB TRATE TEMPERATURE (OC) SUB TRATE TEMPERATURE (OC)
Figure 2: Photoconductivity ratios , upJ/upo and Up2/UpO, Figure 3: Defect density estimated from below-gap absorp-
where the symbols upo and Upl are the initial and final pho- tion coefficients using a constant photocurrent method.
toconductivities under AMl(lOOmW cm- 2) 3-hour illumina-
tion, respectively. The symbol Up2 is the photoconductivity
after 3-hour annealing at 230·C after the illumination.
sity of the films estimated from below-gap absorption coef-
ficients is low as shown in Fig.3[5]. The defect density de-
creases from 9.0 x lOIS to 2.2 x 10 15cm- 3 as T,ub increases
3.Photoelectric properties of undoped films from 200 to 300·C. It increases, however, to 1.0 x 1016 cm- 3
as T,ub increases to 400·C. In spite of the change in the defect
Figure 1 shows the photo- and dark conductivities of the
density, the photo- and dark conductivities keep the excellent
undoped films. The photoconductivity was measured under
value. Within the deposition conditions we have studi~d the
AMl(100mW cm- 2) illumination. The dark conductivity is
change in the defect density ranging 1015-l016cm-3 ma~ not
as low as 10-l1S cm- I and the photoconductivity is as high
affect the initial photoconductivity. Comparing Figs.2 and
as 1O-5S em-I. Both of the photo- and dark conductivities
3, the defect density is rather considered to affect the pho-
do not show any explicit dependence on T,ub in the range
toconductivity degradation.
of 200 - 400°C. The high photoconductivity of 1O- 5S em-I
suggests that the films have a low density of defects which
capture photo-generated carriers.
4.Photoelectric properties of doped films
Figure 2 shows the ratios upJ/upO and Up2/UpO, where the
symbols Upo and Upl are the initial photoconductivity un- Dark conductivity was controlled by substitutional doping
der AMI(lOOmW cm-2) and the final photoconductivity af-
of phosphorus and boron for n- and p-types, respectively.
ter 3-hour illumination, respectively. The symbol Cr p2 is the
Closed symbols in Fig.4 shows the dark conductivity of the
photoconductivity of the film annealed at 230·C for 3 hours
phosphorus- and boron-doped films as a function of the ac-
after the illumination. The ratio upJ/upo indicates the dura-
tivation energy of the dark conductivity. Both the dark con-
bility of the photoconductivity under the illumination. The
ductivity and its activation energy increased monotonously
durability depends on T,ub, whereas the initial photoconduc-
from lO- 11 S cm- I for undoped films up to 1O-3 S cm- I for
tivity does not depend on it. The ratio upJ/upO is higher
the doped films under the dopant-gas ratio PH 3 /Si 2H6 or
than 0.5 and reaches 0.9 for T,ub of 300·C. The degraded
B 2H6 /Si 2H6 of 10-1[7]. These films showed good linearity
photoconductivity recovers up to 0.95 of the initial one by
on the arrhenius plot for the temperature dependence of the
the annealing. The high durability of upJ/upo=0.9 against
dark conductivity in the range of 30 - 180·C, which shows
the illumination is promising for solar cell applications.
that the activation-type conduction is dominant.
The structure of the films which show such an excellent
Besides the change in the dark conductivity with the shift
stability against the illumination is investigated from in-
?f Fermi level by substitutional doping, there is a change
frared spectra. The hydrogen incorporation scheme in the
In photoconductivity too. The photoconductivity was mea-
film is mainly Si-H. This result is consistent with the re-
sured under AMl(100mW cm- 2) illumination. The results
sults obtained by Kuwano et al. in which the suppression of
are shown in Fig.4 as open symbols. There is a certain differ-
Si-H2 is valid for the suppression of the photoconductivity
degradation[6]. The bonded hydrogen concentration is less ence between the photoconductivity profiles for the boron-
than 9at.% and reaches 5at.% for T,ub of 300 - 400·C. This doped and phosphorus-doped films compared with a similar-
result suggests that the films have a rigid and stable struc- ity between the dark conductivity profiles for them.
ture, which brings about the property of the low degradation The boron-doped films show constant photoconductivi-
in the photoconductivity. Even the films have low hydrogen ties, and the values are almost the same as that of undoped
concentration and mainly Si-H mode, however, the degrada- films. In contrast to the case for the boron-doped films, the
tion rate increases for the film deposited at higher T,ub of phosphorus-doped films show the increase in the photocon-
400·C. In the case of high-T,ub deposition, another mecha- ductivities. To clarify the cause of the photoconductivity
nisms that bring about metastable defects in the films, such mcrease, we examined photoelectric and electronic proper-
as a contamination, are suspected to exist. ties of the phosphorus-doped films.
In spite of the low hydrogen concentration, the defect den- We investigated the photoconductivity of the phosphorus-

421
 10- 2 I·
~
• 09 ' - <b-

:\---:\:- '/
0 ' # /
i
E
u
10- '

~ 0
___0_ 0
>-....
.... 10- 6

..
~
....
()
;:l
Q lO- 1 ~.
Z
0 o PHOTO
()
• DARK
10- '·
•
0 .2 0.4 0.6 O. 4 0.6 0.4 0 .2
BORO DOPED U DOPED PHOSPHORUS DOPED
ACTIVATIO E ERGY (eV)

Figure 4: Photo- and dark conductivity of phosphorus-doped, boron-doped and undoped films as a function of the activa-
tion energy of the dark conductivity. The films were deposited at T.ub of 250·C. The photoconductivity was measured under
AMI(IOOmW cm- 2 ) illumination.

--0--0---0_ o
"".----0"-;. lO) - x
-)(-
'0
x .....
- - +." +"""0. . . .
+., . . ... ~
+
... "
,
I
I
I
I

o THIS WORK
o o X SPEAR ET AL.
U NDOPED
+ OKUSHI

U DOPED lO - ' 10- 2 10· 10- 10 10- 1 lO- 6 10- ' 10- 2
PH, / ShH. DARK CO DUCTIVITY (S em-')
Figure 5: The product 'f/Jl.7 of phosphorus-doped and un- Figure 6: Photoconductivity as a function of dark conduc-
doped films for the illumination of a He--Ne laser (hll=1.96eV, tivity for phosphorus-doped and undoped films deposited
7.0x 1014 photons cm- 2 S-1). The films were deposited at T,ub at T,ub of 250·C. The photoconductivity was measured un-
of 250·C. der the illumination of a He-Ne laser (hv=1.96eV, 7.0 x
lO14 photons cm- 2 S-I) .

doped films under monochromatic light illumination using

a He-Ne laser (hll=1.96eV) with a photon flux of 7.0 x product 'f/jj7 shows an increase by the two orders of mag-
lO14 photons cm- 2 S-I. Generally, the photoconductivity nitude with the phosphorus doping compared with that of
~a(S em-I) is expressed as follows , the undoped film as shown in Fig.5. This increase in 'f/jj7
is regarded as the cause of the increase in the photoconduc-
~a = e'f/w<I>(I- R){l- exp(-ad)}/d, tivity by the phosphorus doping. Anderson et al. explain
that the life time 7 is extended due to the filling of empty
where 'f/ is quantum yield, Jl.(cm V-I S-I) drift mobility, 7(S) recombination centers with the shift of Fermi level toward
life time, <I> photon flux, R reflectivity, a( em-I) absorption the conduction-band side[8]. The photo-CVD films, how-
coefficient, and d( em) thickness. There are three parameters ever, show higher photoconductivity than GD films [8] as
R, cy and 'f/Jl.7, which are expected to contribute to the in- shown in Fig.6.
crease of the photoconductivity. Other parameters are not This result suggests that the empty recombination centers
affected by doping conditions. The reflectivity R was not so are expected to be kept at low density. Thus, we investigated
much altered with the doping. The absorption coefficients the defect density of the phosphorus-doped films.
cy of the phosphorus-doped films were less than that of the Figure 7 shows the defect density of the phosphorus-doped
undoped film, which rather results in the decrease in pho- films, which is estimated from below-gap absorption coeffi-
toconductivity and does not explain the increase in it. The cients measured using a constant photocurrent method. The

422
 Dark conductivity is controlled from 10- 11 to 10-3 S cm- 1
10" for both the phosphorus- and boron-doped films. The
phosphorus-doped films show much higher photoconductiv-
,
ft

ity than the dark conductivity in comparison with the boron-

6
~ doped films. The increase in the photoconductivity is due to
the increase of rJJ.n, which is said to be due to the filling of
empty recombination centers by the shift of Fermi level. The
defect density of the phosphorus-doped films is relatively low
compared with that of GD films, which may contribute to
the additional increase in rJf/-r besides the increase by the
10" shift of Fermi level.

0
.. - STREET ET AL.
References
o THIS WORK
10"
[1) Y.Mishima, Y.Ashida, and M.Hirose, J. Non-Cryst.
U DOPE D 10-' 10-' 10· Solids 59/60 (1983) 707.
PH. / Si,H.
Figure 7: Defect density of phosphorus-doped and undoped [2) A.Yamada, M.Konagai,and K.Takabashi, Jpn. J. Appl.
Phys. 24 (1984) 1586.
films estimated from below-gap absorption coefficients mea-
sured using a constant photocurrent method. The films were [3) E.Kawai, K.Kasatani, M.Kawasaki, H.Sato, and
deposited at Toub of 250·C. K.Hirano, Jpn. J. Appl. Phys. 28 (1989) 247.

[4) T.Fuyuki, K.Y.Du, S.Okamoto, S.Yasuda, T.Kimoto,

M.Yoshimoto, and H.Matsunami, J. ·Appl. Phys. 64
defect density obtained is proportional to the square root of (1988) 2380.
the dopant-gas ratio. This result is consistent with a model
of defect creation based on a mass-action law proposed by [5) T.Shirafuji, M.Yoshimoto, T.Fuyuki, and
Street[ll), which suggests that the same mechanism works H.Matsunarni, Jpn. J. Appl. Phys. Lett. 30 (1991)
for the VUV photo-CVD films. Although there is no drastic L538.
reduction in the defect density, the value is relatively low
[6) Y.Kuwano, and S.Tsuda, Int. Conf. Stability of Amor-
compared with that for GD films[12) as in Fig.7. This low
phous Silicon Alloy Materials and Devices, California,
defect density may contribute to the increase of the photo-
U.S.A., 1987, AlP Conf. Proc. 157 (1987) 378.
conductivity in addition to the filling of empty recombination
centers by the shift of Fermi level. [7J H.Matsunami, K.Tokuda, T.Shirafuji, T.Fuyuki, and
M.Yoshimoto, Proc. 4th Int. Photovoltaic Sci. and
Eng. Conf., Sydney, NSW Australia, 1989, p.289.
5.Summary
[8J D.A.Anderson, and W.E.Spear, Philos. Mag. 36 (1977)
In summary, we have examined photoelectric proper- 695.
ties of a-Si:H deposited by a photo-CVD method using
(9J W.E.Spear, and P.G.LeComber, Philos. Mag. 33
VUV(147nm) light.
(1976) 935.
Undoped films show dark conductivity as low as
lO- 11 S cm- 1 and photoconductivity as high as 1O- 5 S cm- 1 [10] H.Okushi, Philos. Mag. 52 (1985) 33. (Photon flux in
under AM1(100mW cm- 2 ) illumination. These excellent this reference is 9.5 x 1014 photons cm- 2 s-l.)
properties do-not depend on T,ub in the range of 200-400·C,
and neither on the defect density of the films . The photo- [I1J R.A .Street, Phys. Rev. Lett. 49 (1982) 1187.
conductivity degradation after the 3-hour illumination and
[12J R.A.Street, D.K.Biegelsen, W.B.Jackson, N .M.Johnson,
the defect density depend on Toub , and reach as low as 10% and M.Stutzmann, Philos. Mag. 52 (1985) 235.
and 2.2 x 10 15 cm- 3 , respectively, for T,ub of 300·C.

423
 POSTER SESSION 2C

PV Systems Technology
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

NEURAL NETWORKS -
A PROPER APPROACH TO THE ENERGY MANAGEMENT PROBLEM?

Marcus Stoll
Fraunhofer-Institut fiir Solare Energiesysteme
OltmannsstraBe 22
7800 Freiburg
Federal Republic of Germany

ABSTRACT. Neural networks are an emerging computional technology which mimic the computional abilities of biological
systems by using a large number of simple. interconnected artificial neurons. Because of their specific processing advantages.
neural networks are the technology of choice for an approach to the energy management problem. The progress in
performance that neural networks could offer, is demonstrated on the important sub task "state-of-charge (SOC) estimate"
of a lead-acid battery in a renewable energy supply system (RES). The results of theoretical investigations encourage
practical testing and further application of different kinds of neural networks to other sub tasks of the energy management
problem.

Adaptive learning: An ability to learn how to do

L INTRODUCTION tasks based on the data given for training or ini tial
experience.
Autonomous renewable energy supply systems (RES)
usually consist of one or more types of renewable energy Self-organization: A neural network can create its
converters (e.g. wind energy converter, PV-generator), a own organization of the information it receives
back-up component (e.g. diesel generator) and a storage (e.g. a during learning and operation.
lead-acid battery). In order to optimize the utilization of the
renewable and the primary energy as well as the operating Feature discovery: Neural networks provide a way to
conditions for the battery storage, it is necessary to make discover the salient, general features which can be
decisions (e.g. switching of a fraction of the load). used to classify a set of patterns.
In such a deciSion-making situation additional information
such as the state-of-charge (SOC) of the battery storage, the Generalization: The ability to respond appropriately
energy input and the energy demand of a future period (e.g. when the input is noisy. Additionally, the generaliza-
the next day) can help to make the optimal decision. III tion capability enables the production of appropriate
Unfortunately the time series of the renewable energy input, response to new inputs which have similarities to the
the load demand and the state-of-charge include stochastic training stimuli.
portions, which aggravate the use of this information. The
extraction of useful decision quantities from this semi-chaotic
time series is a demanding task. 2. SUBTASK "SOC ESTIMATE" - A GENERAL VIEW
A suitable technology to approach this problem are neural
networks. This emerging computional technology mimics the It is well recognized that the determination of battery
computional abilities of biological information processing SOC is an important sub task in solving the energy manage-
systems by using a large number of simple, interconnected ment problem. 131 The SOC estimate is a useful quantity
artificial neurons. Neural networks offer improved perfor-
mance over conventional technologies in areas like: in planning the utilization of renewable as well as
primary energy. 111
• Pattern classification
in optimizing the operating conditions for the
• Prediction of chaotic time series system's battery. 13,41

• Complex mapping in decision-making within the system. III

• Optimization of complex tasks The SOC has an important role in the planning and deci-
sion-making, and therefore it is necessary to have a reliable
• Decision-making SOC estimate of the battery. Various methods for the deter-
mination of the SOC are used:
There are many different types of neural networks, each
of which has different strengths particular to their appli- Ampere-hour balance: Limitations arise from the fact
cation. 121 Neural networks offer specific processing advanta- that the charge acceptance efficiency and the self-
ges, which make it the technology of choice in the manage- discharge factor are varying quantities. The resulting
ment area too. For the energy management problem the balance errors are integrated over time and this leads
following features of this technology are important: to large inaccuracies. 151

427
 Measurement of battery voltage: This is a very
simple and widespread practice, but the voltage at the
terminals is not related exclusively to the SOC.
Unfortunately the voltage depends on a variety of
quantities (e.g. temperature, discharge current, etc.).

Measurement of electrolyte density: The electrolyte

density is, in principle, related directly to the SOC.
However, operating under cyclic conditions leads to
spatiaI-temporal gradients of the electrolyte densi ty
and prevents the attainment of equilibrium in the
mass of the electrolyte.

Storage of discharge characteristics: In this case the

parameterization of the complex discharge phenomena
is necessary.

As things stand at present, results obtained with these

Fig.I: The neurode s in the layer A of the TPM
methods, perhaps excepting the storage of discharge characte-
is activated by the input vector v and
ristics, are not yet satisfactory. IS, 61
responds with the output vector W"out.
The vector u would be the optimum
3. NEW APPROACH TO THE "SOC ESTIMATE" response to the input vector v. 171

The basic idea of the new approach is to train a neural

network with a battery model or real data to estimate the The dimensionality of the pattern vectors must corre-
SOC. To reduce the expense, the neural network is trained spond to the dimensionality of the input vectors. The self-or-
only for a small part of the whole discharge state space of ganization results from the lateral cooperative-competitive
the battery. During most of the time the discharge current Id interaction among the neurodes.
in stand-alone systems is very small (e.g. 12so < Id < Iso) and
therefore the training of the neural network is restricted to 5. LEARNING IN THE TPM
this range of the discharge state space.
For training a data set consisting of the discharge current The aim of Kohonen's algorithm is to generate a mapping
Id' the discharge voltage Ud and the SOC under standard of a usually higher dimensional space V of input signals onto
operating conditions (T = 20·C, Id = 110) is used. Additionally a usually two-dimensional discrete lattice A of formal neuro-
the age ( product of discharge cycles and discharge depth) and des r with the pattern vector wr • The learning process is
the mean temperature of the battery could be defined for the summarized as follows:
whole trainig period. This is done with a lot of different data I. Assign suitable initial values to the pattern vectors Wr E
sets to impart the knowledge of the discharge behaviour of V. If no a-priori information is available assign random
the battery to the neural network. Due to the generalization values.
ability, the neural network self-organizes an appropriate
internal representation of its experience with the battery. 2. Select a input vector v E V according to some pre-speci-
In the recall phase the neural network is provided only fied probability distribution P(v).
with the discharge current and the discharge voltage and
responds with the corresponding SOC of the battery. 3. Determine the location s for which II v - w. II is minimal,
The choice of a neural network depends on the application
i.e.
task. Because this is a complex mapping problem, the self-or-
ganizing Topology-Preserving Map (TPM) has been chosen.
II v Ow. II < II v - Wr II for all rEA.

4. Perform a learning step affecting the pattern vectors of

4. THE SELF-ORGANIZING TOPOLOGY-PRESERVING MAP
all neurodes r in the neighbourhood of s (with s inclu-
ded):
The self-organizing Topology-Preserving Map (TPM),
developed by Teuvo Kohonen 12,7/, operates on a vector
matching principle. The basic concept underlying the TPM is
that a set of pattern vectors can be distributed across the
state space of the input vectors so that their arrangement in 5. Perform a learning step affecting all output vectors of
the space mimics the probability distribution of a set of the neurodes r in the neighbourhood of s (with s inclu-
training data. The mapping of input vectors to the set of ded):
pa.ttern vectors preserve the topographic relations which may
have existed among input data. To each neurode of the layer bwro ut = w rout + f2 • h... 2 • ( u - w ro ut )
A belongs a pattern vector W"in and a output vector W"out·
(see fig,l) Every pattern vector will become an example for a and continue with 2.
class. When an output vector W"out is selected, a single neu-
rode s (pattern vector W"in) is activated. This is the one f\, ~ are learning step sizes and h... 1, h... 2 are pre-specified
which the input vector v most closely resembles. adjustment functions.

428
Fig.2 to Fig.4 show the progress in building a map for the 6. RESULTS
discharge state space of a lead-acid battery with 81 cells.
Every node of the lattice represents one pattern vector. The In order to evaluate the reliability of the TPM for the SOC
dots correspond to the applied training vectors of the estimate a RES consisting of a diesel enlline, a wind turbine, a
discharge state space. battery, an inverter and several consumers was modelled.
Based on the approach of Didier Mayer 181 a comprehensive
175.0..,--;-- - - - - -----------. model of the lead-acid battery was developed. The model
considers the following features of battery behaviour:

170.0 dependence of capacity on discharge current, battery

temperature, number and depth of discharge cycles.
> 165.0 dependence of discharge voltage on discharge cur-
rent, battery temperature, number and depth of
;;3' discharge cycles and actual SOC.
160.0
dependence of charge efficiency on actual SOC.

\55.0 -\-_,.-_,.-_,.-_,.-_,.-_.,-_.,-_.-----1 dependence of charge voltage on charge current,

0.00 0.0025 0.0050 0.0075 0.0100 0.0'25 0.0\50 0.0175 0.0200 battery temperature and age of the battery.
Id [eloA J
consideration of gassing behaviour.
Fig.2: Initial distribution of the pattern
vectors in the discharge state space of Fig.S shows the SOC estimate made by the TPM in the
a lead-acid battery with 81 cells. simulation of the RES. The training model for the TPM and
the model used in the RES are identical. Due to the fact that
the TPM is trained only for a cutting out of the state space,
1iO~..,----,--~-~---~--~__, an estimate is done only during the discharge cycles of the

-r--r II
battery.

168.0 -- -- r I IOO•• T"""--,---r-"!"'<T-"""<""------,....-----......- -,

>
-
..
166.0

::I
164.0
-r+I 1- '"
U-
....
162.0
0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 * <4••
Id [ e,oA )
J
Fig.3: Distribution of the pattern vectors 20•• . ..,.
after 300 training steps.

• .•'+---;.- - ___--+--l--....;...-.....;---l--....
0.0 " .0 ~. O 75.0 lOO.O 11$.0 150.0 17),0 :aO.O

time [h I
170.0

168.0 ,-- Fig.5: "Actual" SOC (solid line) and estimated

(dashed line) in a simulation of a RES.
> l66.0

The SOC estimate of the TPM is restricted to the range of

162.0 20 to 80% C1O' This is a consequence of boundary effects in
the array, which define the format of the map in a manner
160.0 +--+----,,---;--,-.,-- .-----,,----,---+----1 somewhat similar to the way boundary conditions determine
0.00lI.OO4 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020
the solution of a differential equation. (see Fig.4 upper edge
Id [eloA I of TPM)
To explore the parameter robustness different Ieey quantities
Fig.4: Final distribution of the pattern in the training model for the TPM were varied. The robust-
vectors after completion of "the trai- ness against variation of the bat tery temperature is shown in
ning phase. The TPM was trained for a Fig.6. The TPM was trained for a mean battery temperature
new battery with a mean battery tem- of 20·e and then used to estimate the SOC in a simulation
perature of soc. of a RES with an average battery temperature of soc.
Despite the great difference of training and recall situation,
the SOC estimate is useful.

429
 ..... 9. REFERENCES

!
.... ! !
/ I / Heinemann D., et al.: Utilization of Weather Forecasts
and Climatic Information for the Control of Renewable

.... --·~J~.~r. )\'

Energy Systems.
2
Proc. of 8th PVSEC. 1988, pp. 249-253
U / 2 / Maren A.• et a1.: Handbook of Neural Computing Appli-
~
!i !- ~, ..l cations.
i Academic Press Inc.. I 990
~ t··~-·--·-,!·-···,·-·--t·--···-··+··,··-·.. ·~-~----~ ___.. _L_ / 3 / Groumpos P.P., et al.: The Importance of Controls for
J ! !
M

h: 1 an Efficient Operation of Stand-Alone PV Power

! i i Systems.
Proc. of 19th IEEE-PV-Spec.-Conf., 1987, pp. 1246-
:!G ••
····-·-·--·\··-·--·-·1·_··--··-1·- ·---·I·-·········-r·--·----I···········-:-----··· 1251
/ 4 / Willer B.: Management of Electrochemical Battery
D"o+---+I_---,f-_+-_-+_~i--.,---_Ii· -_~ Storagein Photovoltaic Energy Supply Systems.
0.0 :Z5..0 .50.0 n.o Q$.O ISO.O m.o 200.0
~. O
Proc. of 9th PVSEC, 1989, pp. 795-798
time [h) /5 / Philipp R.: Untersuchung und Optimierung eines Bilanzie-
rungsalgorithmus zur Ladezustandsberechnung von
Bleiakkumulatoren in Solaranlagen.
Fig.6: Robustness of the SOC estimate Diplomarbeit, Fachhochschule Offenburg. 1990
against variations of the battery tempe- /6 / McCarthy S .• et al.: Operational Experiences with Bat-
rature. teries in the 16 PV Pilo t Plants.
Proc. of 9th PVSEC. 1989. pp. 1142-1145
/ 7 / Ritter H. et al.: Neuronale Netze.
Addison Wesley, 1990
Furthermore the sensitivity of the SOC estimate against the / 8 / Mayer D. et al.: Modelling and Analysis of Lead-Acid
variation of the open circuit voltage of the new, fully char- Battery Operation.
ged battery and t1.e electrolyte resistance was explored. The Proc. of 9th PVSEC. 1989, pp.245-248
SOC estimate is only sensitive to variations of the open / 9 / Berndt D.• et a1.: Bleiakkumulatoren.
circuit voltage, which is related chiefly to the active material Vartafachbuch. VOl-Verlag. 1986
activity and thus to the acid concentration /8,9/. Because the
open circuit voltage can he easily and reliably measured, this is
no serious limitation of the method.

7 . ADVANTAGES OF THE METHOD

Due to the features of the TPM there are some remar1cable

advantal!es of the method:

Great parameter robustness: Even when the training

situation differs a lot from the recall situation, the
TPM is able to make useful estimates.

Generalization ability: As an example, suppose the

battery is getting older and thus the discharge vol-
tage is decreasing. Although the TPM never was
trained for this case, it responds to this phenomenon
correctly by decreasing the capaci ty.

Data compression ability: The TPM self-organizes an

optimum code-book for the input space, which could
he easily implemented in a microcontroller, and thus
leads to inexpensive devices for SOC estimates.

8. CONCLUSIONS

In simulation calculations the TPM has shown a sophisticated

performance in estimating the SOC of a lead-acid battery. In
order to validate the method previously descrihed, an experi-
mental test set-up will be installed. The realization of simpler
and more inexpensive implementation of SOC devices is thus
possible.

430
10TH EUROPEAN PHOmVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

The Effect of Tilt Angle and Voltage Conditions on PV System Performance

An Experimental Investigation

M. Jantsch, M. Stoll, W. Roth, R. Kaiser, H. Schmidt, J. Schmid

Fraunhofer-Institute for Solar Energy Systems
Oltmannsstr. 22, D-7800 Freiburg, Germany

Abstract:
Five PV-Modules at different tilt angles were installed and connected to an automatic measurement
equipment. The I1V-characteristics, temperatures and insolation values have been measured in two minute
timesteps since autumn 1989. Stored into a database, all the data is available for detailed analysis.
The results up to now include
• tilt angle influence on energy yield and its annual course,
• energetically weighted MPP voltage distribution and its annual course,
• fixed voltage adaptation and best fixed voltage, annual and seasonal adaptation efficiency,
• energy yield as a function of fixed voltage and tilt angle compared with MPP tracking at optimum tilt angle.

Introduction of the 0.4 m2 modules of type AEG PQ 10/40 contains

40 serial poly-silicon cells.
The tilt angle influences the amount of energy col- Insolation is measured in the horizontal and vertical
lected by a PV module. However, representations of orientation, and measurements of the ambient tempera-
these effects on the energy yield and the performance ture and the five module temperatures are also taken
of the PV system are commonly based on theoretical (Figure 4). By controlling an electronic load the data-Iog-
calculations, whereas there is a lack of experimental
investigations. To verify the calculations, an experimen-
tal setup has been installed at the institute's test field.

1. Experimental Setup
Located at 48° north and 7° 52' east in central Europe
(Figure I), an installation with five modules facing south
with tilt angles from 10° to 90° have been installed
(Figure 2). Unlike calculations using synthetic data or
theoretical considerations, realistic experience of clima-
tic influences can be obtained here (Figure 3). The
minimum tilt angle of 10° was chosen in order to avoid
rainwater sedimentation on a horizontal module. Each
Figure 2: Five modules at 10· to 90· tilt angle

Figure I: Location at 48· north and 7· 52' east Figure 3: Realistic energy losses in winter

431
The Effect of Tilt Angle and Voltage An Experimental Investigation
Conditions on PV System Performance

ging computer locates 30 sample points into each UV- 2. Tilt Angle Influence on the Energy
characteristic so that their density is increased around
the MPP in order to assure sufficient accuracy under all Yield and its Annual Course
operating conditions. Through connection to the Insti- As expected, the dependence of the energy yield on
tute's computer network the measured data is entered the tilt angle varies in annual course according to the
into a database day by day, and all the data can be sun's azimuth (Figures 5 and 6), Furthermore the energy
accessed by any other user. ratio between the tilt angles with highest and lowest
output is much smaller in winter (ratio between 1.2 in
winter and 2.1 in summer) because of the bigger diffuse
fraction of insolation in our region during this season.
The optimum annual energy yield can be obtained at a
tilt angle of 30°. The output in winter is best at tilt
angles of 50° ... 70°, although in summer such tilt angles
decrease the energy output by 1.1 ... 1.4. The ratio for
energy per month between summer and winter ranges
from 8 for horizontal tilt angles down to 2 for vertically
mulliplexer mounted modules. Certainly, there are big differences
electronic
and lrod possible between the same months in different years
mullimeter (ratios> 2 have already been observed), This has to be
taken into account in designing a system.

Figure 4: Block diagram

6 - '" july
----- ________ '
'-.---
" whole yearll2
...... december
4
II kWh =0,85 cm~ ", august
,......... .......... whole year/12
-:5
c:: 4.5
5'c:: ...... january
,," ''
0
E ----'--'-"_.,-"_...................... 0
E
3
,, ''
,: \ ,,
:c ;:> ,, ,,
~
......
~
3 :c 2
,,,
,
,,
,
~ !
......
·
,
I•
>. ~

...
CD
v 1,5 ............... -................................ >. , ,
c CD I -
...
.................................................... t..
11) 11)
c:: ,,
V '

0
............................. .~ .......
30 SO 70 90 0-
10 12 14 16 18 20
tilt angle [0] MPP voltage [V]
Figure 5: EnerllY yield versus tilt angle Figure 7: MPP voltage distribution

,
6 19 ~upper bound ,,
'V \... I , mean "
"3
5'c:: '8 18 • ______ ._ ....... lower bound /
0
E
4.5
E
......
-,
·. ............-........... --'

, . ', .... -"

:c G 17 '-----.
-
'\ ,
......... ,'
~ 3 , ., , , ., v
......
~
' ... ... ... ... g
CD
16 ...... ,
>. ... ........ ,
,

-0
CD
t..
v 1,5 :> ................................................ . ....................
c:: 0.. 15
v 0.. ........",.
~
oL-,---..---~-~---,.---.-----, 14
Jan Feb Mar Apr May Jun Jul Aug Scp Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct ov Dec

month (990) month (1 990)

907. or cnctQy output at vo111J8C", belwecn bou nds:

Figure 6: Annual course of enerllY yield at different tilt angles Figure 8: Annual course of the MPP voltage distribution

432
 The Effect of Tilt Angle and Voltage An Experimental Investigation
Conditions on PV System Performance

3. Energetically Weighted MPP Volta(le ned by climate and insolation, this adaptation curve is
shifting in voltage, and also the associated energy yield
Distribution and its Annual Course varies. By weighting the adaptation figures with the
The annual course of insolation and temperature has corresponding energetical contribution, the curve of
effects on the MPP voltage distribution (Figures 7 and 8). fixed voltage adaptation can be derived (Figure 9).
The range [mean] shifts from 14.5V ... 16.5V [15.5V] in Its form and value depends on the module's electrical
summer to 16V ... 19V [17.3V] in winter while the characteristics, the climatic conditions (temperatures and
energy yield is altered as mentioned above (The 'range' insolation), and also a'little on the tilt angle (at 30 0 tilt
here marks those voltages at which 90% of the irradia- angle the curve lies by 0.1 V lower, at 90° tilt angle by
ted energy can be obtained at MPP). The figures are 0.2 V higher than at 10 ,50° and 70 tilt angle).
0 0

mainly conditioned by the thermal operating conditions An annual optimum of 98.4 % is obtained at the best
of the modules which are not only caused by climate, fixed voltage of 15.9V/module which is 0.2V lower
but also affected by the module mounting method used than the energetically weighted mean of the MPP vol-
(installed free or integrated; wind or rear airstream tage. Choosing the best fixed voltage leads to misadap-
possible, or none). tation losses smaller than 4% in summer and 9% in
winter (These figures relate to the hitherto measured
data, e.g. at very cold winter days the losS could be a
4. Fixed Voltage Adaptation and Best little bigger).
Fixed Volta(le. Annual and Seasonal Surely, only a small amount of energy at low power is
affected by the losses in winter. For this reason, an
Adaptation Efficiency MPP tracker consuming energy all over the year does
The efficiency of adaptation as a function of voltage is not payoff energetically compared with an optimally
proportional to the module's power/voltage characteri- fixed voltage.
stic with a maximum value of 100% at MPP. Conditio- The adaptation of batteries and electrolysers can be
evaluated now by
FIXED VOLTAGE ADAPTATION considering a vol-
summer winter tage range instead
min .......... overages ........ · max of one fixed vol-
I tage. The momen-
mpp voltage tary voltage is
tilt = J00
(see scale below) distributed within
this range accor-
ding to the battery
type and state of
charge (or electro-
lyser type and tem-
monthly contribution to {
annual energy yield [%] perature), and the
momentary current.
[ %] mppt (without converter losses I This distribution is
c
.2 100 annual ........ also affected by
o
a.
o
-0
o
worst of summer
95+-------------~~~~-+----~~~------+-----~--------- ""- the system manage-
ment behaviour.
The adaptation
'0 worst \
>. figures within this
u 90~======~~~====~==~
C
range have to be
.~
.!:! \ weighted according
Q)
to this voltage
/ \I U; \ distribution to
.2l \ obtain the wanted
11 12 13 1[, 15 16 17 18 19 20 Iv] adaptation value.
module voltage

Figure 9: Fixed voltage adaptation and best fixed voltage, annual and seasonal
adaptation efficiency for a 40 cells module, type PQ 10/40 (AEG).

433
The Effect of Tilt Angle and Voltage An Experimental Investigation
Conditions on PV System Performance

• It yields an annual energy output of 98.4 % of MPPT

5. Energy Yield as a Function of Fixed operation.
Voltage and Tilt Angle Compared with • Lead/acid batteries are therefore matched best by 34
MPP-Tracking at Optimum Tilt Angle solar cells in series per 12 V, i. e. 36 cells for lower
system voltages with usual cabling and charge regula-
The results concerning the influence of tilt angle and tor losses considered.
voltage conditions on the energy yield can be combined • The tilt · angle yielding maximum annual energy
into a function of which the graph is presented below depends on latitude and diffuse fraction of insolation.
(Figure 10) . The best fixed voltage at optimum tilt angle In Freiburg it is 30°.
leads to an annual energy yield only 1.6% lower than • For autonomous PV systems without backup energy
that obtainable by a MPP tracker without any losses. steeper tilt angles and higher voltages enhance
The decrease of energy yield conditioned by tilt angle energy yield in winter.
and fixed voltage is plotted in the diagram. It can • MPP tracking, if it requires additional hardware, does
obviously be seen that a 12V lead/acid battery with a not payoff energetically compared with an optimally
voltage swing inbetween 12V ... 14.4V is not optimally fixed voltage.
matched by module with 40 serial cells. The new
module types containing 36 serial cells are better mat-
ched, in case of higher system voltages an adaptation of 7. Conclusion
34 cells per 12V could advantageously be used.
Representations of the influence of tilt angle and opera-
To obtain the adaptation graph for other numbers of
ting voltage on the amount of energy collected by PV
serial cells than 40, the voltage axis may be rescaled by
the number-of-cell-ratio, if necessary corrected by a modules are mainly based on theoretical calculations,
factor according to cell characteristics. E.g., the scaling whereas there has been a lack of experimental investi-
factor for TST PO 36/45 modules is 0.91 ... 0.93. gations. The work described here contributes consider-
ably to overcoming this deficit, not only by direct
measurements of the energy yields, but by providing an
6. Summary of the Results unbroken and detailed database for further investigati-
ons including simulations and statistics. If care is taken
• The best fixed voltage of IS.9V/module is 1.25 % in considering the corresponding regional climatic
lower than the MPP voltage's annual average (ac- differences as well as the specific thermal conditions
quired from 1990 data). due to module
mounting, the
results of these
ongOing works can
be generalized and
transferred to other
locations.

8. References:
<I> Gonzalez, Hill.
Ross: 'Photovoltaic
Array - Power Condi-
tioner Interface Cha-
racteristics·. Jet Pro-
pulsion Laboratory.
Pasadena I California
1982
<2> Schmid, Dine\:-
lage: 'Concerted
Action on Power Con-
ditioning and Con-
trol'. Fraunhofer-
Institute for Solar
Energy Systems. Frei-
Figure 10: Annual energy yield as a function of fixed voltage and tilt angle compared w!th burg I Germany 1989
MPP tracking at optimum tilt angle for a 40 cells module. type PO 10/40 (AEG).

434
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON. PORTUGAL

Comparison between measured and computed data for a PV system.

G. Ambrosone, S. Catalanotti, U. Coscia, G.Troise, and A. Sarno*

Dipartimento di Scienze Fisiche, Universita' di Napoli "Federico II", Facolta' di In-

gegneria, Piazzale Tecchio 80, Napoli, Italy
* Centro Ricerche Fotovoltaiche ENEA Via Vecchio Macello, Portici (NA) , Italy

ABSTRACT. The problem of correct dimensioning of photovoltaic plants is not

yet completely solved. For this goal it is necessary to have an accurate and
simple tool for the prediction of the fractional load covered by the PV system
as a function of its dimensions.
In this paper we validate a simple but accurate simulation program, based on
daily energy balance, comparing the data measured on a PV plant for a long period
with the forecast of our program.

1. INTRODUCTION. load is satisfied directly with energy produ-

A promising way to satisfy some of the ced by the PV array. The noct~rnal load is
energy needs of developing countries are stand covered by storage. The eventual diurnal
alone photovoltaic systems (SAPV). energy excess can be stored if the storage is
To be acceptable they must have a good not already full.
reliability level and a competitive cost as The program is divided in subrout ines; each
regards conventional energy sources. Both one models one block of the complete system
goals require a correct dimensioning, of the including PV array, the power conditioning
SAPV because its oversizing is expensive and equipment, the battery storage, the battery
the undersizing implies a bad reliability controller and the inverter. This allows to
level. simulate different system configur-ations as
The problem of SAPV dimensioning is solved chosen by the user. The daily results can be
when one is able to determine the fraction of summed to give monthly or yearly values.
load covered on a long period as a function of Our results are significant only as long
PV array and battery sizes. period forecasts (at least > 60 days). Since
To contribute solving this problem we have many hypotheses are based on averaged values
two valid tools: a program that simulates ac- the longer is the period the more accurate are
curaeely a PV system with relatively simple the results.
calculations and a completely monitored PV The program is written in Microsoft Quick
system. Basic (vers. 4.5) for a PC MS-DOS; one year
In the following we describe the program, simulation requires about 10 seconds run time
the PV system and the measurements done. We (for a 12 MHz 80286 CPU with a 80287
conclude with a comparison between simulated coprocessor) .
and measu~ed data.

3. PV SYSTEM.
2. PROGRAM DESCRIPTION. The PV system installed in our laboratory
Our p,rogram requires daily solar energy consists of a PV array and a battery storage.
values (easily available climatic data) and The cell array is formed by 24 modules
needs some detail of the components (efficien- (Ansaldo AP33D); the array area is 8.64 square
cies as a function of power) . meters and the peak power is 760 W.
Our computer program simulates, day by day, The storage is formed by 12 lead acid cells
the energy produced by the PV system using (Tudor ATF230/21), connected in series, with
daily climatic data. The behaviour of the PV a nominal capacity of 280 Ah. The load is DC,
system is simulated in a simple way. Each so there is no inverter and the load is
component is characterized by a transfer directly connected with the battery.
function, constant on a day, but dependent on These three blocks are connected by a power
daily averaged power value (variable day by conditioner (Sipre) which regulates the wor-
day). Thus we obtain the daily output energy king point among array and total utility (load
multiplying the input ener-gy by the pr-oduct of and batter-y) and contr-ols the battery voltage
the transfer functions of all components. to avoid excessive charging or discharging.
In the energy balance we separate the load During the discharge, the battery control-
required in day-time (diurnal load) from the ler opens th~ connection between the battery
night-time (nocturnal load) . Only the diurnal and load (only the PV array is then connected

435
to the battery) when the voltage value program is a good simulation·of long period
decreases below a fixed value (1.83 V for load coverage, which is the datum needed for
element) . This connection has to be reset by PV system design.
the user. Since our simulation is based on daily
During the charge process the battery energy balance we do not expect a good
controller opens the connection between the PV agreement between measured and simulated
array and battery when the voltage is larger quantities day by day.
than a fixed value (2.42 V for element). This However we present some daily curves to
connection is automatically reset when the better understand how our hypotheses fit the
voltage reaches 2.25 V for element. real behaviour of the system.
With an HP 3497A data logger, electrical Furthermore, to understand the reliability
data as voltages and currents and climatic in modelling each component, we made also a
data as the solar radiation, the temperature, partial validation. In other words we compared
thewindspeedetc .. are measured. All data are separately the simulation of the active part
stored as integrated values on 20 minutes (PV array and power condi t ioner) and the
intervals. passive part (battery, battery conditioner
All these instruments are ENEA property. and load) with the correspondent measured
quantities. These separate validations help
also to control that the two part are simulated
4. MEASUREMENTS. with similar approximation.
We did a series of measurements on our PV
system to determine the energy transfer 5.1 The active part.
functions for each component to be used in our We simulated the array by means of a
program; more details are in /1/. transfer function which depends ori the mean
We have also carried out long period (more solar power. Starting from daily solar radia-
than a year) measurements on the system tion impinging on array and using the transfer
working with a resistive load. function we obtained the maximum energy
The resultR can be best presented as a producible by array.
function of two.adimensional quantities rela- This energy is divided between battery and
ting load, system dimension and solar energy. load. This operation is regulated by the power
These two quantities are: conditioner that chooses the working point of
X: the ratio between the mean active energy array considering the state of battery charge.
(the mean solar energy incident on the Then the transfer function of the power con-
array multiplied by the mean value of ditioner depends on the input energy (energy
array and power conditioner efficiency) supplied by array) and the battery energy
and the mean load. level.
T: the ratio between the maximum energy To validate the models for array and power
storable in the battery multiplied by the conditioner we have compared the simulated
mean value of battery efficiency and the active energy and the measured energy output
mean load. of power conditioner (i. e. energy used by
Thus, X represents the days of load covered battery and load) .
by energy produced by the active part of In the simulation we introduce as input data
system; whereas T represents the days of load the measured solar radiation and measured
covered by stored energy. battery energy level, day by day.
For our data the averaged daily values of In Fig. 1 we show the measured active energy
load, solar radiation and parameters X , Tare and the simulated one. The agreement is good,
presented in Table 1. with an averaged relative error of +1.5% .

5.2 The passive part.

5. VALIDATION. We schematize the battery as a box whose
What we really hope to obtain with our energy level or state of charge is always
Solar Daily Covered Covered
SET Day Energy Load Load Load <X> <T>
(kWh) (Wh) (Wh) (%)

1 93 43.5 Measured 2956 2588 87.6 1. 05 2.13

Simulated 2920 2692 92.2 1. 06 2.16

2 87 23.9 Measured 1854 1727 93.1 1. 15 3.40

Simulated 1988 1832 92.1 1. 07 3.17

3 96 43.4 Measured 2882 2672 92.7 1.16 2.19

Simulated 2879 2703 93.4 1.16 2.19

4 104 32.4 Measured 3585 2109 58.8 0.72 1.76

Simulated 3528 2190 62.1 0.73 1.79

TABLE 1 Summary of the results; The mean daily values of energy for all the set

436
4 Level of Storage (kWh)
Active energ (kWh)

9
I~:'rr !\.;
3 -
.
,.'

\ l
A 7

2 -
5

- - Measure - - Bat: Measured

Simulate ... Bat: Simulated
Day Day
1 I I I I 3 ~

o 20 40 60 80 100 0 20 40 60 80 100
Fig. 1: The energy produced by the PV field Fig. 2: The Daily Energy Level of storage vs
vs day, as measured and as computed. day, as measured and as measured.
known. defect of energy is stored in battery or
The state of charge is given by the time supplied by battery if it is possible. The
integral of the exchanged current. The energy nocturnal load is supplied by the battery.
level of the battery is assumed to be its state The battery energy level variation is
of charge mUltiplied by the open circuit obtained from the input energy multiplied by
voltage . the charge efficiency evaluated at mean values
Sirice the open circuit voltage is also a or from the output energy divided by the
function of the state of charge. the state of discharge efficiency.
charge and the energy level are interchangea- To take into account the function of the
ble. battery controller we limited the charge
We define separately the efficiencies for between values Emax and Emin. When the battery
charge and discharge processes: the charge level is Emax, any storable energy is dumped.
efficiency is the ratio between the battery When the battery level is Emin, the load is not
energy level variation and the energy used covered (deficit).
during the charge process; the discharge To protect the battery, after a deep di-
efficiency is the ratio between the energy scharge, the load is not satisfied until the
delivered during the discharge process and the battery energy level does not reach a value
battery energy level variation. Erec = 60% Emax.
Due to .the definition of battery energy We have determined Emax and Emin from the
level, the two efficiencies become the ratios fixed voltage values used by the battery con-
between two battery voltages at the same state troller.
of charge: in charge process the ratio between The level chosen for Erec has almost no
open circuit and charging voltage, in dischar- influece on long period evaluations because it
ge process the ratio between discharging and causes only a shift in occurance of deficit.
open circuit voltage. For our battery we started determining the
Since the voltage in charge and discharge state of charge as a function of open circuit
processes is a function of current and state vol tage . From measurements of vol tages in
of charge, also the efficiencies depend on the charge and discharge processes at c o nstant
same variables. current we obtained the two efficiencies
The program simulates the behaviour of the (charge and discharge) as a functi.on of
battery making day-time and night-time energy currents and state of charge.
balances. When the current is rapidly variable the
Starting from the initial energy level and battery behaviour changes, but on the long
using the active energy, it calculates the new period the mean values are similar.
energy level and the energy delivered to the To validate the battery simulation we
load. compared the measured and the simulated
During day-time, the diurnal load is battery energy level, introducing as imput
compared with the active energy. The excess or data the measured energy exchanged by the

437
 Covered Load (kWh) Covered Load (kWh) - - Measured
. Sim laled
3 4 -.\

3 -
...
2 ,. :

.
2 -

1 ~
1 -
.'
"
Day
o o I
Day
o 20 40 60 80 100 o 20 40
Fig. 4: The covered load for a subset of
Fig, 3: The covered load for Ihe Sel 1 vs the Set 4 vs day, as measured and
day, as measured and as computed. as computed.
battery with load and PV field, day by day. program the battery is reconnected to the load
This comparison is shown in Fig. 2. The results when its energy level reaches a fixed fraction
are in satisfactory agreement; the error is of Emax (60%). In the real system, ' for
almost always low; for instance in the showed different reasons, sometimes the user let the
set the error ranges from -6% to +17% with an battery energy level exceed this value.
averaged value of +2 %. The error in simulated covered load is low
and of the same order of the error in simulated
active energy and battery energy level.
5.3 Validation of system We could hctve lower errors using transfer
Finally we validate the whole system. functions dependent on more parameters (e.g.
We have run our program using as input the taking into account battery temperature) but
daily solar radiation and the load resistance. this is senseless due to the aleatoriety of
In Table 1 we shown the comparison between climatic data. In fact the variation in the
measured and simulated quantities for four yearly average solar radiation make improve-
sets of data; the most important result is, ments in component simulation almost useless.
obviously, the total covered load fraction.
'l'he results show t'hat a fractional covera-
ge larger than 90% is possible only when X is REFERENCE
greater than 1 and T greater than 2. (1) G. Ambrosone et aI, The SAPV Simulation
To give a deeper insight we show also the Program, submitted to Revue De Physique Appli-
comparison between the measured and simulated quee for publication, November 1990.
covered load day by day. We want to underline
that this comparison gives information on the
behaviour of the PV system, but the important NOTE
data is how much load will be supplied by solar This work has been partially supported by an
system and not the detailed daily balance. ENEA contract.
In Figs 3 and 4 we present the daily load
supplied by the solar system as obtained by
measurements and simulation. The small fluc-
tuations in the daily load are due to the
voltage of battery varying with state of
charge.
The agreement between the measured and
simulated covered load is good for inost days.
There is sometimes a shift of deficit days:
this is due to the battery controller; the re-
connection between batte~ and load after a
deep discharge is manual. In simulation

438
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ISEE* - Renewable Energy Information System

Public Database for Components of Solar Energy Supply Systems
Reinald Grebe, Harald Koch
Department of Control Engineering
ISET - Institut fOr Solare Energieversorgungstechnik e.V.
0-3500 Kassel - Konigstor 59
Telephone (0561) 7294-0, Telefax (0561) 7294-100

• ISEE - InformationsSystem Erneuerbare Energie

displayed as a data sheet on the monitor or printed by the

user's printer.

As an example, the following figure shows the input menu

when searching solar modules Cfu-l).
Since the end of 1989 ISET has been operating a central
database containing information about the componen~ of
solar energy supply systems (lSEE). This project is being
supported by the Hessian Ministry of Economics and ISEE - Renewable Energy Information Syste!11
Technology. Function: Selective Search/Print
Input of user definable criteria
Access to this on-line database may be obtained by any Selected Component: SOLAR MODULE
interested party posessing a computer, which is connected 1. Identification mark: sm*
to the network of the 'Oeutsche TeleCom' by telephone or 2. Character: monocrystalline
Oatex-P. The command-language of ISEE is german. 3. Protection Diode: no
4. Nominal Power fW]: min: 40 max: 45
ISET will establish an english version in 1991/92. 5. MPP Voltage [V: mln:unllm. max: unlim.
B - Criteria are logically combined with: AND
F-ready A-stop XXX-end L-delete all criteria

Figure 1: Defmition of the 'Search Key'

In case of succesful search the data will appear in form of

At present, both technical data and information about the data sheets. The following figure <B&...l> shows the first
data sheet of a solar module record to the search key.
costs may be taken from the ISEE database. Information
about the following components used in solar energy
supply systems is available:
Data Sheet - Solar Module
07/27/90 1 st SET 18:04:06
~ Solar modules
Id. marie :sal14043ar
~ Wind energy converters Company :ARCO Type: M 65 UL
~ AC-OC converters Character :monokrystalline
Protection Diode: -0-
~ DC-DC converters
~ Rechargeable Batteries
Nom. Power [W]: 43.0 HPP Voltage [V]: 14.5
~ Charge regulators MPP Current [A]: 2.90
~ Diesel generators Open-circuit voltage [V]: 18_0
Short-circuit current [A]: 3.27
~ Phase shifters TK-MPP-Power [11K]: -0-
TK-Open-circuit voltage [11K]: -0-
TK-Short-circuit current 11K: -0-
NOCTemperature [DC] : 47
The databases of these components can be examined Shunt resistor: -0-
No. of cells serial: -0- •• _ parallel: -0-
according to defmed criteria with the results being either
' - - - - - - - - - - - - - - - Page 1 of 3 -

439
 ~ 'Readers' only have the right to consult and print
Standard test conditions: from the contents of the database. Entering
infonnation (e.g. improvement proposals) may be
Insolation [Y/m']: 1000. carried out using an internal mail-box system.
Ambient temperature rOC] : 2S
Air Mass AM : '.5
~ 'Suppliers' may enter infonnation about the above
Temperature range [DC]:
min.: -40 max.: 90 Humidity [Xl: 100 mentioned components in a predetennined part of the
system. They also have the same rights as readers.
Dimensions [rrrn] :
Length: 1083 Yidth: 330 Height 36
Yeight [kg]: 4.8
Page 2 of 3 ~ The 'administrator' of the database (ISET) is
responsible for: recording the entries of the suppliers
in the current database after they have been checked
Notes: for any inconsistencies, securing the contents of the
- .odule frame - aluminiua profile database, reading from and drawing conclusions
- extreae environmental conditions from the mail-box system.
- integrated bypass diode

Costs

Date of input: 07/2.7/90

Price: 679-798
Currency: OM

Lifetime [year]: 10.

Notes:
An computer in ISET takes charge of the processing for
- Yarrented for 10 years to retain 90X of
the original rated power any computer which have access to the ISEE database
using teletransmission. In this case, only infonnation from
Yarranty [year]: 10.
the keyboard and screen will be transferred by tele-
L.._ _ _ _ _ _ _ _ _ _ _ _ _ Page 3 of 3
transmission (DFO-Datenfernubertragung) so that external
computers work in tenninal operation.
Figure '2: Description of a data sheet
ISET offers external users the following means of access
to ISEE via DFO:

External user = ISET

=
-
1- Telephone Telephone
The system is not only source of infonnation, it also
2- Datex-P
3 - Telephone
= Datex-P
Datex-P (1991)
allows companies and institutions supplying solar energy
~uipment or know-how to draw attention to their pro-
ducts or services. This means that suppliers have the
opportunity to up-date parts of the database themselves. 1 - Telephone = Telephone
Therefore, ISET should not be seen as the sole creator of
infonnation. All companies and institutions involved in The ISEE database is connected to the telephone network
the distrib'ution of components for solar energy systems using two telephone MODEMs (transmission speed 2400
are 'suppliers of in fonnation , . baud and 1200 baud full-duplex) . Access using telephone
MODEM or any suitable acoustic couple result in the
All groups which are interested in installing renewable lowest fIXed costs.
energy systems are considered as 'users of the system',
namely: engineering offices, companies, manufactures, The variable costs (transmission costs) increase with the
public utilities, local authorities, architects and planners, distance from ISET and the duration of the
tradesmen and craftsmen; in fact, organisations of all communication.
kinds with such projects in mind.
Installation costs for the user (assuming telephone
The users of the database are divided into three groups: connection):

440
~ Telephone MODEM (Hayes compatible) / acoustic PC network or by using a single PC with the
couple: corresponding hard- and software.
- 300.- to approx. 2000.- DM
- or as an alternative; renting Modem In cases where Datex-P access to the ISEE database is
being planned, ISET will gladly inform the user about the
If required, ISET is able to offer a communication details of the required installation. At present, an essential
program (OnLAN/PC) free of charge, which is suitable prerequisite is that the user has a PC which is connected
for personal computers (PC and MS-DOS) and access to to the Datex-P network and that a suitable programme for
the ISEE database using the telephone. This program will communicating with the PC network is available. It has
be forwarded with a short description on receipt of a been planned to enlarge this means of access in 1991192.
formatted floppy disk (S ~" or 3 1h" - 2D or HD).

However, it is normally possible to gain access to the

ISEE data base with the help of a standard terminal 3 - Telephone <==<> Datex-P (planned for 1991)
emulation program.
In the future it will be possible to gain access to the ISEE
Communication with the ISEE database via the telephone data base by using the telephone to dial the nearest server
offers the user a high degree of flexibility concerning the computer (DATEX-P20F - NUl - Network User Identifi-
required hard- and software components. The data local cation) in order to get through to the Datex-P connection
end (DLE) of the user can be any computer which is at ISET.
connected to the telephone network, and which has a
standard terminal emulation program at its disposal. Under certain circumstances this variation will work out
The following figure ~ shows the supported cheaper than using either the Datex-P or telephone
communication programs and terminal emulations. MODEM connection separately. Users of this alternative
system will need a Network User Identification number
(NUl) from the 'Deutsche TeleCom' or the local office of
telecommunications.
ISEE - Renewable Energy Information System
Supported communication programs and This "telephone number" will enable the user to get
terminal emulations through to the nearest server computer and thus to the
Datex-P connection of the ISEE database. Concerning the
1. Ampex 230 17. LSI ADM 3A
2. ADDS Regent 18. PC MacTerm costs, the user will pay for the charges arising from the
3. Adds Viewport 19. Prime PSTI00 connection with the server computer and for the Datex-P
4. ANSI Standard 20. Qume (102/108) charges at ISET.
5. Beehive 4/78/DMS 21. Soroc
6. Dasher D410/0460 22. Tektronix
7. DEC VT-52 23. Teleray
8. Dig Hicrosystems 24. Televideo
9. Esprit III 25. Visual V330 (OG)
10. Falco FAME III 26. VT-I00
11. Hazeltine 1500 27. Wyse
12. Hewlett Packard 28. Zenith Z-19 ,. Prospects
13. IBH 3101 29. Zenith Z-29
14. Kaypro 3D. Teletype
15. Kimtron KT-7 PC 31. DnLAN/PC or HAC
16. Link PCTfRH 32. PC x/ASCOH IV
Further development of the ISEE will include the
following aspects:

Figure 3: Available terminal emulations ~ Update of the data base to represent the current
status of all components

~ Increasing performance
2 - Datex-P - Datex P
~ Additional components:
The flXed costs for a Datex-P connection are quite
substantial at the user's end. However the transmission Solar collectors
costs for database search are small because the costs are Measuring instruments for solar radiation and
related to the amount of transmitted data and not to the wind velocity
duration of the communication. Fuel cells
Electrolyzer
Access to the ISEE database via Datex-P can be attained Hydrogen store
by either using the user-friendly connection of a double Pumps

441
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Mathematical models for the construction of a renewable

energy hybrid plant
G.P. Marchetti - M. Piccolo
P.T. Ministry Members, Rome -Italy

ABSTRACT. An investigation on the convenience areas for the utilization of renewable

energy technology lor TLC power supply plants and the SUbsequent development 01 an
optimized sizing project.

1 - Introduction 2) the energies supplied from both generators are known on the
basis of environmental statistical data; hence fuel annual costs
In order to develop renewable energy sources technology in
power supply plants, it is necessary to provide designers with (0,) are calculable;
tools capable 01 recognizing at short notice any areas of conve- 3) maintenance annual cost (Oy) will be conveniently fixed either
nience in its utilization. either as a stand-alone oras a hybrid (with on the basis of a constant rate or by a different function
generator back-up) project. depending on the management structure.
Typical renewable sources considered are solar and wind Fig.l represents a block diagram showing the partial and total
energy, but the results are general where ever energy sources annual costs; they dipend on the variable b.
with random emission are available. Given the total cost arb) = ORO + 0. + ODED + OM + 0,. a
For this purpose it is necessary to know in advance the Cost-Law disconlinous step is applied to this function at points b = bmax
lor this kind of power plant in order to make comparisons with corresponding, along the energy repartition ratio diesel/renewable
the costs for traditional power supply plants, and therefore to x, to the point x '" 0 (renewable generator stand alone complete
be in a pOSition to make a correct decision. with battery).
Here the cost of un-utilized DEG may be deducted.
2 - Economic optimization requirements After this adjustment is made, the function O(b) is now ready to
Block diagram of the plants includes: search forthe minimum valuli!. 1\ is possible to calculate b=b,cor-
- renewable generator!s (RG). responding to the minimum cost and consequentially:
- storage battery (8); and normally used - optimized overall cost O(b,) (money/year).
- stand-by D.E.G. (Diesel Engine Generator). - optimized blocks costs ORO' a •• 0D'o' OM' 0, (money/year).
- optimized dimensional parameters. (battery energy E and
Apart from nominal power and meteorological parameters of pro-
renewable generator power PAQ)
babilistic-statistic origin. the cost also depends on energy ·sha-
ring factors" (one for each-utilized renewable source). 4 - Annual cost versus battery autonomy function
As a result. the total cost can not be determined unequivocably
unless the minimum cost algorithm is utilized. In order that the function Orb) be of an analytic form, it is
This algorithm, on request, must also be able to take into accOl..'ifi important to focalize the attention to:
all external factors (available plant space, maintenance period - system nominal power. (Pu)
etc.). - RG and B unitary costs qRG' qB' (money! power or energy!
year) and DEG cost aCED (money/year)
3 - One renewable generator design guide line - annual interest rate, i
- battery efficiency 1)., but above all to
At first, the authors developed an analysis based on a mathema-
• environmental parameters.
tical model that included only one renewable random source.
Environmental parameters are responsible forthe power random
This model relates environmental statistical factors to the sizing
generation.
parameters and consequentially to the costs 01 each block.
The very significant random variable able to deduce the percen-
However, since the sizing is not univocally defined, the model
tage time (x) at which, the battery equiped renewable generator
takes the battery autonomy as "variable" and executes on it an
subsystem Is not on, alias DEG supplies load, is the quantity low
optimization process finalized to minimum overall cost.
power persistence t. fls distribution law (after a mathematical
In conclusion, the resulting output gives either the optimized
manipulation is implemented with the variable b), in conjunction
sizing parameters or the costs of each block. In the case 01 a
with the average low power percentage E factor and the charge
system with one renewable energy generator (battery assisted)
law, gives the required formula.
and stand-by D.E.G., one energy sharing factor alone (x)defines
Some hypothesis have been set:
both balance and energy contribution from both sources.

The battery autonomy b (and to a certain extent the renewable 1) Only two levels for poweremission lie taken into account:
energy generator sizing) influences this factor. It spreads from low (specified by a suitable idle factor E) and normal (enough
x = 0 correspondent to an equivalent system with renewable for the load)
generator stand-alone (*) (battery capacity will be maximum b = 2) The design is developed on the assumption that the battery
bmax) and x = 1 corrispondent to an equivalent RG + DEG will be kept in a full charge steady-state and that the time
system without battery (b =0). between two following low power persistences is extended
The. ~inimum cost routine analyses each point of this interval. enough to allow the generators to charge the battery.
Optimization criterion is easily set up: once environmental
3) The data concerning variable t and its distribution law are
statistical par.amete~~ are established, each value of battery
not Significantly affected by the average power level chosen
autonomy b Yields a deslgn-poinf' of the plant. As a consequen-
ce (b and therefore x being fixed): hi the project. This hypothesis can also be considered as a
corollary of hypothesis 1.
1) gener~tors ~renewable and diesel engine) sizing and their
S • Mathematical approach to the environmental Influence
costs are lneqUivocally determined; therefore annual loan repay-
ments of renewable generator (OAG1 DEG (°""01 and battery The lollowing symbols are adopted:
(0.) are also calculable provided their life and annual infe'-ast - x '" unavailability of the renewable-battery subsystem (time
rate are known; fraction); .

442
 ~Consl. 4 4-Cons\.·b-+ '-f(b)---+

LEGENDA
INITIAL
COSTS =:) OEG BATIERY RG RG. Renewable GeneratOl
B. Ball9f}'
CEG. Diesel Engine Generator
b. Battery Autonomy (Independent variable)
x. Tlme fraction when CEG suppties the load
INTEnEST i 01 UnavallebliKy of RG + B$Ubsystem
(x ts uf nction of b)
o (b) . Overall anooat cost

w
u.
::::;
n
-'

FUEL = MAINTENANCE
~ f (b)
Const.·X Const . X =I (b)
ANNUAL
COST
a (b)~ a DEG + as + ORG + OF OM
(MINIMUM COST)
FOR b =bl

Fig. 1 HYBRID PLANT DESIGN GUIDE LINE: OPTIMIZATION CRITERION

- t = tow power persistence random variable in continous 1II-0verslzlng factor sip: 1+ _tL.. (T CH =charge time)
approach (days); lleTcfl
- t, ~ low power persistence random variable in discrete
approach (days); Sizing power is deduced multiplying each of them by
- 1O~ observation time (days) the others: ()
- b = battery autonomy (days) PRG~P-1..:K... I+_b_ Pu
1· EX l1e TCH
Discrete approach
7- Data base utilized and Its analysis
After b is setitis possible to calculate lor each t, > b the residue
t,-b and to take into account each oftheirweights n,ln order to In its experimental phase the project used the following dala
arrive at value x; n, is the namber of the evenls I, along the collected for about one year:
observation time 1;,with I days of low power. persistence. I) Solar energy measurements electrically converted by the
In conclusion, the design will be according to the following photovoltaic generator within an experimental renewable
formula: power supply installed in Rome c/o ISPT in 1988;
Time fraction of (RG+ B) Idle 2) Solar radiation data and wind data (intensity and
direction) measured by instruments installed at the same
place and year.
Continous approach
The analysis, developed on this data base, gave the
In case of a continous distribution curve of the low power following:
persistence probability the equation becomes: 1) n. 9 montly histograms of photovoltaic energy output with

1"" 1
28-31 daily energy columns;
2) a set of Ii variables (days) obtained on the previous
x(b) .. ...!l
N
10
l~b (1- b)\I (t)dl .. t:!..To b
(\ - b) P (tjdt _.L 00

tAl b
(1- b) p (tid diagrams by interpolations with horizontal lines drawn at
presumed load power level, in order to recognize any time
where it is set: interval equal to or greater than a day;
3) a table of ti values and its distribution histograms.
· P (t) = probability densily of 1 (days'); 8- (One renewable generator) model configuration
- v (I) = probability density of I> b (days');
·1 = low power persistence variable (days); Data base typology has convinced the authors to select a
· t m = average value of Ihe variable t (days); model with a probability density distribulion curve of
• N = number of events t along observation time To compound type, consisting of:
• Nb= number of events t > b; ·a decaying exponential law shared at t=lo, with probability
· To = observation time (days); area G, able to represent almost all events;
- I = RG low power daysllj,; ·a Dyrac law with probability area 0 that represents a rare
long persistence event t1, whose caracteristics (repetition
and the following identity is verified: N = 1.. period T and persistence t1 ) are not adequately known.
To tm
Obviously G+D=l.
The overall procedure can be so resumed: In this case we may adopt an equivalent T years observation
the designer will adequately fix the distribution diagram olthe low time which approximately corresponds to NTITo events,
power persistence and consequently find x(b). respectively, NTlTo -1 by exponential and 1 by Dyrac
The same x(b) indicates the time fraction when DEG supplies function. Therefore the probability areas assume values:
ttie load.
G =t:U:Il2.. ; 0= ...IlL.
NT NT where: G+D=l
6 - Sizing criteria of the renewable generator
The sizing of the renewable generator originates from 3lactors: 9- Model algorithms
I - Sizing average power PRG" k P u (Pu=load power)
I-EX The mathematical formula:s that can be set up by these
suppositions are reported below:
11- Peak faclor p: it defines the ratio PRdPRG

443
 USER:
ElECTRIC PERFORMANCES
Pu
I SITE: SITE:
GEOPHYSICAL PERFORMANCES DEDICATED
MEASUREMENTS

APIlEG FOR BATTERY

RECHARGE lBAT~~
AP FOR I
RECH.I
ESTIMATES
&1' IF DNA
FINANCE DEG: DIESEL ENGINE GENER. REN.GEN B.BAT. UNAVAILABLE
:,a
MAiNTEN.:, FUEL II DAILY ISOENEAGY MAPS
t-
~
:Q
,It
i ~
:
, II MINIMUM ENERG. MONTH
~~ n
o ,,~::r~ 1 ~
i~i ~ ffi i ~ " ~ ~ w!!.~
~ ~ :!2 8; : ~ 8 8 ~~
1
U.JC-
kWhlm2!days

~ ~~~~~BBBB ~ I~T7.I~
~- - -'--1
II T £ INPUT'
GH (cis) (d.) DATA:
+----,

(5),,.-L-lL..L.,· -
RG ANNUAL
6)
I
,-L-------L--,
P ESTIMATE
ICCITT PROCEOURE
ONLY FOIl PHOTOVOL EN)
I
(0)
ANNUAL COST!
·h
J
COST! kWh
kW
FIRST PROCESSING (1)
!------------------------------ --• ----1
Pu ~ x[b). ~ s ~ b[x) P 10 ;~~~~~;i~~~~~!

(3)
DEG
LIFE
TIME
BETW.
(YEARS) IHTERV. (4
b (2)
I,
,L DIMENSIONAL PARAMETERS IN FUNCTION OF
____________________ _ I
Lp-R~G~~===Q~81--~~E~-r----t--------~
l(x) &\(x)

o±~±_~.~.r-~±-~----.~~~. + +
DEG MAINT.II FUEL COST_IIRG LOAN REPAYM. B LOAN REPAYM ~----------
LOAN REPAYM.
ODEG(7.1)
C8~T II OF = X • CF II ORG qRG . PRG OB = qB EB (7) = ANNUALCOST:
~-----------

L-_-,+ ! +.-__+-_-.J
I SUM:Q (b)
TOTAL ANNUAL COST
(
8)

OPTIMIZATION
aQ9b =0 ~ bl (9)
,t---. --. --. ---------' -----' ---------------. ---------------------------------------. -----.-
b

I C ( b) bl b) PRG ( I I
EB ( b )11 b ) ~-~-~~U_T_~':'0~ _D_A!~ j
X (

Fig. 2 FLOW DIAGRAM OF DESIGN CALCULATION

444
 p calculalion lor pholovollaic energy see CCITT
where: Gas 4 par 3.5.3.1
P=~
(1 +A) GH
tm = fo~ t . p (t) dt = t average ~. I•• x = x(b) ; b. x·'(x) calculalions

t, = low persistence (rare event)

T = repetition period of I,
/)(1) = Dyrac function
1(t) = unitary step fuction
!:l!(8'
I",
: ..!m.)+!L:.!9.
TI T
lor b S I"
13 - I T-f -tl average persislence of the x(bl' [ (I, ·I,)IT
- m T.f-t m (1 +In ~~) exponential events;
o
for lo~ b S I,
lor t, 5b<oo

I. = 13 . In (f.Tltm); Ihreshold separating for I> 10 a surface

corresponding to the probabifity lrr on Ihe gamma curve.
b(x) _ x' (x) • ~ In ( T ) ~ lorbs t.
xT+P+lo·I, In

Taking inlo account Ihe general expression in section 5 we obtain Dimensional parameters calculation
the finallormula of x (b);

113 _l! 1m)

- (e~-- 1-1
f
1 0
+-- for b Sio
1m T-f T DEG IiI. LoEG = (years)

x (b) = ~ for t. ~ b < I,

InleIVenlion period 61 = ¥
T
o lor b ~ I,
DEG annual cost and RG. B unitary annual cosl

K'G (Block 5,
The costs are assumed. following the laws indicated below;
q=~ A whereA= K, (Block 6)
(1 + ill. ,
C a proportional to the nomin. battery storage .El.JlQ~- E: ~ = Ke CO<G (Sub block 7.1 i'
G Blo~k7l
GRa renew. peak pow. P 00: pO = KRG except...!.:!) Components-annual costs
• R'
COEG fixed within the requested range by Ihe construction sian·
dardization.

The subsequent annual cost originate from the annual loan re-
payment lormula. taking into account the principal vafue A.
annual interest rate i (in percent) and component life L (year$;
.
q = 1(1 + 1/100)
. l
A
(Block 9 J Search lor minimum and consequenl b, ~~ - 0 :>b,

(1 + il100)l - 1

Regards lile L it is set: FIG. 3 MATHEMATICAL CALCULATION OF BLOCKS

- lixed values for battery L. and renewable generator L RG ; T090FFIG2
- value inversly proportional to x lor DEG being a subject-wear
component.
The structure of the model originates directly from the "Flows
diagram 01 design calculation" (fig. 2)
Data comes Irom the data blocks: FIG. 4 UNITARY COSTS AGAINST BATTERY AUTONOMY
- load electric performance;
- environmental performance: geophysical characteristics and c{O)
dedicated.measurements
fhese l.ald values, together with other data (p etc.) to be """"'" I
Coeo •
. 12%
$ 17657
lo
Pu
- 8y"",s T - Iyear
elaborated or estimated, including battery recharge criteria, ge- II
- 0.2 kw - 0.05
Y - O.5year
nerale actual values to be used as data input for subsequent cal- 6Y - 500/8760 year
TCHIb .2
Latilud - 41 "40' N
'I.. - qso
culations 01 blocks 1 to 9. The mathematical calculations in C, • $ 1.000 GH .2.2 kwh /m> .'
blocks 1 to 9 are shown in fig. 3,. c" - SO.7111 ·SlaO p - 7.2
Koo • $ 10.820 IKW I -0.493 .1

Instead of Q(b) being referred to as an annual value it is used as

a proportional quantity refered to as a kWh:
L..:l
Ko
- 20 years
- $ 317IKWh
1m
It
- 3.53 days
- 46cbys ..
"

.....9.~= C (b)
P u 6760h
where: [Pu) = [kW) .,
."
indicating the cost/kWhl(fig. 4) . This is preferred. since it permits .1

comparison between the energy cost 01 this kind 01 plant with the
price fixed by the energy suppliers.
--- -----
.. .. ... IS• It. >S •
". :os. ... ,s. ' I . b (DAYS)

445
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOVOLTAlC, UTILITIES, AND THE ENVIRONMENT IN THE 21ST CENTURY

Ghazi Darkazalli
Spire Corporation
One Patriots Park
Bedford, MA 01730-2396 USA

ABSTRACT. To insure a healthy and prosperous world for us and future generations, the world cannot rely on fossil
fuels as its only source of energy. Volatile oil prices, energy crises, and the deteriorating environment make it
necessary that mankind look for cleaner and more stable fuels that we can rely on. Photovoltaics represent a vital
source of energy that is reliable, economic, and safe to the environment. Photovoltaic systems represent an attractive
option to many utilities in developing and developed countries.

1. INfRODUCTION effects are felt worldwide.

Global climate changes are due to the presence in the air
Utilities, industry, commercial, and residential conswners of the "greenhouse" gases released from fossil fuel burning.
have realized the value of using renewable energies such as CO, will account for half of the global warming problem over
hydraulic power, photovoltaics, geothennal, wind, and others to the next 60 years. Other greenhouse gases such as methane,
meet their energy demand. Governments of developing coun- nitrous oxide, and chlorofluorocarbons will account for the other
tries now know that the cost to bring electricity to the rural half of this greenhouse effect.
world population will represent over 50% of many countries Nuclear power proponents have pointed to nuclear power
export earnings. Compared with an average cost of $7 to $15 as the solution to the greenhouse effect. However, nuclear
per foot to extend the utility grid, photovoltaics are more attrac- power's contribution in the United States is limited because no
tive for small rural power systems than expanding the grid. plants are scheduled for construction in the near future. The
In the last decade, we have witnessed the rapid develop- pace of nuclear power expansion has slowed almost to a halt in
ment of photovoltaics as a source of energy. A mature industry many countries during the last 10 years. Ail existing reactors
has developed with a healthy growth rate. However, many are scheduled to be retired within the next 40 years, and unless
obstacles and barriers still exist. Barriers such as lack of fman- fundatnental problems are solved, it seems likely that most will
cing, strong commitment of local and nation governments, and not be replaced. This decline in nuclear power use is due to
false costs of energy produced shonld be addressed. Photovol- high costs, inadequate safety measures, atld absence of per-
taics and other clean and abundant sources of energy should be manent nuclear waste storage facilities.
compared equally with other fuels and the real cost associated Photovoltaics is the direct conversion of sunlight into
with these fuels, including their effect on the environment. electricity. A photovoltaic cell is the basic unit used for the
The developing countries energy needs are becoming conversion. The cell is a thin layer of semiconductor material,
increasingly important in the global energy picture. Their share usually silicon. No carbon dioxide is emitted during the conver-
of the .total is expected to rise from over 23% of global com- sion process. The principle attributes of photovoltaics are
mercial energy now to 40% by 2020. Most of this increase is shown to be a socially beneficial technology when compared to
due to added generating capacity needed for economic growth other options. These attributes exhibit both the promise of
and improved standards ofliving. This increase, combined with photovoltaics and its barriers. The fundatnental promise of PV
the developing countries' already large consumption of fossil lies in its characteristics, including environmental benignity,
fuels, will result in environmental problems such as acid rain modularity, universal availability, very low operating costs, and
and global wanning. Serious considerations to alternative fuels over a 30-year life expectancy. To sharply contrast the environ-
are necessary' if we want to save our environment. As the mental comparison, Table I represents emissions of conventional
United States and the world strive to improve public health and and renewable energy sources.
preserve the environment, renewable energy sources will play
a growing role. The U.S. alone can cut a minimum of five 3.0 PV FOR UTILIDES
percent and a maximum of 19% (of the projected annual total
of 6.4 billion tons) of CO, by the year 2000(1) The United States is the world leader in the number of
homes using photovoltaics as their source of electricity. Over
2.0 ENVIRONMENTAL EFFECTS 15,000 homes, mostly in remote locations, have PV systems.
These systems proved economical when compared with the cost
The increase in energy use is needed to raise living of extension of the utility grid. The U.S. Department of Energy
standards and to meet demands due to increased world popul- estimates the new generating capacity needed by electric utilities
ation. This increase should be coupled with certain measures between now and the year 2000 to be approximately 111 GW.
that reduce the emission of greenhouse gases such as switching The major portion of this capacity, 73 GW, has not yet been an-
to. non-fossil fuels, increasing the efficiency of energy and nounced or ordered. I')
materials use, reversing tropical deforestation, and implementing The global primary energy demand is estimated to
sustainable use offorest and agricnltural resources. Such meas- double by the year 2025. To meet this tremendous growth
ures are easily attained and should be seriously considered by requirement, many developing countries will tum to renewable
housing, national, and global policies. Acid rain comes mostly energy sources. This will be in the fonn of local stand-alone
from the emission of sulphur dioxides from fossil fuel burning, power plants. Photovoltaics represent a cost-effective, environ-
patticularly coal. The datnage done by these emissions has mentally safe system. Using photovoltaics will elinlinate the
become blatantly clear with the death of forests and lakes atld need for elaborat~ central power plants and extensive distribu-
the increased corrosion of metals and stone structures. These tion networks.

446
Table I. Pollutants Tons Per One Quad: the Total Fuel Cycle. Table II. PV Prices and Resulting Power Segments.*

Source CO2 NO. SO. Price ($/Wp) 6 4 3 2 1

Volume (MWp) 20 40 100 500 2000
Conventional 4,179,100 11,795 11,735 SEGMENTS (%)
Coal
Fluidized Bed 4.175,150 6.126 1,724 Remote Power 75 40 20 5 1
Coal Rural Power 3 10 30 35 35
IG CC Coal 3,254,800 991 1,327 Suburban Power - - 2 40 50
Nuclear 33,931 134 115 Motive Power 2 25 28 15 9
PV 23,266 32 91 Consumer Package 20 25 20 5 5
Biomass 0 2,425 608 Goods

Geothennal 224,360 0 0 Source: ISEIR

Wind 29,230 0 0
Solar Thennal 14,220 0 0 The potential of photovoltaics in energy productions,
emissions, land and water use (compared to conventional and
Hydropower 25,873 0 0 next-generation coal and nuclear plants) is very favorable be-
cause of the following:(3)

More than 25 megawatts of photovoltaic systems are Photovoltaic power plants produce the
interconnected in the United States with the utility grid. Ap- least emissions of any conventional
proximately 40 U.s. utilities have had experience with photovol- power plant, including the operation,
taic generation. This experience may be classified in three construction, and extraction (Table Ill).
categories; l) experimental test facilities, 2) stand-alone power
plants, and 3) grid-connected photovoltaic powered plants. The Table Ill. Emissions Di.splacement Using Photovoltaics
utility interactive systems range in size from small residential (Metric Tons/GWh).*
systems such as those used by New England Electric system
which are 2.2 kW roof mounted, to large central stations such
as the Pacific Gas and Electric Company Carissa Plains' 6.5 FUEL CO2 S02
MW generator. Notable perfonnance improvements in photo-
Coal (AFBC) 960 10
voltaic-powered plants have occurred. These improvements
include 1) higher availability sunlight hours, 2) efficiencies, and Oil (Steam Turbine) 700 5
3) lower operation and maintenance costs.
Gas (Steam Turbine) 480 Neg
Nuclear 8 N/A
4.0 PHOTOVOLTAlC CONTRIBUTION
Source: US DOE
The photovoltaic contribution to world energy demands
is estimated at 10 to 30 percent by the year 2050. This contrib- * 1 GWh is the annual output of photovoltaics oper-
ution will result from the I) availability of inexpensive cells, ating 4.5 hours per day.
2) economical time-shifting of energy available through storage,
or 3) significant time zone transmission. Many third-world
countries will elect to use photovoltaic teclmology because it
can be manufactured locally. It does not require an extensive Photovoltaic power plants utilize the
energy infrastructure, and photovoltaics will reduce the coun- same land area needed by conventional
try's current dependence on expensive electricity for its develop- energy plants when operation, construc-
ment. tion, and extraction and transportation are
Improved photovoltaic cell efficiencies, combined with accounted for.
larger manufacturing plants currently under construction, are
expected to produce cells that can generate electricity at ap- Photovoltaic power plants use at least
proximately one dollar per peak watt. Even with today's tech- one-third less water than any conven-
nologies, economics of scale employing larger manufacturing tional energy power plant.
facilities could bring the cost of photovoltaic electricity down
to 12-15c/kWh, which is comparable to the current cost of In addition to the environmental advantages of photo-
generati.ng electricity. This price reduction will result in an voltaic power plants, utilities' objectives can be easily met by
increase in the market size to approximately 2000 MWp per using photovoltaics for load shaping and demand-side alter-
year. The major area of this increase will be in the rural and natives such as peak clipping, valley filling, load shifting, and
suburban power sectors which will represent 35% and 50% of others.
the total market (Table II).

447
5.0 CONCLUSION

Over 40 utilities in the United States, Japan, German,

Italy, Greece, and many other countries around the world have
had successful experience with small- and large-scale photovol-
taic power systems. The photovoltaic industry is mature and
growing at approximately 30% per year.
Photovoltaic power systems can be utilized for small
electric power plants. These plants are a natural match to
localize power distribution systems. Photovoltaic systems are
modular, which allows for the utilities to add capacity easily,
as needed. Photovoitaics may be used for residential, village
electrification, and central power stations.
Photovoltaic systems are environmentally safe. They
produce the least emissions of any conventional power plant,
require the same land area as conventional energy plants, and
use one-third less water than any conventional energy power
plant.
The value of photovoltaic electricity from an energy and
environment point-of-view is high enough to justify widespread
use of this technology.

6.0 REFERENCES

(1) N. Rader,"Power Surge, The Status and Near-Tenn

Potential of Renewable Energy Technologies," Public
Citizen, Washington, D.C. (May 1989).

(2) G. Darkazalli, "Future Effect and Contribution ofPhoto-

voltaic Electricity on Utilities and the Envirorunent,"
Proceedings of the Energy and the Environment in the
21st Century, Boston, MA, 1990.

(3) S. Sklar, "Solar Energy Regulatory and Policy Issues in

the United States: A 1989 Photovoltaic Review," in
Proceedings of Ninth E.C. Photovoltaic Solar Energy
Conference, Freiburg, 1989.

448
 lam EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CONTROL CONSIDERATIONS FOR PHOTOVOLTAIC POWERED WATER

PUMPING SYSTEMS

Platon Baltas
Center for Renewable Energy Sources
6 Frati St., Fousia, GR-194 00, Koropi
GREECE
fax: +30 1 662-6462

Paul E. Russell
Arizona State University
Tempe, AZ 85287-6506
USA

ABSTRACT

Overall efficiency of pv water pumping systems depends on system topology, sub-

component efficiency, the proper matching of these sub-components and the system
control configuration. The benefit of optimal control methods for pv water pumping
systems is explored in this paper. Two pumping experiments are described here. The
first is located in Arizona-USA and the second in Attica-Greece. Analytic results are
compared to experimental data. Issues related to, computer simulation of pv water
pumping systems, pumping hardware and borehole modeling and parameter identifi-
cation, optimal system topologies for specific environments, analytic techniques for pv
pumping system control optimization, pumping system hardware seleCtion and
matching, are addressed.

1. INTRODUCTION proposed for systems with both high and low produc-
tivity boreholes.
Overall efficiency of photovoItaic powered water
pumping systems depends in a major way on the sys-
tem's control configuration as much as on the efficiency 2. EXPERIMENTAL SETUP
of the solar converters and other component parts.
More water can be delivered from a borehole by care- 2.1 System Description
fully matching the non-ideal water source and the non- Two pumping experiments are described here.
ideal pump and energy source, perhaps including elec- The first is located in Arizona-USA and the other in
trical energy storage devices. Attica-Greece.
The best match of system components is likely to The first pumping system has been installed in an
be the one that tends to insure the least drawdown of 8" diameter, 45 meter deep borehole drilled in weath-
the water level in the borehole, with the motor and ered granite. Water static level is at 6 meters from the
pump operation at or near the load that can be carried at ground leveL No significant seasonal static water level
maximum efficiency. variations were observed.
In some situations, variation from the ideal is
minimal. Though the available sun energy will change 1".IlI>~::i"7",.--_G",r:..:o~und Leve I
through the (lay, there are applications where the water
source is virtually infinite, the particular pump effi-
ciency may be almost independent of the motor drive
speed, so that, as sunlight varies, the changing motor Static Level
efficiency is the only significant factor to change the sys-
tem efficiency. But, better overall efficiency might be
attained, and more water might be pumped in a typical
day, if the system could include an energy storage and
averaging device such as a battery, so as to cause the
motor operation at higher average efficiency. Sun
tracking by the array might approximate the same.
With many pumping systems, the efficiency of the
motor-pump subsystem may change markedly with BorehOle Casing
sunlight and with lift (head). If the measure of per-
formance is to be to pump the most water on a daily or
weekly basis with the minimum investment, careful
matching of array-sunlight capability with consid- Fig. 1. Borehole.
eration of techniques for load av~raging can be useful.
In this paper, a number of examples are cited to il- The second system has been installed in a 20 cm
lustrate the potential gains of averaging loads and im- diameter, 22 meter deep borehole, drilled in an un-
proving system output and efficiency. Experimental known consisten.ce aquifer. Static level is at 11.5 meters
data from two pumping sites are cited. Battery use is from the ground level. No significant seasonal static

449
water level variations have been observed so far. A 1.55 Observations at the second site differ significantly
meter diameter, 21 meter deep, dug well is located 9 to that at the first one. As shown if Fig. 2., drawdown is
meters from the borehole. It is presumed that water relatively small at an almost constant discharge rate of
stored in this well affects the drawdown characteristics 2.8 m 3 /h. Yet, borehole water level seems to be slow to
of the borehole. A typical borehole illustration is recover to static as pumping stops. To explain this be-
shown in Fig. 1. haviour, we consider the influence of the neighbouring
Submersible centrifugal pumps have been in- dug well. Since this has 'a diameter of 1.55 meters, it is
stalled in both sites. A DCI AC three phase inverter implied that 1.88 m 3 of water have to be pumped for
drives a three phase induction motor in both cases. PV every meter of drawdown assuming that no water is
array peak power is of the order of 700 Wl' for both sites. drawn from the surrounding aquifer. Its close prox-
Installation of a battery bank and an additional 300 Wp imity to the borehole (9 meters) affects drawdown char-
pv array at the second site is in progress. acteristics and complicates parameter identification.
Therefore it is not conclusive from the available
2.2 Data Logging data that aquifer properties differ as much, contrary to
Purpose of both experiments is to what may be presumed after a first glance at drawdown
- Perform long term reliability tests for pumping data. Better conclusions may be reached after borehole
hardware, modelling and "subtraction" of the dug well effect. On-
- Evaluate pumping hardware efficiency, going testing is expected to provide the necessary data
- Comprehend borehole dynamic behaviour, for accurate modelling. Water level data from the dug
- Advance the development of new pv water well are to be collected, pending permission 'from the
pumping optimization techniques. owner.
For this purpose, required measurements include
- DC power to pump hardware, 3.2 System Efficiency Data
- Borehole drawdown, A substantial number of data points .have to be
- Water flow rate. collected from an installed system, to determine its effi-
Radiation and temperature measurements are ciency over a wide range of operating conditions. In ad-
performed as well. Data are collected every 5 min. dition, installed systems may never happen to operate
Frequency, power factor and current mea- at certain conditions, therefore, an incomplete "picture"
surements at the ac side of the system have been for hardware performance may result. For this, it is bet-
planned for the second site. It is expected that this type ter if pv pumping systems are tested in a simulated
of data will provide an indirect way of drawdown and borehole, using a power supply that simulates pv array
flow rate estimation given specific motor-pump behaviour. However in our case, schedule and budget
parameters. limitation did not permit this type of testing.
At the first site, the large head variations provided
data over a wide range of operating conditions and that
3. EXPERIMENTAL DATA allows for an almost complete picture of system per-
formance. Combined efficiency (array DC power to hy-
3.1 Drawdown Data draulic power) reached 33% for this system (Fig;. 3).
Discharge and draw down data from both sites However, most of the time the system operated at effi-
have been collected. At the first site we observed a ciencies bellow or around 20%.
dramatic drop of water level as discharge rates reached
the value of 2 m3/h. Water level drops 30 meters below 50~----~~-----------~
the static during the first hour of pumping. The level
recovers approximately 20 meters, S hours after pump-
ing ceases. A water level close to static is attained 10
hou!"s later. That suggests a low transmissivity, low 40
storativity aquifer. <II
-0 .6 -,------r----,-----r-----, \..
Q}
Qj
E
~
.s
rn
8 -0.8 ~~----4_------t_------r-----~ o
....C1)
<l
W
E. I

-1.0
~0
'tl
~ 0
0:1
~ 0 10 20 30 50
Q
-1.2
Q (in lil ers Imin)

Fig. 3. Head vs. Discharge for pumping system at the

o 30 60 90 120 first site.
Time (minutes)
System efficiency data from the second site are not
Fig. 2. Attica-Greece Drawdown data. yet enough and over a wide range of heads to allow for
a similar picture. Maximum combined efficiency for

450
the existing data is around 29%. It is expected to reach pumping hardware and static level. A check valve is
higher values as operating conditions move closer to present in the system (Le. u is non-negative).
the maximum efficiency point. The input power P to the system may be un-
controllable if it is the dc power from the pv array, or
controllable, if a battery is present.
4. MODELS At a first glance it may seem to be inconvenient to
model a system such way. This is true if our scope is
4.1 Borehole just to predict total volume of water to be pumped for a
Uncomplicated mathematical treatment sets the specific input power profile. As a matter of fact, this
need for linear borehole models. However this is not model is less accurate than simulation because approx-
the best way to describe a distributed parameter system imate linear models of the real system are used. How-
like the borehole-aquifer system. Empirical non-linear ever, our scope is not to predict performance but rather
models (1) have been tried for the first site with excel- to solve for the optimal input power profile function.
lent results. In this analysis a linear model of the fol-
lowing form is used.
5. ANALYSIS

For a given input power function pet), the pump-

d{t) is the drawdown and u{t) is the water discharge. ing system response u{t) can be evaluated. The integral
Parameters may be calculated for every particular of pet) over a period Td (say 24 hours) is the total energy
borehole, given the drawdown data. As mentioned in to the system. The integral of u(t) over the same time
previous work (2), this discrete time model may be period is the total volume of pumped water.
transformed to a continuous time one using a bilinear For a system that does not include a storage ele-
transformation. A state space representation for the ment, not much can be done to change the .power pro-
borehole dynamic behaviour may be obtained. file pet). Different power profiles result if pv array ori-
Linear borehole models performed satisfactory for entation or tilt is changed. Different system parameters
data from both sites. Work is under way to develop are implied for different types of pumps. Pumping sys-
quick and inexpensive borehole dynamic behaviour tem designers may try a variety of array sizes, tracking
testing procedures. modes or pumping hardware to achieve a design that
yields maximum water volume.
4.2 Pump Hardware So far, this optimization process seems very sim-
As expected, non-linear models are more accurate ilar to simulation. In fact, simulation may be easier to
in describing the inverter - induction motor - centrifu- handle and more accurate. Yet, this formulation is
gal pump system. However, these models can be preferred for it is more suitable for mathematical analy-
"linearized" for the area of expected points of operation sis.
if data show that system is far from being linear. Math- Let us assume that a storage device (battery) is pre-
ematical formulation simplicity suggests that a model sent in the system and no energy penalty is paid for en-
of the form ergy storage. Let the average power flow to the system
be constant Po. The power profile pet) may be any func-
H = v{P} - a'u tion satisfying the constraint

is more appropriate. H is the total dynamic head, u is

the flow rate and v{P) is a polynomial of the DC input
power P to the system. Fortunately enough the pump-
ing system at the first site is described well by a model of
this type. The problem is to find a pet) function that satisfies
the above constraint and yields maximum volume of
4.3 System water.
Combination of the above mentioned models for
borehole and system hardware, results to a system
model that is suitable for mathematical treatment. A
block diagram for this system has been mentioned in
previous work and is depicted here as well (Fig. 4).
l.
Td 0
Td
lu{t)·d t = maximum

This problem has been solved for three different

o classes of functions.
1. When no constraints on the final state of the
system are imposed.
2. When the final state of the system has to be the
same to the initial state (periodicity), and the
function pet) is continuous.
3. When the final state is same to the initial one
but function pet) is an on-off function.
Detailed solution can be found in (3). A short de-
Fig. 4. System block diagram. scription of the results is presented here.
For the first case, the optimum function looks as
A, B, C and D parameters (matrices) depend on borehole in Fig. 5. Parameters for this solution have been ob-
characteristics. Feedback constant depends on pumping tained from experimental results at the first site.
hardware characteristics. Function v(P) depends on

451
 1000 power flow to the system and if there exist control

/
schemes that yield maximum volume of water. Simu-
lation is not a sufficient tool for this type of analysis.
800 The analytic approach followed here is preferable.

t/
The conclusion is that controlled power flow to
the pumping system helps improve overall efficiency
.......
~
In

600
in some cases .
01
~ -~ 6.3 Looking Forward
'"'
QI This optimal design may be proven impractical if
~
0
400 theory is to complicated. Yet, analysis helps in problem
Q.,
comprehension and derivation of simplified design
rules. Installers need simple design guide-lines or at
200 most an easy to use integrated computer tool. Even in
this case, it may be hard for installers to evaluate opti-
mal duty cycles, battery sizes etc. Arising questions and
o suggestions on future work needed to overcome these
o 120
240 360 480 600 720 problems are presented here.
Time (minutes) Better understanding of borehole behaviour and
Fig. 5. Optimal power flow. development of mathematical models for it is required.
Experimental data from our test sites are helpful but
For the second case, the optimum power profile is data from other pv pumping experiments are needed
a constant. too.
For the third case, the solution depends on duty Development of simplified and inexpensive bore-
cycle of the on-off operation. Results have been re- hole testing procedures may be necessary.
ported earlier (2). Development of design tools (software) that are
easy for installers to use, and simplified quick design
procedures (array size, battery size, pump type etc.)
6. DISCUSSION given site characteristics.
Finally, we may have to develop hardware fitted
6.1 Motivation for this type of application (e.g. inverter control units
Water pumping has long been considered a suit- able to identify environmental characteristics and adapt
able application for photovoltaics. their operation). It is noticed that storage elements al-
High module cost implies that most application lows for flexibility in design through control (matching
array power be in the order of one to few kilowatts. of pump - borehole), therefore, hardware may not need
Engineers are asked to install a photovoltaic system in a to be designed to fit a wide range of operating condi-
borehole or. a dug well, usually of low capacity. If the tions.
total dynamic head remains almost constant during
pumping, the design task is relatively easy. One has to
assume an average head and select the pump of maxi- REFERENCES
mum efficiency at this head and array power level. The
nature of photovoltaic power necessitates the use of (1) P.Baltas and P.E.Russell, "The Dynamics of PV
pumping hardware that performs efficiently over a Water Pumping," Proc. of the 19th mEE PV Spe-
wide range of input power levels. Introduction of vari- cialists Conference, New Orleans 1987, pp. 1031-
able frequency inverters promotes the use of centrifugal 1035.
pumps. (2) P.Baltas, P.E.Russell and C.I.Byrnes, "Analytical
However, design problems may arise if head is Techniques for the Optimization of PV Water
very sensitive to water discharge. In this case, a more Pumping Systems," Proc. of The 8th European
careful design procedure is required if maximum vol- Photovoltaic Solar Energy Conference. Florence,
ume of water is desired. Valuable photovoltaic energy Italy 1988, pp. 401-405.
may be wasted if pumps operate inefficiently or if water (3) P.Baltas, Optimization in Control for Photovoltaic
level inside borehole drops below acceptable limits. Water Pumping Systems. Ph.D. Dissertation, Ari-
Experimental data from our test site helped us zona State University, December 1987.
comprehend the problem. However, we have no good (4) UNESCO, Ground Water in Hard Rocks. Studies
estimate on the number of applications where this and Reports in Hydrology 33, Project 8.6 of the In-
problem may appear. Data on hard rock wells, reported ternational Hydrological Programme, Prepared by
in (4), have been a motivation for our work as well. the Project Panel, Ingemal Larsson: Chairman,
1984.
6.2 Methodology
Analysis of the data from our experimental sites
convinced us that simulation is a useful tool for pump-
ing system analysis and design. Batteries were included
in simulations for optimization purposes. Since they
are used for temporary storage, their size may be com-
paratively small and maintenance requirements low.
They are introduced here for the sole purpose of system
subcomponent matching.
The arising questions are how to manage the

452
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-I2APRIL 1991 LISBON. PORTUGAL

SPATIAL AND TEMPORAL CHARACIERISTICS OF SHORT TERM FLUCTUATIONS IN

SOlAR RADIATION FOR PV-PlANT APPUCATIONS

Hans Georg Beyer, Burchard Decker(*), Joachim Luther, Robert Steinberger-Willms

Renewable Energy Group, Dept. of Physics, Universitat Oldenburg, P.O.Box 2503, D-2900 Oldenburg,
F.R.G_
(*) Institut flir Solarenergieforschung (ISFH), Sokelantstr. 5, D-3000 Hannover, F.R.G.

ABSTRACf
The characteristics of short time fluctuations of solar radiation play an important role in the assessment of
the interactions of PV-generators with electrical supply systems. PV capacity in the range of 100 kW will
naturally be dispersed over some area thus making an investigation into the spatial characteristics of the
radiation field necessary.
In order to analyse these characteristics on a scale of some 100 m we have recorded 10 sec averages of
solar radiation at five measurement sites. Sequences of the data are used here to demonstrate some me-
thods for the analysis of the spatial and temporal structure of radiation data in the sub-hourly time
domain. Results from the extrapolation of a set of single point measurements to the radiation field over a
two-dimensional PV-generator array are presented.

1. INTRODUCTION 800m

In an assessment of the interactions between PV-generators

tN •2 •
.
I
and electrical supply systems detailed knowledge of the
characteristics of especially short time fluctuations in PV-
power output is necessary. As far as the coupling to a grid E l
;
of PV-capacity in tbe range of some 100 kW is considered §
••
~

(e.g. a PV-power plant or a conglomeration of PV-systems

mounted on roofs in a residential area) the PV-generation
capacity is naturally dispersed over a certain area. Since
solar radiation is not spatially homogenous when high
•
~
fluctuations caused by passing cloud shadows occur the time ~

pattern of the power output of a given PV-array not only

depends on the temporal fluctuation characteristics but also .;
on the spatial characteristics of insolation, i.e. the radiation ~
H',
field. "
A description of the output of spatially dispersed PV- Fig. 1: Measurement setup and geographical location. Five
generation capacity was published amongst others by Jewell measuring stations are placed at the indicated points.
and coauthors /1/, Garrett and Jeter /2/ and Kern and
coauthors /3/. A main problem of this research lies in the uo ~------------------~-----------------,
lack of sufficiently time and space resolved data
necessitating the use of synthetic radiation data based on
rrl
IIlO
r
cloud models. ill»
I .,."'"
1\
In order to improve the available data base and to evaluate
methods to establish two-dimensional radiation fields we
have performed a measuring campaign during 10 days in
November and December 1990 in an area of approx. 600 m
/
.
j
by 600 m in North-West Germany. This campaign is part of
a project aiming at establishing a meteorological data set
0
o •
....... " .
....... "
for penetration studies on various temporal and spatial
scales. Fig. 2: Time series of solar radiation on a 500 south inclined
2. DESCRIPTION OF THE DATA SET plane measured at site no.1 on Nov. 30th and Dec. 4th,
1990. The time resolution in these graphs is one minute.
The measurement site is set in an open plane to the north
of Oldenburg in the north-west region of the Federal "':
Republic of Germany about 60 km from the North Sea 0,--------------------------,
coast. Five measurement masts were placed in an
approximate quarter-circle (fig. 1) at distances of 200 to 300 ;~
m. The solar radiation on a horizontal and a 500 south
inclined plane is recorded as 10 sec averages. ~~
Fig.2 shows the time series of solar radiation on the inclined
plane recorded at one site on Nov. 30th and Dec. 4th, 1990. 0.0 0.2 0.4 0.6 0.8 1.0
During these two days high fluctuations in solar radiation kt
occur. An analysis of the clearness index for three hours Fig. 3: Frequency distribution of the instantaneous
taken from these sequences shows the dominance of two clearness index (10 sec averages) from three time
states of atmospheric transmittance. They may be identified sequences of 1 hour duration. These sequences are
as 'cloud' and 'no-cloud' situations (see Suehrcke and characterised by high fluctuations and average clearness
McCormick /4/). index values in the range of 0.5 to 0.6.

453
It has to be kept in mind in the following analysis of the In ord_er to obtain the characteristics of the spatial structure
radiation fluctuations that the sensors used (Kipp&Zonen of the radiation field the cross-correlation structure of the
CMll) have distinct response characteristics. Due to their time series' recorded at the measurement sites is inspected.
finite response time gradients in solar radiation are Assuming that the larger fluctuations in solar radiation are
smoothed in the recorded data. A laboratory test of the caused by clouds drifting over the measurement setup
instruments shows that the response to an instantaneous information on size and velocity vector of the cloud
step change in radiation (0 to 700 W1m") results in a value shadows may be gained from the cross-correlation
of about 60% of the step height for the first and of about functions. If a cloud field consisted of rigid structures larger
95% for the second 10 sec average after the step change has than the size of the measurement setup the cross-
been effected. correlation functions should reach maximum values close to
unity. On the other hand maximum values of cross-
Despite of these shortcomings the data sequence may be correlation noticeably smaller than '1' imply a cloud field
used to demonstrate some methods for the analysis of the whose elements are smaller than the measurement setup
spatial and temporal structure of insolation in the sub- size or which is not totally rigid. The' drift of a rigid cloud
hourly time domain and the extrapolation of a set of single pattern will in general cause a systematic lag in time
point measurements to a radiation field on a two- between the individual time series' recorded. This is
dimensional PV-generator array. reflected in the corresponding time lag at which the cross-
3. METHODS OF ANALYSIS correlation functions take their maximum values. Given the
time lag of all pairs of stations the drift vector of the cloud
Sudden changes in PV-power output are a point of special
interest in the analysis of the interaction of PV-generated shadows may be derived under some simplifying
assumptions (e.g. that the cloud layer is purely two-
power with electrical networks (see for example Kern and
dimensional).
Coauthors 13/). In order to assess fluctuations in power
output, a description of the fluctuation structure of the The relation of the distance vector between two sites, the
radiation field has to be established. cloud drift vector and the time lag is given by:
An overall characterisation of the short time fluctuations of ~/ I!I * 'tij = Ivl 6t ij (2)
solar radiation is given by the frequency distribution of with : v : cloud shadow drift vector
gradients defined by: .
....rij : dlstance vector between two
..1G/n4t = [G(ti + nAt) - G(ti)]/(nAt) sites
where G(t): solar radiation (1) 6tij: respective inter-site time lag
.A t: time resolution of the
A least square method may now be applied to find a
measurements (10 sec) v
velocity vector that best fits the time lags cS\j obtained
n: a given integer from the cross-correlation functions.
for various time increments n.4t. This analysis only renders The drift vector may be used to extrapolate the set of single
indirect information on the structure of individual 'ramp site measurements to the radiation field observed bJ a two-
events' constituting the most severe changes in radiation dimensional PV-generator array. The method we describe
and power output. Therefore a special scheme is applied for here is limited to arrays whose size perpendicular to the
the identification and characterisation of these events. It cloud drift is less than the respective projection of the
has to be mentioned that this scheme is partly based on measurement setup; i.e. we do not extrapolate outside the
subjective judgement. spatial range of our measured data. Furthermore the
Radiation fluctuations of one or more adjacent time steps temporal and spatial resolution of the radiation field
may be combined to a ramp of duration I to k, if the depends on the time resolution of the measured data and
following main criteria are satisfied by all intermediary time the number of measuring points. FigA gives a sketch of the
steps (l < i,j < k) : geometrical configuration of measurement setup and drift
vector for the modelled PV-array. As shown in the sketch
1. IAG(ti) I = IG(ti) - G(tpAt) I > 10 W/m2 the geometrical configuration and the drift vector define
2. signAG(ti) = signAG(t·) 'stripes' that are represented by the radiation series at one
3. l/f* IAG(tj) I < IAG(tid < f*IAG(tj) I of the measuring sites. The radiation at a given instant to
for a given point P(x,y) in stripe i is then given by the
with f; = 3 radiation at the respective measuring point i at time to +~ t.
Criteria 2 and 3 secure the homogeneity of the ramp. In The time lag ~t is determined by equation (2) for the
order to avoid interruption of a ramp by a single distance vector ~p'
intermediate value IAG(ta)1 < 110 G(ta _l)1 that does not
satisfy criterium 3 or critena 1 and 2 simultaneously a test is
made whether the value of A G at step (a + 1) again
corresponds to the previous ramp data. If so, the ramp is
continued. Ramp events may then be classified by their

and k
* duration Tramp = LA t
1
We have applied these methods to sequences of measured Fig. 4: Scheme for the extrapolation of a radiation field.
data that are characterised by the passing of cloud fields. In The direction of the cloud shadow drift vector 'V and the
the analysis we compare the characteristics of measured geometrical configuration of the measuring points define
single site insolation and of calculated PV-array or PV- stripes orientated parallel to the drift vector. The radiation
system data. For the latter we use spatially averaged time time series at a point P(x,y) is given by the radiation series
series'. The main effects of the transition from single site to at the corresponding measuring point shifted by a time lag
spatially averaged insolation are for example discussed by &t resulting from the respective spatial distance and the
Longhetto and coauthors [5] and Beyer and coauthors [6]. magnitude of the- drift vector.

454
4. ANALYSES OF SPATIAL CHARACfERISTICS
Fig.5 shows a section of the synchronous time series' at the tto ~------------------------,
five sites (data recorded Dec. 4th, 1990). During the first
half-hour the changes in radiation are not common for all .- ....~ -
~_3_
sites. This implies that the field of cloud shadows consists of .....l!!!..L
structures smaller than the measurement setup area. --1!!LL

Fig. 6 shows cross-correlation functions for the time series

depicted in fig. 5. The functions shown represent the results
for combinations of site no. 1 with all other sites. In these
combinations time lags (location of the maxima of D+------r----~------r-----~
correlation) of one or two time steps occur. The maxima o Q.25 QSO 015

show a tendency of decrease with increasing distance

(compare fig.l). The time lags imply a drift vector pointing
south-east. A cloud front is first detected at site nos. 2, 3 Fig. 5: Time series of solar radiation (inclined plane) at the
and 4, later passing 5 and 1. 5 measuring points from 11:30 to 12:30 hrs., Dec. 4th. The
time resolution is lOsec. The systematic deviations occuring
Fig. 7 and 8 give the respective functions for a data at high insolation levels are partly caused by albedo effects.
sequence recorded on Nov. 30th, 1990. During the time
span depicted the inter-station time lags show a slightly
different pattern. This pattern may be identified with a drift
of the cloud array from north-east to south-west influencing
.
.;

the sites in the order no. 1 and then nos. 2, 3, 4, 5.

5. EXTRAPOlATION
With the extrapolation scheme described above data from
the measuring sites are used to derive the radiation field
.
.;
affecting a (hypothetical) PV-generator field. We give here n

O~~~--~--r---r---~~
the example for data measured on December 4th. The drift -2 -I 0 1

vector of the cloud shadows was derived on the basis of the lim. lag [min]
time lags identified for all pairs of stations in the cross- Fig. 6: Cross-correlation function for the time series given
correlation functions calculated for the time span from in fig. 5. The correlation for the combination of station 1
10:00 to 14:00 hrs. with all other sites is shown.
The resulting drift vector pointing south-west has a mo ~------------------------,
magnitude of about 15 m/s. Fig. 9 gives an example of the
calculated radiation field for one point of time. The plot
shows the radiation over an area of approx. 1000m by
1000m.' The Y-dir.ection is parallel to the drift vector. The
resolution in Y-direction results from the time resolution of
the measurements and the drift velocity. The resolution in
the X-direction reflects the width of the stripes as sketched
in fig 4.
This radiation field is used to derive the time series of the
average radiation affecting aPV-array. For simplicity we
here chose a regular array covering the area described Fig. 7: Same presentation as fig. 5 but for the hour 10:30 to
above. The spatial resolution in these calculations is 10m by 11:30, Nov. 30th.
10m. In this two-dimensional field the array averaging
process consists both of a moving average in the time
domain within one 'stripe' (see [6]) and an ensemble
.
.;
average over the data from the different measuring sites. .,
.;
Fig.10 shows a comparison of single point and array
averaged radiation series' for the same time section as in
fig. 5. .
.;
...
';+---~--~--r---r---r--1
-1 -2 -I a 1
lim. tog (mirij
Fig. 8: Same presentation as in fig. 6 for the time series
given in fig. 7.

~
-
.. -
,!;
<
~
.<
"'"
Fig. 9: Spatial radiation field at a given point of time over O. U 0 . )0
ho<n
0 . 1:1-

an area of approx. 1000m • lOOOm; this field is extrapolated

from data measured Dec. 4th. The Y-direction is parallel to Fig. 10: Time series of the average array radiation and the
the cloud shadow drift vector. The depicted point of time radiation at a single site for the sequence given in fig. 5.
coincides with the last downward transition in the time The array rad~ation refers to an homogeneous array
series shown in fig. 5. covering the area represented in fig. 9.

455
6. ANALYSIS OF GRADIENTS IN SINGLE SITE AND
ARRAY AVERAGED DATA
Q.OI

Figs. 11a and 11b show frequency distributions of gradients iii 4

AG(t)/nAt in single site and average array radiation for the I -:'~-' I !
I rI~ !:n
om AtaG_
time span 11:30 to 12:30 hrs. on Dec. 4th. The frequency
distributions refer to time increments nAt of 10 and 30
seconds. Both figures show the reduction in the occurence =
i ... .,
~
of extreme gradients in the spatially averaged insolation. In
comparing the single site curves in both figures it is evident
that the 10 sec data display gradients of greater magnitude. "
[!id'i' h-
~rL1:'!!~:;
This incurs that these extreme gradients do not persist over "'"-JOl -:JlIl -Q) 0 Q)
...,.. ~o-.ll
:JlIl 300
consecutive timesteps.
These characteristics are visible more clearly in the results
of the analysis of ramp events. Fig 12a and 12b show the
occurence of ramps in single site and array averaged data
for Dec. 4th (hours 9:00 to 10:00) ordered according to
their magnitude and duration. The large number of small
ramp events ('ripples') occuring in the single site data is to a
large extent averaged out in the array data. In addition to
these ripples ramps with great magnitudes and short
duration Le. with a high average gradient are detected for
the single site data. These ramps reflect direct transitions
between cloud and no-cloud states. Their magnitude does
not decrease in the array data due to the temporal and
spatial size of the cloud and no cloud states. On the other
hand the duration of these transitions is prolonge.d: Fig. lla,b : Frequency distribution of gradients in single site
7. CONCLUSIONS and array averaged radiation. Fig. 11a refers to a time
The specimen analysis presented here indicates that the increment of 10 sec., fig.11b to an increment of 30 sec (see
spatial solar radiation field may show very small structures eqn. 1). The distributions are calculated on the basis of the
when short term data are considered. A more detailed time series given in fig. 10.
analysis along the sketched out lines has to be based on
data gained from solar radiation sensors allowing for time
resolutions down to 1 sec. This would also enable a more
precise test of the applied extrapolation scheme. It should
be mentioned that this scheme may also be applied to
heterogenous PV systems, e.g. a number of domestic
systems dispersed over a given area. It is not necessarily
limited to averages over continuous surfaces.

ACKNOW1EDGEMENTS
Part of this work was funded by the German Ministry of
Research and Technology (BMFf).
We would like express our thanks to Hans Gerhard BIoos
for data aquisition and measurement setup maintenance.
REFERENCES

/1/ Jewell,W.T., RRamakumar, S.Hill: A study of

dispersed photovoltaic generation on the PSO system.
IEEE Transactions on Energy Conversion, 3(1988), 473-478
/2/ Garrett,D.L., S.MJeter: A photovoltaic voltage
regulation impact investigation technique: part I - model
developement. IEEE Transactions on Energy Conversion,
4(1989), 47-53
/3/ Kern,E.C., E.M.Gulachenski, G.A.Kern: Cloud effects
on the distributed photovoltaic generation, slow transients
at the Gardner, Massachusetts photovoltaic experiment.
IEEE Transactions on Energy Conversion, 4(1989), 184-190
/4/ Suehrcke,H.,P.G.McCormick: The frequency
distribution of instantaneous insolation values. Solar
Energy 40,1988,423-430
/5/ Longhetto,A., G.Elisei, C.Giraud: Effect of correlations
in time and sl?atial extent on performance of very large
solar converSIOn systems. Solar Energy, 43(1989), 77-84 Fig. 12a,b: Histogramms of the frequency of occurence of
/6/ Beyer,H.G., J.Luther, RSteinberger-Willms: Coupling ramp events in radiation time series ordered according to
distributed PV arrays to a mains grid - Simulation their duration and magnitude. Fig. 12a refers to single site,
calculations at high penetration rates. Proc. 9th E.C. fig. 12b to array averaged radiation. The histogramrns are
Photovoltaic Solar Energy Conference,Freiburg,1989,1127- based on measured and calculated data for the time span
1130 9:00 to 15:00 hours, Dec. 4th.

456
HYfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

INSEL - A Simulation System

for Renewable Electrical Energy Supply Systems

J. Luther and J. Schumacher-Grahn

Department of Physics, Renewable Energy Group
University of Oldenburg, D-2900 Oldenburg, FRG
P. O. Box 2503, tel. ++49-441-798-3544, fax. ++49-441-798-3201

ABSTRACT: We report upon the block diagram oriented simulation system INSEL for the interactive
§.imulation of renewable clectrical energy supply systems. Blocks defined by input/output relations
and parameters are interconnected to form a model of the system under investigation. Writing an
INSEL simulation program means to transform a grafical block diagram into the simulation language.
A macro concept allows for different levels of complexity of block structures.

Introduction
In the last two decades several programs for the
simulation of renewable energy systems have been
input
developed. From a general point of view, theses file
programs can be classified into three different cat-
egOrIes:
(i) programs that use statistical information in
1+----+1 model
order to predict the longterm performance of the library
system - an example using the utilizability con-
cept is PV F-CHART (Klein and Beckman 1985).
(ii) programs that calculate a sequence of states
output
of a predefined system structure (allowing for sev- files
eral options) in constant time steps - for exam-
ple PVFORM (Menicucci and Fernandez 1988) and
SOMES (Blok and ter Horst 1987). report
(iii) simulation systems, which give the user files
great flexibility in modeling different system struc-
tures - examples are TRNSYS (Klein and Beckman
1976) and INSEL (Schumacher-Grahn 1991). Figure 1: Typical structure of a simulation system. A simu-
lation model formulated in a descriptive language is entered
Simulation Systems into an input file via an editor. Parser and setup perform the
syntactical analysis and the sorting of the statements.
Figure 1 shows a typical configuration of a simu-
lation system. All editor provides the possibility Another important function of the setup is the
to formulate a model in a simulation language. sorting ofthe statements of the input file. This al-
In contrast to algorithmic languages - FORTRAN, gorithm constructs the final calculation sequence
Pascal or C, for example - simulation languages and in particular has to identify iteration loops.
are descriptive, i. e. theses languages allow the user Thus, the order of statements in the input file is
to formulate a problem without giving a concrete arbitrary and has no influence on the order of cal-
algorithmic solution. The sequence of calculations culations. The result of this procedure is a simu-
is found automatically by the simulation systems lator, i. e. a version of the model that can be exe-
compiler and sorting algorithm. cuted.
When an input file is executed it is read sequen- INSEL is a block diagram oriented simulation
tially by a scanner. The syntactical structure is system for the interactive §.imulation of renewable
analysed by a parser, whose task is the prepara- rlectrical energy supply systems. The simulation
tion of intermediate code and/or lists. The setup language of INSEL mainly consists of instructions
identifies the elements of the input file and trans- to describe the structure or topology of a simula-
forms them into their internal representations and tion model and instructions to define the parame-
data structures. ters.

457
 INSEL is written in ANSI FORTRAN 77. The The INSEL-model - i. e. the block diagram of
program includes blocks for electrical system com- figure 2 - has to be transformed into INSELs sim-
ponents as photovoltaics, wind turbines, motor/- ulation language. The following lines of an input
generators, batteries, electrolysis and fuel cells, file give a complete description of the model struc-
power conditioning units, load characteristics etc. ture (see figure 1).
Furthermore, there are blocks for file handling, the 100 TIME
S
generation of plots, blocks solving iteration pro- S 200 LOAD 100.1
cesses, blocks converting meteorological data and S 300 METED 100.1
blocks performing parameter variations. S 400 PVMPP 300.1 300.2 300.3 100.1
S 500 DC DC 400.1
S 600 SUM 500.1 -200.1
A Simple Application of INSEL S 700 BAT 600.1

As an application of INSEL we discuss a model An "s" standing for "structure" is followed by

of a renewable energy system that consists of pv an arbitrarily chosen block number and the blocks
generator, power conditioner, battery and load. name. Input connections are given by the corre-
This standard system may also be simulated with sponding block number and the output number of
the other programs mentioned above. Figure 2 that block, separated by a period. For example,
shows a corresponding INSEL-model. block 400 uses the pt output of block 100 and the
1st, 2nd and 3rd output of block 300, na.mely block
METED. Parameters are entered into the input file
in a similar manner (see below). The sequence
of statements in the input file is completely arbi-
time trary and has no influence on the sequence of cal-
culations in the simulator. This sequence is gen-
erated automatically by INSELs sorting algorithm
power (see "setup," in figure 1).
to load

The macro concept in INSEL

power to 600 dc power
(dis- ) charge Most blocks in figure 2 are macro blocks, i. e. they
battery
are composed of more elementary blocks. For ex-
ample, block PVMPP has to solve different tasks:
(i) the simulation of the voltage/current charac-
teristics of the pv array, (ii) the simulation of the
temperature of the solar cells and (iii) an itera-
tion to find the maximum power point. The block
Figure 2: INSEL-model of a pv /battery /load system.
diagram of macro PVMPP is shown in figure 3.

The simulation run is controlled by block TIME

which provides the simulation time. The length
of the simulation time interval and the time step
may be fixed by means of parameters of TIME. The
meteorological data (time series) for global radi-
ation, wind speed and ambient temperature are
read from a file by block METED. If necessary, the
radiation data may be converted to a tilted plane.
The photovoltaic generator operates in its max-
imum power point, which is determined by block Figure 3: Internal structure of the macro PVMPP of figure 2.
PVMPP. The array power output is weighted with The maximum power point is calculated in an iterative loop
the efficiency of a dc/dc inverter by block DCDC. controlled by block MPP. This block generates a sequence of
voltages in order to find the maximum of P = U I. The "top
The difference between the generated dc power
of loop" block TDL is used by the sorting algorithm.
and the load demand (e. g. given by a daily load
profile) is used to charge/discharge the battery Block PVI simulates the pv generator. The
BAT. Outputs of this block are the battery state of voltage/current characteristic is calculated using a
charge and the amount of backup or excess power. two diode model; the differential equation des crib-

458
ing the array temperature is based on an energy Flexible Outputs
balance. Inputs to this block are voltage U, global
radiation G t to the tilted plane, wind speed V w , The model in figure 2 does not produce any nu-
ambient temperature Ta and the simulation time. merical or grafical outputs. In general, each signal
Output is the corresponding pv generator current. of interest can be observed by connecting one of
Block MPP iterates the voltage to find the max- the following INSEL blocks: SCREEN displays infor-
imum product of U I. The algorithm of this block mation upon the console, WRITE writes data to a
is based on a successive division of the voltage in- file, PLOT produces grafics. For example, a time
terval into three according to the golden section. series of the backup/excess power can be plotted
A voltage output of MPP is transfered via the (see figure 4) simply by adding the record
"top of loop" block TOL to PVI (the block TOL is S 1000 PLOT 100.1 700.2
essential in order to sort nested loops automat-
ically). Block PVI calculates the corresponding to the input file. The inputs of block PLOT are in-
value of the current and returns this value to MPP. terpreted as (x, y)-coordinates of the data points.
Depending on the new product U I block MPP de- If no parameters are specified INSEL uses default
termines the next value of U. In this way MPP values in order to have a quick look at the results.
determines the maximum
1500 . 0 , - - - - - - - - - - -_ _ _ _- - ,

max {f(U) = UI}

UE[Ut,uz]
1000 . 0
where U1 and U2 are given as parameters. Block ~
.,.,
MUL multiplies voltage and current to yield thear- II>
Ii
II>
ray output power P mpp • As a consequence of the .........
500 . 0
a.
sorting algorithm the multiplication is performed J

after the iteration process has stopped. 1 0.0

The calculation sequence of such iterations is

controlled by an internal Jump-Parameter, which
-SOO . O+-~~-._~~-r_~~-,......~~--,J
is determined by INSEL automatically. Thus, a 90 . 0 180,0 .0

user does not have to care for these details. time [dJ
The definition of PVMPP as a macro is as follows: Figure 4: Hourly mean values of backup (negativ) and execss
power (positive). The layout of the simulated system'is as
follows : The mpp tracked pv array has 40 panels of type
macro PVMPP AEG PQ10/40, i. e. a nominal power of 1.6 kWp, slope is 60 0 ,
inputs: 4 Y. Gt vv Ta t south oriented. The mean load is assumed to be 100 W, the
outputs: 4.1 Y. Pmpp daily load profile is taken from PVFORM. The battery consists
parameters: 25 of 12 VARTA bloc cells in series, each cell having a nominal
strings: 0 capacity of 200 Ah equivalent to a maximum energy content
Y. structure of 4.8 kWh. Meteorological data base: TMY Bremerhaven ,
S 1 TOL 3.1 Germany (Beyer et al. 1991).
S 2 PVI 1.1 IN(l) IN(2) IN(3) IN(4)
s 3 MPP 2.1
S 4 MUL 2.1 3.1
Yo parameters Mathematical processing of outputs
P 2 PP(l) PP(2)
With INSEL one has access to the outputs of each
block in a given model. Figures of merit can be
The macro definition starts with the name of calculated by processing the system variables in
the macro followed by the number of inputs, the the block diagram language. As an example we
macro outputs, the number of numerical para- show how the renewable fraction of the system in
meters and string,parameters (% introduces com- figure 2 is determined in an INSEL model.
ment). The definition of structure and parameters The renewable fraction F may be defined by
of the marcro is the same as in all INSEL models
with two exceptions: F = Eload - Ebackup
Eload
(i) macro inputs are defined by the "IN" func-
tion, (ii) macro parameters are defined by the "PP" Energy to the load Eload and backup energy Ebackup
function. When the macro is used in a simulation are integrated over the simulation period. The
a macro expander replaces the variables by their calculation of F may be performed by the source
actual values. code

459
Yo calculation of the renewable fraction Menicucci, D. F. and Fernandez, J. P. (1988) User's manual
S 801 CONST for PVFORM: A photovoltaic system simulation program for
S 802 MIN 700.2 801.1 Yo P_backup < 0 stand-alone and grid-interactive applications. Sandia Report
S 803 CUM 200.1 Yo E_load SAND 85-0376· UC-276, Albuquerque, NM, USA
S 804 CUM 802.1 Yo E_backup Schumacher-Grohn, J. (1991) Digitale Simulation regenera-
S 805 SUM 803.1 804.1 tiver elektrischer Energieversorgungssysteme. Dissertation,
S 806 DIV 805.1 803.1 Yo F Universitiit Oldenburg, FRG
S 807 SCREEN 806.1 Yo display F
Yo parameters
p 801 0
Acknoledgements
This work was funded by the German Ministry of Research
and Technology (BMFT).

backup
excess

Figure 5: INSEL model to calculate the renewable fraction.

The corresponding block diagram is shown in

figure 5. In this model block MIN determines the
backup power Fbackup < 0, i. e. the minimum of
the two inputs. .Road and Fbackup are cumulated
over the simulation period by block 803 and 804,
respectively. The sucessors of these blocks (blocks
805-807) are executed only once at the end of the
simulation. Blocks SUM and DIV calculate the re-
newable fraction which is displayed upon the con-
sole by block SCREEN in this case.

Conclusion
The simulation system INSEL represents a flexible
tool for the analysis of renewable electrical energy
supply systems. Typical applications are complex
design issues rather than the simulation of simple
systems like the one given in figure 2. Because of
the modular structure of INSEL it is possible to
refine the models and the program output at the
main points of interest.

References
Beyer, H. G., Luther, J., Schumacher-Grohn (1991) Compar-
ision of some standard pv simulation programs with the sim-
ulation system INSEL. Proc. of the Biennal Congress of the
International Solar Energy Society Denver, CO, USA 1991
(to be published)
Blok, K. and ter Horst, E. (1987) SOMES - A simulation and
optimization model for autonomous energy systems; descrip-
tion and manual Version 1.1. University of Utrecht, Nether-
lands
Klein, S. A. and Beckman, W. A. (1-976) TRNSYS - A Transient
Simulation Program. ASHRAE Transactions 82, 623
Klein, S. A. and Beckman, W. A. (1985) PVF-CHART User's
manual. Wisconsin - Madison, USA

460
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8--12 APRIL 1991 LISBON, PORTUGAL

SENSITMTY ANALYSIS OF ENERGY REQUIREMENTS FOR TlDN Fll.M

PHOTOVOLTAIC MODULE PRODUCTION

K.M. Hynes, N.M. Pearsall and R. Hill

Newcastle Polytechnic, Ellison Place, Newcastle upon Tyne, UK

ABSTRACT. The determination of energy input requirements for the production of thin mm
photovoltaic modules is discussed, using CuInSe 2 based cells as an example. Two fabrication
technologies, thermal coevaporation and electrodeposition, are compared, assuming a 10 MW
p.a. production facility. The sensitivity of energy requirement to the process parameters
assumed is investigated. The effective carbon dioxide emissions for PV modules are shown to
be significantly lower than those for conventional electricity generation.

1. INTRODUCTION

In view of the present concern over the A) Materials Energy (MtE) - the amount of energy
environmental impact of energy supply technologies, it is consumed in manufacturing the materials which appear in
clear that it is insufficient to consider only the economic the finished product;
cost of a candidate supply. Several recent studies have B) Machine Energy (ME) the amount of energy
addressed the issues of environmental impact, during both consumed in the manufacture, of the production equipment
production and operation, in terms of such aspects as used in the step;
discharges, material toxicity, waste disposal etc [1,2]. C) Process Energy (PE) - the amount of energy used to
However, the manufacture of energy supply hardware perform the process step; this may also include materials,
requires an energy input, which itself has environmental such as process gases, which do not appear in the
consequences, and this "indirect" impact must also be finished product.
considered. For module production at present and in the
near future, this energy input must be assumed to The total process step energy then becomes the sum of
originate from conventional, high environmental impact these three values.
technology and, therefore, the minimisation of energy
requirement is of both economic and environmental
importance. 3. PRODUCTION SEQUENCE
This paper will discuss the calculation of energy
inputs to thin mm photovoltaic module production, taking In this analysis, the production of thin film
as an example the Cd(Zn)S/CuInSe 2 process. The Cd(Zn)S/CuInSe 2 modules will be considered, although the
sensitivity of the derived energy input to assumptions of results may be expected to be generally applicable to
pro~ss technology and parameters will be addressed. Even most thin mm technologies. Several techniques for the
for a process which is not yet fully determined, this deposition of the semiconductor layers have been
approach can allow critical steps to be defined, in order investigated in this work. This paper will restrict the
to meet energy input criteria, whether economic or discussion to two production sequences, representing the
environmental. This information can then be used to upper and lower limits of energy requirement as
identify beneficial development routes. It should be noted determined in this work. In the first sequence, the
that the routes for reduction of cost and energy input are CuInSe 2 layer is deposited by thermal coevaporation, as is
not necessarily the same and so energy analysis is a a CdZnS layer. In the second sequence, electrodeposition
valuable complement to economic analysis for any given is used for the CuInSe 2 layer and the CdS window layer
process. is produced by a chemical dip process. Subsequently, the
window layer is completed by the magnetron sputtering of
ZnO for both sequences.
2. ENERGY ANALYSIS A flow diagram of the production sequence, divided
into the process steps considered, is shown in Figure 1.
The' technique of energy analysis is widely used for For these analyses, all steps other than those relating to
determining energy flows in industrial processes and allows the deposition of the absorber and window layers are
the calculation of the required energy input for unit assumed to be common to both candidate processes. The
output of product. In this case, unit output is defined in module uses 3mm thick float glass for both substrate and
area terms (i.e 1 m 2 of module area) and a linear front encapsulation, whilst rear and front contacts are
production sequence is assumed. This allows the sequence formed by molybdenum and aluminium respectively. Whilst
to be broken down into individual process steps (e.g. it is recognised that patterning of the deposited films
deposition of CulnSe 2 film), for which the primary energy would occur between each deposition step, in order to
inputs are determined. The energy input for the whole effect interconnection of cells, the energy requirements of
sequence is obtained by summation of the inputs for each these operations have been included with those of the
step. This may then be compared with the energy output final encapsulation step for ease.
from the same unit area during its lifetime.
Problems can occur whilst performing energy analysis,
due to the variation in nature of the inputs. The work 4. ASSUMPTIONS
reported here conforms to IFIAS Convention Level 3 [3],
which includes process energy, material input and an This section will outline the assumptions used to
allocation for the energy content of the capital equipment arrive at the energy content of the modules considered.
required. All final energy values are given in kWh(thermal)/m 2.
The input data for each process step were analysed Energy values ior each step have been calculated using
to determine the following quantities in each case: the following procedures:-

461
Figw-e 1. Flow diagram of production aspects would require a detailed consideration of the
sequences used for energy analysis production facility which would be inappropriate at this
stage.
In the initial case, energy values for a particular
Glass Substrate 3mm thick float glass process step were calculated assuming maximum
throughput per machine, based on cycle time and for
continuous three shift operation for 300 days per annum.
Solvent clean The assumed parameters' for the process steps are shown
in Table 1. Clearly, in a true production facility, some
machines will be operating below maximum capacity to
Molybdenum eliminate backlogs in any particular area. Therefore, the
Magnetron sputtering energy values were then corrected, assuming an annual
production of 100,000 m' of modules meeting the
required specification (equivalent to 10 MW at 10%
Deposition of 1) Thermal coevaporation efficiency). This involved corrections for reduced
CulnSe, 2) Electrodeposltlon throughput for some process machines and for cumulative
process yields. In subsequent calculations on the sensitivity
1) Evaporation, CdZnS of assumed process parameters, the total production
Deposition of Mag. sputtering, ZnO sequence data were used.
Window Layer 2) Chemical dip, CdS
Mag. sputtering, ZnO
5. BASE CASE ENERGY REQUIREMENTS
Deposition of Aluminium
Top Contact Thermal evaporation Figure 2 shows the base case energy requirements for
the two production sequences by process step. It is clear
that the most significant variation between the two
Patterning and sequences is the CuInSe, deposition step, which accounts
Encapsulation for almost all the difference in total energy requirement.
Minor differences in the other steps are mainly due to
yield assumptions. The window layer processing is
Front Glass 3mm thick float glass dominated by the ZnO deposition and, thus, little
difference is observed between the two options. This
situation would be altered if this vacuum process could be
Edging Aluminium removed from the electrode position sequence.
Figure 3 illustrates the energy breakdown within the
deposition steps and provides a comparison for the' two
processes. In both cases, the material energy contribution
A) Materials Energy: Where the material input is in is highest for the Mo back contact and is negligible for
elemental form, the energy content of the element was the window layer and top contact. For all steps except
taken from standard references [4]. Where the material
input was in compound form, an extra allowance was
made for the energy consumed in the synthesis of the Table 1. Base parameters for CulnSe2
compound. For the CdZnS, the energy content was taken module production sequence
to be twice the elemental requirement and, for the other
compounds (e.g. ZnO), it was taken to be equivalent to
the elemental requirement for the whole compound mass. Parameter Thermal Electro-
The amount of material required was calculated from the Evaporation deposition
film thickness, allowing for material yield within the BACK CONTACT
deposition sequence and for the process yield (i.e. the Film Thickness (pm) 2 2
proportion of product from the step of an acceptable Mat. Yle ld ('II) 80 80
quality and allowed to proceed to the next processing Dep. Rate (nm/min) 300 300
stage). . Process Yield (%) 95 95
B) Machine Energy: It is not generally possible to
determine the -energy input of production machinery CulnSe, DEPOSITION
explicitly and it is an accepted practice to relate the Film Thickness (~m) 4 3
energy content directly to the capital cost of the Mat. Yield [Cu,ln,Se] (%) 65,70,50 96,99,95
equipment. The assumption here is that all production Dep. Rate (nm/min) 100 33
machinery has a similar proportion of energy input per Process Yield (%) 80 90
unit cost. Machine energy content values were only
available for the late 1970's [5], when energy analysis was WINDOW LAYER
more widely used than at present. On the basis that CdS DEPOS I TI ON
primary energy inputs in manufacture will have altered at Film Thickness (pm) 0.5 0.02
a much slower rate than cost levels, the value for Material Yield (%) 80 95
scientific machinery was adjusted for inflation to yield a Dep. Rate (nm/mln) 500 2
value of 0.77 kWh(t)! £Sterling. This value was used for ZnO DEPOS I TI ON
all base case calculations, although the sensitivity of the Film Thickness (pm) 2 2
analysis to this parameter has been investigated (see Material Yield (%) 80 80
Section 6). Derived energy values were again corrected Dep. Rate (nm/mln) 300 300
for process yield. Total Process Yield ('II) 90 90
C) Process Energy: Most of the energy inputs for this
category were electrical in nature and a conversion factor TOP CONTACT
of 1 kWh(e) = 4 kWh(t) was used throughout. As before, Film thickness (pm) 2 2
energy values were corrected for process yield. Material Yield (%) 70 70
Any output not reaching the required standard, and, Dep. Rate (nm/min) 500 500
therefore, not proceeding to the next stage of the Process Yield ('II) 95 95
production sequence, was assumed to be discarded at no
energy cost. This is clearly a simplification, since there ALL STEPS
would be an energy requirement for the disposal of Machine Lifetime (yr) 10 10
chemical waste. Also, we have assumed no energy benefit Machine Energy Content 0.77 0.77
from the recycling of waste. Quantification of these (kWh (t) /£)

462
 Base case energy requirements (a) the machine energy content per unit cost,
Figure 2. (b) the amount of material incorporated in the product,
(c) the process yield for the critical CuInSe 2 deposition.
[llj Edging Machine Energy Content : Due to the limited data
available on energy requirements of machine manufacture,
[ill F10nt Glass
the value used in this work was extrapolated from a late
400
D Encapsulation
Patterning & 1970's value. It was aSsumed that the actual energy
requirement remained constant, but that the cost increased
according to the inflation rate for scientific equipment. In
tLZl Top Contact order to determine the effect of this assumption on the
derived energy requirement of the module, a variation in
• Window machine energy content between about 50"A> and 250% of

o Absorber
the base value was investigated.
Material Reauirement: The base values for film thickness
and material yield were derived from consideration of the
W&] Back Contact technical nature of each process. However, it must be
recognised that the semiconductor film thicknesses required
• Glass Cleaning for optimum performance are not yet determined. In the

o Glass Substrate
case of module production, other issues, such as process
yield and cycle time, also become important in the
determination of film thickness. We have investigated the
effect of a variation in film thickness for both 'CuInSe 2
and molybdenum between 50% and 150% of the base
value . Similar effects would be observed for variations in
material yield.
Thermal Electro- Process Yield: For a production sequence, process yields
Evaporation deposition become cumulative and, therefore, may be I\xpected to
have a significant effect on total energy requirement .
Here, we consider the yield for the most influential
process step, the CuInSe 2 deposition . A variation between
the electrodeposition of CuInSe 2' the process energy about 80% and 120% of the base value has been
dominates. Machine energy contents, even for high capital investigated.
cost vacuum equipment, are relatively small. Energy
requirements for steps not shown differ only slightly
between the two ' processes and are dominated by the
materials energy requirement of the glass substrate and Figure 4. SENSITlVITY PLOT (L> - Culn5e, film Ihlckness,
superstrate and the aluminium edging. Indeed, if the
contributions due to the two sheets of glass are added,
*-
X - Mo film thickness, • - Machine energy conlenl,
Process yield, Culn5e, step)

this becomes the largest factor in the module energy 120 ,...--- - - - - - - -- - - - - -- -,
requirement.
11 6

6. SENSITIVITY ANALYSIS OF ENERGY

REQUIREMENT .
:>
112

Following calculation of the base case values, this ...,

'ii
>
to
108

work set out to evaluate the effect of a variation in those ID

--------------
'"0;'"
assumptions on the derived energy value. The results of 104
three parameter variations will be presented here . The c::
III
chosen parameters represent those for which the 100
assumptions are most uncertain . These are: ~
0

..
I-
'0 96

..e.,
01

Figure 3. Energy breakdown within deposition ;: 92

steps
0.. - - Thermal Evaporation
BB
---- Electrodeposltion
1 Thermal Evaporation
54
160 2 Electrodeposition
SO
100 150 200 250
• Process Energy 50

120
o Machine Energy
Percentage 01 Parameter Base Valu e

Figure 4 shows the percentage changes in total

W&I Materials Energy energy requirement for the above variations. Despite being
the most uncertain value, it can be seen that changes in
80 the machine energy content have only a small effect on
the energy requirement, giving a result of around 10% or
less even for a 250% variation. By contrast, the energy
requirement is much more sensitive to process yield, with
40 a 10% variation observed for a yield change of less than
25% . However, the range over which the process yield
may be expected to change is much smaller than for the
other parameters and so does not cause too much
concern. It is interesting to note the similarity in the
Back Absorber Window Top effect for both the assumed production sequences.
Contact Contact Changes due to the molybdenum film thickness are
also similar for the two processes and are generally
PROCESS STEP

463
comparable with those observed for variation of the As the average daily insolation level increases, the
machine energy content. The parameter showing most associated CO 2 emissions decrease. Comparison of the
effect is that of the assumed film thickness for the values derived here with those derived from a US
CuhlSe 2 film in the case of thermal evaporation. As in Department of Energy study [1] give a good agreement
the other calculations, it has been assurned that a\1 other (9.05 x 10- 3 cf. 5.9 x 10- 3 kg CO ,tkWh(e)) if the
parameters (such as deposition rate, power requirements same efficiency and lifetime assumptions are used. It
etc.) remain unchanged. This leads to a significant change should be noted that the US study also assumed a 100
in the cycle time for the deposition and, hence, in the MW facility and amorphous silicon technology. The
attributable machine and process energies. The material agreement, therefore, is considered to be very close,
energy contribution is, in fact, very sma\1, accounting for considering the different starting points and methodology
only 0.3% of the almost 18% increase for a film of the two studies.
thickness of 6 jIlO. For the electrodeposition process,
however, the sensitivity to CuInSe 2 thickness is very low,
since this step represents only 2.8% of the energy 8. DISCUSSION
requirement for the base case, and is not shown in
Figure 4. This study has considered the energy requirement for
the manufacture of a thin film CulnSe 2 module and
investigated the sensitivity of that requirement to various
7. CARBON DIOXIDE EMISSIONS processing parameters. The purpose of this work was to
identify which aspects of the production technology were
One of the most important environmental issues is influential in the energy content of the finished product,
the release of carbon dioxide due to industrial processes, in order to feed into present development work. As such,
transportation and electricity generation. In order to the processing conditions were extrapolated from existing
determine the likely contribution from photovoltaic research parameters for a relatively smal1 (10 MW p.a.)
electricity, the carbon dioxide emissions associated with production facility.
the energy requirement in manufacture must be The energy analysis showed the deposition steps to
determined. This study considers only the module be mostly influenced by process energy, with the material
production, although it must be recognised that extra energy requirement a minor factor in al1 cases. Since the
energy requirements would ensue from overheads at the amount of energy required to perform the process step is
manufacturing facility, balance of systems hardware, subject to variation with the detail of the processing
instal1ation and maintenance. To examine the worst case, equipment, this means that there is considerable scope for
it has been assumed that al1 the energy inputs are reduction of energy requirement by careful choice of
electrical in nature and that they are supplied by a technology and processing conditions. This work also
conventional mix of power stations. Thus, a value of 0.3 assumed glass at both front and back of the module, due
kg CO /kWh(t) has been used. to the present substrate requirement for CulnSe 2 ce11s.
The effective carbon dioxide emission per kWh for Since the material energy of the glass is a major
the photovoltaic system depends on the module lifetime contribution to the total energy requirement, there is
and the insolation regime. Figure 5 shows the value clearly considerable scope for energy reduction by
determined as a function of average daily insolation, replacement of the glass substrate with a lower energy
assuming a 20 year lifetime and a system efficiency of content material.
9% (using modules of 10% efficiency). It can be seen It was found that the energy content was particularly
that, even for a low insolation location and the most sensitive to process yield, although this parameter has a
energy intensive production sequence, the associated CO 2 restricted range of variation, and to the CuInSe 2 film
emissions are a factor of 5 below that of a conventional thickness for the thermal evaporation process. This latter
supply mix. In that location, of course, it is unlikely that sensitivity was a function of the increased cycle time
the system would be economical1y viable. associated with the deposition of a thicker film rather
than the material content. Hence, development of a low
energy process should concentrate on the cycle time and,
Figure 5. Effective CO 2 errusswn per kWh as hence, throughput which can be achieved, whilst
a function of average daily maintaining a reasonable process yield.
insolation Calculation of the CO 2 emissions associated with
module production demonstrated that the photovoltaic

-....
system would have significantly lower levels than
conventional power generation, even for very low
0.06 Thermal insolation locations. Of course, as the electricity supply
~ .... Evaporation mix begins to include a higher proportion of .renewable
~ energy technologies, the associated CO 2 emissions wil1

-·S
-...
~
0.05 reduce still further.
~
9. REFERENCES
0.04
(I)
\ 1. "Energy System and Materiel Requirements" ,
~ \
\
2.
Meridian Corporation, Alexandria, VA, February 1989
O. Hohmeyer, "Social Costs of Energy
~
0.03 \\ Consumption", Springer-Verlag, Berlin, 1988
3. I. Boustead and O.F. Hancock, "Handbook of
! \ Industrial Energy Analysis", Ellis Horwood,
C 0.02 Chichester, UK, 1979, p. 184
"""
I~
4. Ibid, p. 309
S. O. Jenkins and R. Harrison, "Energy Analysis of
Electro- ,If ' ......
Wave Energy Concepts", Final Report, Section 2,
0.01 depositiorl ......................
----- (1981), WESC ElSNCONI6321172/099

[Part of this work is supported by the Commission of the

1 2 3 4 5 European Communities under Contract No. JOUR-<104S-C
(MB)]
Average Insolation [kWh/m2/day]

464
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE IH2 APRIL 1991 LISBON. PORTUGAL

ITE-BOSS
A NEW SOFTWARE TOOL FOR PHOTOVOLTAIC SYSTEM DESIGN

A. Bosch, H.P. Hoenes, A. Jossen, H. Karl,

G.Lehner, G.Saupe, A. Zahir
Universitat Stuttgart
Institut fUr" Theorie der Elektrotechnik (ITE)
Pfaffenwaldring 47, 7000 Stuttgart 80
Phone: 0711/685-7250 Fax: 0711/685-7222

ABSTRACT. ITE-BOSS (ITE-Block Orientated Simulation Software) is a

software tool for the design and optimization of photovoltaic (PV)
systems. It uses a block orientated modelstructure that allows the
simulation of various plants. Furthermore the development of new
component models is supported. A standard window environment per-
mits a quick and easy input of the required simulation data. A
post-processor provides the analysis of the numerical results. Op-
timization is done by variation of specified parameters in order
to minimize a criterion function. This paper describes the struc-
ture and facilities of ITE-BOSS as well as our present efforts in
battery modeling.

1. INTRODUCTION identified by their number.

In former projects at our institute a

simulation program (ITESIM) [11 was devel-
oped for the analysis of PV systems. Based EDITOR
on that program we now work on a software
tool for phot6voltaic system design named MODEL·
ITE-BOSS. FilE
The first step in the development of
ITE-BOSS was the study of techniques and PRE-
programs for PV system simulation. This PROCESSOR
study provided the following results:
- A modern software tool should dispose of
a comfortable standard user interface.
- A modular model structure is the basis for
the capability to simulate all imagina-
ble system configurations. SIMULATION
- A library of standard PV components fa-
cilitates model configuration.
- Different kinds of meteorological data
must be processed.
A sufficently precise computation of
physical and economical system data must POST·
be performed.
"- An exhaustive analysis of the simulation PROCESSOR
results is "required.
- An optimization algorithm must be imple-
mented to find the best system para-
meters.
Actually ITE-BOSS runs on IBM compati-
ble PCs under DOS supporting the most im- PLOTTER
portant video adapters. Another version of C'" uN '
the program runs under the operating system
VMS on VAX-computers and VAX-stations.

Fig. 1 Structure of the simulation program

2. STRUCTURE OF ITE-BOSS

Fig. 1 gives a survey over the struc- All the model informations are collec-
ture of ITE-BOSS. ted in a model file. The file is stored in
a specific format using ASCII-characters.
2.1 Description of the simulation model 1'01
The required modular model structure is

[}I~
realized by the use of a block orientated
description language, which is illustrated
in Fig. 2. Blocks are described by a num-
ber, a name indicating their function, the
numbers of their input blocks, and optional
parameters.
The timing conditions are given by the
duration and the timestep of the simula-
tion. Blocks desired for output have to be Fig. 2 Block diagram

465
2.2 Editor teractions via keyboard or mouse are sup-
The configuration of the modelstruc- ported.
ture is supported by special dialogwindows
t~ guarantee a quick and easy input of this
data. For corrections and changes in the Simulation Graphic Optimize
modelfile a texteditor was implemented in
ITE-BOSS. Of course it is capable to handle Model
other textfiles like, e.g. meteorological Blocks
datafiles. Therefore all important simula- Plotblocks
tion data can be managed out of the pro- Timing
gram.

2.3 Pre-Processor
Before a simulation run the modelfile
is checked for syntax and structural errors
like e.g. undefined inputs. If an error oc-
curs, a detailed message containing correc-
tion hints is displayed.
Subsequently the blocks are sorted in Fl=Help F2 =Sav e FIO=Menu/Editor
order to determine the computation se-
quence. A special treatment for algebraic
loops is necessary due to the different Fig. 3 Main menu of ITE-BOSS
computation algorithms required. An al-
gebraic loop is due to a feedback in the
blockdiagram without time delay. On VT-terminals connected to the VAX
The pre-processor disposes the basic computers at our institute no window envi-
datastructures for the simulation. ronment exists. Therefore the described
user interface was transferred to make ITE-
2.4 Simulation BOSS available on these machines.
The simulation progress is organized Running on VAX stations ITE-BOSS is
in an equidistant time pattern. At each capable to use the DEC Windows standard,
timestep the whole simulation model is com- which is based on the,X-Window standard de-
puted. First the blocks that could be sor- fined by MIT.
ted in the computation sequence are com-
puted. Next the computation of the al- 3.2 Computation algorithms
gebraic loop is performed. The calculation As mentioned above, ITE-BOSS computes
process for the algebraic loop will be de- the simulation model in a pattern of equi-
scribed later. distant timesteps.
The numerical results are displayed in At each of this timesteps the blocks
a special window and are stored on a file of the model are treated in a certain se-
for further analysis. quence. Therefore the blocks have to be
It is possible to interrupt the simu- sorted in that sequence. Precondition is a
lation at each time step. After an inter- definite temporal relation between them.
rupt the computation can be restarted as Algebraic loops in the blockdiagram prevent
well as continued. the determination of the computing se-
quence. A simple example for such a con-
2.5 Post-Processor struction is shown in Fig. 4.
The post-processor implemented in ITE-
BOSS offers various facilities to analyse
and display the simulation data as well as
diverse other datafiles. Only a columnwise
format with up to six datacolumns of ASCII-
characters is prescribed for these files.
Statistical evaluation and graphical
presentation of the data is provided. Basic
mathematical operations with datacolumns
are supported. load
The results can be saved on another
datafile or sent to the output devices Fig. 4 Example for an algebraic loop
graphic screen, printer, or plotter.
The calculation of algebraic loops requires
3. FEATURES OF ITE-BOSS an additional algorithm. For that purpose a
n-dimensional Newton-Raphson algorithm [3]
3.1 User environment is used in ITE-BOSS, where n represents the
A primary intention during the de- number of blocks in the loops.
velopment of ITE-BOSS was the implementa- The Newton-Raphson algorithm is a com-
tion of a standard user interface in order mon method to solve multi-dimensional non-
to reach a wider range of applicants. linear systems of equations. In this case
In the domain of PC applications the the system of equations is formed by the
SAA-standard (2) defined by IBM gains more block functions. ~ecause of the large
and more importance. As illustrated in variety of block functions it is very dif-
Fig. 3, ITE-BOSS makes use of pull-down ficult to find an implementation of the
menus in order to facilitate and struc- algorithm matching all kinds of systems of
turize the work with the program. Dia- equations. The main problems are :
logwindows provide a nearly errorfree and i) Determination of an' appropriate
well formatted model input. Depending on initial value.
the program status an online help facility ii) Computation of the derivatives.
offers additional information. User in iii) To guaranty a steady convergence.

466
 i) The theory of the Newton-Raphson algo- fine a different scaling in order to change
rithm only guarantees convergence in a the output range. Statistical evaluation
limited range around the fixed point. like computing means, standard deviation
Therefore it is important to determine an etc. is installed. For the presentation of
initial value for the iteration inside that the distribution of the simulation results
domain of attraction. At the first timestep a histogramm can be displayed.
the block inputs are set to sustain the
block functions inside their normal region 3.4 System Optimization
of operation. In the following timesteps A main task in PV system design is the
the previous operating point is used as the optimization of the plant performance to
inital value for the iteration. If the it- achieve the best utilization of available
eration doesn't converge with this' initial energy. In the simulation model the opti-
value, an algorithm embedding [4] the mization refers to the variation of user
system of equations is activated. defined block parameters in order to get
ii) The analytical derivatives of the block the optimum value of the criterion func-
functions are computed as far as possible. tion. Up to eight parameters can be varied
Otherwise a specific differential quotient simultaneously. To check the criterion
is constructed. function a complete simulation run must be
iii) To achieve a steady convergence the performed.
Newton-Raphsonalgorithm is damped. The re- Optimization can also be used to match
laxation factor is computed according to a model parameters to measured data. Figure 6
criterion which relates to the degree of shows as an example a configuration to de-
convergence [5]. In case of fast conver- termine the parameters of the battery block
gence the relaxation factor approaches 1, BAT. The criterion for the best parameter
which leads to the standard Newton-Raphson match consists in the minimization of the
algorithm. If the convergence decelerates, error squares for the simulated and meas-
the relaxation factor decreases in order to ured battery voltage. Block number SUD cu-
expand the convergence domain of the oper- mulates the error squares.
ating point. The computation time grows si-
multaneously because of the increasing num-
ber of iteration steps needed to reach the
fixed point of the iteration. Irregular
block functions and large fluctuations of
the input data like, e. g. the solar irra-
diance, aggravate the computation process.
A well known problem prohibiting a
steady convergence of the Newton-Raphson Fig. 6 parameter optimization
algorithm is the occurence of oscillations
dur~ng the iteration. This comes from cer-
tain ,constellations in the characteristic Two different algorithms for the para-
curves of the system. Figure 5 illustrates meter variation are disposable in ITE-BOSS.
such an constellation. - A simplex method [6],
y - a combination of a random search
algorithm and the simplex method.
Centroid

~--r-;---------------+-----+-~x

Fig. 5 Oscillations in Newton algorithm

These oscillations are eliminated by the

damping. Nevertheless the variation of the
relaxation factor may cause again oscilla-
tions. This must be considered in the de-
velopment of the variation algorithm.

3.3 Result Representation

The simulation results are stored on a
datafile. In the leftmost column the course
of time is reproduced. Fundamental mathe-
matical operations with the datacolumns are
supported. Viewing and printing of the nu-
merical data is provided with the implemen-
initial simplex
ted editor. Graphical presentation of the
results can be performed on the screen as
------ resulting simplex
well as on a HPGL-plotter. A default sca-
ling between the minimum and maximum value reflection axis
is implemented. The user is capable to de- Fig. 7 Princ'iple of the simplex method

467
 Optimization means minimization of the - computation algorithms without too strong
output value of the criterion block. The demands on the mathematical knowledge of
simplex method is a conventional algorithm the user,
for function minimization. Figure 7 il- - extensive facilities for data output and
lustrates the principle of this method. analysis,
In the n-dimensional space a simplex is a - capability to find the best system con-
geometric figure which is generated by n+1 figuration.
points. First the user defines the initial Consequently ITE-BOSS is a universal pro-
simplex. The maximum point of this simplex gram ~or the design of all kinds of
is determined and reflected at the centre systems. It is easy to handle and requires
of gravity of the remaining points. This no further knowledge of the implemented
procedure is repeated until no further algorithms.
minimization can be performed. Convergence
in a secondary minimum is the main problem
of this method. . 6. ACKNOWLEDGEMENT
As an attempt to avoid this, the sim-
plex method can be combined with a random This project is supported by the German
search algorithm. The random search covers Ministry of Research and Technology,
a wider range to find the initial !!!implex contract no. 0328020C.
for the determination of the absolute mini-
mum. Of course a longer computing time is
the consequence. REFERENCES
[1] H. Karl, Handbuch 'ITESIM Simulations-
4. COMPONENT MODELING programm fUr zeitabhangige und zeit-
unabhangige Systeme', Institut fUr
4.1 Component Libraries and Macros Theorie der Elektrotechnik der Uni-
To facilitate the formulation of the versitat Stuttgart
simulation model a library of standard PV [2] IBM, 'Systems Application Ar~hitecture
components is implemented. The component Common User Access Basic Interface
model i tsel f is represented by the block Design Guide', 1989
function. In the library the specific para- [3] J. M. Ortega, W. C. Rheinboldt, 'Iter-
meters for the different types are stored. ative solution of nonlinear equations
. It is possible to combine common block in several variables', Academic Press
configurations into macros. To achieve a New York, 1970
well-ordered arrangment in the block dia- [4] H. Schwetlick, 'Numerische. Losung
gram the macros are represented as single nichtlinearer Gleichungen' ,
blocks. Before the simulation run the pre- Oldenbourg Verlag, 1979
processor replaces the macros by their [5] R. E. Bank, D. J. Rose, 'Global
block configuration. The macro concept fa- Approximate Newton Methods', Journal
cilitates the handling of user defined com- 'Numerische Mathematik', Springer
ponent models. Verlag, 1981
[6] J. A. NeIder, R. Mead, 'A simplex
4.2 Efforts in component modeling method for function minimization',
Since the PV system dynamics is mainly The Computer Journa.1 Vol. 7, Jan;
due to the dynamic behaviour of the storage 1965
battery, modelling this component is an im- [7] Jossen A., Bosch A., Hones H. P.,
portant task for correct simulation. There- Karl H., Lehner G., Saupe G.,
fore one object of our present work con- Zahir A., 'Battery Control Unit with
sists in the development of a battery model State of Charge Indicator', Proc. of
which describes also the ageing [7] besides the 10th European Photovoltaic Solar
the dynamic behaviour. The operating con- Energy Conference 1991
ditions in the past play an important role
for the behaviour of the battery. Another
problem is the correct determination of the
overvoltage. It depends on the state of
charge, the temperature, and the battery
current. Extensive description of these ef-
fects leads to complex model structures.
This confl icts with the intention to de-
velop a component model with few and easy
to determine parameters.
Present efforts consist in the sepa-
ration of the battery model in two parts.
First the model for the state of charge,
second the model for the overvoltage. The
description ",f the state of charge refers
to a current balance and is easy to deter-
mine. Separating the dependencies of the
overvoltage facilitates the determination
of the overvoltage parameters.

5. CONCLUSIONS

ITE-BOSS is a simulation software for the

design and optimization of photovoltaic
plants. It's main fea~ures are:
- standard user interface which is easy to
handle,
- universal modelstructure,

468
10TH EUROPEAN PHOTOYOLTA1C SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

BEHAVIOUR ANALYSIS OF EXPERIMENTAL PHOTOVOLTAIC SYSTEMS

Paes P. and Rodrigues C.

LNETI - Departamento de Energias Renovaveis
Estrada do Pa~o do Lumiar 22, 1699 LISBOA Codex - Portugal
Tel. 351-1-7582712 / Fax. 351-1-7580901

ABSTRACT. In this paper we describe two experimental

photovoltaic systems (lighting and water pumping) installed and
operating since May 1990 at the campus of LNETI-Lisbon, as well
as the monitoring system implemented for the evaluation of its
performances. The lighting appl ication is based on a classic
stand-alone system, del ivering ac energy to night-time load.
IhTormation on system behaviour includes energy quantification
of the different system components and effect of climatic
seasonal variations on system operation. As for the PV-powered
water pumping system (direct coupling dc application), a test
procedure for characterizing the whole system is proposed,
based on correlations envolving the daily amount of water
pumped, the solar irradiation and the static pumping head.

1. I NTRODUCTI ON applied in view of detailed evaluation of

the system performances, in accordance
The use of photovol tai c (PV) systems with the guideline requirements proposed
"for de central ised electricity generation by the DG XVII-C.E.C. /1/.
(stand-alone systems in remote sites) is Such monitoring method allows, in
increasing in Portugal since the last 3 addition to characterizing systems from
or 4 years. This situation is the result an input-output .standpoint:
of the government's energy policy i) to perform fine studies on the
implemented since 1982, with the different component parts;
elaboration of the "first coherent ii) to state cor-relations between
National Energetic Plan (1982 and 1984 characteristic parameters of the system
versions) which determined the increase for modeling purposes;
o"f R,D&D activities in the area of New iii) to detect unusual conditions on
Technologies, including the renewable system operation or situations generally
energy sources. The entry of Portugal in not predicted by the design methods, ·and
the EEC in 1986 and the consequent access to quantify its effects on the system
to the programs for energy incentives has performances;
definitely contribute to the development iv) to detect any faults or accidents on
of photovoltaics in the country. system components.
It is commonly accepted, today, that The results obtained so far are
independent PV systems are competitive presented and discussed. Information on
with conventional electricity generators systems behaviour includes:
in many appl ications. In particular., they i) quanti"fication o"f their performances
are a quite attractive alternative ill a (useful and dumped energy, efficiency,
country like Portugal, with a good etc.), correlating system parameters with
climatic situation and where there is a meteorological data;
real need to diversifying the energy ii) effect of the climatic seasonal
sources (Portugal imports about 70% of variations on the systems operation; .
its total "final energy consumption).
One of the photovol tai c demonstrat ion
projets currently in progress was issued 2. THE EXPERIMENTAL FACILITY
from a scienti"fic and technological
cooperation programme between Portugal The PV array is installed at the
and Germany, leading to the installation, campus of LNETI-Lisbon and comprises 17
in 1990, of a 2.2 kWp PV generator at the c-Si based modules tilted at about 50' and
campus of LNETI, a state owned facing South. Each module, with a tot~l
Laboratory. The main aim of this project area of 1.5 m2 (apperture area = 1.39 m2 ) ,
is to study the behaviour of different is composed of 144 4>10 solar cells and
experimental PV stand-alone systems (and has the following STC characteristics:
simultaneously to per"form tests on the PV Isc= 9.0 A, Voc= 21.4 V, Pmax= 130 W,
equipment available in Portugal), in '1max= 9.3%.
order to gain the necessary experience At present, only a part of the array
and competence, at a national level, is used for the mentionned appl ications
allowing to support the portuguese of lighting and water pumping.
industry and PV users in general.
In a first stage, two typical 2.1 Lighting system
applications for demonstration purposes The schematic configuration of the
were considered lighting and water system is shown in Fig. 1. It is composed
pumpi ng system. Analytical mon i tor i ng was of /2/:

469
 A 780 Wp PV generator (3 string It is based on a simple
hydraul ic
connected in parall e I , each with 2 circuit that was set 10 meter
up in a
modules in series); depht well and includes a turbine
A 5 stage battery charge regulator, flowmeter and 5 on-off valves allowing
without maximum power point tracking; extraction of water at different levels.
- 12 lead-acid battery elements ( C to = 600
Ah) 2.3 Monitoring system
- A dc/ac inverter. The configuration adopted, shown in
Fig. 3, is the one required for
, I ,
analytical monitoring.
-,o,~ JJ!IIIIjf ?BIl
I
PV
lip
ARRA Y

JJ!IIIIjf lJ
BATTERY CHARGE
-I ....... < > REGLUTOR
24V/EIlIlAh 6eIl-~a w

D
CC/llC CONVERTOR OCI AC
D CC/lVERTOA
24 VIX I 228 V",
24 Vde I 12 Veil:
31il1il W

DC LOAD
D AC LOAD
D
Data Logger ligh t ing
f"U 1 ,~ '" 31il1il W
Figure 3: Monitoring system configuration
Figure 1: Schema of the PV lighting system
An automatic data acquisition system
based on an autonomous data logger
The system was designed to supply ac
(presently supplied by the PV system
power to PL-type lamps for night-time
battery), analyses data every 10 seconds
illumination at part of the campus of
and stores average values at 10 minutes
LNETI. No back-up energy system is
intervals. This unit constantly monitors
provided. The dai ly load energy may vary the electrical data of both PV syst.ems,
in the course of the year. The load
the meteorological data (horizontal and
operation is controlled by a programmable tilted solar radiation, air temperature
timer. and wind speed), the temperature of the
PV array and the water flowrate.
2.2 Pumping system The data logger is connected to a PC
The system installed corresponds to a
computer, specially for data transfer for
commercially available configuration,
further treatment and real time
including: visualization of the measured parameters.
- A 650 Wp PV generator composed of 5 PV
The treatment of data is first
modules (similar to the ones described
performed on a daily basis (in a second
above) connected in series; stage, month ly resu Its are computed for
A floating d.c. motor / centrifugal
the lighting system). The input and
pump set (P =
450 W, max. head = 11.5 m). output power/energy/efficiency at each
The testing facility is shown
system component is calculated. Climat is
schematically in Fig. 2. characterized by the daily solar
irradiation received by the PV array and
Ts.s the average ambient temperature.
The software for data treatment also

+i
includes several graphic routines,
'ii, enabling daily plots of instantaneous
.•
7.6

i values of parameters (power, flowrate,

:T5.7 .!i• meteorological data), as shown in Fig. 4.
The visual ization of these curves
•
1)
constitutes, by itself, a quick,
! 1 practical way to analyse the systems

".~ ", ... t .!P'

.j. ~."
iI ..•
.!l behaviour, each day, allowing to detect
not only any possible fault occurred on

t
en system components, but also situations of
1 •. 9
energy dumping or loss of load .
------ .:. e C"ota r _ .. . n

Figure 2: Testing facility for the PV

pumping systems

470
 BeTore discussing oT these results, a
Meteorological data Tew comments are necessary:
60 1000 i) The main objective OT this project,
~7 .0~~1~9_0J besides being a real, useTul application,
is to gain experience in the operation of
~ so 800
E such PV stand-alone systems and, at the
'0 same time, to test the reliability oT the
of 40
~ ""'
N equipment available in Portugal. For this
600 I
E reason, during the Tirst months oT
~30
~ ~ operation, no particular strategy was
E 400 adopted concerning load proTile, except
"
>- 20 t5 that the system should be able to provide
g a daily load demand oT about 2 kWh in
"tl
0 10 200 winter, without back-up energy system.
...E That is why Eload changes in the course
oT the year.
0 a i i) Data Tor December 1990 and January
0 3 6 9 12 IS 18 21 24
lime (hr) 1991 had to be excluded due to indesired
changes on the system components
conTiguration, caused by damage oT the
PV system - lighting inverter.
iii) Edum is indirectly estimated Trom
1000 10 the correYation (illustrated in F.ig. 5)

•
,...... Psol
3 800 ~~~.:~oJ 8
oT Eph and Esol (or Hi)' considering only
the days in which all the generated PV
'-'"
energy was useTul.
""0 600 6 ,......
0
3 PV SYSTEM Lighting.
•
0 Pload
a::: 400 4 ~

-
6 ~============~~~1
200 2
'-'"
(5
IE ph = -0.03+ 0.71 • H;I
.a
0
a..
1/1
a.. ,..... 5 r = .99
0 0 >-
.c .g 4 Err = 2.570
Q. -2 00 -2 ..........
c... .c
-400 -4 3 3
0 J 6 9 12 IS 18 2I 24 .::.::
'-'"
time (hr) .c 2
a.
w
Figure 4: Plots oT daily measured data.
o~~~~~~~~~~~~~~
023 456 7 8
3. RESULTS AND DISCUSSION
Hi (kWh.m-2/day)
3.1 Lighting system Figure 5: Daily PV energy versus solar
The most signiTicant results obtained irradiation
so Tar are resumed in Table I (values Tor
each month are daily averages):
Dayl ight average ambient temperature As expected, the system produces in
Tamb CC) summer more energy than the load demand,
Dayl ight average modules temperature - or than the one required to keep the
Tmod CC) battery charge bellow the upper allowed
· Solar irradiation in the array plane limit. The excess oT energy results in
Hi = Esol/A (kWh.m- 2 /day) dumping (by the limiting eTTect oT the
· Energy del i vered by the PV generator charge regulator), penalizing the
Eph (kWh/day) eTTiciency. As winter season approaches,
· Energy to load - Eload (kWh/day) this eTTect tends to decrease (Fig. 6).
· Dumping oT PV energy - Edump (kWh/day)
Energy losses between PV generator and
load - Eloss (kWh/day)
Average PV eTTiciency - ~pv (%)
Average system eTTiciency ~sys (%)
Capacity Tactor - Fc

Table I - Results OT the PV lighting system

Month lamb lmod Hi Eph Edum Eload Eloss -=tJlh ,sys Fe

May 21.6 28.8 5.58 3.3 0.5 2.1 1.2 7.2 4.6 0.13
Jun 20.9 27.7 4.99 3.2 0.2 2.4 0.8 7.6 5.7 0.12
Jul 22.7 29.5 5.40 3.4 0.3 2.4 1.1 7.5 5.2 0.13
Aug 25.1 33.6 5.94 3.6 0.5 2.4 1.2 7.2 4.9 0.14
Sep 22.6 31.6 5.27 3.3 0.2 2.5 0.8 7.6 5.8 0.13
Oct 19.1 26.9 4.45 2.9 0.1 2.0 1.0 7.9 5.3 0.11
Nov 14.7 24.9 4.44 3.0 0.2 2.1 0.9 8.2 5.7 0.12
Feb 11.4 19.1 3.72 2.7 0.0 1.9 0.8 8.5 5.9 0.11
Mar 14.2 21.3 4.48 2.9 0.2 1.9 1.1 7.9 5.0 0.11

Figure 6: Distribution oT"'the available

PV energy

471
 This graph shows the monthly evolution PV PUMPING SYSTEM
01' E'ph (=Eph+Edump), subdivided in ".ll 30 r------------------------------.
useTul load energy, dumped energy and
lost ·energy (including losses in cables,
v = a(h) + b(h) . HI
protection diodes, charge regulator, .!!, 20
battery and inverter). E'ph is the energy ~_~*~__~~~~_b~_)~
the system would have produced iT there
were no dumping. * • ....
From the analysis 01' Table I, it
Tollows also that:
- The mean PV eTTiciency (related to the
pannel apperture area) varies seasonally
- Trom 7.2% in summer to 8.5% in winter -
, as a consequence 01' the doubl.e eTTect
01' dumping and temperature on system
operation.
This tendency is also val id Tor the 2 3 • 5 6 7 8 9 10
system overall eTTiciency(5-6%). Static head - h (m)
In average, more than 30% 01' the PV
energy generated is lost. Figure 8: Regression parameters versus
The reI iabil ity 01' the system is solar irradiance
strongly dependent on the correct
operation 01' the BOS (Balance 01' System) Even though the results obtained are
equipment. In Tact, the only problems valid only Tor this particular system,
occurred were due to damage 01' the dc/a.c this seems to be an interesting,
converter and to inaccurate regulation by relatively accurate method (computation
the battery charge controller. 01' the system output within 7% error) Tor
evaluating the perTormances 01' PV-powered
pumping systems.
3.2 Pumping system
The characterization 01' the system
described above is made by determining 3. CONCLUSIONS
the daily output (amount 01' water pumped,
in m3/day) as a Tunction 01' the solar The results obtained so Tar lead to
irradiation, Tor various static heads 01' the Tollowing preliminary conclusions:
the pump (Fig. 7). i) A detai 1 ed eval uat ion 01' PV systems
perTormances requires analytical
monitoring, as it allows to perTorm Tine
PV PUMPING SYSTEM studies on its diTTerent components.
120 ii) The lighting system operates roughly
h = 1.9 m in accordance with the design
,..... 100 predictions, the main Taults being with
h = 3.8 m
>- the BOS equipment. The monthly average PV
C
110 eTTiciency varies seasonally (Trom around
-C 7% in summer to 8.5% in winter). In
'- h = 5.7 m average, the panel eTTiciency is about
~ 60 20% below the peack characteristics. The
E
'-' h =7.6 m system ove raIl eTT i c i ency is aroud 5-6%
and the mean energy loss' between PV
'0
> h = 9.5 m generator and load may rise up to 35% 01'
the energy delivered by the panel.
20
iii) The study 01' a water pumping system
suppl ied by PV generator, using a simple
0 hydraulic test Tacility, was perTormed. A
0 2 3 • 5 6 7 S I 10
method Tor characterizing the system
Hi (kWh.m-2/day) output was applied, based on correlations
envolving only 3 parameters: the daily
amount 01' water pumped, the solar
Figure 7: Daily output vs. solar irradiation irradiation and the static pumping head.

The parameters V (m 3 /day) and Hi

(kWh .m- 2 /day) are well Titted by 1 inear REFERENCES
regression (r>.93) resulting, Tor each /1/ G. Blaesser, K. Krebs, "Guideline Tor
value 01' h (static head), an analytical Photovoltaic System Monitoring", DG
expression 01' the type: XVII/CCE, J.R.C. Ispra (1987).
/2/ P. Paes and C. Rodrigues,
V =a + b.H i "Mop i tor i.zacao e anal i se do comportamento
de sistemas Totovoltaicos experimentais",
where a and bare Tunctions 01' h. Fig. 8 V Cogresso Iberico de Energia Solar,
shows the dependence 01' these parameters Madrid, Sept. 1990.
(a,b) with h (linear Tits, in Tirst
approximation, with r=.99).
Once these correlations obtained, the
system output can be characterized by
only 2 parameters (a and b), given the
pumping head and the solar i rrad iation.
An aditional inTormation is provided by
the rat i 0 -a/b, represent i ng the
threshold irradiation Tor a given h.

472
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

PV SYSTEM PERFORMANCE AS COMPUTED WITH SYNTHETIC RADIATION DATA

R. Aguiar' and M. Collares-Pereira2

'LNETI- Departamento de Energias Renovaveis

Estrada do Pa~o do Lumiar, 1699 Lisboa Codex - Portugal
2CCE - Center for the Conservation of Energy
Estrada de Alfragide, Praceta 1. 2700 Amadora - Portugal

ABSTRACT Therefore, one must ressort to synthetic data. Synthetic radia-

Numerical simulations of a PV system with hourly radiation tion data obtained from stochastic models can emulate the probabi-
data are the most accurate method for evaluating its expected listic and sequential properties of real. observed data. And, in prin-
performance, and dimensioning its components. However, tilted ciple, data can be generated with just the information on average
hourly radiation data bases suitable for these simulations are monthly insolation - a parameter which is readily available from
almost never available. The objective of this paper is to investigate the National Meteorological Services.
the ability of synthetic radiation data (generated by' adequate The main purpose of this work it to investigate if synthetic
average regressive models and/or stochastic autoregressive models), data Can indeed substitute observed data for engineering purposes
to substitute real field radiation data. This ability seems to be - i.e. less than 5% error in the design of the PV system using
confirmed in the present work, since the results of simulations of a synthetic data, compared with using observed data.
typical PV system, with constant load. show that design errors Section 2 describes the scheme for generating synthetic data;
bellow 5% can be obtained. when comparing the results of using Section 3 is a brief description of the PV system simulation method
observed and of using synthetic radiation data. used; Section 4 shows results of simulations of a typical PV system,
with observed and with synthetic data. Finally, Section 5 presents
the conclusions and directions for further work.
1. INTRODUCTION
When the objective is to perform (very) accurate dimensioning 2. GENERATING SYNTHETIC DATA
of PV systems, detailed hourly simulations must be conduced, for
The scheme for generating synthetic radiation data consists on
many sets of design parameters. However, there are almost never a string of models - correlations, i.e. average models or autoregres-
available radiation data bases meeting all the desired conditions, sive, stochastic models - which has as input the monthly number
namely: i) hourly time scale; ii) without faults and spurious values; of sunshine hours [IJ and goes down the time scale, from monthly
iii) for every desired collector. tilt angle; iv) for each possible to daily [2J, and then, hourly values [3,4,5J. At this point the hourly
location of interest; and v) for a time long enough (say, more than radiation is transformed from the horizontal to the desired tilted
10 years). plane [6J. The whole scheme is detailed on Fig. 1.

Models Output of model/Input for next one

• Minimum instrumentation:
Campbell-Stokes heliometers Monthly no. of sunshine hours
• Angstrom-type correlations Monthly clearness index
• Markov Transition Matrix model Monthly sequence of daily clearness index
• a) deterministic model -
(average hourly kt value for each hour)
or b) stochastic model - Daily sequences of hourly clearness index
(Transient, Autoregressive, Gaussian
model - T.A.G.)
• Average diffuse fraction Corresponding diffuse fractions
(Hollands model)
• Tilted radiation model Daily sequences of hourly global l'adiation
(Perez model) in the tilted plane of the collectors

Fig. 1 - Scheme of the algorithm for generating synthetic hourly tilted data.

473
 Stochastic radiation models are develloped to generate sequen- The control simulations, done with observed daily and obser-
ces of values with the same statistical characteristics as those of ved hourly radiation data are represented in Figs. 3 to 6 as solid
sequences observed in Nature: probabilistic characteristics - avera- lines. Note that a logarithmic scale must be used for the AUX axis,
ge, variance, etc., and probability density function - and sequen- to enlarge the details of the region of high and very high solar
tial characteristics - essencially, the autocorrelation function. fractions. The region bellow, or A UX= 0.1 % load corresponds to
Therefore, the synthetic data should be able to substitute observed stand-alone PV systems.
data for simulating any solar system - including PV systems.

4. RESULTS AND DISCUSSION

3. PV SYSTEM SIMULATION PROCEDURE For the current work, four synthetic data sequences were
A typical PV system was simulated (see Fig. 2). Average considered (14 times the month of January, input <K t > = 0.44
efficiency values are taken for the various components. Details on drawn from the observed data):
the validation of the simulation procedure can be found in Ref. [7]. I. observed daily values, average model for hourly radiation
Although the load profile can be specified at the hourly time (average Kt values for each hour)
scale, for the results in this paper a constant load was assumed. II. observed daily values, stochastic model for hourly
radiation (TAG model)
III. stochastic model for daily values (MTM model), average
(DUMP) (AUX) model for hourly radiation
r ! IV. stochastic model for daily values (MTM model), stochastic
PV ARRAY --CONTROLLER -+ INVERTER -+ LOAD model for hourly radiation
Tl PV simulations for each of these types of synthetic data were
STORAGE run as explained in Section 3, and are displayed as dashed lines in
Figs. 3 to 6, respectively.
Fig. 2 - Scheme of the PV system simulated

1.000--------------,

Note that the question of the usefulness of the input data is at

0.100
stake, not the simulation procedure itself. Therefore, if a sufficiently
detailed simulation program (such as the one employed) is used, the
conclusions to be taken from the output of the simulations are not
0.010
dependent of the accurate validation of the simulation program.
The simulations were run for the month of January, in Lis-
bon, at a tilt angle of 40 0, which is roughly equal to the latitude.
14 years of hourly global horizontal data (1975-1988) were availa-
1.0 1.5 2.0 2.5 3.0
ble, which have been transformed to the tilted plane as explained in SLR
Section 2. Fig. 3 - PV simulations with synthetic data type I (observed
lO5 simulations were conduced for each type of radiation data, daily, average hourly values)
namely for every combination of the two PV system design para-
meters M and SLR. M is the nominal battery capacity times the 1.000------------,
allowed discharge depth and times the inverter efficiency, divided
by the daily load. It is essencially the storage size, in days of load;
values used were M = 0.5, 1.0, 1.5, 2.0, 4.0, and 8.0 . SLR, the 0.100
solar-to-Ioad ratio, is the monthly available radiation in the plane
of the PV array times the overall average efficiency of the system,
divided by the monthly load. It is proportional to the area of the 0.010
collectors. To construct the curves on Figs. 3 to 7, for each storage
size, 16 values of SLR were used: SLR = 0.5, 0.7, 0.8, 0.9, 0.95,
1.00, 1.05, 1.1, 1.15, 1.2, 1.35, 1.6, 1.9, 2.2, 2.6, 3.0 . The output of 0.00~.5 1.0 1.5 2.0 3.0
a simulation is, for each day, the backup energy needed (AUX), SLR
which is I-solar fraction, the dumped energy, and the current
Fig. 4 - PV simulations wit.h synthetic data type II (observed
stored energy.
daily, stochastic hourly values)

474
 However, standard software for PV dimensioning using synthe-
1.000--------------,
tic data, already could (and should) be constructed for implementa-
tion in PC computers, thus providing the non-specialists which
want to design a typical PV system with a reliable dimensioning
0.100
tool.

0.010
REFERENCES
1. G. Lewis, The utility of the Angstrom-type equation for the

o.oo~.& 1.0 1.& 2.0 2.5 3.0

estimation of global radiation. Solar Energy 43, 297-299 (1989).

SLR 2. R. Aguiar and M. Collares-Pereira, A simple procedure for the

generation of sequences of daily radiation values using Markov
transition matrices. Solar Energy 40, 269-279 (1988).
Fig. 5 - PV simulations with synthetic data type III
3. M. Collares-Pereira and A. Rabl, The average distribution of
(stochastic daily, average hourly values)
solar radiation. Solar Energy 22, 155 (1979).
4. R. Aguiar and M. Collares-Pereira, TAG: a time-dependent,
1.000------------,
autoregressive, gaussian model for generating synthetic hourly
radiation. Solar Energy (will be published during 1991).
5. K. Hollands, A derivation of the diffuse fraction's dependence in
0.100
the clearness index. Solar Energy 35, 131-136 (1985).
6. R. Perez, R. Seals and R. Stewart, Modelling irradiance on
0.010 tilted planes: a simpler version of the Perez model. In "Proc. of the
ISES Solar World Congress 1987", Hamburg, Germany (1987).
7. R. Aguiar and P. Paes, Testing PV design methods with field

1.0 1.& 2.0 2.& 3.0 data. In "Proceedings of the ISES Solar World Congress 1991",

SLR Denver, U.S.A. (1991).

Fig. 6 - PV simulations with synthetic data type IV

(stochastic daily, stochastic hourly values)

5. CONCLUSIONS
The usefulness of synthetic hourly data for the accurate
dimensioning of PV systems is investigated. Synthetic data was
generated with some of the best available stochastic daily and
hourly models, and used for simulating a typical PV system with
various combinations of the design parameters PV array area and
storage size. Cbmparison of the results obtained with observed and
with various types of synthetic data seems to show that:

- present-day synthetic data can substitute observed data for

most PV dimensioning problems;

- daily stochastic models should be improved;

- current state-of-the-art hourly models are already good

enough for PV design purposes.

Further work in this are should focus on the improvement of

daily models, and on spotting the special PV design problems which
need stochastic hourly data instead of simple average clearness in-
dex data for each hour (eg., tropical locations, PV centrals).

475
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

HYDROGEN ENERGY STORAGE FOR PHOTOVOLTAIC POWER SYSTEMS

P. S. Kauranen and P. D. Lund,

Helsinki University of Technology, Department of Technical Physics,
SF-02150 Espoo, Finland.
J. R. Leppanen,
Neste Corporate R & b~P.O.BOX 310,
SF-06101 Porvoo, Finland
D. J. Spiers,
Neste Advanced Power Systems, P.O.Box 83,
Abingdon, Oxon, OX14 2TB, UK

ABSTRACT. This paper describes the status of a three year photovoltaic

hydrogen energy system development project launched at Helsinki University
of Technology in 1990. The project is a part of the Finnish nat.ional
research program on advanced energ¥ systems (NEMO) financed by the Ministry
of Trade and Industry. The scope of the project is to demonstrate the
technical feasibility of a 100 % self sufficient photovoltaic hydrogen
energy system consisting of a PV array, an electrolyzer, hydrogen storage
and a fuel cell. When succesfully demonstrated the system concept could be
used in remote power applications and advanced low energy houses. Component
tests for an experimental system of 1 kWh daily load are being carried out
and mathematical models have been developed for approximate system sizing
and detailed system analysis. Finally, the economic potential of the concept
is discussed.

1. INTRODUCTION
Seasonal energy storage is a necessity
for wide spread solar energy utilization at
high latitudes. Hydrogen offers a promising
alternative for seasonal storage of solar
photovoltaic (PV) electricity. The
objective of the study is to demonstrate
the technical feasibility of a 100 % self
sufficient PV hydrogen energy system in the
northern Finnish climate. The experiences
of the project are used for more exact
analysis of the economical feasibility of
the concept and possible future component
development. In the future, mass produced
PV hydrogen energy systems could compete -Ol<cflicily
with PV battery and PV diesel hybrid - waltll

systems in remote power applications and in - °2

- HZ
advanced low energy houses.

2. SYSTEM CONCEPT Figure 1. System configuration.

In a PV hydrogen stand alone energy
system excess DC electricity from a PV 3. EXPERIMENTAL
array in the summer is converted into
hydrogen for winter use. The key components In the experimental part of the
of the system are a PV array, an project, the key components of the system
electrolyzer, a custom made steel vessel are being tested and the component
for pressurized hydrogen storage and a fuel interfacing is studied. A 600 W alkaline
cell. As the round trip efficiency of the electrolyzer operating at an elevated
hydrogen storage is considerably lower, pressure of 28 bar and a 225 W solid
typically 30 - 40 %, than the efficiency of polymer (SPE) fuel cell were selected for
a lead-acid battery and as the electrolyzer the first component tests.
and the fuel cell should be preferably The PV-electrolyzer and PV-battery-
operated for long periods for efficient electrolyzer interfaces are being studied
operation, hydrogen is used only for using a computer controlled DC power source
seasonal storage and a battery is used as a as a PV array emulator and the battery
short term buffer storage. Additional charging characteristics of the fuel cell
components which are needed depending on have been measured.
the system design comprise control The experimental data has been used as
electronics, feed water treatment, water input for the numerical models developed
circulation, hydrogen drying, purification for the system analysis.
and compression as well as oxygen drying, The complete system is scheduled to be
purification, compression and storage. A operational in late 1992.
simplified system configuration is shown in
Fig 1.

476
4. MATHEMATICAL MODELLING where n is the amount of hydrogen in the
storage (Nm J ) and PH' CH and elIH are the
4.1 Approximate system sizing rates of hydrogen produced in the
A stand alone PV hydrogen energy electrolyzer, consumed in the fuel cell and
system is characterized by the PV array leaked through the storage walls (Nm /s),
3

size (Wp ) and the hydrogen storage capacity respectively.

(kWh, Nm 3 , H2 ) which are functions of the PH and CH are proportional to the
levels and annual distributions of solar currents passing the electrolyzer and the
insolation and the load as well as the fuel cell, respectively. ~ depends on the
efficiencies of the system components. pressure and material properties of the
A straight forward method for storage and is almost neglible in most
approximate sizing of the PV array and the cases /2/.
hydrogen storage capacity is a spreadsheet For other components of the system,
calculation using monthly distributions of e.g. PV array, battery, the original models
solar insolation and the load as well as of PHOTO are used 13/.
constant efficiencies for the battery If the voltage levels of the
(Tlbat) , the electrolyzer (Tl.,.) , the gas components are carefully matched to each
storage (Tlstor) and the fuel cell (Tlto). A other the system can by simply controlled
conservative estimate is that all the PV by switching the components on and off
and fuel cell production is fed to the according to the voltage and the state of
battery and thus the load is fed directly the charge (SOC) of the battery /2/, see
through the battery (short term storage, switches in Fig. 1. The basic idea is that
Tl.t) and through the hydrogen storage the electrolyzer is connected when the
(seasonal storage, Tl •••• on.') having the battery is fully charged and the fuel cell
following efficiencies: is connected when the battery voltage or
charge drops under a certain threshold
Tl.t = Tlb.t 2 (la) limit. If the limits are carefully chosen
11seasonal := 1lele11stor1'lfcT'lbat (lb) the electrolyzer is utilized only in the
summer and the fuel cell only in the
The electrolyzer has to be at least winter.
big enough to produce the required amount
of hydrogen under continious operation 5. SIMULATION EXAMPLE
during the summer and the fuel cell has to
be able to cover the avarage load even if As a case example, one year energy
the losses in the battery are taken into balance of a system for 1 kWh daily load in
consideration. However, the optimal sizing the climate of Helsinki, Finland, 60 "N, is
of the electrolyzer, the fuel cell and the simulated USing one hour time step and
battery is a complex problem which depends Helsinki Test Meteorological Year (TMY)
on the component properties and the choice weather data as input /5/. The system
of the system control strategy. configuration is shown in Table. 1. The
electrolyzer and the fuel cell are the
4.2 Numerical simulation components being tested. However, the
Accurate electrochemical and thermal number of series connected cells in the
models of an electrolyzer and a fuel cell fuel cell has been increased in order to
/1/ as well as a model for a pressurized match it to the system voltage level. The
gas storage /2/ have been derived and size of the battery was selected to ensure
incorporated in a program package PHOTO sufficiently long operating periods for
/3,4,5/ which is designed for the both the electrolyzer and the fuel cell.
simulation and analyses of PV and hybrid
power systems. Table 1. Simulated system.
The following equations are used to
describe the electrochemical and thermal
behaviour of an alectrolyzer or a fuel cell Location: Helsinki, Finland, 60 oN
Ill: Solar insolation: 930 kWh/m 2 /a, (horizontal)
PV-array: 750 Wp , (2*8 panels)
U = Ns[a o + a,T + bln(I) + (ro/T) I) (2) Battery: 24 V, 500 Ah,
(5 kWh effective capacity)
Ceff(~T/~t) = N.I (U-U o) + Qe. - K(T-T.mb ) (3) Hydrogen storage: 3.6 m3 , 28 bar, 360 kWh
Electrolyzer: 24 V, 25 A, 600 W
where ao,a"b,ro,C.tf and K are constants, U, Fuel cell: 28 v, 15 A, 420 W
I and T are the voltage (V), the current Load: 1 kWhlday, constant
(A) and the temperature (aC) of the
electrolyzer or the fuel cell, N. is the
number of series connected electrolysis or The simulated annual energy balance
fuel cells to form the electrochemical and average component efficiencies are
device, ~t is the length of a time step shown ln Table 2. The energy balance
(s), ~T is the change of temperature over obtained by the spreadsheet sizing method
the time step (OC) , Uo is the and nominal component efficiencies are
electroneutral (open circuit) voltage of shown for comparision. It can be noted that
the electrolyzer or the fuel cell (V) the spreadsheet calculation gives a
and Qex is the power of external heating or slightly large PV array size because it is
cooling (W). not taken into consideration that part of
The gas storage is modelled by a the PV production is fed directly to the
simple differential equation /2/: load or the electrolyzer and part of the
fuel cell production is fed directly to the
dn/dt = PH - CH - elIH (4) load. For the same reason, somewhat less
hydrogen is produced in the electrolyzer
according to the spreadsheet calculation
than in the hourly simulation. However, the
differences are small and the spreadsheet

477
calculation is shown to give a good first
approximation for the system sizing.
A closer look at the monthly energy
balance, Fig 2., shows that the kWhimomh
electrolyzer is operating from February to 120
October and the fuel cell from October to
March. The overlapping operation in
October, February and March could be
avoided by a more accuratE! cont.rol of' the
system.
Finally, the daily energy balance in
the beginning of June, Fig. 3., shows the
buffer operation of the battery. The
battery is first fully charged and then
discharged to the electrolyzer. The total
cycle is approximately 3 days in June and
becomes longer as the insolation levels
become l~wer.

Table 2. Simulated and calculated energy

balance (E, kWh) of a PV hydrogen energy
system and simulated avarage and nominal
component efficiencies (~,%).
- Epv - - Eele

simulated spreadsheet Figure 2. Simulated monthly energy balance.

Location Helsinki, Finland 60 oN.
Epv 658 688
E e1e 337 317 Whlday
E stor 270 kWh 25.4 kWh
8000
E fc 127 127
E 10ad 365 365
~el. 80 80
~bat 82 % 80 %
~fc 51 50
l1stor 99 100

6. ECONOMIC CONSIDERATIONS

A PV hydrogen energy system should be

able to compete with both PV battery and PV
diesel hybrid energy systems. The potential
cost reductions over a PV battery system
arise from the reduced sizes of the PV
array and the battery and over the PV
diesel hybrid system from the fuel and
maintenance costs.
To study the effects of the component,
fuel and maintenance costs, a comparison
was made between these systems in the Figure 3. Daily energy balance in the
location of Helsinki, 60 oN, for loads up beginning of June showing the battery
to 4 kWh/day /6/. operation (cha: charged, dis: discharged).
As small electrolyzers and fuel cells
are at a prototype stage of development, a
cost estimate based on manufacturers'
target costs was made for a future mass
produced hydrogen storage system. The 3.5
effect of possible anvances in battery
"
.~
C. 3
technology were taken into consideration, E
as well.
in" 2.5
The systems compared were:
'"'"
1. Conventional PV battery system , "
.2
"0
battery allowed to discharge 30 % ~
over the winter for seasonal
storage.
2. Improved PV battery system,
battery allowed to discharge 60 %.
3. Conventional PV diesel hybrid
system.
2
4. Prototype PV hydrogen system,
hydrogen storage efficiency 36 %. LOAD [kWh/day]
5. Mass produced PV hydrogen system, *' PV+bottery (OOO(max)=30%. C10) - - PV-Hydroc;len prototyPl! sy!tem
efficiency 36 %. ..... PV+battery (OOO(mox)-60%. C10) -+- PV-Hydrogt!n moss.prod. sysh!m
+ PV+bottery+dieset

The first year costs (investment) of

the systems are shown in Figs. 4. and 5. Figure 4. Comparison of the PV hydrogen
The costs are, for all but the smallest energy system with PV battery and PV diesel
systems, lowest for the PV diesel system. systems. Location Helsinki, Finland 60 oN.

478
 PV hydrogen systems are shown to be REFERENCES
competive with PV battery systems for loads
larger than 1 kWh/d even at today's /1/ J. Vanhanen. Mathematical
prototype prices. Modelling of an Electrolyzer.
The estimated total costs over a 30 Report TKK-F-Cl19. Helsinki
year lifetime with 0 % interest rate are University of Technology, Dept.
shown in Fig. 6. The figure shows that Technical Physics 1990.
there is potential for a mass produced PV
hydrogen system to compete with the PV /2/ E. Holttinen. Dimensioning of a
diesel system if the hydrogen system can be Seasonal Storage for PV Power
designed to operate reliably enough to make Systems. Master's thesis. Helsinki
savings in the maintenance and fuelling. University of Technology. Dept.
However, more reliable information of the Technical Physics 1990. (In
costs and performance of the hydrogen Finnish) .
storage components are needed to make more
exact conclusions of the real market /3/ L. M. Manninen et. al. PHOTO - A
potential. computer simulation program for
photovoltaic and hybrid energy
systems. Document and user's
guide. Report TKK-F-A670. Helsinki
University of Technology. Dept.
Technical Physics 1990.
..
u
/4/ L. M. Manninen and P. D. Lund .
.~ Design Tool PHOTO for sizing of
E Hybrid Power Systems: Program
... Verification. Proc. 10th European
in
>- Photovoltaic Energy Conference and
...
II>
Exhibition. Lisbon 1991. (Poster
"~ 3C.07) .
o0;
0:: /5/ J. P. Vanhanen. Simulation of
Solar Hydrogen Energy Systems.
Helsinki University of Technology,
Dept. Technical Physics. To be
published.

Figure 5. First year investment costs of PV /6/ J. Leppanen et. al. Energy Storage
hydrogen, PV battery and PV diesel energy for Photovoltaic Power Systems.
systems. Location Helsinki, load 1 kWh/day . Neste Corporate R&D, 1991.

...
OJ
";:
Co

E
...
in
.>-
II>

"~
o
0;
0::

000_
PV- botl~'y
XI I:
">"'''"''
60'I.
PV-
diesel ..-
H2 -

')'H
systtm,
- .
,.,w.

Figure 6. Total costs over a 30 year

lifetime of PV hydrogen, PV battery and PV
diesel energy systems. Location Helsinki,
load 1 kWh/day, interest rate 0 %.

479
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ANALYSIS OF PHOTOVOLTAIC ENERGY POSSIBLE APPLICATIONS FOR

SUPPLYING DRINKING WATER TO RURAL ZONES
E. MICHEL
COSTIC (Comite Scientifique et Technique
des Industries Climatiques)
Rue Lavoisier - Z.I. St-Christophe
F - 04000 DIGNE LES BAINS

ABSTRACT. For satisfying the drinking water supply in isolated

sites, the contracting firms are often facing difficult problems
of power supply. This might concern, for example, supply of
electroval ves, dosing pump systems for spring water chlori-
nation, measurement system of water levels in the water-towers,
remote control for filling up the water-tower ... Requirements
bound to remoteness, difficulties regarding access and environ-
ment may render excessive the connexion costs to the power
distributing main grid. The conventionnal substitution solutions
(electric auxiliary generating sets, batteries, cells, ... )
impose, as far as they are concerned, important costs for opera-
tion, staff and fuel. A solution to these compulsory conditions
is quite often to use photovoltaic generators. The surveys have
enabled to index more than fifty installations in France proper
and in the French Overseas Departments. We have selected seven
application types where photovoltaic energy may be indicated.
The applications concern the following fields pumping,
electrovalve, chlorination, booster compressor, purification
station, flow rate measurement, remote measurement. Every
application has been the object of writing up a file. In every
file the application is shortly described. Power as well as
energy requirements are afterwards determined. This study has
been funded by the French Ministry of Agriculture.

1. MAIN RESULTS OF OUR SURVEY 2. EXAMPLE OF AN EXISTING INSTALLATION

Among more than fifty installations Department CORSICA
in France proper and in the French Overseas System owner Syndicat d'Electri-
Departments, seven application types of fication Rurale
photovoltaic energy to the supply of Sud de la Corse
drinking water appeared at the end of our 2, cours Napoleon
survey. 20000 AJACCIO
They are put together in the table Contractor TOTAL ENERGIE
below. DEVELOPPEMENT
7, chemin du plateau
INSTALLATION PHOTOVOLTAIC APPROXIMATE 69570 DARDILLY
TYPE NUMBER PEAK POWER COST IN
FF (TNE)
Study carried out by : TOTAL ENERGIE
DEVELOPPEMENT
Pumping of 7 in Metro- 40 to 3840 Wp 13 000 to
drinking water litan France 552 000 Object of installation : water pump~ng for
16 in Guyana
1 in Guade-
cattle breeding. The project a1ms at
loupe demonstrating how a solar pump can solve
1 ln The
Reunion
the problem of water supply for cattle
10 in breeding in a mountainous area.
southern
pacific
Overall cost 552 000,00 F
Financing regional fund of energy
Dosing pump 12 40 to 160 Wp 6 500 to mastering
for water 43 000
chlorination 239 200,00 F (43 %)
european community
Rotative strai 2 50 to 160 Wp 25 000 and 220 800,00 F (40 %)
ner for purifi 55 000 SER of Southern Corsica
cation station
60 000,00 F (11 %)
Pumping set 1 common at user
for purifica- purification 32 000,00 F (11 %)
tion station station
Expressed needs pumping of 20[m3 /day]
Valve controls 4 50 and 200 Wp 15 000 and at a height of 80 [m]
40 000
that is to say 6 000
Water level > 10 20 to 66 Wp 7 000 to
[Wh/day]
transmission 20 000 Installed peak power: 3 840 [Wp]

Flow rate 2 30 wp a 500 Advantages of photovoltaic application :

measurement pumping is more reliable and maintenance
simpler than with a DIESEL set. Connection
to EDF grid was not to be contemplated
because of its remoteness.

480
The Zigliara installation is outlined Energy :
below.
The dosing pump ADY is rated with a power
of 50 [W]. For 120 fifteen seconds pulses
a day the necessary energy will amount to
25 [Wh/d].
Use of the electrovalve will require very
little energy, a few [Wh] per day.
Manufacturers : C.I.R. - ADY
Final set-up contractor
PAVELEC - S.D.E.T. (Lyon)
Sizing :
Sizing with 25 [Wh/d]
Example : installation in St-Malon
sur Mel (35)
Installed peak power 40 [Wp]
voltage 12 [V]
storage 108 Ah (1,296 Wh)
Fig.l Water pumping for cattle breeding in Corsica
Cost
The photovoltaic modules are of the
PHOTOWATT (40 [wp]) trade-mark. A dosing pump of ADY make with ist soda
The DC/AC converter was designed by hypochlorite vat comes to about 7 000 FF.
Total Energy Development. It makes the The cost of the photovoltaic installation
output current frequency vary for adjusting is in the range of 4 000 FF at 6 000 FF
the energy required by the pump to the according to the area.
energy generated by the solar panels.
The pump is a multicellular one,
bearing Leroy-Somer trade-mark. Its rated Installation type
capacity is 2,2 [kw].
TED has put on this installation a
measurement system provided for several
parameters. Since June 1988, the
measurements are effected automatically and
the data are transmitted via a satellite.
It was thus possible to observe, for
example, that the energy supplied to the
pumped water was at best one third of the
energy delivered by the solar collectors.
This
service
installation was
in March 1987 and has
put into
been
! I
l
~
3-
-----'
@~ ~~'eedbOCk mech-;;-nis;;;r - , --.,.----
operating without any problem since then.
3. APPLICATION TYPES ,
,
, I! "~ ,

1 riow Toto
3.1 Different applications mot...
j
I
Seven application types have been
selected : dosinq pump or elec rovol'wle
- pumping
- electrovalve
- chlorination Fig.2 Chlorination installation type
- booster compressor
- purification station
- flow rate measurement
- remote measurement 4. AREA OF PROFITABILITY FOR PHOTOVOLTAIC
POWER COMPARED WITH EDF GRID
3.2 Example of the chlorination
1. The peak power is calculed on the basis
Presentation : of 150 FF/Wp. The extremes go from = 110
FF/Wp (factory price) to 180 FF/Wp (maximum
Drinkability of pumping water often passes price including installation).
through chlorination. It is effected by 2. The available energy is calculated on
dosing soda hypochlorite or by the basis of radiation, amounting to 3,6
hydroejection of gazeous chlorine. The [KWh/d.m2 ] (Southern France in December).
first application goes through the use of a Energy is calculated by the formula :
dosing pump, the second one through the Ed = 0,6 x Ei x Pp.
presence of an electrovalve.
In both cases, energy needs are small. The
photovoltaic solution may prove profitable
for numerous isolated installations.

481
 /
Capital coat
(KF)
CD Wp @Wn/ d
1500
JOOO

/
210
t
I
I
leo 1000 I
2000 I

j~m.'.
110 700 1500

1
/
60 400 900
800
50
700
JOO
600
40
Area of
profitability 500
JO 200
400

20 JOO
16.55
100 200
10 70 Chlomatlon
100
J.8 tElaClTo""l""
*
50
(
I ) Romot.
m.... romfllt
30 100 200 300 400 500 1000 ,~ 2000 2~ Distance
to grid (m)

Fig.3 Area of profitability for photovoltaic power compared compared with EDF GRID

5. CONCLUSION
Indeed, some economic studies,
conducted according to the case implied,
show that photovoltaic generators may be
definitely competitive in terms of capital
cost as well as of operating cost. It
appears, therefore, of importance to
analyze this technology as a potential
solution for the power supply of isolated
equipment destined to furnish drinking
water. The cost of the generated 'Watt II
being relatively high, it is important to
limit the power of the installed equipment
and so to wa-tch with care the consumption
of the terminal units. in fact, from the
energy viewpoint, the choice of some
efficient equipment enables to render
completely profitable the photovoltaic
solution where as it might have appeared
impossible in a first approach.
In a great number of cases for
drinking water supply in rural zones, the
electric power needs are very small, and
using photovoltaic energy seems quite
indicated.
It is especially the case for spring
chlorination systems, safety valve control
or flow rate measurements.
Other applications more questionable
in Europe, as for example, pumping and
purification stations may absolutely be
adapted for Developing Countries.

482
HITH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &--12 APRIL 1991 LISBON, PORTUGAL

PV SYSTEM WITHOUT CHARGE CONTROLLER FOR GRID-CONNECTED AND STAND-ALONE OPERATION

D. Reismayr
Ludwig-Bolkow-Systemtechnik GmbH
DaimlerstraBe 15, D-W8012 Ottobrunn/MUnchen

ABSTRACT. The design conditions, especially a possible

solar energy surplus, leading to the idea to omit the
charge controller are presented. The operation of the
system with the grid acting as a variable load for
managing the batterstatus is explained and as a result of
the first measurements some proposals for further
improvements are pointed out.

1. INTRODUCTION The advantage of grid-connected photovoltaic

systems, normally without any battery, is to be able
1.1 Description of the project to utilize the maximum amount of energy producible
The presented photovoltaic system referring to by the solar generator by means of a maximum-power-
a pig-keeping installation is part of a project for point tracking device.
the application of photovoltaic technologies to A combination of both versions would allow the
agriculture being set up within the Agricultural utilization of their advantages, i.e. this yields
Schools of Central Franconia at Triesdorf Germany, a system generally working in stand-alone operation
and consisting of 5 small plants in the range of 2 (on principle capable to be changed into a
to 6 kW. completely autonomous system) and when there is a
These plants together with an overall measurement surplus or deficit of energy to be switched to the
system are being installed besides the above mains.
mentioned one for cattle-keeping, domestic science
school, agricultural machinery school and fish- 2.2 Problem
farming. The rated values of the photovoltaic plant Because of minimazing costs no shunt-type
for the pig-keeping installation are: charge controller working at a frequency of some
solar generator 6/8 kWp , summer/winter kilohertz has been chosen but a controlle.r shunting
battery voltage 120 V Le. switching off several strings of the solar
battery capacity 200/360 Ah, summer/winter panel when the battery voltage has reached a certain
inverter power 6 kW level. This means the problem would be not to be
autonomy about 2 days at a depth of discharge of 50% capable to deliver a surplus of energy to the
at 0° Centigrade inverter.
consumption 4,2/6,6 kWh; summer/winter
2.3 Solution
1.2 Design conditions for the partial project From that problem arose the idea to omit the
The design had to be made by finding a charge controller and to control the battery status
compatibility of almost inconsistent conditions. On by the inverter itself.
the one hand the plant should be suitable for stand-
Schemallc Diagram 01 the PV - System without charge controller
alone operation relatively far away from the village

..... --..... ,.... ..........---

on the fields because of the bad smell nuisance, of
course at minimum cost, especially for the storage

r-----l-r---"T"""l-----it>
"....,......_-,.--
battery. On the other hand the aim is to exploit to ~...-..-

full advantage all the surplus of energy produced by

the solar generator when there is no load and at the
same time the battery fully charged. This may happen
in summer but also during bright winter days. The
latter condition would lead to the necessity to use
this surplus either for e.g. additional water
heating in a real stand-alone system or to deliver

--
it to the grid. But that means to get back
temporarily- to a grid-connected operation.
... -
2. APPROACH
Figure 1: Schematic Diagram of the PV-System
2.1 Stand-alone and grid-connected operation without charge controller
A real stand-alone system needs a charge
controller for charging the battery e.g. according
to a constant current - constant voltage strategy or Fig. 1 shows the schematic diagram of the
at least a device for protecting the battery against circuit. The function of the charge controller is
over-charging and deep discharging. adopted by the inverter and the mains and/or a
The disadvantage of such a system is not to controlled variable resistance.
utilize the solar energy when there is the battery For a completely autonomous system, of course,
charged and at the same time is no need of energy. the version using the resistance is necessary.
Another disadvantage will be a noncomplete autonomy The grid (via the inverter) or the resistance
when there no large annual storage battery is will act as a controlled additional load for
provided. managing - the battery charge or in other words: that

483
is also identical to the utilization of the surplus Furthermore energy or charge balances, i. e.
of energy. current measurements (panels, battery, inverter
The controlling is done by switching the input), and variations of the threshold voltages
inverter in addition to its stand-alone load to the have to be made.
mains and back to the stand-alone operation between Due to time delay of the whole agricultural
two threshold voltages of the battery, Le. the project some measurements have been done only since
inverter input voltage. February of this year. Systematical registrations
An example of the behaviour of the battery will take place in May 1991 after setting in
voltage control at an almost constant value of rated operation of the pick-up systems.
capacity of 80% and different loads is shown in fig. But the informations obtained until now give
2. first impulses for further improvements.

U"" threshold voltage 3. RESULTS

Uo battery no load voltage
I; inverter input cunent
Il load QJrrent (as pan of 3.1 Battery behaviour
inverter input cun.)
IG ,;Jrid curr~nt (as part of The U-I-battery curves have been determined
Inverter Input curr.)
I", battery charge current out of voltage versus time lines at constant
currents (see fig. 3).
Operation point!.:

stand alone
grid connected '····l u +curves
'····f Iatlery.,.,.:lOcelllof40Pz5200.aooAh
....jPar......: eonmnt percentapof,...ct a , .
•• j

150
i

Uthrl

U""2

80%

.. ..
IG
,.
" "
Figure 3: U-I-battery curves at constant
percentage of rated capacity
li1 Ii)
90-i-----r----~~--,_--~r_--_,----_r--~
_1-
These curves are to be considered as to be
10 20 30 A rather static ones. Dynamic curves generally have a
slighter inclination within the charge-discharge-
System behaviour (changing operation points) oholar generator, battery range close to the open circuit battery voltage. The
and converter at fast inueasing load current ILl .Iu and almost constant dynamic behaviour at different frequencies of load
battery capacity of 80% of rated value variations should be subject of further
investigations.
WT"rnOI90I02,'
An example of the behaviour during a winter
Figure 2: Behaviour of the battery voltage control day is shown in fig. 3, i.e. the arrowed line from
at different loads point A to E. The corresponding current distribution
(panels, battery, inverter input) and insolation is
The first step of the consideration is the seen from fig. 4 and 5.
operation point Ao without any control and the load CImnt_rv_ _
current ILl (transformed to the input side of the en---t7.2.11t1)

inverter). After switching on the control the

battery voltage will drop (opertion point Ad and
the grid current 10 will be delivered to the grid. A
sudden increase of the load current III to I., will , /
T '"
cause a further drop of the voltage to the operation 'V
point A,. Then the grid connection will be released
and the stand-alone operation point A, will be ., \
reached. The rest currents from the operation points \
\ /
---+
to the solar generator U-I-curve are the charging

...
currents.
,
In that case of control for safety reasons the
inverter has to be sustainable to a relatively wide
range of input voltages (up to the open circuit
Azm.dI:8ouIh
EIwabI: 50 •
11:$1
...
voltage of the solar generator).
Figure 4: Current distribution PV-System Pig-
2.4 Measurements keeping on February 7th, 1991
For a further prediction and for an
optimisation of the behaviour of that system the U- Between 13.40 hours and 14.55 hours the panels have
I-curves of the battery at constant percentages of been switched off for simulation of darkness.
the rated capacity have to be determined. In
addition to that it is necessary to register the
acid density (volumetric weight) for observing the
charge in the battery.

484
 Insolalion PV - System At the grid voltage of UN = 220 V the constant
Pig - Keepi ns load Pv of 1 kW is fed by the inverter and the mains
in the range of the inverter output voltage Uw from
217,8 V up to 222,2 V. This point is the ideal
stand-alone operation . At higher voltages energy is

,,--r- ...........
delivered to grid •

--' ~ Another investigation necessary for further

optimization of the' system would be battery

1 I '-." measurements for testing the cell behaviour

(voltage, temperature) on different frequencies of

-...
1 '" charge - discharge cycles (see above). It may also
J become useful to enlarge the battery capacity of the
system design.

4. CONCLUSION

14:00 14.311 16'" Irr-a The proposed PV-system without charge-

controller and with the charging management done by
delivering solar surplus energy to the grid is just
being investigated. For further optimization of the
Figure 5: Corresponding insolation to figure 3 and system it is necessary to make more measurements
4 on February 7th, 1991 concerning the battery behaviour on various
conditions. Additional support of the charge control
Sometimes a continuous discharge of the by e.g. the battery current, the acid volumetric
battery haa been observed perhaps caused by a too weight or the load power seems to be useful either
short charging time, Le . a too low total charge in a simple way with a switch to the mains as done
during some days, but to find out the reason would now or in a more sophisticated manner using a
be subject of the optimization period to come. continuously working AC-power distribution
Such an unbalanced charge situation may also controlled by the inverter output voltage.
have been the case on February 7th (see fig. 4).

3.2 Proposals for improvements 5• ACKNOWLEDGEMENT

The first findings of the few observations
just having made prove some proposals for further The author wishes t.o give many 'thanks to Mr.
improvements to be implemented in the near future. F.X. Mittermaier, Mr. G. Hess, Miss S. Wallis and
The next one would be to increase the upper Mr. Chr. Heinlein for interesting discussions.
threshold voltage (for switching to the grid) from
2,3 V per battery cell up to 2,4 V.
Furthermore the sequence of the priorities of
the operations might be used as follows.
1. only charging (at no load)
2. stand-alone load
3. grid-connection load
In addition to that further quantities such as
the acid volumetric weight, the battery current and
the load power might be used besides the battery
voltage for supporting the charging management. But
in that case the simplicity of the system proposed
formerly would be left.
The use of more qua~tities for the control of
the battery status would give the opportunity to
control the inverter output voltage in dependence of
the battery status in such a way that at constant
load the inverter output is changed continously
without any break to the mains.
An example is shown in fig. 6.

Power distribution versus Inverter output voltage

'.

--_....

- ....
Figure 6: Power distribution versus i nverter
output voltage

485
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Methodological Approaches for the Estimation of a

Capacity Credit of a Grid-connected Photovoltaic
Electricity Generation
Martin Kaltschmitt. Alfred Voss
Institute of Energy Economics and the Rational Use of Energy
University of Stuttgart
Pfaffenwaldring 31. Tel.: (int.-49) 711/685-7575
D-7000 Stuttgart 80. Federal Republic of Germany
Abstract
Renewable energy sources are getting more and more important. Therefore, studies of an integration
of these options in the existing electricity supply system are neccessary. On this background it is the
aim of this paper to present and discuss different methodological approaches for the estimation of a
capacity credit of a grid connected photovoltaic power generation system. First only the solar energy
supply is considered. Then a mixed conventional -renewable power plant system is investigated.
Afterwards the interrelation between the demand-side behaviour of the electricity consumption and
the photovoltaic power generation is analyzed. Later on the given solar energy supply during annual
peak load periods observed during the last 25 years is investigated.

1 Introduction Thl'rcfore, a temporally and geographically high resoluted si-

mulation of the photovoltaic power generation is performed
Current discussions on energy policy are more and more do- based on measured meteorological data (d. [2]). The utiliza-
minated by issues dealing with renewable energy sources. A tion strategy for the renewable energy is characterized by a
major reason for this is the debate about the effects of 00 2 - total of 600 MWp power installed in photovoltaic modules
emissions on global climate. Thus. the utilization of renewable [2]. In order to verify and relativize the different approaches
energy sources especially for electrical power generation has applied for one particular year the climatic conditions given
been investigated in numerous projects. On this background, during yearly peak load periods in the last 25 years are finally
the determination of the share that renewable energies can examined.
contribute to the "dependable power supply" of a power plant
system is increasingly important.
The "dependable power supply" of a power plant system 2 Solar Energy Supply
is defined as the available power output that can be generated
Due to the high energy density and the simplicity to store
by the system with a high probability (typically approx. 97
fossil fuels, primary energy is available with a high proba-
%) [1]. If renewable energy systems are considered. the strong
variations of their energy source (e. g. solar beam radiation bility in a conventional power plant system. This does not
variations on weather conditions with moving clouds) needs hold for solar energy respectively the solar radiation. Here,
to be accounted for. Since the "dependable power supply" for wide variations of the energy supply occur as a result of pe-
riodical - day/night and summer/winter - fluctuations and
a mixed conventional-renewable system should be the same as
for a conventional park, the installed photovoltaic power has random meteorological alternations. Thus, for renewable en-
to be much larger than the substituted conventional power. ergy systems it is not only necessary to account for a system
The "capacity credit" of the installed photovoltaic power is reliability probability but also for the unavailability of the
then given by its portion that contributes to the "dependable energy source when the availability of the generated power is
power" output of the mixed power generation system. studied.
The output power of a tilted (30' tilt angle) southfacing
It is the aim of this paper to compare different approaches
photovoltaic surface of one square meter is shown in Fig. 1.
for determining the capacity effect of a photovoltaic power ge-
The power output is divided into two parts. The stochastic
neration. At first, the photovoltaic power production is stu-
share represents the power generation potential where the
died separately and the share of its power output that is
actual power output depends on the alternating cloudiness.
available with a high probability (97 %) is computed for lo-
The deterministic share of the output power is created only by
cations in South Germany. Secondly, the complete power ge-
diffuse radiation which is available during the whole daytime.
neration system - a combined conventional-renewable power
The actual power output which is a function of random
plant system - is analyzed. The share of the installed renew-
meteorological conditions is usually higher than the determi-
able power that can be used to replace conventional power
nistic power output. Nevertheless, the reliable power output
under the same reliability conditions is determined. A third
of a photovoltaic system (reliability probaility 97 %) is of-
approach studies the relationship between a photovoltaic po-
ten looked upon as not useable throughout the year. Reasons
wer generation and the current demand-side behavior in the
for this are the neglectible power output during night hours
public electricity grid. Each approach is discussed and then
and the serious alternations during the day. Another reason
applied to the conditions found during a particular year in
is the possibility of snow covering the modules at locations
the south-western state of Germany (Baden-Wiirttemberg).
in Central Europe. In these cases, there is no power output
resp. electricity generation at all. Then during the daily peak
load hours aroun,d noon in South Germany [3] even the pho-
tovoltaic power production caused by diffuse radiation is not

486
 wlnterday summerday
.. 80
E stochastic
:i: share
~ 60
oS
a
a.
40
::I
III

e[ 20
o
t~ll2~sAlIl.=
4 8 12 16 20
detennlnlstlc
share

5 9 13 17 21
MET·tlme In h MESt·tlme In h

-
lowest technical ... 01 supply highest technical
po"lble supply
"-- 4---
during 1 _ days
....... ---
possible supply

Figure 1: Time behavior and the deterministic resp. stochastic share of the the photovoltaic power
generation during a winter and summer day

given. wer that contributes to the dependable power output of the

In Central Europe however, it is not too likely that the conventional-renewable power plant system.
power output of all photocoltaic modules distributed in a lar- The non-availability of conventional power stations in a
ger geographical area will cease during daylight hours because power plant system can be described by a model that uses
of meteorological conditions. Hence, a minimum photovoltaic two states (state one: power plant supplies its installed po-
power generation during the day of at least the magnitude wer with the corresponding availability probability; state two:
observed at the winter solstice during complete covered s- power plant does not supply power with the corresponding
ky (deterministic share of the photovoltaic power generation unavialability probability). Because of simple and safe fuel
caused only by diffuse radiation; cf. Fig. 1) can be assumed storage it is generally not accounted for non-availability of
throughout the year. The corresponding portion oCthe instal- the used primary energy. This mathematical model, howe-
led photovoltaic power that is available at these worst meteo- ver, is not capable of describing the fluctuating photovoltaic
rological conditions depends on the technology of modules power generation systems. The reason is that the prerequisite
and DC/ AC converters. Between 11 am and 1 pm (Middle of the two state model - timely and spatial independence of
European Time; MET) this fraction equals approximately 0.4 the availability probability - is not given for the solar energy
to 0.7 % of the installed DC power. Other studies [4] find a source.
minimally available share of 1.6 % of the installed power. In order to determine the availability of a combined
These considerations, however, do not account for the se- conventional-renewable power plant system the unavailabi-
vere periodical alternation of the solar flux during the ye- lity probability function of the renewable energy supply can
ar. The minimum available power for each month during the be calculated first. Secondly, a convolution of the cumula-,
noon hour between 12 am and 1 pm changes from 2.5 % in tive unavailability function of the renewable energy supply
December to 21.2 % in June. It is to be noticed that this peri- with each unavailability function for the individual conven-
od occurs one hour later during daylight saving time (Middle tional power plant is performed. To determine the capacity
European Summer Time; MEST). This means a reduction of credit of the studied conventional-regenerative power plant
the minimurri available photovoltaic power output from 20.8 system the conventional power available with a probability
to 19.2 % during the most likely daily peak load hour between of 97 % is determined first. Then the availability probabili-
11 am and 12 am. ty function of the renewable power generation is calculated.
The dependable power output of the conventional-renewable
power plant system available with the same probability of
3 Conventional and Renewable 97 % can then be computed. The resultant increase in de-
Energy Conversion System pendable power output is then equal to the capacity credit
of the renewable energy elements in the park [5].
In addition to the considerations of chapter 2 - where the By applying this approach the capacity credit of a pho-
solar energy source and the power output avialable with a tovoltaic power generation was calculated also for the sa-
high probability are discussed - aspects occuring in interac- me south-western state of Germany (Baden-Wiirttemberg; cf.
tion with a conventional power plant system will be included [2]). The one-year-average value for the capacity credit is esti-
now. Thus, the mixed power generating system consisting mated to be 0.9 % refering to the installed peak power (Wp).
of renewable and conventional components will be addres- This rather low result is a consequence of the unavailability of
sed. The "dependable power supply" for the mixed system photovoltaic power output during night hours (approximate-
must be the same as for the conventional park. On these pre- ly 50 % of the year). The time periods around noon deserve
mises, the installed photovoltaic power has to be much lar- further at tention as they are usually the peak load hours of
ger than the substituted conventional power because of the the public electricity grid under current demand-behavior in
strong variations of the solar radiation. The capacity credit South Germany [3]. They were analyzed separately and the
is then given by the portion of the installed photovoltaic po- capacity credit around noon hours was found to be a porti-

487
on of 2.2 to 2.9 % of the installed peak power (Wp ) of the 5 Climatic Conditions During Ye-
photovoltaic modules.
The solar energy source is characterized by broad varia- arly Peak Load Periods
tions during the year. Therefore additionally the monthly a-
verage values for the capacity credit are calculated during The investigations presented in chapter 2 to 4 were based on
individual hours of the day. This investigation showed that the analysis of one specific year. The meteorological conditi-
ons in South Germany during the yearly peak load periods
between 8 am and 9 am the capacity credit changes from
zero (during the winter months) to a maximum of 1.8 % in of the last 25 years were studied in order to be able to gain
more general results.
summer. In contrast to this for the hour between 12 am and
The investigation of measured daily mean temperatures
1 pm a capacity credit from 1.2 % in December to nearly 4
at the yearly peak load days shows a more or less clear co-
% in July is given.
herence between days with a high load demand and those
with significantly low temperatures. During moderate winter
4 Demand-side Behavior and Re- periods (e.g. winter 88/89 with Ta ••r • ge = 3 ... 6'C) the inter-
relation of high electricity demand and low temperatures is
newable Energy Supply less distinct. In contrary to this during cold winters (e.g. win-
ter 84/85 with T•• er • ge = -9 ... - 20'C) there is a stronger
The approaches presented so far studied isolated so-
correlation. The capacity credit of a renewable energy supply
lar electricity generation (chapter 2) as well as a com-
will be technically relevant especially during extreme load de-
bined conventional-renewable electricity production in a
mand periods when the power demand is significantly above
conventional-renewable power plant system (chapter 3). The
following investigation studies the potential of load reducti- the yearly average. Hence, the daily sunshine hours reporte-
d by several measuring stations in Baden-Wiirttemberg were
ons for the conventional power plants as a result of renewable
examined more closely during eight selected winter peak load
energy contributions to the public electrical grid. In order to
do this the possible photovoltaic electricity generation is com- periods with significant low temperatures. In many cases the
pared to the demand-side behavior in the public grid in South insolation was found to be rather high few days before the
Germany. particular yearly peak load day (no or only very few clouds).
The hourly mean values of electricity generated during This is a result of the high air pressure period that is usual-
one year for the public grid in Baden-Wiirttemberg in 1988 ly responsible for extremely cold periods in winter in South
(48 665 GWh/a [6]) can be simulated based on measured Germany. The following days, however, the temperature in-
meteorological data and measured values of the grid load for crease until the yearly peak lo.ad day is reached. This process
several time periods. The same meteorological data is use- is usually accompanied by a change of the general ~eather
d to calculate the possible photovoltaic electricity generation conditions characterized by an increasing cloud covering of
from 600 MWp installed and spatially dispersed photovoltaic the sky. Therefore, most peak load days are characterized by
modules (cf. [2]). The corresponding simulated photovoltaic no or very few sunshine hours. This tendency cOJitinues after
power generation is then integrated into the demand-side be- the yearly peak load day. This correlation would have cause-
havior. Figure 2 shows the highest daily load demand and d only a very low theoretical photovoltaic power generation
the corresponding power reduction due to the integration of during the studied eight yearly peak load periods. Hence, the
photovoltaic power generation during each day. Additionally, capacity credit of photovoltaic modules must be expected to
the resulting daily load reduction and the duration curve of be close to the lower range of the possible spectrum during
the load reduction are presented. yearly peak load periods in South Germany.
During the summer months high insolation values result
in partially significant reductions of the remaining load to be 6 Conclusions
generated by the conventional power stations. Fig. 2 shows a
notable load reduction for the conventional power stations for Objective of this paper was the description and discussion
some winter days. This is usually the case for working days of possible approaches for estimating the capacity credit of a
when the daily peak load during noon hours matches the pos- photovoltaic power generation. The analysis shows
sibly high photovoltaic power generation (especially during
• that there are several different approaches to determine
high air pressure periods with significantly low temperatures,
the capacity credit of a photovoltaic power generation.
high insolation and no clouds). Nevertheless load reductions
The variation range of the results is correspondingly
are not expected during certain working days when the daily
large and variies between zero and approximately 20
peak load occurs during the early morning or evening hours
%. As the attention being given to the photovoltaic e-
and during weekends which are characterized by a relatively
nergy source is increasingly expanding, it should be a
balanced demand. Therefore, 80 days (approximately 22 %
serious aim of energy policy to find an approved consen-
of the year) show no daily peak load reduction in the public
sus of general validity about the future methodological
grid in the presented case. On the other hand, during nearly
approach.
two summer months a capacity credit of approximately 20 %
based on the installed photovoltaic power can be expected. • that yearly mean values are not appropriate for an ana-
The hypothetical assumption of a policy introduced shif- lysis of the capacity credit of photovoltaic power gene-
ting of the peak load to the noon hours would lead to different ration. Due to the supply characteristics of the solar
results. In this case only 9 days (2.5 % of the year) without energy source it is suitable to employ hourly mean va-
daily peak load reduction in the public grid under the then lues for each month.
given demand-side behavior would have to be expected. This.
would lead to a capacity credit for the photovoltaic electricity • that - despite the differences observed between the re-
generation of about 0.1 %. sults of different approaches - the capacity credit of
photovoltaic power generation will be in the range of
only a few percent of the installed photovoltaic peak
power (Wp ).

488
 .........
~

6000
2
~OOO

4000
. I
JOOO
A

2000
d.tt, 8 01. 1. 0l1li i ll t. t , 11 •• it ,,1.111
1000
rt l . "'.il, • ••. l lud . ilh PV-pltt. r It:.Cr,t i""
0
0 JO 60 90 120 150 180 240 270 JOO JJO

....
....
~

' 400

'500
, • • " 'tllllu:llofl .1(1111" If lhl PV-tl,drlc: ll , . f t • .
d,roli.fI ' I ' WI Ir thl ,. , er r,4uclll"
' 600~0--~JnO--~6~0~~9"'0'--71~20'-~I'SftO~1~8~0'-~2~10'-~2~4ftO~2~7~0~~JO"O'-~J'l~0--U
L ,. . i flo •

Figure 2: Daily peak . load during one year in one of the ferderal states in Germany (Baden-
Wiirttemberg) and the corresponding load reductions achieved by a photovoltaic power generation
under the same meteorological conditions

• that low insolation values and correspondingly low pho- (5) Bouillon, H.; Machate, R. D.; Tetzlaff, G.
tovoltaic power output values are to be expected during Leistungseffekt von Windkraftanlagen
annual peak load periods in winter urider South Ger- Elektrizitiitswirtschaft 88(1989), 24, S. 1752 - 1759
man weather conditions. The capacity credit will, there-
(6) Ministerium fiir Wirtschaft, Mittelstand und Technologie
fore be close to the lower edge of the possible spectrum (Hrsg.)
during these times under the given meteorological cir- Energiebericht 1988/89
cumstances. This result was found for one reference year Statistisches Landesamt Baden-Wiirttemberg, Stuttgart, 1990
and also for past time data of the last 25 years.

References
[1) Vereinigung Deutscher Elektrizitiitswerke (Hrsg.)
Begriffsbestimmungen in der Energiewirtschaft (Elektri-
zitiitswirtschaftliche Grund begriffe)
Verlags- und Wirtschaftsgesellschaft der Elektrizitiitswerke,.
Frankfurt, 1981
(2) Kaltschmitt, M.
Moglichkeiten und Grenzen einer Stromerzeugung aus Wind-
kraft und Solarstrahlung am Beispiel Baden-Wiirttembergs
Dissertation, Universitiit Stuttgart, Dezember 1990

[31' Vereinigung Deutscher Elektrizitiitswerke (Hrsg.)

Statistik fiir das Jahr 1988
VWEW-Verlag, Frankfurt a. M., 1989

[4] Riiuber, B.; Holland, M.; Holder, A.

Photovoltaische Energieerzeugung in Baden-Wiirttemberg
Fraunhofer Institut fiir solare Energiesysteme, Freiburg, Ok-
tober 1987

489
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE
8-12 APRIL 1991 LISBON, PORTUGAL

FIABILITY COEFFICIENT MAPS FOR PV SYSTEMS DESIGN

M.A.Egido, E.Lorenzo
Instituto de Energia Solar
Universidad Politecnica de Madrid
Phone: 34-1-2441060; Fax: 34-1-5446341

Abstract

The reliability of stand-alone photovoltaic systems is analyzed in

terms of the Loss of Load Probability, LLP. With the idea of simplifier
the design of PV systems a new analytic mode l, as accurate as the
numerical models and as simple as analytic models, is proposed. For each
location, the model requires as input 4 different coefficients, that can
be obtained using a set of coefficient maps.

lJ.A.G(O) G(O)
1.- Introduction C~ = ---L-- = CA· - · -
G
The merit of a stand alone pv-system should be where G(O) is the mean daily irradiation on an
judged in terms of the reliability of the horizontal surface. In practical cases,
electricity supply to the load. This is usually pv-designers can directly obtain the value of lJ.A -
quantified by the concept of Loss of Load and consequently, the required number of pv modules
Probability, LLP, defined as the ratio between the from C~. Meanwhile, from CA they need to
energy deficit and the energy demands both on the
load, over a long period of time. Because the calculate the ratio G/G(O) witch requires some
random nature of the solar radiation, the LLP of, expertise and expense.
even a trouble free, pv-system is always greater In other hand, the long term averages of daily
than zero. irradiation are usually available in terms of
monthly means. This allows to associate a'different
The size of a pv-system is a general concept
including the dimensions of the pv-array and the value of the pv-array capacity for each month.
accumulator. A useful definition of such dimensions lJ· A. Gdm (0)
C'
relates to the load: In a daily basis, the pv-array Am (m=1, ... ,12)
capacity, CA ' is defined as the ratio between the L
mean pv-array energy production and the mean load
energy demand. The storage capacity, Cs' is defined where G and G (0) are, respectively, the monthly
dm dm

as the maximum energy that can be taken out from averages of the tilt and horizo?trl daily
the accumulator divided by the mean load energy irradiation. Moreover, it has been show that the
demand. So: location dependence of the results can be reduced
by correlating the collector area to the
lJ.A.G irradiation worst month, defined as the one showing
C
C= and the lowest relation between solar irradiation and
C=
A
L s energy consumption. For simplicity, in this paper,
L
we concern ourselves with loads exhibing a constant
daily energy consumption all along the year. Then,
where: A is .the pv-array area, lJ is the pv-array
efficiency, G is the mean daily irradiation on the the worst month is the one corresponding to the
pv-array, L is the mean daily energy consumption lowest value of G (0). Hence, in that follows we
dm
and C is the useful accumulator capacity. use for CA and C~ the lowest values of C and C' ,
Am Am
Note than CA depends on the meteorological
respectively. Other cases can be easily derived
condi tions of the location. That means that the from this one.
same pv-array for the same load can be "big" in one
si te and "little" in another site with lower solar 2. The reliability map.
radiation.
The task of sizing a pv-system consists of
finding the better trade-off between cost and Each point of the C~- Cs plane represents a
reliability. Very often, the reliability is a size of a pv-system. This allows to map the
priori requirement from the user, and the reliability as figure 1 shows. The lines are the
pv-engineer problem is formulated as follow: Which loci of all the points corresponding to a same LLP
is the pair of CA and Cs values leading to a given value. Because of that, we call them isoreliability
LLP value at the minimum cost? This paper propose a lines.
new method that attempts to be accurate and simple Note that the definition of C~ and Cs implies
to use. that this map is independent of the load and
Because long term averages of daily depends only on the meteorological behavior of the
irradiation incident on surfaces other than location. We will see that the isoreliability lines
horizontal are not generally available, is useful are, very nearly, hyperbola with their asymptotes
to define a new parameter parallel to the x and y axis, respectively.

490
 SOC'" lIC
j s
~ E
AUXJ
a

and

- - LLP - 01
- - lLP & 0.05
• llP - 0.01 where EAUXj is the energy supplied by the auxiliary

1"1
generator in the day "j".
If the simulation is carried-out over a great
number of days, N, in order to be statistically
meaningful, then the LLP value corresponding to the
stand alone pv-systems (i. e, the system of Fig . 2

foot
........ excluded the auxiliary generator) is given by

-
j=H

--- ~
LLP
L=l EAUXj

021 'l' 15' . . '1'2 ., ...• '(6

N. L
S o'oge Copoci y This equation implies the use of a constant
pv-array efficiency. The validity of such
assumption for the present purpo~j?' is supported
by the results of Ambrosone el al ,who found the
Figure I. Reliability map.
same LLP values using daily timesteps and
l)=constant as they did using hourly timesteps and
considering the 1) dependence on solar irradiance,
3.- Determination of isoreliability lines
ambient temperature and battery state -of yharge,
The LLP for a given pair of and Cs values, C~
particularly for Cs'" 2. Some authors (4-5 have

is calculated by means of a rather detailed presented results supporting the opposite idea,
simulation of the pv-system behavior. To explain it i.e., that simulations in hourly basis always leads
in a simple way, let us imagine a pv-system as to more accurate LLP values than in daily basis.
represented in Figure 2 and let us assume that all Nevertheless, a careful analysis of such resul ts
the consumption occurs during night and that the discloses that the differences are due to the use
battery is free of energy losses. Then, the state of different and not exactly· equivalent
of charge of the battery, at the end of the day "j" transposition models from horizontal to tilted
radiation, in such a way that the daily irradiation
is given by: on the pv-array is different when they calculate in
hourly and in daily basis. This is a collateral
C' .G
effect that cannot be attributed to the timestep
SOC min (SOC + A dj 1)
selected for simulation and that can be take into
j j-l C. G (0)
s dm account in more simple ways.
Since the simulation is run for a long time
where Gdj , is the daily irradiation over the period, the initial state of th~ battery has no
effect. Moreover, as Gordon 6) noted, the
collector surface on the day "J" and C, is the
assumption of a 24-hours distributed load lead to
nominal capacity of the accumulator. It is
the same LLP value that obtained from our nighttime
calculated from the corresponding horizontal va(;ue
using the algorithm proposed by Hay and McKay 2) . load if Cs is reduced by I .
Furthermore, let us assume that the auxiliary The minimum LLP value is determined for the
generator is managed in such a way that, at the end length (?f reference data sequences. Klein and
of the day "J", it fulfills the battery if the Beckman showed that with the at present
stored energy is lower than the load requirements . av~lable historical data (~ 20 years) only LLP >
Then, 10 can be validated . Below this value there is a
need to rely on extrapolations. The reason for that
is the natural variability associated to the
climate . That is, equally probable series of 20
years solar radiation data, can lead to different
LLP values for LLP < 10- 2 .
4 . - The new proposed model

Two ideas come out from the previous section:

a) Numerical models can be adequate for the
estimation of LLP values. However, their use for
~ I - - --l sizing is rather complex and consuming in terms of
T computing time, to the extent that they are beyond
the skill of the great maJori ty of current
STORACE I,.OAD
pV-installing companies.
b) All the analytical models proposed up today
allow sizing of pv-systems in a very simple way by
AUXIUARY
CENERATOR
means of straightforward hand calculations.
However, they exhibit significant lack of accuracy
Which disfavors their general use .
The motivation of our work has been the
development of a sizing procedure as accurate as
the numerical models and as simple as the
Fig 2. Schematic of the photovoltaic system used analytical mode.1s. To that end, we have done an in
for simulation.

491
dept~ analysis of the solar radiation Spanish data
base 8) and we mapped the reliability for 42
different locations and for 11 LLP values uniformly
distributed between .5 and 5x10- 4 . So, 462 160
isoreliable curves have been studied. We found all
of them can be well represented by the expression
c
C'
A
= m.C-
S
n .~ 1 20
.~
III
where m and n are depend on the LLP value and on
the location. Fig 3 shoWS the relative error due 8 080

to this geometrical regression for all the curves.

The mean value of the coefficient of determination,
R2, is 0.98 and it varies between a minimum of 0.93 040
and a maximum of 0.999.
For a given LLP , m can be defined as the C~
value corresponding to Cs = 1 . n takes into account ooe '0 0.40 060- 0 0
Clearness ndex 0 the Worst Mon h (K )
the effect of the persistence of the solar
radiation . Note that C decrease when n increase ,
s
while keeping constant all the other parameters . Fig 4. Coefficient m versus Ktm

Fig 5 and 6 show Spanish coefficient maps for

0.08 two different values of LLP. For a given LLP value,
the calculation of the C~ value corresponding to a
o particular C value is a straightforward task using
006 S
W
these maps .
QJ
> 6. Conclusions.
.- 00.
o An analytic model f or accurate sizing
Q)
a::: stand-alone photovoltaic systems have been
I
deve loped. For each location, the model requires as
002
1 , input 4 different coefficients that are obtained

000- . +. i~.i~
~~~~~~L4 J ! from the study of large sequences of historical or
generated daily irradiation data .
00001 0001 001 0.1
Loss of Load Probab·litie Maps of such coefficients have been presented
for Spain. From these maps, the task of pv-sizing
Fig 3. - The Loss of load Probability versus the becomes very simple and can be accomplished by
geometrical regression error. means of straightforward hand calculations.
References
Moreover, we found that for each location, the
values of m and n are related whit LLP, through the (1) R. N. Chapman. "Design consideration for
simple regressions stand-alone photovoltaic systems. Proc. Symposium
on Applications of Solar and Renewable Energy-86.
m m + m .10g(LLP) Cairo, Egypt. March 1986
1 2
(2) S . E. Hay and D.C.Mc Kay . "Estlmatlng Solar
Irradiance on Inclined Surfaces: A review and
Assessment of Methodologies". International Journal
of Solar Energy, vol 3 , 203-240. 1985 .
R2 val'ues for both regressions are greater
that 0.93.
The main benefit from these equations and is (3) G.Ambrosone, S.Catalonotti, U.Coscia and
the use of only 4 numbers for each location instead C.Trolse. "Comparison between power and energy
of the complete sequences of radiation data , methods of analyses of photovol taic plants" . Solar
Energy, vol 34(1), 1-8. 1985.
without accuracy losses . This facilitates the task
of the pv-sizer .
Furthermore, we explored the possibility of (4) LL.Mora, M. Egido. M.Sidrach. "Utilization of
obtaining m and n from meteorological information s1mulated hourly global radiation series for sizing
stand-alone photovoltaic power systems".
of general availability . As an example, Fig 4
shows the variation of m versus Kt ,. It is clear
(5) F. Lasnier and T.C. Ang "Photovoltaic
that the use of any correlation linking both Engineering Handbook" . Adam Hilger. ISBN
parameters leads to significant accuracy losses. 0-85274-311-4. 1990 .
Similar comments merit all the correlation we
analyzed between m or n with other radiation (6) J . M.Gordon."Optimal sizing of stand-alone
parameters (Gdm , etc.) and between m and n. This photovoltaic solar power systems". Solar Cells, vol
20, 295--313. 1987.
research is still being undertaken by including
other wldely available meteorological data like the
number of rainy days, etc. Meanwhile,the present (7) S.A . Klein and W. A. Beckman. "Loss-Of-Load
results suggest that the location dependence of the Probabilities for stand-alone photovoltaic
systems". Solar Energy, vol 39 (6), 499-512. 1987 .
pv-size method cannot be neglected without paying a
significant price in terms of accuracy . This agree
(8) Radiaci6n Solar en Espana : Mos 1972 a 1984.
with the fact that the monthly clearness index over
a period of 20 years shows different distributions Centro de Es1udios Meteorol6gicos. Ministerio de
Transportes. Turismo y Comunicaciones.
for the different locations.

492
~\ COdrIU('IIL (1.1'1' () \) N eOefflCI nl (1.PP 0 \)

Fig 5. Coefficient Spanish maps for LLP

M e00[[1 'Ienl (LPP 00\) codflC"lcnl (LPP U 0 \)

Fig 6. Coefficient Spanish maps for LLP 10- 2.

493
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

STOCHASTIC MODELLING OF .n PY .sy.sT~,1 ,IIl'H SHi~L COUCEIJT1UTION

V.L.FGr~ and R.GriGoreccu

PhY>Jics Dept., Poly technical Institute of Bucharest
Spldiul Independen'~ei 313, Bucharest, l{omqnia

i\BSTI1ACT. The purpose of this study is the optimum dimensioning of a

PV system (ccpture urea-electric storuce Oapi.lCi ty), \1hich supplies fij
COnBuner (a radio-relay or a radio-phone). Two ensential conditions
ure imposed: the presci:ibed system reliability for continous load nupply
::lnd the estOlblisi1ment of an optimum from the technic-economic point of
viev;. j\ re<Jlist evaluation of the enerGY oapture, stor~lge and delivery,
considered an r8ndom processes, is obtc;ined by stochastic modelling \'Ihich
correl~tes duy-to-day solar radiation variation to loss-of-power of PV
sys'l;em. This method offers pairs of vclues of capture area-stor8se capa-
city \'ihich denominate the indifference curve for a prescribed system
reliobility; it is rep~'e3ented by the probability of the lo\vest chargine;
state. ~he economic optimum corresponding to minimal cost of the system
is est3blished by the economic modelline; of PV system, correlated with
the indifference ourve obtuined previously. It is emphasized that solar
concentr~torG utilization can determine f~vourQble economic effects.

1. STOCIL',ZTIC MODELLDTG - '1he ;oecond stute is symbolized '0'1 b,,;

in tilis O::lse th8 ::-~ored enel.'GY LlVel J
u

~; Dtatistical method is developped ~ie;;; ,Ii tllin -(;l1e l'anse j, bnt the }?:::-e'lriOlW
(Bucciarelli, /2/); it is assumed the c.l3y see In incre:.;Ge in stored one:;:'s:".
effeclf~y-to-day correlation in solar
radiation on the probability of 108s-of- The con&tr:::int t::c trc:nsition prob:.;bi-
pOYier for a stand-dlone PV sys!;cm. Thus lity matrix for t11i2 system we need me3-
a realist enercy capture, storage and 8u:!:'es of the probcbility of stoppine: up
distribution prooess, as a random walk or do~n one level in Dtor~ce space riven
may be obtnined. that the previot1<S d<;y Was "'bed" (m.~ 0:;:'
It is considered ~~e existence of a "coo a" (b,.), /f2.'. ':Ie lot: ae.' be theJproba-
constant oonsumer of enereY L; the mean bi:J..ity o::J:; ':;~>,:mi':;ion of +~, civ,9n th3
values of incident solar enert;y, JA+. and 'p2.'cviou,J dey \18f:< ''.wod''; t" be tho ).1.'00<:1-
of its variance, ~~ are determined, bility OI Cl t;:':;m:ition of -:.,1 ,o;iven the
t8kinG into account evc:::y day 8S a st'lc;e p2.'evious d,1Y W28 "bc;d"; J3 be the p:!:'ob;:;-
in 8 stationary Murkov process,/1/-/3/. bili~ of G t~"ndtion of -.L1, Given the
1:here are tno pos~itle events: prev~ous day Yii./S "good"; d be '(;he p-,.'oba-
- on enerGY i" which is deli vercd by bility of a trclm,ition of -A , Given the
solLlr panels, exceeds a required enerGY, the previous d:::;y \'I'Jl:: "bad".
L, th2t is "; > L ([;n event ,'jith probabi- We h:;ve
lity p); d.+fl= 1; (+0= 1; 'P =""'p +i"q; . (2)
- 8n ener~y j, is insufficient for :3 q "') p + cfq; ~= l-Clho)=~+n-l.
I'equiJ~ed enerGY L, that is ,', '" L C~n event From tho matrix equation
wHh probabiUty q).
Two Dt:ces of t~li.B modellinc;, /2/ ~re the tr::.n:::ition1 [the stGtlOtlary ]
[ prob8bility st:-ite veotor ;:
considered: matrix
(I) N possible charce utates of tho
cner;;y utol'.j~e cJpacity (lre considered = [tr..e. st"te
of" system
vector]
and c<Jc~ sGste (step) is of ::",nc;o A ,th;;:t
iB N :: A ' ..here G is the Q;Jximum stored :;ucci~lrelli, /2/ obt:::dned the probabili-
eae].... . c~r.. It is 88Dumed ~-j oor'::lc18tion ties of occupyin~ of the 2 N charGe
doy-to-..i::ly oi.' t;he enerGY, -,Ihich iG delive- stutes of system.
red by Gol~~ p~nalG, cun't; be ezt~bli~~ad. A reliability coeffiden-:;, Y/£, can be
"b,., 1'\~o"''']bi' i"-v!if, 0'" o ""u!pin'" t'w " til in'(;roauoed; it repre::ients the probability
~~~~~;o~~t~~qW1G ~lvc~ b~~ ~~ J - 0 of ocoupyine; of the "ba'lest" charl"e ot",te
1'1.1.
9f; = ~J'-~_\l-:..u
U

(1)
Q 1 _ :AN
\.ho~')e :A:; pi r; ~~tl:d p+ ("~ =
1 • 2 • .sYSTEM DIME1JSIONHG
(II) J; GOl'reLt:i,on in d",y-to-dJY on
s4Ystem st~i:;es ~G(...Ui:"'C5 donblin~ ti:.e·num- 2.1. Computin0 alzorithm
bar or s'::,:t,:;;: in -::tOl':::::-3 ~::p~,ce (2N). Normalized v8:::-iubles are eonsidere<l:
a ;: 1>/L; d == DIL==(A .....L)/D=8-1
GO ~;c~ ~l~~l'~~S~~~f,"tJt;~~'e CO/2}~:"cnlin~ (l~)

- '1he i:irst "- tde :is s::;T£boli~ed .;y mc j .Pb =Jl'J - 1; cJd == eTa =eT,;IL (5)
in tId!] CalJe the storGd enerGY level liSs ne == elL; !l'= A/L == C/NL (6)
i"li thin the renee j, and the previous day
see " deore~1i3e in stored ensI'SY.

494
 'We assume the energy, which is 3. SJ.!1'\lm-1~LONE pv SnlTEMS OPTIMUM
delivered by solar panelD,is proportio- DmENSlmrr NG
nal to the incident solar enerGY, that
is ;; .1. S:/stem "tates
O',/.Pa =cr:;Iy4>; era =Pa((J<i-P.p) (7) ':2he influenoe of doy-to-dDY solar
radiation variation on different states
The purpose of our calculation is to of a PV system is analyzed. Principal
establish oorrelation between oapture oalculation duta are: ~ 0,01;=
area (measured bY)1. ) Dnd storaGe oapa- (1'41 IN = 0.4; P4>;.: 3.kVjh~m2day. Three
oity ( measured by Uno) ~r'an imposed oa:;;es ,ire considered: ;,(,9= 0,2);
reliGbility ooefficient,9l~. B(9 = 0); 0(9 = -0.2). In 1'i:;.1 is
Computing alr,orithm ~is the presented the distribution of occupyine
following; d" probubility of u~r;,;tem st;,tos fo::, these
- read,input datu; JI Z ; S ;(~p.p ); CbBes. I!e C6n r:Jmu,J}:: t;he fol1o·~·Jinr;:
- beGin cyole a1'ter y.. ; - if!> 0, then should be necesz8ry
to incro8se the storoge c~p3city for
-'oompute: the sc;me obtdned !f~,; ...
d

1. Yd; <18 ; CJ"d; A' AIL = - a nec~tive correlation decreases

2. p; q ; ~ the stol.'c;::;e 0,,1'801 ty, but it is sensible
3. oL; J3 ; t' ; J
I
increased the probability to find system
in the " . lWl5'[; chare;;, 5tete;
4. H; nc = n A ,., - if 1':= 0 or p.~ 0, then should
9l'lb' ••• , !1lJm' 9l 11'b
5. 1l11l ' be also sensible inore3sed the probDbi-
li ty to find the s;-;rstem in the lO,W'8st
- end cyole a~ter.Pa; oharsin.,9stutej
- obtain pri~cipal dataJUa = 1'(no ) - t;ne e:<ccossive overdimensionino;
for g[ E const. = oontributes to the deore8sinr:; system
inde:pendence.
2.2. S;ystem optimum dimensioning. • 3.2. Indifference eurve and straight
The curve.pa =
f(n o ) for 'J{E ;;;: oonst. line of oonstcnt aoct.
is jenominated the indifferenoe ourve For the C3se ". ( p := 0.2) it is
obt[:ined '/;he indifferenoe curve from
and gives pairs of values ~a' nc) for Figure 2. The straight line of constant
a prescribed system reliability,jfE. By cost is civen by:
choosinG a p~ir of v~lues \Pa' no) on (12)
indifferenoe curve, we have
;, =PnL; C = ncL (8)
The oapture ~ea is determined from where 0-Jl re~resents annex installations
coeffioient. We oonsidered p~:509 $/m2 ;
lic :.. 1I1'p~ llo (9)
where rz
0 is the mean efficiency of Pc = 100 S/kYThj fc = 0.09; --r.; 10. =
solar cells. Thon:
''Je assume a line,:,r dependence of , c~ =
l696.6jla + 100 nc (13)
inatollation oost, c s ' ,'lith .~ Cind C; The minimum villue of the system sDecific
it is obtained cost corresponds to the pair of v;lues
(ya :: 1,1; ne =A.35)j VJe bet 9~ :
c~ = Pa .rt + Po 0 = PeAcC 1c.p«i'
)+PcC (10)
= 2319.9 $/kWh. Tnkins into aooount
Pe and Pc are specific costs for sol~r radiCition oonoentrato~s utilization

.
whe~e
~'. !-,no. c. ¥
the equation of s;;stetp co"t bocomes;

c~
3y norwelizotion of eQ.. (10) we find:
CalL = =
PaPa 'l- Peno (11)
c.'::P~Jl+p~) h .f1;:
.p c
.I ,.t\.. :}\:/Pc
... nc t:\
(14)

E(l. (11) re:;;Jref:ents fl straiGht line vlith In eQ.. (14) the follo\JinC not"tions
nec;c"dve slope, in coordinates Cpa' no)' are introduced: K - concentration ratio·
M '
for o~ == con",!;. It 02n be ShOVih the PI;: - Boler conbentrutors co;.;t,Q.uota from
stroi:;ht line which is to:m.:;ent to the '.;he.rJhole COGt of solar cells. For 1',_ =
inditference curve determines the miniMum
v~lue of c _, rw.d satisfieG t",o import8nt
= ~
0.15 Clnd 1, - 10 (smnll conccntration)
y - ,..,., "'.

it is obt~dn~d 'l;h0 8Quution:

~eouirements: Supp~ c~ = 1950.9 fa + 100 no (15)
- it is ensured, the reli<,blSVOf loa.d,
ClccordinC; to imposed paramete::'~;; 'lhe optL urn point corresponds to' the
- it i" re:"lizcd t;lle mininiz;,tion pah' of vulucs (Ya 1.3; no =
3.05); =
of it:;; cost, V!5.t:l f::;vourc.ble effect on we cet in tnis case 0;=2740.17 $/k'!!h.
sy3tem eGonomic Gonpct:.tivit;i-.
By oonvenient placement of PV cells
tovlard the fOCal plClne, in the condition
of season orientation, an imposed Bini-
mum oonoentration ratio is achie .... cd.I'hot
implies both the obtaininG of :, C;l'e~t
Qcce:ptanoe anele and the .USc 0::: passiVe
ooollng system, and the I·cducinc of the
necessary number of PV ceIlj~~uvious
favourable eoonomic effects.

495
 _ A
_. -. - 8
_ - -. C

Fi::;.1. '1'he oooupying pl'obability difitribution

of sY9tem st8tes for the ouses ii, B
I.Jnd C.

jJo
2

1,1

#.7

',0

IS

4,2

4.f - - - - - _

F1;.2. :ehe indiffe:::ence olU've for the Oal3e 11,.

496
 4. OONCLUSIONS

, , prinoipal results obtc;ined in

~:1e
tn~~ paper are;
- it oan be established the ai;ora36
oapacity 0 ~ind 'i;hc captarc area :for
a constllnt daily lOAd L. '.l'hus, tho mec~n
n,
dolivered ene~'gy,)J. ~nd its v:.:rionce,
era andl;he p:..'obab!e evolution , i of
tIlE) foLlowing day allow to insure this
load for a prescribed system reliability
9fE~ a computin~
algorithm Can be utili-
zed for the obtaininG of indifference
curves llnd of the optimum p«ir 'fa' n c)'
\~hich characteriZeS the minimum cost
of the system;
- the economic modeHinr; of I'V sys-
tems i;:; ancll;::zed for t·.IO OG;:.es~ vlH,h
concentrato:.:s and \C,ithout t:1AII.
HE::n.JRIDTJ}~S
/1/ .JoLeGtionnc, ;,pplic'Jtion du mod~le
M"rl:ovien ;:;inmlific a
I' ~tude du
compo:'tement du Gteok8C;~ d' une con-
t:C:'31e ,']ol£lire, Hev.Phy" ••,ppl. lL>,
139, 1979. -
/2/ L.L.:Llucoiarelli, The effect of dr,y-
to-day correl~tion in solar r~dia­
tion on the pl'ob:;;0ility of' lOGs-of-
po,'Jer in '" s'!;snd-81one photovoltc:io
0118~ ~y sys tom, .solcJr ~n8J"'G~J 22. 11,
198~.
/3/ HoIJ oGhapmJn, J)evelopl:lGnt of .Jizing
nomot;;I';,o.s foo.' .st;.1r..d-olone photovol-
t::Jic/Gto'::'::::ue I DJ,~ste1:13, ;301ar Ene:."'cy
:}.2., 71, 19890

497
 SESSION 03

Crystalline Silicon - Materials and Devices

10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE &-12 APRIL 1991 LISBON, PORTUGAL

THE DEVELOPMENT OF THE WAFER COST AND AVAILABILITY FOR

THE PHOTOVOLTAIC INDUSTRY; RIBBON/SHEET APPROACHES AND
ITS COMPARISON WITH WAFERS: EFFICIENCY-COST TRADE-OFF.

Heinz Herzer
Wacker-Chemitronic GmbH
Postbox 1140
8263 Burghausen
Germany

ABSTRACT. The Photovoltaic Industry is a young industry which has

not yet matured to handle its business in a profitable way. One of the
main reasons is the conflict between operating technologies, real
costs, and diversified applications under loss-generating market
conditions and the big visions to make Photovoltaics become a
renewable clean energy source for the future. A driving force always
has been the projection of low-cost metallurgical sand reduction
combined wi th r i bbon/ sheet approaches if c-S i is concerned, and the
advent of a-Si and thin film technologies if alternatives and c-Si
replacing materials are concerned.
Today, we recognize a concentration towards c-Si as the basic
material for power PV modules and systems. With regard to the
scientific/technological state of the art, even here, a wide range of
methods are presently investigated as there are EFG, S-Web, RAFT, RGS,
SSP, MS, thin film c-Si, and others. The potential in terms of
efficiency and cost-advantages/disadvantages will be commented.
Looki~g at the industrial status of large-scale production commercial
and economical aspects are dominating bringing everything to the
classical production of monocrystalline and multicrystalline wafers,
both in connection with ID or multi-wire cutting.

1. INTRODUCTION
This consideration does not differ too far from the
semiconductor grade silicon. The expectations in
1.1 General remarks the past to rapidly develop a solar grade material
In 1990, the world PV module shipments have at low cost could not be realized within the
achieved approx. 50 MWp which is equivalent to a projected time frames except the macroscopic
total value of 300 to 400 mio US$. Compared to the crystal perfection of cast multicrystalline silicon
semiconductor silicon industry, the semiconductor due to the fact that the casting technology and the
device industry, and the electronics industry, material processing well advanced covering a
respectively, this value stands in a ratio of growing share of the total actual PV production.
1 : 10 : 200 : 2000.
In Germany, some projections request that in 1.2 The market history and market development
the year 2005 10% of all energy production should The PV market history from 1976 to 1990 is
be covered by solar hydrogen /1/. In other represented in Fig.1 showing the very beginning of
countries might be similar anticipations. How to initial pilot production followed by a step to a
produce the huge amount of PV substrate mater i a 1s
and modules and what will be the price we have to
pay? There are many calculation showing how many IMWpl
poly factories we need to cover the demand of 50
feedstock material if hyperpure silicon will be 45
the dominating material. We easily arrive at
hundreds of factories and the above given 40
comparison would bring the PV industry close to the 35
electronics industry, indeed. Contrary to these 30
visionary figures, in 1990, the theoretically 25
available amount of hyperpure silicon from the

~~~.~.II
semiconductor silicon manufacturing· industry for 20
terrestrial solar application has been utilized by 15
only one third. To come to the point, we have to 10
leave the visions for the time being and 5
concentrate on the existing technology and its
impact on the existing business of the PV industry. 0
The today's PV industry is strongly interlinked 1976 1978 1980 1982 1984 1986 1968 1990
with and dependent upon the semi conductor s i 1icon
manufacturing industry. With the view from a
hyper pure silicon producer the essentials of
silicon for solar application can be devided into Fig. 1 The market history of PV production /2/.
the technical key essentials: purity, crystal
perfection, geometry and the commercial key
essentials: quality, availability, price.

501
 installation of more multi-wire cutting equipment
is evident and it should be mentioned that Wacker-
_ Amorphouo Multi D Mono Chemi tron i cis the on 1y s i 1icon producer among the
II<WDJ
top ten in the world who is continuously and
60,0 substantially supplying poly/remelt, as well as
mono- and multicrystalline ingots/blocks and wafers
50,0 as semifinished substrate materials to the PV
~O ,O
module/system manufacturers.
30,0

20,0

10,0 Company Technology Integratfon

0,0
1985 1986 1987 198B 1989 1990
iii,
II.. 0 Cl
U I: l- I: I:
0
.D
Q; 0 '3 Cl E
:e E E E
'"

--
I: ::J Q)
II)
Q)
(ij (ij E::J 0
'3 E
o ._~ 1i5 ~ ~ (J) Q)
Fig. 2 The market development by technology /3/. §"§~ ,[1'0 0
1;i
~
1;i
c:- U
~
~
Q;
'0
0 1;i
Q E >-
E E '" <GUU a. u ~
0 (.) II)

first large scale production in 1983. In Fig.2 the

Siemens Solar
- Industries
• . 0

market development from 1985 to 1990 is shown

broken down into amorphous sil icon technology (a-
Si), multicrystalline (multi c-Si), and mono-
-GmbH
•
crystalline (mono c-Si) silicon technology.
The graph is rendering a-5i in a decline
TST
•
whereas c-Si is gaining, both multi and mono. Bayer 00 PQQ< !XX
Assuming a further continuous growth of c-Si the
question arises: Can we meet the corresponding Nukem 00 !SO
demand of silicon materials and what will be the
price we have to pay? Ansaldo
••

--
Helios
•• I I
IS/Wpl
lIalsolar
••
25,0 BP Solar
•• 0
I I
20,1 R&S
•

-
20,0

15,0
Photowatt
•
15,0
Kyocera
• •
•
11,5 11,1
10,0
10,0 Hoxan
B,l 8,6 8,4
I I
1,1
Sanyo/FuJifTaiyo
• I I
• •
5 ,0
Solarex

-
I
0 ,0
1916 1978 1980 1982 1984 1986 1988 1990
Chronar •
Weslinghouse 0 KX; ~
Mobil Solar
•
Fig. 3 The selling price development
PV modules /4/.
Astropower
•• ~
CEL 00<
• 0
Heliodinamica •• I I I I I I
1.3 The cost/price development of PV modules M-Setek •
In line with the continuous volume increase a
steady cost/price reduction for PV modules could be
Wacker •• I
realized in the past as demonstrated in Fig.3 where
the annual average selling prices of PV modules are Legend: • Production o ~ Planning
plotted. We recognize an upwards trend since 1987
reflecting the need to cover the production costs
and, from the view of the European manufacturers,
the adverse currency changes of the US$ versus the
European currencies. Fig. 4 A technology and competitors scenario for
the manufacturing of PV materials, cells,
1.4 Technology and competitors scenario modules, and systems.
A technology and competitors scenario is
exhibited in Fig.4 distinguishing between the basic
technologies and the degree of integration from the
raw materials to the finished modules/systems. A
trend of backwards integration can be observed, the

502
2. THE PV SILICON MATERIALS In that respect, the floating-zone technique is
the most favorable one, since it is only in contact
2.1 The feedstock silicon material (purity) wi th the inert gas in the vesse 1 and as far as
The present supply of feedstock s i 1icon raw auxiliary materials and energy consumption are
materia 1 for the PV industry comes from the concerned it is a 1so the most economica 1 techn ique
hyper pure poly silicon production as a by-product and results in a mono c-Si with the highest purity
and the remelt from Czochra1ski and float-zone and crystal perfection. The major disadvantages are
crystal production; a survey of the production unfortunately the high and costly requirements to
techno log ies and compan i es is put together in the the bulk and surface morphology of the poly
Table 1. feedstock material. For both, Fl and el, the
economical limits in large area wafer production
Table 1: Production of hyperpure poly. silicon and the neccess ity of mater i a 1 waste and cost for
for semiconductor and solar silicon the squaring are a significant handicap. The high
crystals in 1990 (Source: Dataquest). achievable PV efficiency, however, and the fact
that large area wafer processing requests the
installation of new production equipment which will
Production Company Nameplate/ Remarks not be implemented in the PV industry too soon are
technology production making these materials so attractive, today.
capacity[t]
Multi c-Si:
Trichloro- Hemlock , 8500/5900 Semi- Looking at the multi c-Si the gap in PV
si lane Highsil , conduct.- efficiency which is typically around 2% compared to
Tokuyama grade mono c-Si is just compensated by the more
Soda, qua lity econonomical casting process. Apparently, we
Wacker- observe an i ncreas i ng demand for augmented crysta 1
Chemitronic Solar grade perfection which is difficult to comply with the
quality same demand for high volume output at low cost. A
difficulty here is the quality specification and
Si lane Komatsu/ 1250/900 dito measurab i 1ity wh i ch is genera lly done by the f i na 1
Asmi effiency of the solar cell/module. It should be
mentioned that efforts are concentrated at national
Silicon- Ethyl 1250/900 dito and governmental supported university and public
tetra- Corp. Small research organizations to investigate into material
fluoride granulates characterization and process related phenomina with
the aim to develop the economical potential of
multi c-Si for high-efficient solar cells and PV
power modules/systems /6-7/.
The total production of poly silicon in 1990
was at 7700 tons out of which an estimated 690 tons 2.3 Slicing of wafers (geometry)
have been consumed by the PV industry. Th i s For the majority of wafer slicing the inner
estimate is calculated for an average PV efficiency diameter (10) saw cutting is carried through in
of 12% and a wafer thickness of 350 and 450 11m. most of all cases; for the slicing of thinner
Assuming a modest increase of the average PV wafers the multi-wire (MW) saw' cutting is in a
efficiency to 14% and the processing of 200 and 250 continuous progress /8/.
11m wafers, in 1995, the demand of feedstock silicon
materials to produce an estimated 140 MWp c-Si will 2.4 Direct sheet/ribbon approach
rise to 1500 tons, accordingly. This would be in As already indicated above, the direct sheet/
line with a 10% share out of the semiconductor ribbon production combined with a low-cost
grade hyperpure poly silicon demand which is anti- feedstock sil icon would be the utmost advantageous
cipated to total to 15000 tons in 1995 (Source: substrate production methode if we were able to
Dataquest). eliminate the metal impurities and accomplish a
sufficiently high crystal perfection.
2.2 Crystallization (crystal perfection) Provided a crystallization with the today's non
Mono c-Si: electronic grade poly/remelt silicon quality the
As we have the raw material in hyperpure results in terms of output and efficiency for the
quality we have solved the first problem circle or various methods are shown in Table 2 (from Eyer et
technical key essential: the purity. Although big al /9/). This survey makes evident that the ez/ Fl
efforts have been undertaken to introduce low-cost as well as the cast multicrystalline reference
solar-grade silicon by direct metallurgical material by output times efficiency comes out best
refining these approaches failed so far because of if the PV efficiency is given a higher weight. The
the purity requirements which do not deviate very difference will even increase if lower grade
much from the va lues requested for a standard Ie poly/remelt feedstock silicon is used because of
substrate material. the refining effect by segregation in the case of
Of course, the final bulk purity level in the Cl, Fl, and cast multi c-Si and the lack of
solar cell is the determining factor and has to be segregat i on in the case of all direct sheet/ri bbon
seen as the result of feedstock silicon, crystal crystallization procedures.
growing and wafer processing. Evidently, the Hence, for the actual size of production volume
convent iona 1 crysta 11 izat ion procedures of bu lk c- and consumpt i on the we 11 estab 1i shed procedures of
Si production have the important advantage ,of mono and multi c-Si are the most feasible ones.
purification by segregation, a matter of fact WhlCh This is in line with the conclusions of Ciszek /10/
all direct sheet/ribbon procedures are more or less and DeMeo et al /11/. Regarding the most advanced
missing because of the much higher solidification techn i que by means of output, the RAFT techn i que,
velocities /5/. Another crucial feature is the it shou 1d be ment i oned that th i s program has been
distance between the solid-liquid interface and any abandoned because of the lack of commercialization
surrounding part of a crucible, a die-block or a of a substrate material with an efficiency below
ramp material which will be more or less in a 10%. The most viable potential of direct sheet/
physical/chemical reaction with the liquid silicon ribbon production still has to be seen in the
herewith giving a deteriorating quality impact to presently operating techniques of D-Web and EFG at
the solidified crystalline bulk material. Westinghouse and Mobil Solar, respectively, as well
as the S-Web at Siemens.

503
 Table 2: The various methods of direct sheet/ribbon approach compared
to sliced bulk c-Si (Eyer et al /9/).

NAME COMPANY METHOD PRODUCT CRYSTAL STRUCTURE O~TPUT EFFICIENCY PROBLEMS

(em /min)
Cast Hoxan growth from melt ribbons columnar gra~ns 200 $ 10
Ribbon Japan shaping by a 100 11111 wide of some 1 mm
flat C-cavity
D-Web Westing- growth from melt ribbons monocrystalline 10 $ 17 high skill
house shaping by 50 mm wide (twin) needed
USA two Si-dendrites
EFG Mobi I growth from melt octagonal large elon~ated grains 160 $ 15 carbon, SiC,
Solar shaping by C-die tubes 800 11111 of many cm stress
USA circumference dislocations 104/cm2
RAFT Wacker growth from melt square sheets columnar grai2s 20.000 $ 10 dislocation
Germany on re-usable ramps 50 x 50 mm of appro 1 mm density
dislocations 107/cm2 ramp reuse
RGS Bayer growth from melt on ribbons columnar grai2s 6.000 not rep. grain size,
Germany re-usable substrate 100 mm wide of appro 1 mm defects
dislocations 104/cm2
Spin Hoxan spinning melt into square sheets columnar 2rains 400 $ 12 oxygen,
Cast Japan flat C-cavities 100 x 100 mm of < 1 cm dendritic
structure
SSP FhG-ISE zone melting of ribbons large elon2ated grains 20 $ 13 bowing
Germany powder layer by 100 mm wide of some cm width growth
optical heating dislocations 10 3 - 10 6 rate
S-Web Siemens growth from melt on ribbons columnar g~ains 1.000 $ 12 flatness,
Germany a C-net 100 mm wide of some mm impurities
dislocations 10 4 - lO6
TSE Academy growth from melt on ribbons
Sci USSR a graphite foil
through a feeder
For reference
CZ/FZ different growth from melt 150/125 mm monocrysta 11 ine 200 $ 18 1) slicing
sources diameter 14 - 15 2)
SILSO growth from melt 400 x 400 mm multicrystalline 1.600 12 - 13 2) slicing

1) Production scale for space application

2) Large scale production for terrestrial PV application

Table 3: Correlation between the Si-wafer

3. EFFICIENCY-COST TRADE-OFF quality and the Si-wafer selling
price tolerance in a PV system.
3.1 Silicon substrate selling price tolerance/
PV efficiency (a case study)
The cost structure of a PV system can be Si-wafer multi mono(CZ) mono(FZ) brand"x"
di vi ded into 20% s il icon, 20% process, 25% modu Ie, qual ity
and 35% system /12/. The cost structure of a PV
module can be divided into 30% silicon, 30% PV efficiency 12% 14% 18% 10%
process, and 40% modu Ie, respect i ve ly. Assumi ng a per 100 cm 2 1,2 Wp 1,4 Wp 1,8 Wp 1,0 Wp
market selling price of OM 18,50 per Wp for the
system and OM 12,00 per Wp for the module the Selling price OM 18,50 OM 18,50 OM 18,50 OM 18,50
selling price for 100 cm 2 system or module will per Wp
amount to the price values listed in Table 3 and 4
distinguishing into silicon substrate qualities of Selling price OM 22.20 OM 25,90 OM 33,30 OM 18,50
10, 12, 14, and 18% efficiency. Related to a per 100 cm2
reference cost/price of OM 4,20 per 100 cm 2 multi PV system
c-Si wafer the acceptable cost/prices for the other
substrate materials then amount to the values Selling price OM 4,20 OM 7,90 OM 15,30 OM 0,50
listed. per 100 cm2
Si-wafer

504
 Table 4: Correlation between the Si-wafer s.~~------------------------------------,
quality and the Si-wafer selling PtIC. Monoe.yetalllM Wal ••
price tolerance in a PV module. [ DIIIIPC}
10 _1Oem2 0ul 0' DIll. 125mm
'4.00

Si-wafer multi mono(CZ) mono(FZ) brand"x"

qua I ity 12 .00

PV efficiency 12% 14% 18% 10% 10.00

per 100 cm2 1,2 Wp 1,4 Wp 1,8 Wp I,D Wp
• .00
Selling price OM 12,00 OM 12,00 OM 12,00 OM 12,00
per Wp
• .00
5 .80
Selling price OM 14,40 DM 16,80 OM 21,60 OM 12,00
per 100 cm 2 4 .00 350/ 250"", 200,.n
PV module
Selling price OM 4,20 OM 6,60 OM 11,40 OM 1,80 '1180 '884 ,8ee '882
per 100 cm2
Si-wafer
(a)

The example demonstrates the weight of the sil icon

substrate quality in the total cost structures. It
also makes evident the already existing difference p,,..
s.~
Multleryetanln. Wa'.,
in price acceptance for the modules compared to the \. 10_10cm 2
systems. Certainly, any significant cost reduction ( OM/ pe l

in the system or the module fabrication will change • .00 '"

these realations markedly. • .00 ~5 .• 0
~
3.2 Materials availability 5 .00
As a 1ready ment ioned above, in 1990, the tota I
consumption of feedstock sil ico~ raw mat~rials of " .00
the PV industry for the productlon of C-Sl and the
equivalent amount of PV modules in MWp and the
resulting average weight in gram per 1 Wp has been 3 .00

as follows:
1990: 690 t - 33 MWp 2 .00 350Jo"'l 2S0,0m 200,.m

21 g - 1 Wp.
Provided an improved overall PV efficiency from 12% "ao ,084 'eBa
to 14% an increase in crystallization yield from
70% to' 75% and the reduction of wafer thickness
from 350/450 ~ to 200/250 ~m a projection could be (b)
given as follows:
1995: 1500 t - 140 MWp
11 g - 1 Wp. Fig. 5 Selling price development for (a) mono c-Si
and (b) multi c-Si.
Hence, the avai labil ity of the raw materials
will not be the problem. The dominating question
wi 11 be: Can the PV industry cope with an increase
from 33 MWp to 140 MWp or 33% AGRs durinQ five con-
secut i ve years and a 11 the neccessary 1 nves tments the solar silicon substrate materials are concerned
into the correlated production equipment? we have to concentrate on the most feasible
improvements as they are: thinner wafers and, as a
3.3 Advanced solar silicon materials consequence, the imp lementat ion of the appropriate
I n the semi conductor VLSI product ion, handling and processing equipment.
especia lly the Mega ORAMs, the advent. of advanced
silicon materials have made blg progress 3.4 Silicon wafer cost/price development
introducing oxygen controlled, bulk micro defect Fig.5 (a) and (b) are showing the selling price
controlled internal gettering materials. Together deve lopment for so 1ar grade mono c-S i wafers and
with the progress in the wafer geometry and multi c-Si wafers, respectively. Starting with
cleanliness and the steady increase of wafer semiconductor prices we went down to too low prices
diameter the semiconductor device industry took tak ing advantage of cheaper po ly/reme It silicon as
advantage through high yields, high outputs and a by-product from the semiconductor silicon
high economical processing. The higher price of the product i on; the product i on of mono c-S i was
wafer was less important. favoured by low-cost equipment for CZ pulling and
In that respect, advanced solar silicon was and ID cutting and the production of multi c-Si was
still is seen with cost/price reduction only. favoured by governmenta 1 support of R&O programs.
Although tremendous efforts have been spent to ~ind As the vo 1ume increased we had to ra i se se 11 i ng
a low-cost solar grade silicon the expectatlons prices in order to cover our product ion costs and
could not be fulfilled because of the lack of the neccessary investments to expand production
eff i c i ency. On the other hand, the process i ng of capacity. How can we continue?
the estab 1i shed mono and mu 1t i c- Si we 11 advanced Comparing mono with multi c-Si we have to see
and the theoretical I imits have been nearly the price difference in relation to the difference
reached. What conclusions can be drawn? As far as in the PV eff i c i ency and its impact in the tota I

505
system/module cost structure. In a short term /6/ Workshop "Defekte in multikristallinem
consideration the mono c-Si may offer some Silicium.", Fraunhofer-Institut fUr Solare
add i tiona 1 advantage due to the access to id 1e FZ Energiesysteme, Freiburg/Glottertal,
and CZ capacity and the introduction of reduced September 12-14(1990)
wafer thicknesses and MW cutting. In a long term
cons i derat ion the mu 1t i c-S i wi 11 offer the
advantage of the large rectangular wafer area. /7/ 2nd Symposium on Photovoltaics Materials,
This is a conservative view showing that with A.F.M.E., CNRS-PIRSEM, Centre de Recherche
the existing technologies for the silicon substrate Nucleaires Strasbourg, Fraunhofer Gesellschaft,
materials the key essentials efficiency and Nice France, December 10-12(1990)
availabi 1ity can cope with the foreseeable growth
of the PV industry within the next five years. /8/ C. Hauser, Proceedings of the 9th E.C. Photo-
voltaic Solar Energy Conference, Freiburg,
September 25-29(1989)
4. OUTLOOK
/9/ Silicon Sheet Materials for Solar Cellsj
The PV confl ict A. Eyer, A. Rauber, A. Goetzberger,
We tried to figure out that even with a Optoelectronics Vol.5, 239-257(1990)
conservative estimate on the today's existing
technologies of Czochralski and/or casting /10/T.F. Ciszek, Silicon Materials Quality and
crys ta 11 i zat ion together with advanced cutt ing Throughput, IEEE 0160-8371, 31-38(1988)
technique there is enough sil icon for solar
application at economical prices. However, we have /ll/E.A. DeMeo, F.R. Goodman Jr., T.M. Peterson,
to distinguish between a normal business J.C. Schaefer, Photovoltaic Power: A U.S.
development, even at high growth rates, and the Electric Utility R&D Perspective, Proc.21 IEEE,
visions to replace fossile and nuclear energy Photovoltaics Specialists Conference, Orlando,
production by photovoltaics. At the time being, the Florida (1990)
PV industry is in a conflict with itself. The
present situation is: /12/B. Voss and J. Knobloch, Fraunhofer Institut
- a relatively small market seize of worldwide fUr Solare Energie, Freiburg, Final Report
300 to 400 mio US$ BMFT 0328711 A (1990)
- annual growth rates which are continuously
overestimated
- giving low/no return on investment.
The future anticipations from the PV industry are:
- to serve as a renewable energy source
- at a "competitive" cost/price value per kWh
- by huge capital investments to enlarge
production capacity.
How to get out of the confl ict? We have to
bring the PV production to profitability, we have
to prope 1 equ ipment manufacturing and process
engineering for continuous improvements 1n the
segments of high efficient solar cell processing
and PV systems manufacturing.

Where is the market place? Taking into consider-

ation that 80% of all people worldwide do not have
any electricity at all it is a challenge for all of
us to introduce PV energy as an already proven
technology, technically and economically, to
improve living conditions in the third world, e.g.
growing food, preparation of food, education and
medical care. To imply photovoltaics as a
renewable energy source in our industrialized
nations will be a longer way to go.

5. References
/1/ Investigations for the Enquete-Comission of the
German Bundestag "Consequences of Technology -
Estimates and Valuing." DIW, DLR, EWI, FfE,LBS,
University of Oldenburg, Stuttgart, June (1990)
/2/ Source: Photovoltaic Insider Report, 1011 W.
Colorado Blvd., Dallas.
/3/ Source: NEDO Report 1(1990) and PVIR 2(1991)
/4/ A. Rauber, Fraunhofer-Institut fUr Solare
Energiesysteme, Freiburgi presented in "Trends
der Photovoltaik-Entwicklung." at the German
National Symposium of Photovoltaic Solar Energy
in Kloster Banz/Staffelstein, March 5-8( 1991)
/5/ J.A. Burton, R.C. Prim, W.P. Slichter, J. Chem.
Phys. 21, 187(1953)

506
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

BEHAVIOUR OF MULTI CRYSTALLINE SILICON MATERIALS

UNDER HIGH EFFICIENCY PROCESS CONDITIONS

W. Schmidt, W. Ebner*, K.-D. Rasch, G. Wahl

TELEFUNKEN SYSTEMTECHNIK GmbH, Solar Cells
Theresienstr. 2, D-7100 Heilbronn, Germany
* TELEFUNKEN SYSTEMTECHNIK GmbH, Space and New Technologies
Industriestr. 29, D-2000 Wedel, Germany

ABSTRACT. Based on the analysis of high efficiency processes

and theoretical calculations, a catalogue of requirements was
formulated that should be fulfilled by a multicrystalline
silicon substrate material to allow cost-effective processing
of large-area cells with high efficiency. A simple test
process was applied to elucidate if nowadays commercially
available silicon materials can at least partially match these
requirements. Two different high efficiency cell processes
lead to the same results: Two of the five tested materials
show negligible degradation of electrical material properties
after additional high temperature (1000 'C/1100 'C) treatment
and do not need hydrogen passivation. One material needs
hydrogen passivation to achieve best results. Two materials
show distinct degradation of electrical properties if high
temperature treatments are applied.

1. INTRODUCTION 2. EXPERIMENTAL

High efficiency modules (1, 2) were The following commercially avail-

fabricated from single and multi- able multicrystalline silicon materials
crystalline silicon cells with 17.8 % and were investigated:
13.8 % average efficiency under encapsu- BAYER (Bayer AG,Germany) material,
lation conditions, respectively. These S = 0.8 - 1.5 Ocm
cells had no silicon oxide passivated - HELIODINAMICA (Heliodinamica, Brasil)
surfaces. To achieve improved cell effi- .9 = 1.0 - 2.0 Ocm
ciencies, development was started in our - HEM, (Crystal Systems, USA)
laboratories to realize a cell structure J~ = 1.5 - 3.0 Ocm
shown in principle in Fig. 1. Recently, - OTC (Osaka Titanium Co.,Japan) material
19.3 % conversion efficiency (19.0 % jP = 1.0 - 2.0 Ocm
average) were obtained under AM 1.5 - SILSO, (Wacker-Chemitronic, Germany)
condi tions (reference cell calibrated at f = O. 5 - 3. 0 Ocm
SERI) with 25 cm' large cells, fabricated
on CZ, <100>, 10 Ocm boron-doped substra- The wafer thicknesses were between
tes with the structure of Fig. 1, except 300 and 400 J.lm.
the back side, which was completely Since silicon oxide surface passi-
metalized. vated solar cell structures are not
It was also tried to apply this and compatible with the hydrogen passivation
similar structures (3, 4) to mul ti- process of material defects, and the high
crystalline standard SILSO substrates. efficiency processes often include a
Though a 8 cm' SILSO solar cell with over thermal treatment (oxidation) at 1000 'C
15 % efficiency was achieved, it failed or higher (3,7), the influence of such
until now to realize this efficiency for process steps was investigated by
large-area SILSO cells with a cost-effec- applying the following simple test
tive process sequence in large quanti- process:
ties. One reason can be, as M. Green et Alkaline etching to remove saw
al reported (5), that standard SILSO is damage and an additional proprietary
not the optimum choice for the fabri- cleaning step
cation of high efficiency mUlticrys- Oxidation at 1000 or 1100 'C for 1 h
talline silicon solar cells. in an oxygen/TCA atmosphere without
Based on our experience an analysis ramp up or down steps, followed by
of the high efficiency processes and on oxide etching
theoretical calculations (6), a catalogue Heavy phosphorus diffusion (30-
of requirements was formulated (Table I) 40 o/a )
that should be fulfilled by a multicrys- Screen-printed Ag back contact
talline substrate material to allow cost- Evaporated TiPdAg front contact with
effective processing of large-area cells 8 % shadowing
with high conversion efficiency. Since Annealing at 400 ·C for 15 min in
our company has no in-house multi- reducing atmosphere
crystalline silicon material production, Hydrogen plasma treatment for passi-
investigation was started to prove how vation
far those materials, currently available
for us on a commercial base, can meet the
high efficiency process requirements.

507
 For the exp~riment, typical groups comparing the results for the two
of about 10 wafers were selected from processes, we find similar results for
each material. Bayer and SILSO material. The differences
To prove the test results, solar for the HEM material can be explained by
cells were produced from the same lot of different resistivity values of the
each material, using the standard buried- wafers within the tested lot. However,
contact solar cell (BCSC) process for FZ the BCSC sequence leads to distinctly
material (8) or a conventional process lower solar cell characteristics for the
sequence (2) for a n+p structure without OTC material. From our test process
surface passivation and high temperature results, we can conclude that the
steps, but including hydrogen passivation material was degradated during the high
treatment. temperature steps in the BCSC process.
Cell performance was measured at Solar cells from Heliodinamica material
25 ·C under air mass 1.5 simulated irra- had the lowest solar cell efficiencies
diation. fo: our standard test process sequence.
Th~s and the observed degradation by high
temperature treatments explains the
obtained comparatively low values.
Altogether, the results for these
two processes are in good accordance with
3. RESULTS AND DISCUSSION the findings of the test process
experiments.
The test process without oxidation
step, but with hydrogen passivation
treatment is defined as a standard 4. SUMMARY AND OUTLOOK
process. The electrical data of
corresponding solar cells are used as A first investigation of commer-
reference. The differences due to omitted cially available multicrystalline silicon
hydrogen passivation and the additional wafers has indicated that the material
oxidation process at 1000 and 1100 ·C are properties of wafers from Crystal Systems
shown in Fig. 2 for the open-circuit and Bayer AG fulfill important require-
voltage and in Fig. 3 for the short- men~s. for cost-effective large-area high
circui t current. A remarkable influence eff~~~en~y ~ell processing: Hydrogen
on the fill factor was not observed in pass~vat~on ~s not necessary, negligible
any case. After the oxidation step some degradat~on of electrical material pro-
wafers. exhibited a strong wa~page, perties at high temperatures of 1000 • C
depend~ng on the crystalline structure
and nearly homogeneous resitivity distri-
which is troublesome in following bution over wafer area. The OTC material
automatic processing steps. The probable also needs no hydrogen passivation for a
reason of the warpage is the relief of conventional process sequence but
internal stresses incorporated during the degrades during a high temperature treat-
crystallization process. ment. The same holds for the Helidonamica
The results in Fig. 2 and 3 can be material which currently. suffers from
summarized as follows: minor electrical materia~ properties,
Only the SILSO solar cells show compared to all other mater~als. SILSO is
dist~nct. improvements ~ue to hydrogen sufficiently stable under high tempera-
pass~vat~on treatment ~n the standard ture. tr 7atments but needs a hydrogen
process without high temperature pas~r~'at~c:m treat~ent to achieve highest
oxidation step. eff~c~enc~es wh~ch imposes severe
For BAYER, SILSO and especially the restrictions on possible cell structures
HEM material, the oxidation step causes a and fabrication process sequences. It
minor reduction of the cell parameters. als,? .suffers from large resistivity
The reduction is negligible for the var~at~ons over the wafer area. The
temperature of 1000 ·C. :esults found with our test process were
The HELIODINAMICA and especially the ~Il accordance to those of the high
O~C samples degrade distinctly in the efficiency processes. It should be
h~gh temperature step. This degradation
can part~ally be compensated by hydrogen stressed that all these findings hold for
nowadays commercially available multi-
pass~vat~on.
The open-circuit voltage and short- crystalline silicon materials. From some
circuit current data for the solar cells suppliers, experimental multicrystalline
fabricated on the various materials with silicon samples were received which
tl.le co~ventional and the BCSC process are exhibited an already improved adaption to
g~ven ~n Table II.
the cataloque of cost-effective high
A comparison of the data for the efficiency process requirements. Anyhow,
conventional process reveals that SILSO the investigations will be continued to
solar cells have lower open-circuit achieve in cooperation with the suppliers
voltages as cells from other materials a multicrystalline silicon material which
with similar resistivity. The same result allows the cost-effective fabrication of
was attained in our production line for large-area multicrystalline silicon solar
screen-printed solar cells. A possible cells
explanation for this finding is the
normally observed large resistivity
variations of up to ± 25 % within most 5. ACKNOWLEDGEMENT
SILSO wafers, which were not measured for
the other materials. It is also worthwile This work is partially supported by
to note that rather high short-circuit the Bundesministerium fur Forschung und
currents are attained for the Bayer Technologie
material with about 1 ncm resistivity.

508
6. REFERENCES HEM OTe SILSO

(1) H. F1edl et aI, Techn. Digest 4th

Int. PVSEC, Sydney, 1989, p. 455
(2) K.-D. Rasch, W. Schmidt, Proc. 9th
E.C. PVSEC, Freiburg, 1989, p. 175
(3) J. Knobloch, A. Aberle, B. Voss,
Proc. 9th E.C. PVSEC, Freiburg, 1989
p. 777
(4) A. wang, J. Zhao, M. Green, Appl.
Phys. Lett. 57 (6), 1990, p. 602
(5) M. Green, A. Blakers, S. Wenham,
Proc. 9th E.C. PVSEC, Freiburg,
1989, p. 301
(6) J. Knobloch, A. Aberle, W. Warta, B.
Voss, 5th Int. PVSEC, Kyoto, 1990
(7) A. W. Blakers et aI, Techn. Digest
4th Int. PVSEC, Sydney, 1989, p. 801
(8) H.-W. Boller, W. Ebner, Proc. 9th
E.C. PVSEC, Freiburg, 1989, p. 411 Fig. 3: Relative differences in short-
circui t current due to high temperature
oxidation at 1000 ·C or 1100 ·C for 1 h
and hydrogen passivation treatment.

':',,$ II ]te, Mult,(.f,S oll,ne Silicon Solar Cell

Table I Requirements on multicrystal-
Contact line silicon wafers for cost-
AAI Passivation effective large-area solar
Loyer cells
Diffusion length in the base about
200 ~m or more for 1 Ocm resistivity
p
Hydrogen passivation not necessary
( H+ -+ 0)
p. No degradation of electrical or
mechanical material properties due
mmDllIWi!iIEI!i!III!IillI!lIEill!ll!ji 1G1lI!i!Ei!l!l!ffi!OOlJ1l1wmum- Passivation Loyer to treatments at high temperatures
(~ 1000 ·C) or with large tempe-
Contact rature gradients (» 100 ·C/min)
Low EPD « 10 3 cm- 3 )
Fig. 1: Schematic diagram of front and Low oxygen and carbon content
rear surface passivated solar cell Homogeneous resistivity distribution
structure. over wafer area

HEUO-
BAYER
DINAMICA
HEM OTe SILSO Table II comparison of the data of solar
cells fabricated on various

II
0

f%~ ~~
I I multicrystallines silicon
mil I I materials fabricated by a
I I ~ I I conventional cell process.
-4
I I I I
~
~ ~
I I I I Material Mean conventional
a
."
c 0 0 I I I I resistivity process

"--
g -6 0 0
a 9 (ncm) J sc
> I I I I V?c
<l (mV (IlIA/em' )
I I I I
I
-1 2 I I I
atter Ii"" BAYER 0.9 593 29.8

-\ 5
0 before H'"
I
I
I
I I
I

HEM 2.6 574 30.0

Fig. 2: Differences in open-circuit vol- OTC 1.7 583 29.8

tage due to high temperature oxidation at
1000 ·C or 1100 ·C for 1 h and hydrogen SILSO 1.0 580 28.7
passivation treatment.
SILSO 1.8 571 29.5

509
Table III Comparison of the electrical
characteristics of solar cells
on various multicrystalline
silicon materials fabricated by
the buried contact process.

Material Mean Buried-Contact

resistivity
(ncm)
(~~r
J,c
process

(mA cm")

BAYER 0.9 592 29.7

HELIO-
DINAMICA 1.7 558 27.8

HEM 2.6 564 30.4

OTC 1.7 565 29.3

SILSO 1.0 583 28.9

SILSO 1.8 570 29.8

510
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

RECENT ADVANCES IN SILICON INVERSION LAYER SOLAR CELLS AND

THEIR TRANSFER TO INDUSTRIAL PILOT PRODUCTION

R. Hezel
Institut fUr Werkstoffwissenschaften 6
Universitat Erlangen-Niirnberg, 0-8520 Erlangen, FRG

W. Hoffmann and K Jaeger

NUKEMGmbH.
Postfach 11 00 80
D-{)450 Hanau, FRG

ABS.TR~cr. The potential. of 1.0 x lOcm2 MIS inversion layer solar cells for high
~fficle~cles w~ demonstrated uSl!lg simple, low-temperature and short-time processing in a
mdustnally onented laboratory Ime. Peak efficiency values exceeding lS% at 1()()()W/m2
could be obtained remaining almost constant down to an irradiance of 600W/m2.
Thin-silicon bifacially sensitive inversion layer solar cells with high effective efficiencies on
monocrystalline and multicrystalline silicon are presented and the conditions for low-
temperature back surface passivation by plasma silicon nitride deposition are outlined.
Result~ of outd?Or measureme~ts for a ?if~cial inversion layer cell module are presented
both With and Without sun tracking. A gam m output power of more than SO% relative to an
one-sided module was observed.

1. INTRODUCTION 2. CELL STRUCTURES AND OPERATION

Very high conversion efficiencies for today's dominant In our laboratory both the standard type MIS inversion
solar cell material silicon have recently been achieved by layer solar cell with a continuous ohmic back contact was
sophisticated cell structures of the diffused p-n junction type on developed as well as the bifacial Iype inversion layer solar cell
crystalline wafers of conventional thickness /1/. In order to reach with a gridded back contact in combination with back surface
the cost level needed for large-scale use of terrestrial solar cells passivation. Since in both cases the front side configuration is
both low manufacturing costs and high efficiencies are required. identical only the cross section of the bifacial cell is shown in
As an alternative to obtain high conversion efficiencies at low Fig.I. As an induced-junction device the inversion layer (IL)
production costs solar cells based on induced junctions are solar cell is based upon the formation of an inversion layer at the
considered promising. There are several different approaches to semiconductor surface, which provides a conducting path to the
this cell type in the literature !2~/. adjacent metal-insulator-semiconductor (MIS) grid fingers for
In the present paper some recent advances in crystalline the minority charge carries generated by the impinging light.
silicon inversion layer (IL) solar cells with metal-insulator- According to Fig.l the inversion layer is induced by fixed
semiconductor (MIS) contacts using aluminum as grid positive charges present in the transparent insulator film on top
metallization and plasma silicon nitride as charged dielectric are of the semiconductor substrate. This film also has to serve as
reported together with the present status of the technology passivation and antireflection layer. Silicon nitride has been
transfer from the laboratory to large scale fabrication. proven to be particularly suitable for this crucial part of the
In addition to the simple processing for this cell type device due to its unique properties such as very high density of
further significant reduction in cost per watt are achievable by fixed positive interface charges to obtain a low inversion layer
reducing the thickness of the silicon substrate below 100llm as sheet resistance, long term stability of these charges, excellent
well as by utilizing also the light impinging on the rear side of AR coating and surface protecting properties /8-11/.
the cell. For this purpose a simple back surface passivation The electrical properties of the cell are determined by two
technique also based on silicon nitride was introduced which is main regions: (i) the tunnel MIS contacts and (ii) the thin film
compatible with the low temperature processing characteristic for system and its interfaces in the active area between the collecting
inversion layer solar cells. Thus very efficient bifacial solar cells contacts.
without any highly doped region and completely processed at For an ideal AI/SiOx/p-Si tunnel MIS contact, the band
temperatures below SOO°C could be fabricated on thin bending q'l's at the silicon surface is given by the work function
monocrystalline and multicrystalline silicon wafers. Data of difference of the metal and the underlying silicon. For 0.8!lcm
these bifacial inversion layer solar cells under outdoor conditions p-silicon and aluminum a barrier height q'l's = 0.87eV is
are presented and their future potential is discussed. resulting, indicating that in thermal equilibrium the silicon
surface beneath the metal is strongly inverted /12/. The ultrathin
FIXED II'o'TERFA E oxide layer (dox - l.3nm) blocks majority carrier (hole) currents
CHARG MI GRID (AI) injected into the metal at forward bias and reduces surface
~ recombination. Therefore very high open~ircuit voltages are
achievable using MIS contacts /13,14/.
The charge distribution and energy band diagram for the
system silicon nitride/thin-()xide/p-silicon in the active area of
the solar cell is discussed elsewhere /S,9/.
Details about the theory of MIS contacts as well as about
ILICO ,p-TYPE modelling of inversion layer solar cells are published in /lS,16/
and /17 ,18/, respectively.
Compared to conventional diffused or implanted p-n
junction solar cells MIS inversion layer solar cells are distin-
OHMIC
guished by several basic advantages. Simple and short time
CO ITACT processing at low cost is possible. Due to the presence of an
(AI) induced junction, high-<ioping effects such as bandgap narrow-
LIGHT ing and Auger recombination are eliminated. Since there are low
temperature (!> SOO°C) steps only, no lifetime degradation occurs
Figure 1: Schematic cross section of the bifacial MIS in bulk silicon resulting in a high sensitivity for long
inversion layer solar cell. wavelengths. In the case of multicrystalline silicon the problem
of enhanced diffusion along grain boundaries is eliminated and
the glow discharge hydrogen passivation of the silicon substrate
can effectively be performed prior to cell processing. Thus no

511
restrictions are existing concerning the hydrogenation parameters
such as temperature, processing time, plasma power etc .. Even MI - IL CELL
with multicrystalline silicon, surface texturing turned out to be 15 10 x JOcml ~r--

very effective for inversion layer solar cells /19/.

3. FABRICATION PROCESS
-
The simple low temperature fabrication process both for
r--
the standard type and bifacial MIS inversion layer solar cell is
illustrated in Fig.2. As can be seen, the fabrication process of the ,.........
bifacial type MIS IL solar cell only slightly differs from that of

snn
the standard type cell (i) a much thinner substrate is used (ii) the
back contact is gridded and (iii) plasma silicon nitride is
deposited onto the back surface at 450°C.

flll.
Details of the fabrication process are given in /20/ and
0
In
<12 dl.! d .) dJJi <14 <14..5 <-15 cl5.S
ANDARD ~ LL 8IFACI AL CELL
EFFle l E CY [%)

p-Si
I U pJlJ.,,-lvllion I p-Si Figure 3: Efficiency distribution of 10 x 10cm2 mono-
Surr.t:~ luturina: crystalline silicon MIS inversion layer solar cells (AM1.5,
Cleanlnl,l: lOOOW/m2, 25°C).

Ililick (onla('1 form.llliion

1.02
n:ack 1ur(lIC'e p...ssiv.::Illon
by plDJm.:I sitkon nilride ~
dfpllldlion « 4501l C) SIN
1.00

Tun nrl o"ide growth

"'1.4n m.SOO"C ~
!: 0.98
~

C ELL AR EA 10 l< IOcm 2

MI rronl ('onlacl l"id 0.96 • MilL
CuiulII con l:lmimillion
a N+ P

0.4 0.6 0.8 1.0 1.2

Pb rna filltOn n ltrlde
dtposilion CM1 front :ddt IRRADIA C E [ 103W/m2]
(-ISO"C)
Figure 4: Experimental light-intensity dependence of the
efficiency of 10 x lOcm2 MIS inversion layer and conventional
Figure 2: Fabrication process for silicon MIS inversion layer N+P solar cells. 11 is normalized to its value measured at
solar cells. lOOOW/m2.

4. LARGE-AREA STANDARD MIS INVERSION

LAYER SOLAR CELLS power plant, since insolation levels much lower than lOOOW/m2
are by far prevailing during normal terrestrial operation.
For the standard (monofacial) inversion layer solar cells Several modules consisting of ten MIS-IL 10 x lOcm2
of laboratory size (2 x 2cm2) presently total area efficiencies of solar cells, as shown in Fig.5, were fabricated and successfully
about 16% are achieved on monocrystalline silicon and 14% on passed the specified endurance tests /27t. An innovative encap-
multicrystalline (Wacker Silso) material. Using a more sulation technique developed at Nukem was applied offering the
sophisticated fabrication process with improved front surface possibility of producing large area modules up to 1.5 x 2.Om2
passivation active area efficiencies up to 17.9% were resulting /27/. Currently a pilot plant is under construction for manufac-
on monocrystalline silicon /22/. Furthermore, the effects of UV turin~ of MIS inversion layer solar cells with IMWp annual
radiation on cell performance were studied and means for their capacity.
elimination could successfully be developed /23/.
The simple technology and minimum number of
processing steps associated with MIS-IL solar cells are
particularly advantageous for future economic large-scale
automated manufacturing. Since a prerequisite for this is an
increase in cell area at least up to 10 x IOcm2 , the feasibility of
the crucial production steps was demonstrated for these large
area cells including homogeneous distribution of the insulator
charge density and thus inversion layer sheet resistance,
homogeneity of the ultrathin tunnel oxide and formation of a
narrow metal grid fl4,25/.
The distribution of the efficiencies obtained for
10 x IOcm2 monocrystalline silicon MIS-IL solar cells in a
industrially oriented laboratory line is shown in Fig.3. As can be
seen, efficiencies higher than 15% could be achieved whereby
43% of the cells are ranging between 14% and 15%.
An important result could be obtained regarding the
light-intensity dependence of the solar cell data /26/. In Fig.4 the
efficiency, normalized to its value at l000W/cm2 , is plotted as a
function of the irradiance for 10 x IOcm2 MIS-IL and conven-
tional N+P solar cells. It can clearly be seen that the decrease in
efficiency with decreasing irradiance below lOOOW/m2 is less
pronounced with MIS-IL solar cells compared to N+ P cells.
This advantage of inversion layer solar cells leads to a sub- Figure 5: Inversion layer solar cell module (cell size
stantial increase in the overall yearly efficiency of a photovoltaic lOx IOcm2)

512
 s. THIN-SILICON BIFACIAL MIS INVERSION
LA YER SOLAR CELL
120
®
5.1 Back Surface Passivation 100
The inversion layer solar cell technique is particularly
suitable for the application of very thin selfsupporting silicon
substrates due to its simplicity (handling!) and due to the
~
"I
.. 80
absence of high temperature treatments. E
For high bifacial sensitivity apart from the drastically -'"' SILICO.
reduced substrate thickness (saving material costs!) an excellent "§: ,------.oL,
NITRIDE
QN
back surface passivation is of prime importance. As shown in 60
Fig.1 plasma silicon nitride was used by us as back surface r:S F;;;;;;;:;:~f1i:~~N
passivation film in the area between the ohmic contacts which 40
fNVERSI6N
LAYER
DII
simultaneously provides a very good antireflection and
protective coating.
By the nitride film both the reduction of the surface state
density and the presence of strong inversion accomplished by 20
TN ~ 4S0' C + SOOOC ANNEALING
fixed insulator charges ("charge-induced passivation") can be
utilized for the suppression of surface recombination /8/.
In Fig.6 the dependence of the interface state density Oil 0
and recombination velocity So at the depleted surface for
structures with plasma silicon nitride on silicon are plotted as a ®
function of the nitride deposition temperature. A minimum value
of 3 x 1010cm-2ey-l for Oil and an So value in the 1000m/s 150 TN =270' C
range could be achieved for the nitride film deposited at 450°C
and annealed at 5 IOoe /8/. 7
Dit So E
o AS DEPOSITED '"' 100
....
0

.... ....
• 0 AFTER AN 'EALINC
AT sure, J5min

.::- loll .. ~

o
2:;

....
50
o
I
;>
I

1
Q- 0
1011 4 6 8
CE TRATION [10-5 moUlj

Figure 7: Midgap interface state density (a) and fixed insulator

charge density (b) as a function of the concentration of the
cesium-contaming solution for structures of plasma silicon
1010 ~_ _--::-:!-::-___ ~-;;- __~*,,_ _ -' 10 I nitride on thin-oxide-covered p-silicon (100), 0.5!lcm. Nitride
200 300 400 500 deposition temperatures are TN = 270°C and TN = 450°C,
DEPO ITION TEMPERAT RE lOCI respectively.
Figure 6: Effect of deposition temperature an.d ann~aling of
plasma silicon nitride films on silicon upon mldgap mterface electrons in the inversion layer, induced by the positive insulator
state density 0it and recombination verocity at the depleted charges, is smaller by a factor of 1.7 at high cesium concen-
surface So. trations for the 450°C ftlm compared to the 270°C nitride film.
The drastic effect of the nitride deposition and annealing
In order to obtain efficient charge induced passivation a temperature on the output current of the cells under rear side
high density of fixed positive insulator ch.arges i!1 conjunction illumination is demonstrated elsewhere /8/. A minimum value of
with a low density of fast interface states IS req~l1red. Insulll:tor 400crn/s was evaluated from spectral response measurements for
charge densities in the order of 1O.13cm-2 ~ achiev~ ac.cording the surface recombination velocity at the back surface, including
to a well established procedure mcorporattng cesIUm mto the the olunic contact areas /8/.
plasma silicon nitride ftlm /28/. It's important to know, how the
cesium contamination affect the density of fast interface states at 5.2 Cell Performance
different nitride deposition temperatures. In Fig.7a for two Up to now, for 2 x 2cm2, 100~m thick monocrystalline
nitride deposition temperatures .TN = 270°<; and ~50oC, respec- bifacial MIS inversion layer solar cells AMI efficiencies of 15%
tively, the interface state .denslty Oil at rru~gap IS plott~ as. a for front side and 13.2% for rear side illumination could be
function of the concentratton of the alcohohc esCI solutton, m obtained /8,21/. The corresponding efficiencies achieved for
which the wafers are dipped prior to the nitride deposition. As multicrystalline (Wacker Silso, 130J.l.lIl thick) bifacial cells are
can be seen, with increasing cesium concentration, Oil is drasti- 13.5% and 7.8%, respectively /29/. From the measured and
cally increasing for structures with silicon nitride deposited at calculated base thickness dependence of the efficiency for
TN = 270°C whereas only a slight increase, starting from a multicrystalline MIS-IL solar cells as shown in Fig.8, the high
much lower ~alue, can be observed for the case of sili~on nitride potential for efficiency improvement under rear side illumination
deposited at TN =450°C. Apparently,.a large n:acuon o.f ~e can be seen. Further thickness reduction below 130llm should
cesium-induced interface states are passivated durmg the mtrlde be possible.
deposition process at TN =450°C and d~ng the postannealin.g In our laboratory a small test panel consisting of
step performed at 500°C. Thus, as an Important r7sult, pasSI- 8 bifacially sensitive 2 x 2cm2 monocrystalline silicon MIS-IL
vation of the silicon/insulator interface at higher cesIUm concen- solar cells was assembled.
trations and thus at higher fixed charge densities QN/q is much In Fig.9a the current output of ~e bifacial panel relati~e
more effective for the 450°C nitride film. However, improve- to a one-sided panel as measured dunng a cloudless day (m
ment of the silicon-insulator interface by increasing the nitride May at Erlangen) is shown, both stationary and with sun
deposition temperature from 270°C to 450°C is paralleled by a tracking. As indicated by the insert, the panel was simply
decrease in the density QN/q of positive charges, as can be seen arranged in some distance of a white painted floor. As can be
from Fig.7b. For both nitride films. a strong incr~ase of Or:</q seen, a gain in current of more than 50% is achieve~ by the
with increasing cesium concentratlon occurs with saturation bifacial panel, for the stationary mode in the mornmg and
values of about 9 x I0 12cm-2 and 1.5 x 1Q13cm-2 for the 450°C afternoon even up to 60% more current is generated compared to
and the 270°C, respectively. As a consequence, the density n of a conventional panel. It is evident that the absolute current output

513
 o. EXPERI IE T L This work was supported by the Bundesministerium ftiT
CALCULATED ( ~ 40Ocmls, L. ~ 115J.1 m) Forschung und Technologie (BMFT) of the Federal Republic of
IS
Germany.
••• •
FRONTlLL
•
REFERENCES:
/1/ M.A. Green, Technical Digest of the Internal. PVSEC-5,
p.603 (1990)
/21 G.C. Salter and R.E. Thomas, Solid St, Electron. 20, p.95
(1977)
/3/ P. van Halen, R.P. Mertens, R.J. van Overstraeten,
AMi 25°C R.E. Thomas and J. van Meerbergen, IEEE Trans El. Dev.
ED-25, p.597 (1978)
°O~--~~1~
00~-J--~ 20~O~~--~300 ~------~~ /4/ R.B. Godfrey and M.A. Green, Appl. Phys. Lett. 34,
CELL TH1CKNESS [""'] p.790 (1979)
/5/ R. Hezel, Solid St. Electron. 24, p.863 (1981)
Figure 8: AMI efficiency as a function of cell thickness for /61 R. Hezel and R. Schtlrner, J. Appl. Phys. 52, p.3076
bifacial multicrystalline (Wacker Silso) MIS inversion layer (1981)
solar cells under front and back illumination. Calculated curves m R. Schomer and R. Hezel, IEEE Trans. El. Dev. ED-28,
are given for a diffusion length of 1151lm and a back surface p.1466 (1981)
recombination velocity of 400Cm/s. /8/ R. Hezel and K. Jaeger, J. Electrochem. Soc. 136, p.518
(1989)
/9/ R. Hezel, Proc. 16th IEEE Photov. Spec. Conf., p.1237
(1982)
1.7 /10/ R. Hezel, Proc. 18th IEEE Photov. Spec. Conf, p.180
(1985)
/11/ K. Jaeger and R. Hezel, IEEE Trans. El. Dev. ED-32,
p.1824 (1985)
! 1.6
/12/ S.M. Sze, Physics of Semiconductor Devices, J, Wiley,
-; 1 New York 1981
_i 1.5 Tl<ACK'NG TIlE SU~
/13/ R.B. Godfrey and M.A. Green, IEEE Trans. El. Dev.
ED-27, p.737 (1980)
/14/ M.A. Green and A.W. Blakers, Solar Cells 8, p.3 (1983)
/15/ M.A. Green, F.D. King, J. Shewchun, Solid St. Electron.
1.4 17, p.551 (1974)
1000 ® /161 J. Shewchun, R. Singh and M.A. Green, J. Appl. Phys.
48, p.765 (1977)
/17/ C.E. Norman and R.E. Thomas, IEEE Trans. El. Dev.
ED-27, p.731 (1980)
<800
.§.
/18/ P. de Visschere, IEEE Trans. El. Dev. ED-30, p.840
... (1983)
/191 R. Hezel and L. Hu, Technical Digest of the Internat.
1J 600
'"'" PVSEC-5, p.701 (1990)
U
(20/ R. Hezel and R.-P. Vollertsen, Prot. 5th EC Photovoltaic
400
Solar Energy Conf., p.l113 (1983)
TRACKJ cn Hi51.1
STAT'ON"PV (SOUTI ' lO")
(21/ K. Jaeger and R. Heze1, Proc. 19th IEEE Photov. Spec.
Conf., p.388 (1987)
200
OUR N'OO;r..l (22/ W. Bauch and R. Hezel, Proc. 9th E.C. Photov. Solar
Energy Conf., p.390 (1989)
8 10 12 14 16 18 (23/ R. Hezel, Proc. 21th IEEE Photov. Spec. Conf., p.239
TlME 1£52:) (1990)
(24/ W. Hoffmann, K. Jaeger and R. Hezel, Proc. 4th Internat
Figure 9: Current output of a bifacial MIS-IL test pan~l Photov. Science and Engineering Conf. (PVSEC-4),
positioned in front of a white back~ound on a cloudless day In p.725 (1989)
Ma¥ at Erlangen, stationary and With sun tracking a) relativ~ to (25/ W. Hoffmann, K. Jaeger, G. Luthardt and R. Hezel, Proc.
an Identical one-sided cell b) bifacial mode and frontslde 9th E.C. Photov. Solar Energy Conf., p.677 (1989)
contribution only. (261 K. Jaeger, W. Hoffmann and R. Hezel, Paper presented at
this conference
in the stationary mode is considerably lower than that obtained (27/ W. Hoffmann, Metal144(7), p.658 (1990)
with tracking the sun. This is demonstrated in Fig.9b for the (28/ R. Hezel, K. Blumenstock and R. Schomer, J. Electro-
front side only as well as for the front and rear side current chem. Soc. 131, p.1679 (1984)
contributions. Referring to Fig.8, effective efficiencies over 20% (291 R. Hezel and K. Jaeger, Proc. 20th IEEE Photov. Spec.
should thus be achievable under real outdoor conditions for very Conf., p.1560 (1988)
thin bifacial multicrystalline silicon inversion layer solar cells.

6. CONCLUSION
With MIS inversion layer solar cells relatively high
efficiencies also with large area cells were obtained by simple
low-temperature processing in an industrially oriented labora-
tory line. Bifacially sensitive inversion layer solar cells are
promising, since very high effective efficiencies are achievable
without complicating the fabrication process. There is a large
potential for further improvement of inversion larer so~ar cells,
mainly by optimizing the tunnel MIS contact and InVersiOn layer
properties as well as by improving the surface passivation on the
front and rear side. Using thin multicrystalline silicon with
proper bulk passivation both for the standard cell and for the
bifacial inversion layer solar cell efficiencies as high as for
monocrystalline silicon should be achievable and thus a further
significant cost reduction will occur.

514
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

INCREASING CELL EFFICIENCY OF S-WEB SILICON RIBBON MATERIAL USING

SYNTHETIC SILICA CRUCIBLES

R. Falckenberg, J.G. Grabmaier, A Lerchenberger

Siemens AG, Corporate Research and Development
Otto-Hahn-Ring 6, D-8000 Miinchen 83, Germany

R. Schindler
Fraunhofer-Institut fUr Solare Energiesysteme
Oltmannsstr. 22, D-7800 Freiburg, Germany

ABSTRACf

For solar cell application Si ribbons 12 em wide and up to 150 m long have been grown with a pulling
speed of 1m/min by the S-web technique. The ribbon material is polycrystalline with grain sizes in the
order of mm2. This should allow to process cells with efficiencies up to 15 %. Efficiencies of 12 % for
test solar cells of 20 mm x 20 mm in size have been reported earlier.
In order to find out if impurities are the real limiting factor for higher efficiencies, a neutron activation
analysis was carried out on Si ribbon material, carbon fiber net, silica crucible, and Si starting material,
respectively. The silica crucible was identified as the source for Zr, Mo, W impurities. In the case of Fe a
clear statement as to its origin was not possible. Replacing most of the natural silica by synthetic silica as
crucible material eliminated some of the impurities. Material grown from this crucible resulted in a
maximum cell efficiency of 13 %. In order to achieve efficiencies exceeding 13% it is necessary to either
completely replace natural by synthetic silica crucibles and/or to keep the time of contact between melt
and crucible low.

apparatus was 12 % (cell size 20 x 20 mm).

1. INTRODUCTION
In order to find out if this discrepancy could be
The S-Web technique has been developed in order to explained by the presence of impurities, a neutron activation
produce a low cost Si sheet material suitable for solar cell analysis was made of Si ribbon material, Si starting material,
processing without additional preparation. In this technique carbon fiber net, and silica crucible.
[1-4] Si melt is used as starting material; it is contained in a
long shallow silica crucible. A carbon fiber net - in contact
with the melt - is pulled horizontally across the melt surface 2. NEUTRON ACTIVATION ANALYSIS AND
with a high pulling speed v. At the lower side of the net a Si EVALUATION
layer crystallizes, the thickness depending on the contact
time between net and melt. Thus, at a given final ribbon In order to identify the impurities in the ribbon and
thickness d or contact time t, the pulling speed v = L/t(d) is their sources a neutron activation analysis was performed.
determined by the length L of the Si melt. With a pilot Besides the Si ribbon the potential sources Carbon fiber net,
apparatus Si ribbons 12 em wide and up to 150 m long have Si starting material, and quartz crucible were examined. The
been grown with a pulling speed of 1 m/min. impurity concentrations (atoms/cm3) determined are listed
in table I.
The material quality is largely defined by the use of a
C-fiber net with relatively wide meshes. Because of its low To facilitate a comparison of the impurity
density the net floats on the melt. Crystallization proceeds concentrations in the ribbon with the concentration of
from only few nucleation points spreading quickly across the possible sources, the following strategy was pursued:
melt surface. Consequently, a material of relatively coarse
grain structure is obtained which should allow cell i. Impurities in the ribbon are grown in only through the
efficiencies of up to 15 %. The efficiency actually obtained Si melt. The impurity concentrations of the different
with test cells of ribbon material grown with a laboratory sources were converted into concentrations in the melt.

515
ii. From the impurity concentration of the ribbon the by comparing the mean and highest concentration of the
impurity content of the melt from which the ribbon had earlier work /6/. The values thus obtained are listed in table
been grown was calculated. II.

iii. The comparison of the melt concentrations calculated ad iii: For the purpose of easier comparison the impurity
as described in (i) and (ii), respectively, allows concentrations from the different sources are normalized to
conclusions on the impurity sources in question. the respective value for the Si ribbon (Table III).

For the calculations of the impurity concentrations in The results obtained by comparing the values of table
the melt following assumptions have been made: III can be summarized as follows:

ad i: Via diffusion impurities of the C-fiber net are dissolved 1. Carbon fiber net: With the exception of Cr all
in the Si melt. The diffusion continues until the impurity impurity elements could possibly account for less than 20 %
concentration in the melt is equal to the concentration in the found in the final ribbon. Only Cr could provide more than
carbon fiber net. the amount necessary to grow into the ribbon. However, the
time the graphite net is in contact with the melt is very short.
The impurities contained in the quartz crucible are Also, on contacting the melt the carbon fibers are covered
integrated into the Si melt by two different processes: by by a very thin layer of SiC formed immediately after
diffusion and by dissolution of the silica. The diffusion contacting. SiC is considered an effective barrier against
continues until the concentration in the quartz is equal to diffusion. For these reasons it is unlikely that the Cr stems
the concentration in the melt. For dissolution of the silica from the carbon fiber net. Its origin remains unclear.
two different velocities of convection are considered: v = 0
em/sec and v = 2,5 em/sec. These different convection 2. Quartz dissolution: The elements Zr, Mo, Ta, and W
velocities lead to silica dissolution rate of 10 m/h and 20 may contribute by some 20 % to 50 % depending on the
m/h, respectively /5/. The impurity content of the silica dissolution rate of the quartz. The other elements may
dissolved was recalculated for the volume of the Si melt. contribute only a few percent and are not considered to be
of importance for cell efficiency.
The concentrations in the starting material are
unchanged. 3. Diffusion from the silica crucible: If one considers
outdiffusion from the quartz crucible into the melt Cr and
ad ii: The calculation of the impurity concentration in the Fe may contribute some 50 % or 70 %. Zr, Mo, Ta, and W,
melt from the impurity content of the Si ribbon was however, can account for the entire concentration found in
performed according to results of earlier investigations /6/. the ribbon, as the concentration in the melt is higher by an
Fig. 1 demonstrates for Fe as an example the U shaped order of magnitude.
impurity profile with respect to the ribbon thickness. The
steep decrease of the impurity concentration from the 4. Si starting material: The impurity concentrations
surface connected to the carbon fiber net is due to the found are small compared to all other possible sources.
decreasing growth velocity for increasing ribbon thickness, Their contribution to the impurity concentration in the
which is associated with an increase of the effective ribbon is negligible.
segregation coefficient. The initial highest impurity
concentration of the ribbon therefore is equal to the melt
concentration. 3. SIUCA CRUCIBLE MATERIAL

The increase of the impurity concentration towards the Some of the impurities in the ribbon may be present in
second surface is due to diffusion out of a thin film dragged amounts deleterious to solar cell application. In table IV the
out of the Si melt along with the ribbon. It is associated with concentrations are compared with threshold values above
a stagnant layer and has to be taken into account when which an influence on the cell efficiency was found /7/. Fe,
calculating the melt concentration. Zr, Mo and W exceed these threshold values.
Zr, Mo and W quantitatively were traced back to silica
The impurity content of the ribbon is determined by as their source, Fe may come from there, too. So
integration of the concentrations across the ribbon thickness. experiments were aimed at clarifying the influence of the
For each impurity element the concentration found in the crucible material. Therefore natural silica contacting the Si
ribbon was used to determine the concentration in the melt melt has been partially replaced by synthetic silica. About

516
90 % of the natural quartz crucible was covered by sheets of This work was supported under the technological
synthetic silica which was considered to contain less program of the German Federal Ministry for Research +
impurities. The quality of the ribbon material grown in a Development, Project No. 03E-8434-A. The authors alone
essentially synthetic silica crucible was tested by producing are responsible for the contents.
test solar cells.

REFERENCES
4. PREPARATION AND EFFICIENCY OF SOLAR
CELLS /1/ Falckenberg, R; Grabmaier, J. G.; Hoyler, G.:
6th E.C. Photovoltaic Solar Energy Conf., London
Si ribbon grown with a crucible made of synthetic silica April 15-19, 1985, pp. 980-984
was used as substrate for 20 x 20 mm2 test solar cells. A /2/ Falckenberg, R.; Grabmaier, J. G.; Hediger, F.:
standard process was applied to the material: Following a J. Cryst. growth 82 (1987), pp. 10-16
deep acid etch a phosphorous diffusion at 820 °c was /3/ Falckenberg, R; Grabmaier, J. G.; Eisenrith, K.-H.;
applied. After removing the n + layer on the backside by Freienstein, B.; Lerchenberger, A.; Schindler, R;
acid etch, hydrogen passivation by implantation from the Lutz, F.
backside was applied. Subsequently AI was evaporated and Proc. 21st IEEE Photovoltaic Specialists Corn.,
sintered at 450 °c forming the backside contact. The front Orlando FL, USA, 1990
contact consisted of an evaporated grid formed with Ti-Pd- /4/ J. -G, Grabmaier, R Falckenberg:
Ag. Double layer antireflection coatings with Ti0 2 and MgF Technical Digest of the International PVSEC-5,
finally were applied. Kyoto, Japan, 1990, pp. 297-301
/5/ Hirata, H.; Hoshikawa, K.:
The solar cells had efficiencies ranging from 12 % to Jpn. J. Appl. Phys. 19 (1980), 1573
13 % with an average of 12.8 %. Compared to material /6/ Falckenberg, R; Freienstein, R; Hoyler, G.;
grown from natural silica crucibles an average increase of Grabmaier, J. G.; Hediger, F.:
1% absolute in efficiencies was observed (fig. 2). This is 8th Photovoltaic Solar Energy Conf., Florence, May
mainly due to an increase in short circuit current and hence 19-23 (1988), pp. 1295-1300
in diffusion lengths. The increase in diffusion lengths may be /7/ James R McCormick:
attributed to the absence of appreciable amounts of Zr, Mo Silicon processing for Photovoltaics, edited by C. P.
andW. Khattak and K. V. Ravi, pp. 1-47, 1985

5. CONCLUSION

The influence of the crucible material on ribbon

growth has been demonstrated by replacing natural silica ~
with synthetic silica. An increase in efficiency of solar cells Impurity Ribbon C-fiber net Crucible Silicon

has been observed which can be linked directly to the [em-3] [em-3] [em-31 [em-31

different quality in crucible material. For the incorporation

Cr 2.3xI0 13 4.4x10 16 1.9x10 14 2.2xIO I2
of crucible impurities into the Si melt not only dissolution
Fe l.Ox10 15 23xlO l6 6.Ox10 15 1.2x1013
may be of importance, but also diffusion from the quartz
Ni 3.2xI013 2.lx1015 < 1.8x1014 1.2x1012
may be of importance. In order to achieve efficiencies
Co 9.3xlO ll l.Ox10 14 < 1.9xI0 12 7.2xIO ll
exceeding 13 % it is necessary to either completely replace
Cu 9.8x10 14 5.8x1015 <4.4xIO I5 6.3x10 12
natural by synthetic silica crucibles and/or to keep the time
Zr 2.2xIO I4 7.1x10 14 3.lx1016 2.3xI0 13
of contact between melt and crucible low. A short contact
Mo 1.7x1013 2.OxIO I5 2.4x10 15 1.1xlOll
time between crucible and melt will be automatically
Ta 2.5xIO IO 1.5xlO l2 4.lx10 12 < 1.2x109
realized in continuous ribbon growth equipment.
w 93xlO l2 2.5x10 14 1.2xlO 15 7.5x1O ll

ACKNOWLEDGEMENT Tab. I; Impurity concentrations measured by means of

neutron activation analysis.
We gratefully acknowledge the work of E. W. Haas
and R Hofmann, who did the neutron activation analysis.
The skilful work of H. Lautenschlager, Ch. Schetter and E.
Schaffer in solar cell processing is appreciated.

517
 Impurity qN qQL qR
'IoD 'lsi
[em-3] [em-3] [em-3] [em-3] [em-3]

Cr 3.4xl014 7.4xl0 13 1.5xl012 2.2xl012 1.0.1014

Fe 1.8x1014 2.4x1015 4.8xlO 13 1.2xl013 5.0.1015
\\
Ni 1.7xl013 <7.0.10 12 < 1.0.1011 1.2xl012 1.5xl014
\
8.0.10 11 <7.0.10 10 < 1.0.1010 7.2xl011 \
Co 4.6x10 12 \
Cu 4.6x10 13 <1.8x10 15 <2.0.10 12 6.3xl012 5.0.1015
\
\
\
Zr 5.7xl012 1.2xl0 16 2.5x10 14 2.3x1013 1.0.1015 \
\
5.0.1014
\
\
Mo 1.6xl013 9.6xl0 14 1.9xl013 l.lxl011 6.8xl0 13 \
4.0.10 13 \
\
\
Ta 1.2xl010 1.6x10 12 3.3xl010 < 1.2x109 1.0.1011 \
6.6xl0 10
W 2.0.10 12 5.0.1014 9.6xl012 7.5xl0 11 3.7xl013
2.0.10 13

Tab. II: Impurity concentration in the melt [atoms/cm 3) to be 10'2.~__~__~____T-~~~~~

expected for different sources: diffusion from the C-fiber o 100 200 300 400 SOO=dF
net (qN)' diffusion out of the silica (qQD)' solution of the d[J-lITI1-
silica surface (qQL) for two different convection veloci-
ties. From the ~ctually measured concentration in the
ribbon the concentration in the melt is calculated (qR)' Fig. 1: Fe-concentration CFe in heavily
contaminated S-Web versus
ribbon thickness d; - measured by
Impurity qN qQD qQL + qSi
R
--
R R
NAA, --- calculated for segrega-
tion and diffusion, ... for diffusion
Cr 3.4 0.74 0.037
from stagnant layer
Fe 0.036 0.48 0.012
Ni 0.11 0.04 <0.0067
Co 0.174 <0.15 <0.16
Cu
Zr
0.009
0.005
<0.36
12.4
<0.0013
0.27
0.54
1
N [%1
,- ,--,
Mo 0.23 14.1 0.28
60 I I
I I
0.56 I I
I
Ta 0.12 16 0.33 I
50 I
0.66 I
W 0.054 13.5 0.27 I
I
0.54 40 I
I
I
I
Tab. III: Impurity concentration normalized to I
the Si ribbon 30 I
r-- I
I
I
I
I
Impurity qR qmax Source 20 I
I
[em-3] [em-3] I
I
10 - I
Cr 2.3 x 1013 1.3 x 1014 - I
I
1.0 x 1015 2.5 x 1014 not clear I
Fe I
Ni 3.2x 1013 5.0x 1015 - o Tj[%l-
1.1 x 1015 11 12 13
Co 9.3 x 1011
11.5 12.S
Cu 9.8 x 1014 4.1 x 1017 -
Zr 2.2x 10 14 3.6x 1011 quartz
D melt crucible of natural Silica
Mo 1.7 x 1013 6.0 x 1011 quartz :::::: melt crucible of synthetic Silica
Ta 2.5 x 10 10 2.3 x 1011 -
W 9.3 x 1012 1.2 x 1012 quartz
Fig. 2: Efficiencies obtained for natural
Tab. IV: Comparison of the spatial density qR of impu-
and synthetic silica crucible
rities found in the ribbon with literature values
qmax /7/ believed to be threshold limits for material.
influencing cell efficiency. In the last column,
the likely source of the impurities is given.

518
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

On-line manufacturing of Si solar cells compatible with high speed

Si-ribbon growth

J.G. Grabmaier, H.P. Urbach, K.-H. Eisenrith

Corporate R&D Siemens AG, Otto-Hahn-Ring 6, 0-8000 MUnchen 83

K.A. MUnzer
Siemens Solar, Frankfurter Ring 152, 8000 Mtinchen 45

Abstract

For the terrestrial photovoltaic application in the gigawatt range

new on-line processing steps are necessary. Applying a new produc-
tion process for the pn-junction large area (10 cm x 10 cm) mono-
crystalline solar cells have achieved conversion efficiencies of
13 % and polycrystalline S-Web solar cells efficiencies of more than
10 %. With a new grid contact preparation technique solar cells of
monocrystalline wafers achieved 12,4 % efficiency and S-Web solar
cells achieved up to 8 %. Both processes can be done in a conven-
tional production line and include a reasonable potential of improv-
ing the solar cell quality and cost. The methods and the results are
discussed in detail.

1. Introduction type) cut from S-Web ribbonS were

employed.
Today's photovoltaic power application
is based on crystalline silicon solar The natural 2 rum to 3 rum thick oxide
cells. The applied standard processes layer was removed by HF etching. In
for cell production are batch processes contrast "as grown" S-Web Si material
(crystal pulling or block casting, wafer has no oxide layer and therefore an
cutting, cleaning and etching of wafers, etching step was not necessary. For
P-diffusion, screen printing). preparing solar cells (10 cm x 5 cm in
size) etched Si wafers were covered on
For a future large scale and economical only one side with amorphous Silicon in
production of solar cells in the a capacitively coupled RF glow discharge
gigawatt range these processes must be apparatus (13,56 MHz). The addition of
substituted by on-line processes with PH 3 (3 %) to SiH 4 in H2 provided an n-
high throughput IFig. 11. type layer. Variations of the deposition
time caused different thicknesses of the
In the Siemens research laboratories a a-Si:H,P layer. The crystallization of
continuous Si-ribbon growth process (S- the a-Si:H,P layer was carried out
Web-technique) has been developed, either in a conventional resistance
achieving up to 150 m long, 12 em wide furnace or in an inductively coupled
ribbons at a growth speed of 1 m per furnace. The time and the temperature of
minute I 1 I. To avoid the usual crystallization varied from 0,5 h to 4 h
cleaning and etching of the Si and from 550 0 C to 700 0 C, respectively.
substrates for cell proceSSing and to Subsequently the temperature was raised
use the highly pure as-grown ribbon to the diffusion temperature, which
surface, processes have been developed ranged from 750 °c to 1000 °C. During
for an on-line production of the pn- the diffusion period of 3 min to 10 min
junction as well as for grid production P atoms diffuse into the substrate by
on the front side of the solar cell 12/. which the crystallographic interface and
These processes are compatible with the the pn-junction are separated. The
high throughput of the S-Web-technique crystal structure of the n-Iayer was
and tolerate a slight surface roughness. investigated by TEM microscopy and the
This contribution deals with the sheet resistance was determined by 4
investigation of a new method to prepare point probe measurement.
pn-junctions and with a new technique of
producing the front grid contact. Best results were employed for preparing
Results obtained are related to the solar cells (10 cm x 5 cm in size) in a
quality of solar cells prepared by conventional manufacturing line. The HF
conventional techniques. etched Si wafers were covered with a 10
to 30 nm thiC~6 P-doped a-S~:H layer
2. Experimental ([P] = 7 x 10 P atoms cm- ). The
crystallization was carried out at 620
2.1. Preparation of the pn-junction °c for half an hour. The P diffusion
took place at 850 °c for 5 min. To
Fig. 2 depicts the process flow chart examine the quality of the pin-junction
for on-line cell fabrication based on metallization was performed by
the new techniques. Monocrystalline Si- sequential evaporation of Ti/pd and Ag.
waferS(1,4 ncm, p-type) cut from The solar cells were characterized under
Czochralsky grown rods and standard test conditions (AM 1,5; 25 °G;
polycrystalline Si wafers (1,4 ncm, p- 1000 W/p).

519
2.2. Preparation of the front grid the amorphous silicon starts at the
contact interface of the monocrystalline
substrate resulting in an epi-layer (10
The on-line manufacturing of Si solar nm thickness). Using S-Web Si material
cells requires a metallization process as a substrate a 25 nm thick amorphous
compatible with the high silicon sheet silicon layer is deposited by similar
production rate. In addition the deposition conditions. Crystallization
metallization process has to be adopted of the a-Si:H,P layer under the same
to the characteristics of the S-Web conditions results again in an epi-layer
material, i.e., to the uneven surface. A (Fig. 4 b). The TEM micrograph reveals
study of different metallization that the surface is not as planar as in
techniques led to a computerized direct the case of a monocrystalline Si
writing system (trade mark: Micropen substrate due to the characteristic
(Dage), USA), that was originally surface of the substrate.
designed for the hybrid technique. This
system is able to allow the writing The thinner the a-Si:H,P layer is the
unit, including a writing pen, to follow shorter is the crystallization time. So
height differences of the substrate far, the crystallization of a 10 nm to
surface (± 0,5 mm). The writing unit 30 nm a:Si:H,P layer takes 0,5 h. Recent
extrudes nearly any material fluid results indicate that optimized
regardless of the viscosity (from water conditions reduce strongly the time
to putty). needed .
For experiments a pen tip with a Second, the P-diffusion process is
diameter of 100 ~m and a silver paste characterized by preparing test solar
used in the screen printing technique cells (Fig. 5) and measuring their I(V)-
were employed. The Ag paste was written characteristics. Short diffusion time
on bare S-Web material (Fig. 3a) and for (3,5 min) and low diffusion temperature
comparison on monocrystalline silicon
wafers. Each of the variables as writing (T < 850 °C) do not lead to measurable
speed, pumping rate of the Ag paste, pen results. In general, low temperatures
force, deposition thickness and and short diffusion time create a thin
deposition width were individually n+-layer with relatively high sheet
programmed and more or less complex resistance. Here, the crystallographic
patterns were written. interface and the pn-junction are not
separated far enough; a high density of
Best results were employed to produce recombination centers is expected (low
test solar cells with a written grid Isc and ff). A moderate increase of
contact. For comparison test solar cells temperature and/or time decouples
(2 cm x 2 cm in size) were made from S- crystallographic interface and pn-
Web material (Fig. 3b) and junction and the n+-layer thickness
monocrystalline Si as well. The pn- expends slightly. The sheet resistance
junctions on the substrates were decreases and the other cell data shift
prepared by a conventional P diffusion t~ a maximum. A further increase of the
p~ocess and the sheet resistivity of the n -layer thickness causes an increasing
n -layer ranged from 45 % to 50 0/0 recombination of light generated
. The front grid was written onto the carriers and fewer minority carriers
n+-layer with the Micropen system. take part in the photo current.

After grid contact drying at 100 °c the Best results are obtained at a diffusion
back side was metallized by screen temperature of 850 °c and a diffusion
printing. Finally the solar cells were time of 5 min.
fired for some minutes at a temperature
of about 700 °c and the front side was Solar cell preparation in a well
covered with an AR coating. optimized manufacturing line leads
within the error bar of measurement to
3. Results the same level of quality as the well-
known conventional process has
3.1. pn-junction established. Fig. 6a shows an I(V)-
characteristic of a monocrystalline
The preparation of the pn-junction by solar cell (10 cm x 5 cm in size), that
deposition of a P-doped amorphous Si achieves a conversion efficiency of
layer onto a p-type monocrystalline 13,0 % and in comparison the I(V)-
substrate has been investigated in two characteristic of a S-Web solar cell
steps. (Tab. ~) demonstrates the corresponding
values for Voc ' Isc' jsc' 'YJ and FF.
First crystalline silicon substrates are
covered by a 2 nm to 3 nm thick natural Fig. 6b shows the spectral response of a
oxide layer. Deposition of a P-doped monocrystalline (c-Si) and a
amorphous silicon and subsequent polycrystalline S-Web silicon solar
crystallization result in a polysilicon cell, prepared by the new pn-junction
film with randomly oriented process. The curves demonstrate a steep
crystallites. Thereby produced solar decrease of the spectral efficiency in
cells show very poor photovoltaic the blue area. This indicates a strong
activity. Prior to the deposition of "dead layer" in the n+-layer;
amorphous silicon the oxide layer has to diminishing of the width of this dead
be removed, for example by a HF dip. layer by creating a steeper P
Fig. 4a shows that the interfacial oxide concentration profile should result in
is broken up into individual oxide better values for the spectral
particles (2 nm to 3 nm in size). Under efficiencies in this range. The lower
this conditions the crystallization of values of the spectral response of the

520
5-Web-cells compared with the values of 4. Conclusion and Perspectives
the monocrystalline cells in the area
above 500 nm can be explained by lower Encouraging results are obtained by the
minority carrier diffusion length. new methods to prepare the pn-junction
Compared to the monocrystalline wafers, as well as the grid contact for Si solar
5-Web-wafers are polycrystalline type cells. The preparation of large area
wafers containing grain boundaries and a solar cells (10 cm x 5 cm) indicates
lot of inner grain defects. This that the conventional production line
polycrystalline character of the for the deposition and crystallization
material is typically connected with of P-doped amorphous silicon for p/n-
lower values for the diffusion length. junction production is well optimized.
Conversion efficiencies of solar cells
3.2. Front grid contact achieve 13,0 %, that is comparable to
efficiencies of conventionally produced
Grid lines produced by the screen solar cells. As a result the quality of
printing technique show a width of the pn-junction prepared by the new
120 JLm to 150 JLm and a thickness of app. method is excellent.
10 /lm. A reduction of the line width
would desirably decrease the shadowing Though the conditions have not been
losses. Therefore, the first goal was to optimized so far, test solar cells based
write 100 /lm wide li~es with the on S-Web material have nearly achieved
Micropen system on monocrystalline 5i 8 % conversion efficiency (more than 10
wafers. The use of a pen tip with a low % with AR coating). Further experiments
cross section provides about 110 JLm wide aim primarily to a shorter process time,
lines that are not continuous. It is that is, a reduced deposition time
assumed that there is too little combined with a thinner a-5i:H,P layer
material being written. If more Ag paste and a shorter time of crystallization.
is applied (higher pumping rate) the
material spread to a 140 JLm - 150 /lm Furthermore there is a potential of
width and continuous lines are produced. improving the quality by creating a
This result is obtained at a writing steeper P concentration profile.
speed of up to 35 rom/sec.
Monocrystalline solar cells with a
The same observations are made for S-Web written front grid contact achieve 12,4%
material. When the width of the lines is conversion efficiency, polycrystalline
low about 150 JLm there are contact 5-Web cells up to 8 %, both prepared in
spaces and many opens. Again it is the production line. This values
assumed that theLe is simply not enough indicate, that a written front contact
material to "fill" the gaps as the pen is of the same quality as the
tip moves across the uneven S-Web conventional screen printed contact. It
surface. seems that a better harmonization
between the employed Ag paste and the
Increasing pumping rate and pen force writing conditions could increase the
yield wider lines (200 JLm to 250 JLm). In writing speed from 35 rom/sec to 50
addition, more material makes the lines rom/sec. Meanwhile, the surface quality
continuous (Fig. 3). This result is of the S-Web material has been improved
obtained for a writing speed of up to 20 much; it is expected, that the writing
rom/sec. The thickness of the lines speed on S-Web substrates will achieve a
ranges from 5 JLm to 10 JLm, that is similar value.
comparable to lines produced by the No work has been done to optimize the Ag
screen printing technique. paste for the writing application.
Because the employed Ag paste contains
The resulting resistance of the 50 JLm large particles a save writing of
gridlines written with the writing lines is possible down to a width of
system in comparison to those of screen about 150 JLm. If 100 JLm lines are a true
printed gridlines are shown in Tab. 1. requirement, a paste with smaller
For broad gridlines (150 JLm-200JLID) the particles and higher viscosity would
results are identical. For fine help. A further reduction of the width
gridlines the pen tip writing method is achievable by the use of metal
seems to be in advance because the cross organic inks.
section of the gridlines can be defined
more exactly. This work has been demonstrated that an
on-line manufacturing of solar cells at
There is a tight relation between the high throughput is possible without any
quality of the lines and the loss in cell quality. Moreover, there is
characteristic data of solar cells. For a reasonable potential of improving the
monocrystalline cells with a diffused cell quality and the cell cost.
pn-junction i.e. the increase of the
line width from app. 110 JLm to 220 JLm Acknowledgment
raises the conversion efficiency from
11.3 % to 12,4 %. The increasing This work has been supported under the
efficiency is dominately related to a Technological Program of the Federal
raising fill factor (ff: 64.9 % to Department of Research and Technology of
71.4 %). A similar result is obtained the FRG. The authors alone are
for cells made from S-Web material, that responsible for the context. The authors
achieve efficiencies of up to 8 % (the are indebted to thank Mr. Kausche for
grid design provides about 100 % more the depositon of the P-doped amorphous
shadowing I ) . silicon layers.

521
 Reference
[1] Falckenberg, R.; Grabmaier, J.G.· [2] G7abmaier J.G., H.-P. Urbach, K.-H.
Eisenrith, K.-H.; Freienstein, B:: E1senrith and K. A. Munzer
Manufacture of Si ribbons by the S- On-line manufacturing of Si solar
web-Technique cells compatible with high speed Si
21st IEEE PV Specialists Conference, ribbon growth 5th Intern. PVSEC,
Kissimmee, Florida (USA), May 21-25 Kyoto (Japan), Nov. 26-30, 1990,
(1990) p. 695 p. 509
Voe Ise Jse FF 11
tmV] [Al [mA/em'] [%l [%l
grid line
width
100 150-180 200-250
methode [11m] [11m] [11m]
I, S-Web 545 1.25 27.8 62 8.5
pen tip
0.5-0.8 0.25-0.35 0.15-0.25 II, c-SI 598 1.53 34.0 71 13.0
writing
....................... ........... ............ ............ ...........
screen III, c-SI 595 1.53 34.0 72 13.1
0.8 0.3 0.2
printing

Tab. 1 Resistance (O/cm) of gridlines written with the Tab. 2 Typical data of the I-V-characterlstlcs of solar celis
pen tip in comparison to those of (10cm X 5cm) based on monocrystalline 01 c-Sj)
screen printed gridlines respective polycrystalline (I, S-Web) silicon' wafers
prepared by the new pn-Junction formation In
comparison to datllS of mono crystalline solar celis
prepared by the conventional diffusion process (Iii)

not

d
nozzle

Q
(j

continuoul ribbon pulling PE CVD direct palt. writing

Fig. 1 On-line manufacturing of SI solar celis

1......_ _ _ _ _ ~t--SI-SUbstrate

Deposition

Crystallization iS200C, At
1'"II\I!IiI\\lllillii\lilili[lIir~~~:; columnar

Diffusion .I8so°c, AI
ls min. 1 m
Ji rnm

~FlZ/!lIl\lIlI\IJ:c:;t\l1l:U:~lrjl\lI'O:I~U/~I2t/IlDI~IIII--_ pn-junction
b)

Paste writing ~

Ell'!:i:["'ffi"'0\1i, Grid line Fig. 3 a) Grid line written on S-Web Si material

b) S-Web t•• t .olar cell with written front grid contact
Back contact

FIrIng l7000 C
Fill mill ffi \II 5]<,/1' Solar cell

Fig. 2 Process steps of an on-line solar cell preparation

522
 er)'II.a llf zed a""'S I layer

SIOI-parl lc l e!li
1.5t-----~~~
substrale [AJr-_ _ _ ___

12 nm
1.0

Fig'. 4a TEM cross section of a crystallized P-doped

amorphous silicon layer on monocrystalline AM 1.5
silicon TJ = 25°C
0.5 A = SOcm2

c:rysta llize d
.-Si layer

Fig. 6a I-V characteristics of c-SI and S-Web

solar celis (1 Oem X Scm) with the new
pn-junction preparation method
28 nm

Fig. 4b TEM cross section of a crystallized P-doped

amorphous silicon layer on S-Web SI material

8
[%] E .5
::I
C
o'"
::I

I 7
........+.,..
••
,
.
'
o~--~~~~---~~---~~~
400 600 800
Wavelength
1000 [nmJ

..........
Fig. 6b Spectral efficiency of a monocrystalline
6 \ (c-Si) and a polycrystalline S-Web silicon
solar cell prepared by the new
pn-junction process

a)

\±5~O--~---8~5-0--~---9~50--[~1
T-
Fig.S Influence 0' oreparatlon conditions (T, t) on the
quality of" p~iuncfiOns Tn test solar cells
( D _ : . - 3,5 min., - 5 min., · - 1 0 min.)

523
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

CONlROLLED NUCLEATION: A NEW APPROACH TO IMPROVE

THE CRYSTALUNITY OF RAFT-FOILS

I. Reis, A Hurrie, and A Rauber

Fraunhofer-Institut fUr Solare Energiesysteme
Oltmannsstr. 22, 7800 Freiburg, Germany
D. Helmreich, and A Beck
Heliotronic GmbH
8263 Burghausen, Germany

ABSTRACf. A variety of substrates has been tested for use in silicon

foil casting according to RAFT (Ramp Assisted Foil-casting Tech-
nique). As an alternative to the currently used PyC-coated graphite
ramps monocrystalline silicon ramps with a structured surface are
utilized to prepare RAFT-foils with improved crystallinity. Structuring
is done either by tooling (for regular, symmetric geometries) or by
chemical etching (for statistically distributed structure elements}.
'Orientation of monocrystalline silicon ramp «100>, <111» IS
transferred to the crystallizing melt film.

1. INlRODUCTION
The characteristic feature of the Ramp Assisted Foil-
casting Technique (RAFT) compared to other methods for
silicon sheet growth is the fast and direct solidification of
silicon wafers which are directly usable for solar cell pro-
cessing (1). This is achieved by use of a substrate as a tem-
porary shaping element - referred to as the "ramp" - which
has sufficient heat capacity for the absorption of the latent
heat of crystallization. This substrate is guided past the
meniscus of a silicon melt. During the time of contact, the
top of the ramp surface is coated with a silicon melt film
which then crystallizes by rapid conduction of the crystal-
lization heat into the ramp.
The corresponding heat transfer rate, which governs
the crystallization of the silicon melt film and the resulting
external geometry of the silicon foil, is determined by ramp
and process characteristics. Basic material and surface pre-
paration of the ramps are highly relevant to the process
Itself and to the properties of cast silicon foils. The ramps
should meet the following basic requirements:
- High temperature stability
- High purity to eliminate the transfer of electrically
active impurities into the crystallizing silicon film
- Low chemical reactivity with the silicon melt
- Controlled nucleation of silicon melt film to improve
the grain size of the crystallized silicon foil
- Favourable separation behaviour of crystallized
silicon foil from ramp surface.
Several ramp systems have been tested so far; best
results have been obtained with graphite ramps with a
pyrocarbon coating (2).
RAFT-foil grown on these substrates has been charac-
terized through its grain structure, defect situation and pho-
toelectric parameters. The foils have a quasi-columnar
grain structure with grain boundaries perpendicular to the
surface (fig. 1). Grain size ranges from severallO I'm to a
few 100 I'm. depending on the local heat transfer into the
ramp. Dislocation densities up to lOrcm-2 within the indi-
vidual grains are caused by steep temperature gradients
necessary for high growth rates. Generally, the dislocation
density of large grains is about two orders of magnitude
higher than that of small grains. Nevertheless the local
short circuit current, as revealed bv EBIf' :Inri LBIC
measurements, shows a clear inl.:rease Will mcreasing grain
size. This result shows that diffusion length Lo in RAFT-
foils grown on pyrocarbon coated graphite ramps is limited
rather by grain boundaries than by dislocations. Figure 1: Typical grain structure of RAFT-foil grown on
In order to improve the solar cell parameters of pyrocarbon coated graphite ramp
RAFT-foils, a new approach has been tried with ramp Top: cross-section, Secco-etched
systems that allow a well controlled nucleation and crystal- (thickness: 300 I'm)
hzation. Bottom: surface, Sirtl-etched

524
2. NEW APPROACH Spacing and height of the contact surfaces have been varied
within the limits given in table 1. With a pulling s£eed of
Monocrystalline silicon should be an ideal ramp ma- 100 mm/s and substrate temperatures near 1000 C opti-
terial provided that the foil can easily be separated from mum values of spacing and height of the contact surfaces
the ramp after crystallization. This can be achIeved by suit- are in the range of 200 reseectively 100 I'm. The surface of
able surface structuring. By using such ramps we expect the a mechanically structured SIlicon ramp is given in figure 2.
following growth process:
- The crystallographic orientation of the substrate
( < 100 >, < 111> ) is transferred to the crystallizing
silicon melt film.
- A regular or statistical pattern of small pyramids
leads to elevated contact surfaces where the crystal-
lizing silicon melt film rests and controlled nu-
cleation takes place.
- Due to the hot silicon melt the pyramid apices are
molten back to a sectional area, which is large
enough for the optimal removal of the crystalliza-
tion heat.
- The apices act as nucleation centers and the foil
overgrows the pyramidal grooves.
- During the cooling down process, ramp and foil con-
tract differently and the interconnections between
silicon foil and pyramid apices are broken.
- Ideally this break-off occurs at the point of the
smallest cross-section, allowing a high re-usability of
the ramp without regeneration.
Figure 2: Surface of a monocrystalline silicon ramp
structured by mechanical methods (spacmg of
3. EXPERIMENTAL the contact surfaces: 180 pm)

3.1. Mechanical methods

Regular surface structures were machined with 3.2. Chemical methods
suitably dimensioned diamond tools (sawing, grinding) or < 100> -oriented monocrystalline silicon ramps have
by lapping techniques. The structures consisted of quadratic been treated by chemical methods already known for effec-
(truncated) p>,!amids arranged in quadratic arrays (for sili- tive light trapping. Contrary to the mechanically structured
con ramps WIth < 100> orientation). In the case of silicon ramp an irregular pattern of pyramids of differing sizes and
ramps with < 111 > orientation hexagonal pyramids ar- distances from each other has been produced. The height of
ranged in hexagonal arrays are preferred. The total contact the pyramids as well as the distances between them ranges
surface facing the crystallizing silicon layer is in the range from a few microns to about ten microns as demonstrated
of several percent of the parent surface of the ramp. in figure 3. Chemical structuring of ramp surfaces is less ex-
The distance of adjacent contact surfaces from one pensive than structuring by mechanical methods and regen-
another has as its upper limit the value at which the melt eration of used ramps appears possible in an economic way.
sagging between the supporting surfaces fails and the melt
comes into contact with the substrate parent surface.

Table 1: Substrate Parameters and Process Parameters

for the Growth Operation

Spacing of the Height of the Substrate Pulling

contact surfaces contact surfaces temperature speed
[I'm] [I'm] [0C] [mm/s]

720 100 970 25

360 100 970 50
240 100 970 100
180 100 970 100
90 100 970 100 Figure 3: Surface of a moncrystalline silicon ramp
structured by chemical methods
180 100 970 50
180 100 970 100 4. RESULTS AND DISCUSSION
90 100 970 200
60 100 970 300 The results of our experiments confirm essentially the
considerations mentioned above.
Figure 4 shows the as-grown surface of a RAFT-foil
180 100 970 100 cast on a < 100> silicon ramp structured by mechanical
180 50 970 100 methods. The backside of the foil shows small contact
points, where crystallization started. By lateral growth a
quasi-monocrystalline sheet has been formed with low
180 100 920 100 angle grain boundaries about mid-distance between the
180 100 970 100 nucleation centers. Orientation of the foil is the same as
that of the substrate. A cross-section shows higher disloca-
180 100 1020 100 tion densities near the contact points. This is due to steep
temperature gradients in these regions.

525
 - - Figure 5: RAFf-foil grown on chemically structured ramp
Top: back-side, as-grown
~ Bottom: front-side, as grown

Figure 4: RAFf-foil grown on mechanically structured Silicon foils separate easily from silicon ramps with
< 100> silicon ramp sound surface structures. This allows a high re-usability of
Top: surface, as grown the ramps which is the pre-condition of an economic pro-
Middle: back-side, as grown duction process.
Bottom: cross-section, Secco-etched Further investigations have to clarify the influence of
this improved crystallinity and defect situation on the pho-
toelectric characteristics of RAFf-foil (i.e. diffusion length
Only a few experiments have been performed until In, solar cell parameters).
now with chemically structured ramps. The growth mech-
anism is the same as in the case of mechanically structured
ramps, but only the apices of the highest pyramids act as 5. ACKNOWLEGDEMENT
nucleation centers. This can be seen in figure 5 in compari-
son with figure 3. The regular geometric pattern indicates a This work has been supported by the Federal
quasi-monocrystalline structure. Because of the smaller Department of Research and Technology of the FRG
distance of nucleation centers, compared with the mechan- under contract numbers 032-4506-D and 032-8986-A The
ically structured ramps, no (regular) nucleation pattern is authors alone are responsible for the contents.
visible on the front-side and the surface appears smooth
and structureless.
Mechanically and chemically structured silicon ramp's REFERENCES
are suitable to produce quasi-monocrystalline silicon fOIls
by the RAFf-process. Foil thickness depends on the pro- (1) A. Beck, J. Geissler, and D. He1mreich, Journal of
cess parameters (pulling speed: 100 rnrn/s, substrate tem- Crystal Growth 104 (1990), 113
perature: -1000 DC) and is in the order of about 250 /.1m. (2) A. Beck, J. Geissler, D. Helmreich, A Rauber, 1. Reis,
Dislocation density is similar to that in multi crystalline foils and R. Schindler, Proc. 11th Symposium on Industrial
grown on pyrocarbon coated graphite ramps. Crystallization (1990), 791

526
 POSTER SESSION 3A

Compound Semiconductor Solar Cells I

 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

STEM-STEBIC TECHNIQUE FOR DEFECT INVESTIGATION IN

SEMICONDUCTOR PHOTOVOLTAIC MATERIALS

M. Garozzo and A. Parretta

ENEA - Centro Ricerche Fotovoltaiche, 80055 Portici (Italy)

M. Vittori
ENEA - Casaccia, 00060 Rome (Italy)

A. Camanzi and P. Alessandrini

Eniricerche, S.p.A., 00015 Monterotondo (Italy)

ABSTRACT. The Scanning Trasmission Electron Microscopy, when used in the Electron
Beam Induced Current mode (STEM-STEBIC), can allow the correlation, at high spatial
resolution, of structural and electrical properties of individual defects in semiconductor
materials. In this paper we describe the STEM-STEBIC defect analysis in gallium
arsenide (bulk and epitaxial) and polycrystalline silicon with both p-n and Schottky
barriers, in order to put in light some spurious effects associated to multiple scattering of
the primary beam and to asses the potentiality of this technique. A proper device for a
JEOL 200 kV transmission electron microscope was set up with facilities of double tilting
and applying bias voltages. This apparatus allowed us to study correlations between
diffraction contrast, at different orientation conditions, and conductive contrast, at different
junction barrier amplitudes. The typical black recombination contrast at the electrical
defects was not observed in poly-Si, unless very thick regions were examined or a direct
bias voltage was applied. Recombinating electrical defects were easily detected, on the
contrary, in GaAs, also near the thinnest region of the specimen.

1. INTRODUCTION region, for a reliable interpretation.

In this paper we describe experimental results
Polycrystalline and thin epitaxial layer obtained with a device for a JEOL 200 B E. M., with
semiconductor materials have recently received facilities of double tilting and applying bias voltages,
increasing attention within the planar electronic which has been designed in order to study
technology for realizing MOSFET structures, Schottky correlations between diffraction and recombination
barriers and solar cells. It is well known that one of the contrast. The experimental work has been carried out
best non destructive techniques for the on several types of semiconductors, in particular Si
characterization of semiconductor devices is based on and GaAs, with different kinds of junctions, in particular
the collection of charges injected by the electron beam p-n homojunctions and Schottky barriers, in order to
of a SEM through an electron trasparent Schottky clarify the nature of the spurious effect associated to
barrier diode or p-n junction. This technique, called multiple scattering of the primary beam and decide
EBIC, has been widely discussed for the case of bulk when the STEM-STEBIC technique can be profitably
samples (1,2) observed in SEMs operating at low
energies. However, SEM-EBIC suffers from some
disadvantages: the resolution is limited and the 2. EXPERIMENTAL
information, obtained only near the specimen surface,
depends on the beam penetration into the material; 2.1 Tilt device
moreover, spurious effects due to the space-charge The side entry stage designed for studying our
region width can be found particularly with Schottky electronic devices fits perfectly into the electron
barriers (3), when their widths are higher or microscope chamber of a JEOL 200 B, without any
comparable to the penetration depth of electrons. modification of the anticontamination and X-ray
Finally, it is usually difficult to determine on the same microanalysis facilities. A tilt of ±50 degrees around
defect both the electrical and structural properties two ortogonal axes is permitted: the specimen
(4,5). Some authors (4,6,7) suggested to use the movement is obtained by a method of wire
scanning trasmission electron microscope in the transmission through an external knob which tilts the
induced current mode (STEBIC) for semiconductor cartridge; the tilt in the perpendicular direction is
recombination studies: the method should allow the obtained by a rotation of the whole stage. Two metal
correlation of the informations obtained in conductive contacts are glued with silver paint on the specimen
mode and in diffraction contrast. This kind of analysis, surface and are connected to a Keitley current
which requires, of course, a thinned sample and an amplifier which drives the main video amplifier of the
electron beam energy significantly higher than that scanning transmission microscope. A bias voltage can
used in SEM, was applied in particular to the be applied to the specimen in order to modify the
observation of devices (6-9). However, STEBIC contribution of carrier diffusion to the collected signal
contrast often shows new and anomalous features with and/or the influences of electric fields in the depletion
respect to the ones expected on the basis of the
carriers recombination (4,10,11). These effects were 2.2 Specimen preparation
interpreted as produced by secondary ionizing Different kinds of materials were examined in
radiation due to interaction of backscattered or the present work: polycrystalline silicon wafers of p-
diffracted electrons with the sample surroundings (9). type and .5+1 ohmxcm resistivity by Wacker, GaAs
Therefore, in some cases, the combination of the wafers made by MCP Electronic Materials, Nd",6x10 16
different effects results in a contrast which requires an cm-3 and GaAs epitaxial films, grown by MOCVD onto
accurate study, with diffraction techniques combined to (100) oriented commercial GaAs ; these films were
methods of changing the width of the space-charge unintentionally n-doped with Nd=1x10 16 cm- 3 (12).

529
 a b c

Fig. 1: a) STEM image of a grain boundary and of a scratch in Schottky diode on poly-silicon.
b) , c) Corresponding STEBIC pictures taken at slightly different specimen tilt angles,
showing contrast variation at different diffraction conditions.

The disk-shaped wafers were ultrasonically cutted and STEBIC contrast at the triple grain joint in a thick
polished on one side with a chemical jet method. region is observed if a reverse bias voltage is applied
Thinned areas of .5 mm in diameter were generally (Fig. 3).
obtained; in some cases, larger areas were produced
by ion beam milling. Before thinning the wafers were
polished in order to remove any surface mechanical
damage. The preferential etching along twins and
grain boundaries in polycrystal silicon was limited by
modifying the etching conditions generally used for
silicon single crystals (13) as follows: (i) very slow flow
rate (.5+1 cm 3/min); (ii) a nozzle-specimen distance
strictly fixed at 3 mm in order to have a central hole in
the specimen; (iii) a nozzle diameter ranging between
.3+.4 mm; (iv) an argon flux injected onto the
specimens during etching to avoid chemical vapor
contamination; (v) etching solution frequently renewed
in order to mantain an optimum polishing activity. The
best jet polishing solution for GaAs samples was
HN03:H20:HF;3:4:1. The polishing apparatus was
made by teflon. The total time of preparation was 5+6
min for 400 11m thick wafers.

2.3 Device preparation

After thinning, Schottky diodes were made onto
the flat surface of the discs by vacuum evaporating 10
nm of AI for silicon materials and 30 nm of Au for GaAs
ones; a good electrical ground was obtained on the
polished side by depOSiting a gold and an aluminium
ring, 1 11m thick and .2 mm wide, onto silicon and GaAs Fig. 2: a), b) TEM pictures of oxidation induced
respectively, around the periphery of the discs, out of stacking faults and scratches in polY-Silicon
the thinned region. No mechanical stresses were taken in adjacent regions of a diffused
introduced into the thin films by using this deposition homojunction specimen. The orientation
method. Some p-n junctions on silicon polycrystals conditions are changed from a) to b)Jn
were realized, before thinning, both by standard order to show tl)e stacking fault contrast and the
diffusion methods and by phosphorous implantation. bounding partial dislocation, respectively. c),
d) Corresponding STEBIC images showing
recombination contrast.
3. RESULTS

3.1 Silicon polycrystaJ materials

STEM and STEBIC images of a grain boundary
and a scratch in a Schottky diode are shown in Fig. 1.
STEBIC contrast variations due to specimen tilting,
without any bias, can be observed. TEM and STEBIC
images of twins and stacking faults in a diffused
homojunction are presented in Fig . .2. A strong
recombination contrast is observed in STEBIC at the
stacking fault (Fig. 2c), together with a black-white
contrast perpendicular to the stacking fault direction,
while twin boundaries do not show any contrast. The
black line through the stacking fault in Fig. 2d is due to V;QmV V=-100 mV V=-350 mV
recombination contrast at a surface scratch. Black dots
in STEBIC images are probably associated to small Fig. 3: a) STEBIC image of a triple grain joint in a
precipitates as seen in TEM picture 2a. Isolated thick region of an ion implanted junction on
dislocations were found in these specimens, but they polY-Silicon, showing recombination contrast.
do not show any "black" recombination contrast; on the By applying a reverse bias this kind of contrast
contrary "bright" contrast along the dislocation line is decreases (b) until it is inverted (c).
generally observed. Finally, an inversion of the

530
3.2 GaAs materials 4. DISCUSSION AND CONCLUSION
A high dislocation density probably due to
growth stresses is observed by TEM on GaAs bulk STEBIC observations show a different
~ingle crystal (Fig. 4a); the corresponding STEBIC behaviour between polysilicon and GaAs materials
Image shows strong recombination contrast around which can be summarized as follows. Bright images of
these defects (Fig . 4b). Black dot contrast obtained by defects are always observed in intermediate thickness
applying a reverse bias (Fig. 4c) seems to be related to regions of silicon specimens; this kind of contrast is
areas of different thickness, probably due to etch pits. increased by applying a reverse bias voltage. Black
An increasing reverse bias voltage produces an contrast (typical recombination contrast) at .1 11m of
increasing resolution in STEBIC images; at the same resolution is only obtained in thicker regions, or when
time the contrast decreases, but is never inverted (Fig. a direct bias voltage is applied. A large surface
4d~. :rhe same effects are observed along the recombination dramaticaly reduces the STEBIC signal
polishing trace of Fig. 5a in an epitaxial film; however, in the thinnest specimen areas. When GaAs materials
black areas, ranging between .1 and 1 11m in diameter are concerned, a strong recombination contrast is
(Fig. 5b), do not correspond to particular defects generally obtained also in the thinnest region of the
observed, nor to local thickness variations. The specimens and can be decreased by changing the
thickness of the visible regions of these specimens is width of the depletion layer with a reverse bias voltage.
estimated to be about 1 J.1m. At the same time, the resolution is improved. This kind
of behaviour can be interpreted on the basis of the
The EBIC current generally depends on the
junction field, minority carrier diffusion length and
carrier loss processes through recombination during
diffusion at the defects and surfaces. The defect
electronic properties are detected, as known, through
charge collection processes taking place in the
depletion layer, where non-uniform electric fields are
present. It is reported (3) that the contribution from
carrier diffusion to the collected signal would be
a negligible when thin films are observed in STEBIC
mode, because the film thickness is generally
comparable with the width of the depleted layer under
v=OV zero bias conditions for GaAs and Si doped in the
range: 1015<No<1017 cm-3 . For materials and
junctions examined in this work (200 kV), the
observable film thickness ranges from .001 to 1.5 J.1m
while the width of the barriers are ranging between .1
and .5 11m. The carrier diffusion lengths are generally
higher than 111m. Moreover, smaller signals are
expected in the STEBIC mode, because of the smaller
generation volume compared to the SEM case and of
The semiconductor specimen acts as a detector,
not only for the excess minority carriers generated by
c d the focused incident electron beam, but also for
carriers generated by secondary ionization radiation,
produced by the interaction of backscattered or high
v= -tV v= -2V angle diffracted electrons with the sample
surroundings. Moreover, the amount of primary
Fig. 4: a) TEM picture of a dislocation on GaAs bulk. excitation of excess carriers can be affected by
b) Corresponding STEBIC image showing dynamical diffraction effects related to electron
strong recombination contrast at the defect. By channeling. Consequently, the detected Signal, IT, is a
applying a reverse bias (c) the resolution convolution of the generated one, Is, with the
increases, but the contrast is never inverted. recombination function of the defect. If:

~ iT liT == ~ Is lis (1 )

the observed contrast is pratically due to relative

variations of this last signal (15). In samples with high
non radiative recombination rates and high surface
recombination rates we have a decrease in the excess
carrier concentration. It follows that carrier generation
effects introduce some difficulties in the interpretation
of recombination contrast at the defects, particularly in
materials for which both minority carriers diffusion
lengths and depletion layer widths are high (9). In fact,
when the carrier recombination at defect produces
only small effects on the image contrast, this can be
dominated by the carrier generation which is related to
Fig. 5: a) STEM picture taken on GaAs epitaxial film the energy loss of the primary beam per unit specimen
showing a polishing trace. b) STEBIC image of path and to the amount of spurious radiation due to
the same region. The black dots do not interaction of electrons scattered at high angle with the
correspond to extended or linear crystal specimen surroundings. It is well known that both
defects; they are probably due to local off- effects can be related to the excitation of different
stoichiometry or not-resolved impurity clusters. Bloch waves in the crystal which in turn depends on

53 1
the specimen orientation and on Bloch wave scattering like silicon, show a ratio between the barrier width and
at defects. The bright contrast observed in silicon film thickness generally more critical than direct gap
specimens along the scratches in the thinnest regions materials having shorter diffusion lengths, like GaAs.
and the effects of tilting on defects contrast (Fig. 2) are
consistent with an increased fraction of collected
carriers due to this kind of effect. In fact, the stacking 6. ACKNOWLEDGEMENTS
fault observed in Fig. 2 displays a bright-dark type of
image, in the conductive mode, which can be probably We want to thank Dr. P.G. Merli and Dr. C.
correlated to Bloch wave scattering. Only an accurate Donolato (CNR-LAMEL) for useful discussions and
study of the combined electrical and diffraction contrast suggestions.
with a double tilt device will allow to understand these
defect features. Then, on materials with high surface
recombination, the visibility of electrical contrast in
STEBIC mode appears to be very critical. In fact, for REFERENCES
thin specimens the contrast can be dominated by
generation effects, while increasing the film thickness (1 ) D.B. Holt, "Quantitative Scanning Electron
recombination contrast is observed but with the Microscopy", Academic Press, N. Y.,197 4,p. 213.
resolution limitations pointed out by Donolato (16). (2) C. Donolato, J. Appl. Phys. 54 (1983) 1314.
However, when the surface recombination velocity is (3) B. Sieber and J. Philibert, Phil. Mag. B, 55
high, it is also possible to have a mean diffusion length (1987) 575.
parallel to the surface very small (17), improving the (4) P.M. Petroff, D.V. Lang, J.L. Strudel and A.A.
resolution. This picture is consistent with the effect Lojan in: "Scanning Electron Microscopy" 1978
induced by the application of a reverse bias to the Vol. I, SEM Inc., AMF O'Hare (Illinois) 1978, p.
device on the defect contrast (Fig. 3). When the 325.
depletion layer width is increased and becomes (5) D. Fathy and U. Valdre, J. Micr. Spectr. Electron.
comparable with the specimen thickness, the 5 (1980) 1975.
generation effects dominate the contrast and a white (6) D.G. Sparrow and U. Valdre, Phil. Mag. 36
image of defects is observed due to Bloch wave (1977) 1517.
scattering at the defects themselves. By summarizing, (7) D. Fathy, D.G. Sparrow and U. Valdre, J.
the contrast and the resolution in STEBIC are Microscopy 118 (1979) 263.
depending on a very delicate balance of all these (8) D. Fathy, P.E. Batson and U. Valdre, Inst. Phys.
factors: (a) film thickness; (b) surface recombination Conf. Ser. n° 36, 1977, p. 223.
rate; (c) channeling effects; (d) minority carriers (9) P.M. Petroff, EMAG 81, Inst. Phys. Conf. Ser. n°
diffusion length; (e) depletion layer width. When 60, Cambridge 1981, p. 501.
(b), (d) and (e) are small, the recombination effects are (10) A. Camanzi, M. Vittori, A. Parretta and P.
detectable also in the thinnest regions of the film. In Alessandrini, Proc. Xlth Int. Conf. on Electron
fact, on GaAs specimens, a sufficient recombination Microscopy, Kyoto, 1986, p. 395.
contrast is always found also in the thinnest areas of (11 ) P. Alessandrini, A. Camanzi, M. Garozzo and M.
the sample. Moreover, since Ld is small in GaAs, the Vittori, Proc. Vilith Europ. Congr. on Electron
resolution is mainly limited by the beam broadening Microscopy, Budapest, 1984, p. 971.
given by (13): (12) A. Camanzi, A. Parretta, M. Vittori and M.
Garozzo, Solar Energy Materials 17 (1988) 457.
(13) P.G. Merli and U. Valdre, J. of Physics E, 5
b oc [ Z?- r t3 / A ]112 / Eo (2)
(1972) 934.
(14) J.1. Goldstein, J.L. Costley, G.W. Lorimer, S.J.B.
where Z is the atomic number, A the atomic weight, t Reed (1977) SEM 1977, 1, 315, O. Johari ed.,
the film thickness, r the density and Eo the beam IITRI Chicago.
energy. For t=1 11m and Eo=200 keY we have, (15) C. Donolato, EMAG 1981, Inst. Phys. Conf. Ser.
b == .1 11m for Silicon n° 60,1981, p. 215.
b == .3 11m for GaAs. (16) C. Donolato, Phys. Stat. Sol. (a) 65 (1981) 649
Summarizing, when the charge collection and 66 (1981) 445.
efficiency is very small, contrast due to channeling (17) D.B. Darby and G.A. Booker, J. Mater. Science
effects is observed, while, for very high efficiency, a 12 (1977) 1827.
contrast reduction will result, being all the collected (18) M. Garozzo, A. Parretta, G. Maletta, V.
current due to drift current; finally, in the intermediate Adoncecchi and M. Gentili, Solar Energy
cases, dark contrast due to electrical recombination Materials 14 (1986) 29.
will be obtained. Thus, the high collecting factor in
GaAs, with shorter minority carrier diffusion lengths
and lower surface recombination rates with respect to
Si, allows to study defects by STEBIC technique,
without the limits imposed by the generation effects. In
agreement with the previous considerations, an
increase in resolution, together with a reduction in
recombination contrast, is generally obtained applying
a reverse bias Voltage. Our results are in agreement
with theoretical analysis performed by Donolato (14);
in fact, in GaAs, the image resolution, w, appears to be
better than in silicon, when shallow junctions are
realized in epitaxial GaAs films (18). A mean value of
w=.1 11m is found with film thickness values of .1 +.2
11m.
In conclusion, in order to detect recombination
contrast at defects in STEBIC mode, materials with
indirect gap and long minority carrier diffusion lengths,

532
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Improvement of Device Efficiency By Time-

Resolved Photoluminescence Measurements
by

R.K. Ahrenkiel, B.M. Keyes, D.J. Dunlavy and L.L. Kazmerski

Solar Energy Research Institute

Golden. Colorado 80401

ABSTRACT Here R is the recombination rate per cm 3 per second and p and n are
the free hole and electron concentrations, respectively. One may
The photovoltaic technology is based on minority-carrier properties easily show that in low injection, the radiative lifetime is:
of the semiconducting materials under investigation. The improve-
ment of minority-carrier parameters in a cost-effective manner is 'tr = 1/( B N) 2)
the focal point of much current research. The minority-carrier
properties of III -V compound semiconductors are best characteriz,ed where N is the majority carrier density.
by time-resolved photoluminescence in special diagnostic structures.
Here we will describe the measurement theory and measurement The second important mechanism in minority carrier kinetics is the
techniques. Recent developments in the technology relevant to Shockley-Read-Hall(SRH)I,2 or nonradiative recombination
photovoltaic devices will be reviewed. mechanism. The nonradiative lifetime is then given by:

INTRODUCTION -..L =
'tnr
Li (O'i Vth Ni) 3)
The development of high efficiency photovoltaic devices has been Here si and N i are the capture cross section and the volume density
based both on the III-V compound semiconductor and silicon of the ith level. The nonradiative centers that exist at surfaces or
technologies. In this work, we will discuss a characterization interfaces are often called interface states. A solution of the two
technique which has proven to be useful for compound semicon- dimensional problem at the interfacial plane is solved in terms of the
ductors. These devices have been of both single and multijunction interface recombination velocity S(cm/s).
configurations, the la~ter being primarily feasible by the epitaxial
growth of combinations of binary and ternary compounds. Work
has focused on semiconducting materials in the GaAs and S =O'Vth Ns 4)
AlxGal_xAs series although other III-V's have recently been in-
where Ns is the planar density of recombination centers.
vestigated.
One may show that the bulk lifetime is given by adding the
The measurement of minority carrier lifetime in relatively simple
recombination rates. The result is:
diagnostic structures is a quick and efficient step in processing
technology. This preliminary materials assessment is much more
time efficient than fabricating a complete device and then assessing
its performance. Also, the individual components or epiJayers can
be characterized individually in the procedures to be described.
If the active region is free of recombination centers, the minority
carriers will decay with a theoretical or radiative lifetime of l/(BN).
EXPERIMENTAL TECHNIOUE
The cost-effective elimination of deep levels is one of the goals of
materials research.
Our technique for measuring time-resolved photoluminescence is
called time-correlated single photon counting. The data is obtained
DIAGNOSTIC DEVICES
using a pulsed dye laser to excite the PL and photon counting was
used to analyze the results. The PL photons emitted by the sample
The standard experimental technique revolves around the excitation
are focused on the input slit of a scanning monochromator. The
of excess minority carriers (Dp or Dn) by a short pulse oflight such
latter is tuned to the appropriate wavelength which is that corre-
as from a pulsed laser. The photoluminescence (PL) decays with the
sponding to the band-to-band recombination. The single photon
minority carrier lifetime. The total PL output is simply given by
detector is a Hamamatsu model R2809U microchannel plate detector
summing the excess minority carriers (Dp(t) in this case) over the
(MCP) which has a time resolution of 25 ps full-width half-max
active volume and multiplying by the radiative probability. The
(FWHM).
excess carrier minority carrier concentration r is the charge density
in the othe~ise neutral region. Defining:
RECOMBINATION PHYSICS

The recombination mechanisms in III-V materials can be described p =An or Ap 6)

by two basic processses. These are radiative recombination and
Shockley-Read-Hall recombination. Theory shows that the radia-
tive recombination rate Rr of free electrons and holes is given by: p(r,t)dV
7)
Rr = B P r I)

533
 If r(r,t) decays exponentially with the minority carrier lifetime, the 4 Jl DH DeYices
PL intensity can be written in terms of the total initial excess density 100000 ~-----------------------------,

ro:

10000
(f)
8) ........
E
u
The crystal grower makes a confinement structure to prevent mi- 1000
(J)
nority carrier diffusion out of the active region. Here one grows a a
smaller bandgap active region between two wider bandgap layers of
like doping type. The wide bandgap layers act as minority carrier
1 00 ;-~--,-~--,-~--.-~--,-~~
mirrors. One may write a linearized expression for the PL lifetime 700 720 740 760 7BO BOO
in this device as:
T{CO), Growth Temp
9)
Figure 1: Interface recombination velocity at the heterointerface.

where S in the interface recombination velocity at each The interface recombination velocity was found to vary from a
heterointerface. By making two structures which are identical minimum of about 500 cm/s a 775°C growth temperature to 20,000
expect for the active-layer thickness d, one can obtain unique value cm/s at 700°C.
of bulk lifetime and S. If the confinement layer is fabricated of the
same material as used in the real solar cell for passivation or 2. Epitaxial GaAs on GaAs Substrates
confinement, the S value determined above is useful in cell optimi-
zation and modeling. The early work by Nelson and Sobers 6 and later work by 't Hooft7
and co-workers indicated that the lifetime in GaAs DH devices can
This isotype double heterostructure(DH) has long been used as a be described by the radiative recombination lifetime when the 2S/
diagnostic device 3,4,5 for materials in the AlxGal_xAs/GaAs family d term is removed. Thus. that data indicates a monotonic decrease
of devices. Here we use this structure as a means to measure the in the lifetime with doping level going as l/(BN). Recent workS on
minority-carrier lifetime in the normal course of materials research. metalorganic chemical vapor deposition (MOCVD) grown GaAs
By measuring the PL lifetime in these relatively simple, contactless DH structures indicated similar results. The lifetime at No =5x 10 16
devices. one can quickly assess the processing technology. For cm- 3 can be explained by the radiative lifetime and S = 2500 cm/s.
example, one can select the optimum growth temperature by fab-
ricating these devices over a given temperature range. Also one can 3. Ultra-Long Lifetime GaAs
assess the quality of different source materials by making devices
from them in a similar manner. A number of such investigations will Linear transport theory indicates that the maximum lifetime in a
be described here. given region of a device is given by eq. 2. In the absence of
appreciable bulk, nonradiative recombination and interface re-
EXPERIMENTAL RESULTS combination, the minority carrier lifetime approaches the radiative
lifetime. Early workers found that in heavily doped (lx10 18 cm- 3)
GaAslAlxGal_xAs Process Development DH structures, lifetimes exceeding tR were found. Partain 9 and co-
workers measured the spectral response and diffusion length in
1. Passivation and Confinement heavily doped GaAs homojunctions. This work indicated diffusion
lengths that were longer than a "theoretical" maximum obtained
Here we will illustrate a typical example of process development from the majority carrier diffusivity and tR. The explanation of-
using the time-resolved PL technique. These measurements were fered here is that photon recycling increases the lifetime above the
done in collaboration with workers at the Spire Corporation with the theoretical limit producing an enhanced diffusion length.
goal of maxmimzing the performance of a single junction GaAs
solar cell.
200 ,---------------------------,
This device is a metal-organic chemical vapor deposition (MOCVD) d = 4 .0 microns
AlxGa l-xAs passivated GaAs single junction solar cell. The reported
AM 1.5 efficiency of24.8% was a world record for a period oftime.
Process development research was greatly expedited by comple-
mentary time-resolved photoluminescence measurements on double 100 a DetB
heterostructures. The latter simulated the base and emitter regions • Theory
of the solar cell and produced values of the minority carrier lifetime
and interface recombination velocity of the components of the solar
cell. The optimum growth parameters were used in the device 0 +--r_,--~,_~_,--~~~_4
fabrication process. 700 720 740 760 7BO BOO
HCO) , Growth Temp
To simulate the base region ofthe device, diagnostic structures of
composition AIO.3Gao.7As/GaAs/ Alo.3Gao.7As were grown over a
range of temperatures. The doping levels were the same as those of
Figure 2: The PL lifetime of 4.0 mm DH structures as a function of growth
the base and back -surface field ofthe solar cell of. Figure I shows
temperature.
how the interface recombination velocity at the heterointerface
varied with growth temperature.

534
Recent work lO indicated in addition that ultra-long lifetime DH 5. Epitaxial AlxGat_xAs
structures were produced by optimizing the growth temperature.
These devices were made with active layers 4 and 8 mm width and Recent work on devices grown at the Research Triangle Institute
doped n-type to IxlO 17 cm- 3 to 1.5xlO 17 cm- 3. The measured (RTI) describeI6.17.18 the effects of processing on minority carrier
lifetimes were routinely found to be three to five times the radiative properties of MOCVD AlxGa l_xAs. As noted, the earlier work 1Z
lifetime tR of 33 to 50 ns. Some typical data are shown in Figure indicated a rapid decrease of lifetime with the aluminum concen-
2 and shown tPL> tR for growth temperatures above 700°C. tration x. The decrease in lifetime with increasing aluminum
concentration may be the introduction of residual oxygen or water
These ultra-long lifetimes were attributed to photon recycling and vapor associated with MOCVD growth. The source of the con-
substantiated by calculations II. tamination appears to be either the trimethylaluminum (TMA) or
the arsine AsH3 source gases.
One of the best Spire devices has an active layer thickness d is 8 mm,
and a donor density of 1.3xlO l7 cm- 3. The device was grown at The best lifetimes at x - 0040 have been achieved by liquid phase
775°C. The measured PL lifetime of 290 ns is more than 7.5 times epitaxy (LPE) at the Hughes Research Laboratories. The LPE
the radiative lifetime. As expected from theory, the PL lifetime devices were grown by R. Loo and co-workers at the Hughes
approached the radiative lifetime as d was made smaller. The Research Laboratories. These Alo.3SGao.6ZAs layers were grown
analysis of thin structures similarly grown indicated that the interface using the infinite solution LPE technique. The layers were grown
recombination velocity is about 500 cm/s. The photon recycling from a modified multi well crucible housing three AlGaAs solutions
effects must be considered in device design and modeling for high with different compositions and dopants. These devices have
quality epitaxial material. shown minority carrier lifetimes 19 longer than any other reported
in the literature at 40 % aluminum concentration. The confinement
4. GaAs Grown On Silicon layers were made with n-AlxGal_ xAs(x=0.80). The active layers
were n-type and the thickness in the DH structures varied from about
There have been many reports l2 of GaAs majority carrier devices 4~toll~.
grown upon a silicon substrate. The several advantages of the
GaAs/Si heteroepitaxial system are low substrate cost and weight. A composite of AlGaAs data is presented here in Figure 3. The
The cost advantage of using silicon rather than GaAs substrates is lifetime shown here is the bulk lifetime from a number of labora-
still a prime motivating factor for photovoltaic applications. A tories (in epitaxial materials) as calculated by the various means
difficulty associated with the GaAs/Si system is the 4% lattice described in the text. The superior lifetimes of LPE AlxGa I-x As at
mismatch which results in the generation of dislocation densities x-oAO is apparent here. This lifetime improvement apparently
exceeding lOS cm 2 in the GaAs. These dislocations do not seriously results from cleaner starting materials and the purging of impurities
degrade majority carrier transport parameters such as mobility. from the melt during growth. The solid line is a fit through the
However, our studies 13 indicate that the minority carrier lifetime is original data of t Hooft and van Opdorp,12 The subnanosecond
reduced three or four orders of magnitude by recombination at lifetimes expected for x > 0.30 is not found for the samples measured
dislocations. The lifetime reduction is much less when special here. The LPE and the high temperature MOCVD materials exhibit
buffer layers are grown between the GaAs active region and the
lifetimes far larger than expected. The shorter lifetimes inAlxGal_xAs
substrate I4 ,15. The DH diagnostic device is grown on top of the
are process dependent and may be be related to contaminants in the
strain reduction layers. The GaAs buffer region included a thermal
starting materials. The identification and eradication of these
cycle growth (TCG) layer in some cases. The latter is grown by
contaminants is essential to the improvement of III-V minority
several repetitions of a deposition/annealing/cool-down cycle. In
carrier devices. In the meantime, the quality of material based on
other samples, a thermal cycle annealing (TCA) process is used in
AlxGal_xAs (X>0.20) has improved significantly during the last
which the annealing/cool-down cycles are performed after the
several years. The improvements are such that AlxGal_xAs top-
entire buffer layer has been grown. Finally, the DH diagnostic
junctions are the standard in multi-junction solar cells.
structure is grown on top of the buffer region.
GalnP Window Layers
A model of Yamaguchi and co-workers of NIT Opto-Electronics
Laboratories and given by eq. lO.
Recent work by 0lson 2o and co-workers describes DH structures of
composition Gao.s1no.sP/GaAs Gao.s1nO.5P. The measured recom-
bination velocity at the Gao.5Ino.5P/GaAs interface was remarkably
_1_ = B ND + 1t3 0 Nd + 2S. small- about 1.5 cm/s in some devices. They reported aPL lifetime
'tPL 4 d 10) for an updoped DH device(d=1 J.lm) of 14 ms. This is the longest
lifetime ever reported in GaAs. The background doping level in
the active layer is in the range 2-4x10 14 cm- 3 and therefore the ra-
diative lifetimes range from 12.5 to 25 ms. For the several DH
devices fabricated. the lifetimes were independent of active layer
thicknesses (l ~ < d <3 ~) and ranged from 8 to 15
Here Nd is the average density of dislocations (cm- 2). The model ms. The temperature dependence of the lifetime of these devices
have been measured to check for a trap dominated lifetime. The
replaces the radiative and surface recombination terms with the
measurements show that thelifetime decreases as T-I.59 indicative 21
control device lifetime of 30 ns These values span the range of
of radiative rather than nonradiative recombination. The latter
lifetimes found in our control devices and the diffusivity was chosen
as 2 cm 2/s. The NIT model predicts the minority carrier lifetime result can be interpreted that the device is near the theoretical
lifetime. Thus in summary, one can make very high efficiency DH
very well for dislocation densities greater than about 5xl07 cm- 2 as
diagnostic devices with either GaInP or AlGaAs confinement
can be seen from the figure. The model predicts that Nd must be
reduced below 106 cm- 2 in order to get processing limited lifetimes. layers. The passivation technology is very important to photovol-
taic device fabrication.
A variety of experifments are in progress using strained layer
superIattices and other novel structures to reduces dislocation
densities in the active layer.

535
 16 M. Yamaguchi, A. Yamamoto and Y. Itho, J. Appl. Phys. 59,
.000

·· Hoc-II

LP[/Hu,tJet
17
1751(1986).
M.L. Timmons, J.A. Hutchby, R.K. Ahrenkiel, and DJ.
Dunlavy,Symposium Proceedings: GaAs and Related Com-
· "Oevo/JilT! pounds, 00, 289, (1989).
<n
S
.00
· "t)C.YD/V.n.~ 18 R.K. Ahrenkiel, D.J. Dunlavy and M.L. Timmons, 20th IEEE
Photovoltaics Specialists Conference-, Conf. Proc. pp. 611
<l>
(1988).

-
E
....
<l>
19 R.K. Ahrenkiel and DJ. Dunlavy, 1. Vac. Sci. Tecnhol, A&,
822 (1989) ..
- -' '0 20 R.K. Ahrenkiel, DJ. Dunlavy, R.Y. Loo and G.S.Kamath,
10um. Appl. Phys. 63,5174 (1988).
21 J.M. Olson, R.K. Ahrenkiel, D.J. Dunlavy, B. Keyes and A.E.
Kibbler, (submitted for publication).
22 L.W. Molenkamp and H.F. J. van 't Blik, J. Appl. Phys. 64,
. 00
+-------~----~~--~~------~
o. 02 0' 4253( 1988).
A lumi num Content-X

Figure 3: Composite AlxGal_xAs lifetime data.

ACKNOWLEDGEMENTS

The authors wish to thank S.M. Vernon, M. Timmons, J. Olson and

R. Loo for providing the samples and devices used in these studies.
We also wish to thank M M. AI-Jassim for the TEM dislocation
studies of the GaAs/Si system.

REFERENCES
W. Shockley and W.T. Read, Phys. Rev. 87, 335 (1952).
R.N. Hall, Phys. Rev . .81, 387 (1952).
R.J. Nelson, J. Vac. Sci. Technol.U, 1475 (1978).
4 R.I. Nelson andR.G. Sobers, Appl. Phys. Lett. 32, 761 (1978).
5 R.J. Nelson and R.G. Sobers, J. Appl. Phys . ~, 6103 (1978).
6 R.J. Nelson and R.G. Sobers, Joum. Appl. Phys. 49,
6103(1978).
7 G.W. 't Hooft, C. Van Opdorp, H. Veenvliet and A.T. Vink,
Joum. of Crystal Growth 55, 173(1981).
8 R.K. Ahrenkiel, D.J. Dunlavy, J. Benner, R.P. Gale, R. W.
McClelland, J.V. Gormley, and B.D. King, Appl. Phys. Lett.
~, 598(1988).
9 L.D. Partain, D.D. Liu, M.S. Kuryla, R.K. Ahrenkiel, and S.E.
Asher, Solar Cells, 28, 223 (1990).
10 R.K. Ahrenkiel, DJ. Dunlavy, B. Keyes, S.M. Vernon, T.M.
Dixon, K.L. Miller, and R.E. Hayes, Appl. Phys. Lett., 55,1058
(1989).
It K.L. Miller and R.E. Hayes (unpublished).
12 "Heteroepitaxy on Silicon: Fundamentals, Structure, and
Devices", ed. by H.K. Choi, R. Hull, H. Ishiwara, and R.I.
Nemanich, Mater. Res. Soc. Symp. Proc. 116 (Reno, NV
1989) pp. 179-282.
I3 R.K. Ahrenkiel, M.M. AI-Jassim, D.. Dunlavy, K.M. Jones,
S.M. Vernon, S.P. Tobin and V.E. Haven, Appl. Phys. Lett.
53, 222(1988).
14 S.M. Vernon, R.K. Ahrenkiel, M.M. AI-Jassim, T. M. Dixon,
K.M. Jones, S.P. Tobin, and N.H. Karam, Mater. Res. Soc.
Symp. Proc.-1989 (in press).
15 R.K. Ahrenkiel, M. AI-Jassim, DJ. Dunlavy, K.M. Jones,
S.M. Vernon, S.P. Tobin, and V.E. Haven, 20th IEEE Photo-
voltaics Specialists Conference -1988, (in press).

536
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

INVESTIGATIONS OF EMITTER CHARACTERISTICS OF LPE GaAs SOLAR CELLS

H. Welter, A. Bett, A. Ehrhardt, and W. Wettling

Fraunhofer-Institut fUr Solare Energiesysteme, Oltmannsstrasse 22, D-7800 Freiburg, FRG
Phone: -49-761-4014-0; FAX: -49-761-4014-250

ABSTRACT. GaAs solar cells of 22 % AMl.5 efficiency were processed using the LPE
etchback-regrowth method. In order to improve performance and yield of the cells a systematic
study of the emitter characteristics was performed.
By varying the most significant growth parameters, viz. the melt composition, the
temperature and the time of growth, series of cells with emitter thicknesses between 0.5 and 4
microns and with various thicknesses of the AlxGal_xAs window layer were grown. The
emitter layers were investigated by X-ray rocking curve analysis, CV-profiler and
photoluminescence decay measurements. A detailed study of the Zn diffusion process was
performed and interpreted in the framework of the interstitial-substitutional diffusion model.
It was found that the solar cell performance is not critically dependent on the emitter
thickness Xj between Xj = 1 and 3 microns. At smaller values of Xj. however. Jsc and Voc
decrease rapidly with Xj. This is consistent with higher recombination parameters as observed
in the PL decay measurements.

1. INTRODUCTION layers were characterized by various methods to study

the details of the epitaxy process. The samples were
Since the early sixties GaAs solar cells have also processed to solar cells of lxl cm2 area by our
always been the cells with the highest efficiencies /1/. standard process 171. As a result of these
Today's best cells show one-sun AM1.5 efficiencies of investigations solar cells can be processed reproducibly
24.8 % with an AIGaAs window layer /2/. resp. 25.7 % with AM 1.5 efficiencies of about 21 % and with peak
with an GaInP2 window layer 13/. These record values efficiencies of 22 % (see figure O. It will be shown
have been achieved by using the MOCVD growth with in the next section that these efficiencies are obtained
fairly involved layer structures including BSF layers. from layers with optimal emitter thicknesses of 1 to 2
Model calculations and experimental results have shown microns. The design of a LPE-ER cell thus differs
that the most efficient configuration has an emitter of considerably from that of a MOCVD-based cell. These
about 0.5 microns thickness and a hetero-window layer differences will be discussed in section 3.
of about 50 nm thickness 14/. An alternative method of
GaAs growth for solar cells is LPE. This technique is. 30,---------------------------------------,
in principle. also suitable for growing complex layered
structures by using multi-melt configurations 151. 2S
One modification of LPE growth is the "etchback-
regrowth" technique (LPE-ER) which has the advantage
of great simplicity 16/. In this growth procedure a
single Ga melt is used which contains Zn as dopant and
Al for the hetero-window layer. It is saturated with
respect to As at T 1= 850 0 C. The temperature is then
raised to T 2 = 865 0 C so that the melt is now slightly
undersaturated. When the melt is brought into contact Voc = 1015.7 mV
with the n-doped substrate an isothermal etchback I.. = 26.1 mA/cm2
process is started until the melt is saturated again with 5
fF = 83.9 r.
respect to As. The Zn dopant diffuses into the 1) = 22.2 r.
substrate thus forming the p-doped emitter by O+-----~------r_----~----~------~----~
200 400 soo 100 1000 1200
overcompensation of the substrate doping. At the same VOlIAGE ~V.J
time Al replaces the Ga atoms in the region near the
surface to a large extent. thus forming the p-doped Figure 1: I-V characteristics of a 1 cm2 GaAs LPE-ER
Al xGal_xAs window layer. solar cell under AM 1.S illumination.
The etchback-regrowth process depends critically
on the contact time tc' which is responsible for the
emitter depth and profile. on the undersaturation 2. EXPERIMENTS
temperature ~T= T2-Tl. and on the melt volume. The
latter influences the etchback rate and thus the surface 2.1 Zn- diffusion
quality and also the formation of the window layer. In a series of about 30 epitaxy runs the contact
In order to optimize the GaAs cell efficiency a time of substrate and melt was varied systematically
systematic study was undertaken to find the best between 10 and 230 minutes. The carrier (hole)
growth conditions for the LPE-ER technique. The concentration profile of the emitter was subsequently
measured by a CV-profiler. As an example. figure 2

537
shows 3 profiles of different emitter thicknesses Xj. measured emitter depths Xj as a function of XIZn for
The relation between the contact time and the emitter constant contact times of the melt tc= 45 min. The
depth is given in fig}lre 3. It is described by a square solid line gives a best fit and represents a relation:
law, i.e.: Xj IX (t)1I2. This is an indication for a
concentration dependent diffusion mechanism, i.e. for
D= D(C).
..-----------------------------------~.--,
~~---------------'~~~~~II
DIFFUSION llME
. P20mIn
[] H20mln 3
"
10"
~
• P45m1n
o H45m1n
];

I!10"
1 x-
_ P 130 min
:z
" H 130 min
~ 2 ATOM% ZINC

;:----- ~
k
• xz,,= 70mg
• Xz" = 70""90 no AI
"
•
•
•
Oio----,---~r---~.r--~-;.r-~--~~c---~ac---~w~--~~
~+_----~--~----~----r_--~----~
0 3 5
I min 1
D£PTH &un]
Figure 2: C·V profile of three GaAs LPE {ayers grown Figure 3: Emitter thickness xj as a function of the square
under diffusion times of 20. 45 and 130 minutes. root of the diffusion time for a number of GaAs LPE
respectively layers. The solid squares are AIGaAsIGaAs hetero-
The concentration profiles C(x) can be described by window layers. the solid circles are layers grown
an complementary error function (erfc) with a value
without Al in the melt (see text)
C(O) at the GaAsI AlGaAs interface that is almost ~~---------------------------------,
identical for all profiles:
C(O) = 3-5 xlO 18 cm- 3.
These doping profiles can be analyzed in the
framework of a substitutional-interstitial diffusion
model 181. Here we follow the work of Algora et al.
/9/. These authors performed a SIMS analysis of LPE
doping profiles. Their results are in some respect very
similar to our findings, in some respect there is some
disagreement, as will be discussed below.
In the substitutional-interstitial diffusion model the
Zn dopants are either singly ionized negative acceptor
ions on Ga sites (Zn-) or singly or doubly ionized
donors (Zn+, Zn++) on an interstitial lattice site. The
number of the interstitial Zn ions is orders of ~+.D-'----------~----'-------~--~~----~."
magnitude smaller than that of the substitutional ions, ATO~IC ZINC COKTrnT ~ AT 865 c
(therefore Zn acts as an acceptor) but the diffusivity
is much larger on interstitial sites. Figure 4: Dependence of the emitter thickness on the Zn-
This model leads to a solution of Fick's second concentration in the melt
law with a concentration dependent diffusion constant:
The surface concentration Co(Zn) in these samples
(1) D IX Cr+1 / XIAs is, of course, also dependent on XIZ n' It was found,
however, that Co(Zn) is about a factor 30 smaller than
where r = 1 for Zn+ - and r = 2 for Zn++ -diffusion expected from a theory by Jordan 1111. It is assumed
and XlAs is the As concentration in the melt (see that this results from evaporation of Zn atoms during
below). the heat-up time of the melt because Zn has a rather
The diffusion constant can be evaluated at every large vapor pressure.
point x of a measured diffusion profile by a The diffusion of Zn into the substrate is also
Boltzmann-Metano analysis 1101. It is found that at influenced by the aluminum concentration in the melt.
large carrier concentrations ( C> 2 1018 cm- 3 ), i.e. in The phase diagram of the AIl GalAs 3-component melt
the region of the profile near the surface D is shows that with decreasing Al content the As saturation
constant, whereas further in the bulk at the diffusion concentration increases. This in turn leads to an
front the diffusion constant obeys a D IX C2 law increased number of Ga deficiencies in the substrate.
indicating that Zn diffuses as a singly ionized donor. To a large degree the diffusion process of Zn is
To a large extent the diffusion of the substitutional Zn- therefore shifted from the fast interstitial Zn+
acceptor is thus due to the interstitial Zn+. As a diffusion towards the slow substitutional diffusion of
consequence of this concentration dependence of D the Zn- and D becomes smaller with reduced Al
profile is fairly flat throughout the emitter and has a concentration in the melt. This can also be seen in
steep diffusion front, as is shown in figure 2. equ.(1) where D is inversely proportional to XIAs'
There is also a dependence of D on the Zn
concentration in the melt, X1Z n' Figure 4 shows

538
2.2 AlxGal_xAs hetero-window layer A second reason for reduced values of Isc at
In a second series of LPE runs several layers smaller values of Xj is the following: Not all light is
were grown with different undersaturation temperatures absorbed in the emitter if Xj is smaller than 1 micron.
fj, T and subsequently analyzed by X-ray rocking curve Some of the carriers, however, that are generated in
analysis and by photoluminescence (PL) mapping. In the base are lost by diffusion and recombination
table I the results obtained from 3 samples with fj, T = because we have no BSF in our base structure.
2, 10, and 15.7 oC, respectively, are presented. It is These findings are supported by photoluminescence
seen that the AIGaAs window layer thickness decreases decay (PLD) studies on samples of different emitter
from 301 nm for fj,T= 2 °C to 122 nm for fj,T= 15.7 thickness Xj' In PLD measurements a photoluminescence
oC, whereas the AI-content x remains constant: x '" 0.9. transient is recorded that is excited by a short (20 ps)
PL mapping showed, in addition, that the layers with laser pulse. The set-up and the theory of PLD has been
large fj, T are more homogeneous with respect to PL described elsewhere /16, 17/. Here it is important to
intensity /17/ . Since the optimum window layer mention that the PLD transients depend on the surface
thickness of the hetero-window layer for optical and and bulk recombination parameters tb and S,
electrical passivation lies between 50 and 100 nm the respectively and on Xj and D in a fairly complex way.
undersaturation fj, T should be chosen at least 15 °C. The recorded transients are simulated on a computer by
variation of the above mentioned parameters within
2.3 Solar cell parameters reasonable limits.
The influence of the Zn diffusion and therefore of
Xj on the solar cell parameters was studied by growing ------------------------------------------------------
a series of 30 layers with different Xj and processing Table 1: Influence of window-layer thickness and AI-
them into solar cells according to our standard process content on undersaturation temperature fj, T
but without AR coating.
Fig. 5 gives the measured values of 1sc as a fj,T(OC); wind.thickn.( nrn); AI-content x
function of Xj' The solid squares are the experimental 2 301 0.91
values, the dotted and solid lines are calculations which 10 133 0.90
were performed with help of the computer programme 15.7 122 0.92
PC-lD /13/. There is a broad maximum of 1sc values ------------------------------------------------------
between I and 2 microns. For larger values of Xj' Isc
drops because the diffusion length L in this region is J~--------------~====~
equal or smaller than Xj and not all carriers are • x =3.8~
collected by the pn-junction. The experimental values lie .. ~ODE1.

between theoretical curves calculated for surface

recombination velocities S= 103 and S= 104 cm/s, • x .=.12L:!!!
respectively. This shows that the front surface • ~ODE1.
passivation by the AlGaAs window layer is rather good. '!4~
For values Xj below I micron the experiments give a
strong decrease of Isc with decreasing x,. This can be
explained if one assumes larger values 0/
surface (and
bulk) recombination for smaller Xj' Smaller Xj values
mean shorter contact times of substrate and melt. It
can be assumed that the contact of melt and substrate
has a gettering effect of impurities in the surface near
region /14, 15/. It is then plausible that a shorter ~·+_------~------~--~~~------~------7.
M ~ U ~ ~
contact time leads to larger values of recombination TildE [ns)
rate.
~~-------------------------------. Figure 6: Photoluminescence decay transients of 3 GaAs
solar cells of different emitter thicknesses; dotted:
experimental values. solid lines: theory calculated using
the following parameters: For xr3.8 f..LlTI: S=7103 cm/s.
T=3 10.9 s. D=27 cm2ls; for xr2.5 pm: 5=104 cmls.
T=2 10-9 s. D=30 cm2ls; for xj=1.9 f..LlTI: S=5 104 cmls.
T=D.8 10-9 s. D=35 cm2ls

In Fig. 6 PLD transients taken from 3 samples

with different Xi are shown. The theoretical transients
(lines) were calculated using parameters as given in
the figure caption. Note that for the thinner cells larger
values of S are found to fit the curves. This is in
W~O----~O~----'---~U~--~2~---U~---3'---~3~~--~ agreement with the interpretation of the 1sc results as
D.UTl<R lliICI(IoIESS (p~ discussed above.
The values of Voc for the 30 samples of figure 5
Figure 5: J sc as a function of emitter thickness for 30 were also measured. They show a similar dependence
LPE GaAs solar cells (without AR coating); squares: on Xj as 1 sc ' but the maximum is less pronounced /12/.
experiment. lines: theory (PC·1D)

539
3. DISCUSSION /3/ S.R. Kurtz, J.M. Olson, A. Kibbler "High-efficiency
GaAs solar cells using GaInP2-window layers",
The analysis of the Zn diffusion process showed that Proc. 21st IEEE-PVSC, Orlando PI 1990, p. 138
the main contribution of the Zn diffusion is due to Zn+ /4/ P.D. Demoulin, M.S. Lundstrom "Projections of
on interstitial sites. This is in disagreement with the GaAs solar-cell performance limits based on two-
findings of Algora et a!. who have found from a dimensional numerical simulation", IEEE- Transact.
D a C3 dependence that Zn diffuses as Zn++. On the Electr. Dev. 36, 897 (1989)
/5/ Alferov Zh.I., Andreev V.M., Konnikov S.G.,
other hand a Zn+ diffusion from the vapor phase has Larinov V.R., Pushny B.V "Investigation of a new
been described by Roedel /8/. In this work also CV- LPE method obtaining AI-Ga-As heterostructures",
profiling measurements have been performed. So it Kristall und Technik, 1976, No 10, p.1013-1020.
seems that there is a disagreement between CV- /6/ J.M. Woodall, H.J. Hovel "An isothermal etchback-
profiling and SIMS measurements. regrowth method for high-efficiency Ga1_xAI As-
It was mentioned above that the Zn evaporates from GaAs solar cells", Appl. Phys. Lett. 30, 4n (1~),
the melt during the heat-up time on the melt. There is reprinted in Solar Cells 29, 173
an indication that during this time Zn diffuses into the (1990)
substrate even before the melt is brought into contact /7/ A. Bett, F. Lutz, T. Nguyen, W. Wettling "High-
with the substrate. This layer that is diffused from the efficiency isothermal LPE grown GaAs solar cells
vapor phase must be removed completely in the with high fill factor", Proc. 9th EC Photovoltaic
Solar Energy Conf. Freiburg 1989, Kluwer Academic
etchback process after bringing the melt into contact Publ. London. p. 510
with the substrate in order to start the well defined Zn /8/ R.]. Roedel "Recent developments in diffusion in
diffusion from the melt. Therefore a not too small GaAs and Gah~Al,J.s", in "Gallium Arsenide
undersaturation t. T and also a not too small volume of Technology", VoLli, U.K. Ferry ed., Howard W.
the melt is mandatory. Sams and co, Carmel, Indiana, 1990, p. 287
The etchback and the subsequent regrowth process /9/ C. Algora, G.L. Araujo, A. Marti "Diffusion of Zn
are important as cleaning and gettering steps for the into GaAs and AlGaAs from isothermal liquid phase
surface and the bulk region close to the surface. This epitaxy", J. Appl. Phys. 68, 2732 (1990)
technique exhibits an optimal emitter thickness of 1 to /10/ B. Tuck "Atomic diffusion in III-V
2 p.m which is larger than the emitter thicknesses used semiconductors" ,ISBN 0-85274-351-3
/11/ A.S. Jordan "The solid solubility isotherms of Zn
in MOCVD- grown structures.
in GaP and GaAs" , J. Electrochem. Soc. 118, 781
The LPE-ER technique when used in connection (1971)
with optimized solar cell technology has an AM 1.5 /12/ H. Welter "Optimierung von GaAs Solarzellen
efficiency potential of about 23 %. By using a double hohen Wirkungsgrads", Diplom-thesis, University of
layer AR coating we hope to reach this limit in the Freiburg, FRG, 1991
near future. /13/ P.A. Basore, D.T. Rover, A.W. Smith "PC-lD
version 2: enhanced numerical solar cell modelling",
Proc. 20th IEEE-PVSC Las Vegas 1988, p. 389
ACKNOWLEDGEMENT /14/ H.J. Hovel "Solar cells", in "Semiconductors and
Semimetals" Vol. 11 , Academic Press, (1975)
Valuable support by T. Nguyen, F. Lutz, Ch. /15/ V.M. Andreev, O.K. Salieva, U.A. Solovev, O.V.
Sulima "Bulk nonradiative recombination reduction in
Schetter, A. Habich, G. Habermann, and S. Cardona is n-GaAs, due to annealing under GaAl-melt", Cryst.
gratefully acknowledged. This work was supported by Res. Technol. 23, 1531 (1988)
the Bundesminister fUr Forschung und Technologie /16/ W. Wettling, A. Ehrhardt, A. Bett, F. Lutz
(BMFT) under contract number 0328554B. "Transient photoluminescence decay investigations of
LPE GaAs heteroface solar cells", Proc. 21st IEEE-
PVSC, Orlando Fl 1990, p. 357
/17/ A. Ehrhardt, W. Wettling, A. Bett "Transient
REFERENCES photoluminescence decay study of minority carrier
lifetime in GaAs heteroface solar cell structures",
/1/ J.C.C. Fan "Promises of III/V solar cells", Tech. to be published Appl. Phys. 1991
Digest of the Int. PVSEC-5, Kyoto 1990, p. 607
/2/ S.P. Tobin, S.M. Vernon, S.J. Wojtczuk, C. Bajgar,
M.M. Sanfancon, T .M. Dixon "Advances in high-
efficiency GaAs solar cells", Proc. 21st IEEE-
PVSC, Orlando Fl 1990, p. 158

540
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

InGaP, a promising material for tandem solar cells

A. van Geelen, P.R. Hageman W. Gabrielse and L.J. Giling
University of Nijmegen, Department of Experimental Solid State Physics III, RIM,
Faculty of Science
Toernooiveld, 6525 ED Nijmegen, the Netherlands

ABSTRACT. Int-xGa."P can be used as a material for wide bandgap solar cells. The photovoltaic quality of
Int-xGa."P is better than that of Alt_xGa."As, which is used mostly in GaAs based tandem solar cells. This paper
describes the MOCYD growth and characterization of Int-xGa."P epi on GaAs. Perfectly lattice matched In1-xGa."P
epilayers have been obtained at different growth temperatures. These lattice matched layers exhibited excellent optical
and electrical properties. Photoluminescence measurements at 4K resulted in a F.W.H.M. of 9.0 meY while Hall- v.d.
Pauw measurements of the undoped layers gave mobilities of 6000 cm%y-tsec- 1, the best value reported in literature
sofar. The lattice matched layers have been doped n- and p-type over a wide carrier concentration range. Prelimenary
results on In1-xGa."P solar cells indicate an efficiency of 4.5 % (AM1.5). Improvement of this efficiency may be attained
easily.

1. INTRODUCTION portance, control of the layer composition and lattice

matching to the GaAs substrate are discussed first.
Monolithic two junction solar cells based on GaAs Secondly, the growth rate of Int-xGa."P, which is im-
and AlxGal_xAs have been predicted to reach an ef- portant with respect to control of the layer thickness,
ficiency of 30 - 35 % (AMO, one sun). But in spite is given as a function of temperature. Furthermore n-
of this, only recently a GaAs/ AlxGat_xAs tandemcell type and p-type doping of 1no.6Gao.6P have been per-
was able to attain an efficiency slightly higher than a formed. As a result of these investigations InO.6Ga.o.6P
single junction GaAs solar cell [1]. Limitation of the solar cells on GaAs substrates were made and some
efficiency of the tandemcell is generally ascribed to the prelimenary results are presented.
poor photovoltaic quality of the AlxGat-xAs. This is
thought to originate from the sensitivity of AlxGat-xAs
2. EXPERIMENTAL
for small traces of O 2 and H 2 0 present during growth
in the MOCYD reactor. 2.1 MOCYD Growth and Characterization
Int-xGa."P is a IIl/V material which can cover, All the Int-xGa."P layers were grown in a com-
dependent on composition, a wide range of bandgap mercially available low pressure MOCVD reactor. Tri-
values. If it is chosen to be lattice matched to GaAs it methylindium (TMI) and trimethylgallium (TMG)
has a bandgap of about the same value as AIo.4Ga.o.6As: were used as sources for the group III components. Ar-
1.90 eV. However, compared to AlxGat_xAs it has a sine and phosphine, both 100 % pure, were the group
much lower sensitivity for oxygen and water. This sen- Y precursors. Disilane (Si 2 Hs , 100 ppm in N2 ), was
sitivity is of the same order as for GaAs and forms used as n-type dopant while diethylzinc (DEZn) was
no problem for modern, state of the art reactors. Fur- available as p-type dopant. All epilayers were grown
thermore the recombination velocity at the lno.sGao.sp- on (100) 2° off towards (110) oriented GaAs substrates
GaAs interface is approximately 40 times lower than (both doped and semiinsulating). The reactor pressure
at the AlxGat_xAs-GaAs interface [2,3]. So as for the was kept at 20 mbar in all experiments. Hydrogen, pu-
photovoltaic quality, it seems that Int-xGa."P is a more rified by a palladium diffusion cell, was used as carrier
suitable material for tandem solar cells than gas at flow rates of 5 or 7 slm.
AlxGat-xAs. Samples were electrically characterized with C.V.
Control of the Int_xGa."P epilayer composition is profiling and Hall- v.d. Pauw measurements. Optical
of primary importance because of the intended lattice quality of the grown material was determined with pho-
matching to GaAs. For Inl_xGa."P this is less easily toluminescence measurements, either at 4 K or at room
achieved than for AlxGat_xAs. The difference in lat- temperature. The composition of the Int-xGa."P epi-
tice constant between GaP and InP is large compared layers and the mismatch between Int-xGa."P and GaAs
with that between GaAs and AlAs. Therefore a small substrate could be calculated from X-ray rocking-cur-
change in composition of the Int-xGa."P will cause a ves and electron microprobe measurements (EPXMA).
relatively large change in lattice constant. In prac- Layer thicknesses were measured on cleaved and etched
tice both control of composition and lattice matching samples by SEM and optical microscope.
can be achieved without difficulties [4]. So it appears
that Int-xGa."P is really a promising material for tan- 2.2 Solar Cell Processing and Efficiency Measurement
dem solar cells. This is supported by the fact that a After the MOCYD growth front (Au/Zn/ Au) and
monolithic InGaP /GaAs tandem solar cell with an '7 back (Au/Ge/Ni) contacts were made bye-beam evap-
of 27.3% is made by Olson et al [5]. oration. The front contact was defined by a mask and
In this paper we present the first results of our obscured approximately 8% of the cell area. After
Int-xGa."P growth experiments. Because of its im- t
evaporation the contacts were annealed for 4 minutes

541
at 460 'C. The cell's perimeter was defined by cleaving mismatch of 6.5 x 10- 4 at room temperature which im-
the edges. A selective etchant was used to remove the plies a solid composition X G • = 0.507. Part b of fig.
GaAs contacting layer between the grid fingers. The 3 gives a plot of a perfect lattice matched In1-xGa."P
antireflection coating consisted of three layers: MgF 2 , layer at room temperature. The FWHM of 21 arcsec.
ZnS, MgF 2 , with thicknesses of 20,520 and 980 A. for fig. 3a and 21 arcsec. or less for fig. 3b indicates the
The cell efficiency was measured using a simple good quality of the epilayer. The little satelite peak
xenon lamp and a water filter. The short circuit current o. 55 r-r--,--,--r.........-.-.,.,.-r-T""T-.--~--,--..~......,
was calibrated by using a reference cell. The current-
voltage characteristic was measured by a four contact
method. Q

0.50
3. RESULTS AND DISCUSSION

3.1 Lattice Matching

To obtain lattice matched In1-xGa."P epilayers on
GaAs substrates, x should be 0.516 at room tempera-
ture and 0.504 at 900K. Experiments were carried out 0.45 ••••• T=800·C
aaaaa T=840'C
with different gas phase compositions and using differ- uu.T=700·C
ent growth temperatures. In fig. 1 the fraction gallium - - T=800'C
in the solid (Xs) is shown as a function of XG , i.e. the - - - - T=640,700'C
concentration of TMG over the total concentration of
III components (TMG + TMI) in the gas phase, at 0.4 0 O
L..L.->-L---10-'-0->-..........2.. .L0-0..........-'3-0LO-'--'--'-4-'-00-'-..........5...J00
three different growth temperatures. As can be seen
in fig. 1 lattice matching can be obtained in the whole V/111
temperature region used. The gas phase composition
Figure 2: solid phase gallium content (Xs) of the
necessary to obtain lattice matched epilayers depends
Inl-xGa."P epilayer as a function of the V fIll ratio at
however on the growth temperature. At higher growth
different growth temperatures.
temperatures less TMG is required.

GaAs
20" a
t..!..!.tl T=800'C

0.60 =T=1I40'C
~T=700'C

26" ~ FWHM
0.55
FWHM
til
X 0.50
WB (arcsec)
-I

0.45
GaAs 20" b
0.40 LL.............L-I..........................L.J'-L-'-'-....................................................... InI-xGa."P
0.40 0.45 0.50 0.55 0.60 0.65 A
XG
Figure 1: solid phase gallium content (Xs) of the FWHM 21"
In1_xGa."P epilayer as a function of the gas phase com-
position (TMGjTMG+TMI) at different growth tem-
peratures.
The solid composition is found to be indepen-
.J ~ (arcsec)
-
dent of the V JIII ratio (i.e. molfraction of phosphine
over molfraction TMG and TMI) applied at growth Figure 3: X-ray rocking curves for a) slightly mis-
temperatures above 640 'C. At 600 'c instead the epi- matched Inl_xGa."P epilayer and b) perfectly lattice
layer becomes more gallium rich at higher V JIII as is matched epilayer at 300 K.
shown in fig. 2. This might be due to the fact that
TMG decomposes faster in the presence of phosphine observed at higher w in fig. 3a (and possibly in fig.
or arsine. 3b) is due to a more gallium rich composition of the
Examples of X-ray rocking curves are given in fig. In1-xGa."P. This may be caused by a non optimum
3. Figure 3a shows a graph of an Inl-xGa."P epilayer switching procedure between the GaAs buffer layer and
which is slightly mismatched at room temperature, al- the In1-xGa."P epilayer.
though it is nearly lattice matched at the growth tem-
3.2 Growth Rate and Incorporation Coefficient
perature. From this rocking curve we calculate a lattice
The total growth rate and the ratio of the rate of

542
incorporation of gallium and indium species [61 (a) are (peakenergy 1.87 eV), a value which belongs to the best
plotted as a function of reciprocal temperature (l000/T reported in literature.
(K- 1)) in fig. 4. For these experiments the concentra- In all lattice matched samples a P.L. peak en-
tions of TMG and TMI were equal and kept constant ergy shift was observed. The shift indicated a low-
over the whole temperature region. ering of the normal bandgap of 1.98 eV by 20 to 100
The growth rate shows an apparent activation me V (at 4K), depending on the growth conditions. The
energy of 5.8 kcal/mol in the low temperature region. anomalous change in bandgap is caused by spontaneous
The same apparent activation energy is found for growth (CuPt-type) ordering in the material, as is described
of GaAs under similar conditions. This indicates that by several authors [7,8,91.
this temperature behaviour is determined by the gas
phase decomposition of TMG. Under these growth con- 3.4 Electrical Characterization and Doping
ditions the growth rate is in the transition region be- Our undoped In 1- x GaxP layers are typically n-
tween kinetically gas phase limited and diffusion gas type, with a free carrier concentration of less than
phase limited growth. 1016 cm- s . The mobilities of these undoped layers vary
At growth temperatures higher than 650°C the between 3000 and 6000 cm2 V-I sec- 1 • To the know-
growth rate shows nearly no temperature dependence. ledge of the authors these are the highest values re-
This is consistent with diffusion limited growth. ported in literature.
The growth rate was found to be independent Doping experiments were performed by changing
of the V/III ratio for the growth temperatures used, the input molfraction of either disilane (n-type) or di-
although the V /I11 ratio was varied between 50 and ethylzinc (p-type). The growth temperature was 600
900. 'C. We were able to dope Ino.sGao.sp n- and p-type
T(OC) over a wide range starting at a few times 1016 up to
710 11110 1110 MO 8 X 1018 cm- s . Mobilities of the intentionally doped
1.40 0.70 layers are very high with respect to the literature val-
ues.

1.20 0 .60 3.5 Solar Cell Characteristics

'i:'
..c:: A design of our first Ino.s Gao.s P solar cells is de-
..........
ex e picted in fig. 5. The solar cells were grown at 600 'c
1.00 0.50 :t
....... on n-type GaAs substrates (n = 5 X 1017 cm- S ). First a
Cl:: thin GaAs buffer layer was grown (n = 5 X 1017 cm- S).
.5 Successively the Ino.sGao.sp base and emitter were
0.80 0.40 grown and doped respectively n = 2 X 1017 cm- s and
p = 1 X 1018 cm- s . The p-type GaAs contacting layer
was doped p = 8 X 1018 cm-s. No passivating layer was
0.60 1-.<-~~-<-..~.......-,-........................~~.......J 0 . 30 yet applied. The solar cells were made reasonable thin
1.00 1.05 1.10 1.15 1.20
1000/T (K- 1 ) because of their future use as a top cell in a monolithic
tandem solar cell.
Figure 4: ratio of rates of incorporation of gallium and
indium growth species a and growth rate (R) as a func-
tion of reciprocal temperature (1000/T). a is defined p- InGaP emmiter 0.5 J.lm
as rTMG/rTMI. rTMG and rTMI are concentration inde-
pendent growth rates for GaP and InP.
In fig. 4 it is seen that at lower temperatures a n- InGaP base 1.4 J.lm
depends strongly on temperature. This is thought to
be caused by the larger activation energy for gas phase
decomposition of TMG in comparison with the activa- n- GaAs buffer layer 0.1 J.lm
tion energy of TMI. So at these lower growth tempera-
tures the incorporation ratio and thus the composition
n - GaAs substrate
of the solid is determined mainly by the TMG decom-
position. At higher temperatures both TMG and TMI
decompose fast enough so that a levels off. In this case
diffusion of growth species towards the surface deter-
Figure 5: schematic drawing of our Ino.sGao.sp solar
mines a. In this region a is slightly larger than one,
cell.
meaning that the overall rate of incorporation of gal-
lium growth species is somewhat larger than that of Although no effort has been made to optimize
indium species. this Ino.sGao.sp solar cell, already an efficiency of 4.5
3.3 Photoelectrical quality % is reached (AM 1.5, one sun). From fig. 6 it can be
The photoelectrical quality of lattice matched seen that the Voe is 1.16 V, the J se is 6.2 rnA cm- 2
In1 - x GaxP as determined by P.L. is a complex func- and the FF is 0.63.
tion of growth temperature, V /111 ratio and growth The cells still suffer from a rather big series re-
rate. In all lattice matched samples reasonable P.L. in- sistance, causing the poor fill factor. Also the cur-
tensities and F.W.H.M. values were obtained, although rent density is not maximal, probably because part of
some samples did not show any luminescence at room the incoming light is not absorbed in the thin solar
temperature. Our best F.W.H.M. at 4K was 9.0 meV cells. The lack of a passivating layer results in a rather

543
 -e
N

<.I
111
V...=1.16 V
J .. =6.2 mA/cm 2
13] J.M. Olson, R.K. Ahrenkiel, D.J. Dunlavy, Brian
Keyes, A.E. Kibbler, Appl. Phys. Lett. 55, 1208
-........ (1989).
< FF=O .63

'-'
e 10
7)=4.5~ [4] Shigekazu Minagawa, Hitoshi Nakamura and
Hisumi Sano, J. of Cryst. Growth, '11, 377 (1985).
[5] J.M. Olson, S.R. Kurtz, A.E. Kibbler and P.
Faine, Appl. Phys. Lett. 56,623 (1990).
[6] W.G.J.H.M. van
Sark, G.J.H.M. Janssen, M.H.J.M. de Croon, X.
Tang, L.J. Giling, W.M. Arnold Bik, C.P.M. Dun-
-0.5 o 0 .5 1.5 selman, F.J.P.M. Habraken, W.F. van der Weg, J.
v Appl. Phys. 64, 195 (1988).

-& [7] Tohru Suzuki, Akiko Gomyo and Sumio Iijima, J.

of Cryst. Growth, 93, 396 (1988).

[8] Akiko Gomyo, Tohru Suzuki and Sumio Iijima,

Phys. Rev. Lett. 60, 2645 (1988).
Figure 6: I-V characteristic of the in figure 5 depicted
solar cell. [9] G.B. Stringfellow, J. of Cryst. Growth, 98, 108
(1989).
high recombination velocity at the emitter/air inter-
face which diminishes the current. The value of the
open cicuit voltage indicates that the p-n junction is ACKNOWLEDGEMENTS
not optimal yet.
The authors like to thank P.C. van Rijsingen,
Improvements can be carried out easily by es-
G.J. Bauhuis, S.M. Olsthoorn and A.L.H. Stolls for
tablishing the optimal values for layer thicknesses and
their indispensable help.
doping levels. Further benefits are expected from a
thin passivating layer of AlxGal_xAs or AlxInl-xP on
top of the emitter. Also the switching procedure from
GaAs to Inl-xGaxP can be improved,.thereby reducing
the number of threading dislocations at the junction.

4. CONCLUSION

The results of our first investigations on the wide

bandgap material Inl-xGaxP are quite encouraging.
By MOCVD the material can be grown lattice matched
to GaAs over a wide range of temperatures. Both in-
corporation coefficient and growth rate depend in the
lower temperature region (600°C - 650°C) on the de-
composition rate ofTMG. Using optimal growth condi-
tions, high quality Ino.sGa.o.sp epilayers could be grown
with an X-ray rocking curve F.W.H.M. of 21 arcsec, 4
K photoluminiscence F.W.H.M. of 9.0 meV and record
mobilities of 6000 cm2 V-l sec- 1 in undoped samples.
Prelimenary efforts to make In1-xGaxP solar cells
yielded an efficiency of 4.5%. Improvement of the cell
can be obtained easily by finding the optimal values
of thickness and doping levels for the different layers.
Also a passivating layer may be applied. At this time
improvement of the solar cell efficiency seems straight-
forward and there seems to be no basic limitations to-
wards higher efficiencies.

REFERENCES

[1] B.C. Chung, G.F. Virshup, and J.G. Werthen,

Appl. Phys. Lett. 52, 1889 (1988).

[2] G.W. 't Hooft, M.R. Leys and F. Roozeboom,

Jap. J. Appl. Phys. 24, L761 (1985).

544
IOTII EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Design and Measurement of Antireflection Coatings for AIGaAs/GaAs Solar Cells

M. E. Nell, Z. EI-Ankah, H. Eschrich, D. D. Lin, G. Nischwitz, B. Reinicke, H. G. Wagemann

Technische Universitat Berlin, Institut fOr Werkstoffe der Elektrotechnik
JebensstraBe 1, D-1000 Berlin 12, Tel.: ( .. 49) 30/314 - 22442
Federal Republic of Germany

Due to reflection losses the light generated current density in a GaAs solar cell can have a level of only 62.8 % of the available
amount for an AMO spectrum. With an antireflection coating (ARC) characterized by the number of layers, their refractive
indices, extinction coefficients and layer thicknesses the reflection of a solar cell can be reduced. Reflection and transmission
for thin layers are calculated with a matrix formulation. Based on data for the spectral response and literature data for the
spectrum AMO, the refractive indices and extinction coefficients, antireflection coatings are evaluated. Single, double and triple
layer ARCs without extinction are considered forthe choice of optimum materials and the limit current density. ARCs of MgF2,
Si02, Si3N4 and ZnS consisting of 1,2 or 3 layers are optimized. ARCs for GaAs solar cells including an Alo.aGao.2As window
layer are considered. The optimum Alo.aGao.2As layer thickness for a single layer ARC is zero. But for a Si02-ARC of 0.0947
11m it is 0.0266 11m. The reflection of an AIGaAs/GaAs solar cell has been measured and compared with calculations.

1. Introduction •. 2 1t dj
OJ = -",- AI.
IVJ phase thickness of layer j
Due to reflection losses the light generated current denSITY in a GaAs Nj=nj-ikj complex refractive index of layer j
solar cell can have a level of only 62.8 % of the availab!e amount for nj refractive index of layer j
an AMO spectrum. Using thin layers of transparent materials the kj extinction coefficient of layer j
transmission of the incident light into the solar cell can be increased
dj thickness of layer j
and the reflection decreased by destructive and constructive
wavelength of light
interference of the light waves. The reflection and transmission of
structures of thin layers can be calculated with a matrix formulation '"
for the transfer of the electric and magnetic field across the interface
between two layers. An antireflection coating consists of one or more For an assembly of thin layers the characteristic matrix of the system
layers on the solar cell: Every layer is characterized by its refractive has to be calculated by the matrix multiplication of the matrices of the
I layers.
index, extinction coefficient and layer thickness.

Antireflection coatings (ARC) are calculated for a given calculated (2)

spectral response of an idea! GaAs solar cell and the standard
spectrum AMO. In a first step ARCs consisting of 1, 2 or 3 layers of (3)
ideal materials, I.e. materials with no extinction, are evaluated to
determine limits of current densities and the value of best refractive
index and layer thickness. Then ARCs with layers of MgF2, Si02, The reflection and transmission is then given by equation 4 and 5 for
Si3N4 and ZnS are considered in 1, 2 or 3 layer structures and perpendicular incident light.

(~~: g)(~~: gJ
optimum layer thicknesses are determined. To reduce the surface
recombination veloctty GaAs solar cells are covered with a layer of
AlxGal-xAs. II is intended to have a highly transparent composition of
R = (4)
AlxGal-xAs, i.e. a high band gap or a large value of x. Therefore x is
in the order of 0.8 up to 0.9. Because Alo.sGao.2As is partly T = Re(Ni+l)
Re(No)
(~)(~J'
No B+ C No B+ C
(5)
transparent it can also be integrated in the ARC calculations.
No=fl:J-iko=fl:J (6)
The measurement of reflection gives evidence of the quality of the
ARCs. At higher wavelengths the GaAs solar cell becomes
For a given wavelength the optimum antireflection coating can be
transparent. Then the measured reflection can be attributed to the
calculated for an ideal material characterized by refractive index nj,
GaAslAu-interface on the backside.
extinction coefficient kj=O and the layer thickness dj. If the layer
thickness is a quarter wavelength, I.e.

2. The Calculation of Reflection and Transmission

di=~m
, 4 nj , m135
= " , ... , (7)

There are several ways to calculate the reflection and transmission of the condition forthe optimum refractive index assumes. an easy form.
structures of thin films 11, 21. Most appropriate for many layers and
the realization on a microcomputer (PC) is the matrix formulation 111, Single layer ARC
where the calculation of the transfer of the electric and magnetic field
is reduced to the solution of a matrix multiplication. (8)

Every layer is represented by the coefficients of a matrix Mj. These Double layer ARC
are calculated from the properties refractive index, extinction
coefficient and thickness. (9)

'~l
Triple layer ARC
cos 1 I Nj
Mj 8 (1 )
[ i Nj sin8j COSOj (10)

545
 ~

~
E S!02 AFlEA PAliK
~
u
2

z
g
AEfllACIIVE IIfJEX

~
"
l'l
..
~

.
>
....u EXTINCTION COEFfiCIENT
a:
i:la:
o.z 0.4 0 .6 0.8 1.0
W.VELENGTH I IJIII!
tfAVELENGlH I 10'-
Figure 1: Spectral irradiance of standard spectrum AMO. Figure 4: Refractive index and extinction coefficient for Si02.

§
60
"A' SOlAR CELL 11H COl&..E LAYER ARC
ITt<OtIl EXT! IIOH _
HgF2 I ZnS
- ..::.
~ SI3/<. AFTEA PALIK

8
..'"
" ~
2
REFRACT! VE INO£><

~..
z ' .0
0

~
....
~ Z.o
..
..
c
u ~
~
'" >

~
02 0 .4 0 .6
W.VELEnGTH I jJ_
o• 1.0 i 0.2 0 ,4 0 .6 O. B 1.0
WAVE1..£NGTtt I ~.

Figure 2: The spectral response without reflection, for an ideal Figure 5: Refractive index and extinction coefficient for SiaN4.
double layer ARC, the MgF2 I ZnS ARC and without
ARC.
.g
1<
InS ~TER PAlIK

~
z
~
!l
;:
1:j

i.
>
.... EXTlNCTlOH COEfflCIE
u
il! 0
i:l
02 0 • 06 0 .8 1.0 II: 0 .2 0." 0 .6 O.B 1.0
w. ELE GTH I 1.1 .... vELEltGlH I 10'
Figure 3: Refractive index and extinction coefficient of GaAs and Figure 6: Refractive index and extinction coefficient for ZnS.
Alo.8GaO.2As.
Considering thin layers the amplitude for the coherent reflection and The effect of the ARC is completely given by the transmission. To
transmission is important. With thick layers the light waves from the optimize the ARC for maximum current density the transmission has
front and back interface do not interfere, because of too small to be maximized. A computer program is used to calculate many
coherence length by definition. In this incoherent case the intensities different ARCs, i.e. a variation of refractive index for ideal ARC (no
have to be considered for the perpendicular reflection and extinction) and a variation of layer thickness, and to determine the
transmission 13/. values with the maximum resulting current dens~y.
Rj-l,j+ (Tj-l,j Tj,j-l-Rj-l,j RjJ-l) Tj,j+l e- 2rt.dj
R= (11 )
1 - Rj,j-l Rj.j+l e 2rt.c1i

T = Tj-l,j Tj,j+l e-adi

(12)
3. Data for the Calculation
1 - Rj,j-l Rj.j+l e 2adi
The calculations are based on literature data. The spectral irradiance
The indices indicate, in which order the layers j-l, j, j+ 1 have to be is given by the AMO standard spectrum (fig. 1) 14/. The spectral
taken into account. response has been calculated for an optimized GaAs solar cell
without consideration of reflection 151. Fig. 2 shows the spectral
The current density generated by a solar cell is calculated by response with and without reflection determining the limits of a solar
integration of the product of the spectral response of an ideal solar cell with ARC. The optical data of the materials, i.e. refractive index
cell without consideration of its reflection with the transmission and and extinction coefficient, are given by Palik 161 and for Alo.8Gao.2As
the spectral irradiance. ( S(A) spectral response, IA(A) spectral by Aspnes 171. Fig. 3 to 6 show the refractive index and extinction
irradiance). coefficient of GaAs, Alo.8Gao.2As, Si02, Si3N4 and ZnS. For MgF2 n
= 1 .38 and k = 0 have been taken 18/.
J =f 'A(").,) T(A) S(A) cJ). (13)
Spectrum

546
 5 .0
GUs SOLAR CELL WI1H SINGLE LAYER ARC
M.AXIMUH A1 0.0667 ~". 1 9!H: 34 . 1 -.\1,.II1II2
. G..... SOLAR cal -WI1H OOU8l.E LAYER ARC
"AXl""" olT 1 539 . 2.166:- 36 . " u,/c .... 2
"
~ ' .0 I. MAXJf4I..Ao4 _
~ 2 ... XltIUH __
x
"'
~
"'
E 3 0

....
~
a;

2.0

1.0 2.0 3 .0 ' .0 1.0 1.2 ,• , 6 , .0 20 2 .2

LAYER THICKNESS 1 ~. REFRACTIVE IN)(X LAYE~ 1

Figure 7: Refractive index and layer thickness contours of Figure 9: Refractive index contours of generated current density
generated current densny for a single layer ARC under for a double layer ARC under AMO irradiation.
AMO irradiation.
38
5.0 G..... SOL.A~ CELL ITt-! TRIPlE LAVER ARC
1. U.YEA .H;IF2 _._
G.II".I SOlAR C£Ll WITH OOUBl.E LA YEA ARC 2 . LAYER S'3N4 _
HAXlHU1.. 36 . 4 aA/' •• '2 3 LAYEI' InS

~
. LAYER 0.0862 "' .. S. !539
• 0 2 LAYER: 0 039'~" 2 . 766

~
::"' 3 .0
~ ~ "-
~
2 LAV[A
~
a;

~ 20
_ _ 36 ••'/'." 2
_ 3!5 aA/ean 2
32
,
'-'

10 L-_==_J..-_-'-__"-_......._.........J . 10~'
30
o 1.0 2.0 1 .0 • 0 5.0
1.0 2.0 3 0 " .0 5 0 6 .0 LAYER THICKNESS I fI-
lnE~ THIC)( SS I fla Figure 10: The generated current density dependent on thickness
Figure 8: Refractive index and layer thickness contours of for the 3 layers of the MgF2 I Si:JN4 I ZnS triple layer
generated current density for a double layer ARC under ARC under AMO irradiation.
AMO irradiation.
J I mAlcm2
ARC refractive index - layer thickness IIJ-m current d.1 38
laver 1 laver 2 laver 3 mAlcm2 37.6 CURRENT DENSITY WITHOUT REFLECTION

sinale 1.991/0.0667 34.1 37 37.1 TRIPLE LAYER ARC WITHOUT EXTINCTION

double 1.539 I 0 .0862 2.766/0.0394 36.4
triple 1.361 10.0896 2.200 I 0.0569 3.445 / 0.0254 37.1 36.4 DOUBLE LAYER ARC WITHOUT EXTINCTION
36

Table 1: Optimum layer thicknesses and current densities for 35.4 MgF2i SisN.v lnS
ideal single, double and triple layer ARCs under AMO 35 35.1 MgF2i lnS
irradiation.

4. The Determination of Ideal Antireflection Coatings 34 34.1 SI NGLE LAYER ARC WITHOUT EXTINCTION
33.9 S'JN4
When designing antireflection coatings informations on the choice of f 23.6 CURRENT DENSITY WITHOUT ARC
optimum materials and the limit of an ARC are wanted . Therefore Figure 11: Comparison of optimum antireflection coatings for GaAs
ideal ARCs, i.e. no extinction (kj=O) , are considered. Fig. 7 shows solar celis under AMO irradiation.
contours of limit current densities for a single layer ARC . By its
refractive index materials can be selected resulting in high current
densities. With the refractive index the range of optimum layer current density I
ARC material layer thickness IIJ-m
thicknesses are given. Similar calculations have been performed for mAlcm2
double layer antireflection coatings. Fig. 8and 9 show the contours of Si3N4 0.0656 33.9
optimum refractive index - layer thickness combinations for the upper ZnS 0.0550 32.2
single
(1) and lower (2) layer of the double layer antireflection coating. As a Si02 0.0932 31.3
result of the same calculations the contours of refractive index MqF2 0.0997 30.6
combinations of layer 1 and 2 are shown in fig. 10 for choice of
MqF2 I ZnS 0.0893/0.0480 35.1
optimum materials. For triple layer ARCs the calculations are very double
time consuming for the used microcomputer. In this case no contours SiC>.> I ZnS 0.0797 I 0 0. 450 34.9
are given, but the best ideal triple layer is listed with the others in triple MaF21 SisN4 I ZnS 0.077810.0309/0.0280 35.4
table 1.
Table 2: Optimum layer thicknesses and current densities for
single, double and triple layer ARCs under AMO
irradiation.
5. The Design of Antireflection Coatings

There are several materials available for the use as antireflection the optimum layer thickness and the generated current density for the
coating 19,10,11/, but the number of materials, which have the spectrum AMO are calculated (s. table 2). In fig. 10 the generated
optimum refractive index and are transparent in the spectral region current density dependent on thickness is shown for the 3 layers of a
from 0.1 to 0.9 IJ-m, is very limited. MgF2 (n=I .38 at 0.8 IJ-m), Si02 triple layer ARC. The highest current density can be achieved with a
(1.45) , Si3N4 (2.01) and ZnS (2.31) are considered with literature triple layer ARC. See fig. 11 for comparison. The increase in current
data including extinction. For a single. double and triple layer ARC density with a double (triple) compared to a single (double) layer ARC

547
 35 5 .0

.
N
au ..SOLAR CELL COATED WITH HEASUAElENT
S13H4 O.O~ "'. MIl A)G, .. br- 0 . 81 5A1f'lE 88.
5102 0,0947 V. AIC) A1G...s 'x- O. BI _.0 CA!.C\I.. ... T1ON
Al l>o ... (x- O . 8) GUa WJTH
~
; 30 S\JN4 (0 . 07"1 .... OPH .....'
A1G ••• (Jr-O 8) 10 t 11"1
.... 3 .0
~
i ~.O
~
,,
I
25 I
~

1.0
--- ....
20
....

-
- - - - ------
11110-'
1.0 ~.O 3 .0 _ .0 5 .0 _.0 5 .0 6 .0 7 .0 8 .0 9.0
LAYER THICKNESS OF "'l~A. tx- O.el I~. WAVELENGTH I p.

Figure 12: The generated current density dependent on thickness Figure 13: Measured and calculated reflection for GaAs solar cell
of the AIo.sGao.2As layer and the single layer ARC of 881 .
SiD:! and Si3N4 under. AMO irradiation.

8. Conclusion
thickness of current
ARC material layer thickness I Antireflection coatings for AIGaAs/GaAs solar cells consisting of 1, 2
AIo.sGao.2As I density I
um um mAlcm2 or 3 layers have been calculated for ideal and real conditions.
0.10 SiaN4 0.0741 29.8 Ranges for optimum ideal ARC are determined. Real ARCs are
single examined and the triple layer ARC consisting of 0.0778 11m MgF2,
0.03 Si02 0.0947 31 .8
0.0309 11m SiaN4 and 0.0280 11m ZnS is found to be the best for a
0.10 Si02 0.1027 28.5 GaAs solar cell. When an AIo.sGao.2As layer is taken into account the
double 0.10 MaF2 / ZnS 0 .1040 I 0 0
. 601 31 .3 optimum layer thickness wnh an ARC is zero. Only with a Si02-ARC
t~le 0.10 MgF2ISi3N4/ZnS 0.0969/0.0206/0 .0484 31 .3 an optimum AIo.eGao.2As layer thickness exists. The calculation of
measured reflection of AIGaAs/GaAs solar cells is possible
qualitatively. To explain the measurement at wavelengths over .0.9
Table 3: Calculated layer thicknesses for optimum ARCs on 11m the incoherent reflection from the backside has to be taken mto
Alo.sGao.2As coated GaAs solar cells under AMO account.
irradiation.

is at least 3.5 (0.9) %. The gain is small for a triple against a double 9. Acknowledgement
layer ARC. This indicates that n might not be appropriate for
economic reasons to realize such a coating. The provision of AIGaAs/GaAs solar cells by Prof. Dr. Wettllng,
Fraunhofer-Institut fOr Solare Energiesysteme, Freiburg i. Br., and by
Mitsubishi International is gratefully acknowledged.

6. Antireflection Coatings Including the AlxGal -xAs Window Layer

To reduce surface recombination GaAs solar cells are coated with a 10. References
layer of AlxGal-xAs. Lnerature data for AIo.sGao.2As 171 are
considered in the calculations. Because of material instabilny data for 111 H. A. Macleod, Thin-film optical filters, Adam Hilger Ltd, Bristol,
x>0.8 are not available. Extrapolations from x = 0.7 and 0.8 to 0.9, 1986, p. 32 - 43
which is the desired experimental value, do not show signnicantly
different resu~s. Without ARC the calculations show an optimum 121 11/, p. 50 - 52
layer thickness for Alo.sGao.2As of 0.0327 11m (s. fig. 12). For its
extinction an optimum layer thickness of Alo.sGao.2As other than zero 131 11/, p. 67 - 69
does not exist for an ideal single layer ARC . But for a Si02-ARC
(0.0947 11m) the optimum AIo.sGao.2As layer thickness is 0.0266 11m, 141 M. P. Thekaekara, Applied Optics, 13, 1974, 518
which is caused by the inferior match of refractive indices (1 .46 to
1.991 desired) . In 1121 a s m
i ilar result is shown. Also for SiaN4 the 151 M. E. Nell, P. Christ, B. Reinicke, H. G. Wagemann ,
optimum AIo.sGao.2As layer thickness is zero. The results in table 3
Proceedings olthe 6th International Solar Forum 1988, p. 487
concern an Alo.sGao.2As layer of 0.1 11m as a technological limit for
our LPE.
161 E. D. Palik, Handbook of Optical Constants of Solids,
Academic Press, Inc., New York, 1985

171 D. E. Aspnes, S. M. Kelso, R. A. Logan, R. Bhat, Journal of

7. The Measurement of the Reflection of Antireflection Coatings
Applied Physics, 60,1986,754
and Comparison wnh Calculation
18/ 11/, p. 506
The reflection of manufactured AIGaAs/GaAs solar cells has been
measured. It is possible to calculate the measured data qualitatively.
191 11/, p. 389 - 398, p. 504 - 511
In fig. 13 it is shown that the reflection of the solar cell can be further
reduced if the SiaN4-ARC has the assumed properties from the
110/ P. Baumeister, "Interference, and Optical Interference
Inerature. The high reflection at wavelengths over 0.9 11m can be
Coating", in Applied Optics and Optical Engineering, ed. R.
attributed to the incoherent treatment of the reflection from the gold
Kingslake, VoL I, Academic Press, New York, 1965, p.285
covered backside of the GaAs substrate, respectively. The contact
grid at the front of the solar cell has been included in the
111/ G. E. Jellison, R. F. Wood, Solar Cells, 18, 1986,93
measurements, but not in the calculations. Considering appropriate
reflection losses by the grid the reproduction of experimental data
1121 A. Yoshikawa, H. Kasai , Journal of Applied Physics, 52, 1981,
might be improved.
4345

548
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Comparison of an Analytical and a Finite Differences Model Referring to GaAs Solar Cell Performance

H. Eschrich, J. Bruns, M.E. Nell, B. Reinicke, H.G. Wagemann

Technische Universitat Berlin
Institut fUr Werkstoffe der Elektrotechnik
Jebensstr. 1, 0-1000 Berlin 12, Germany
Tel.: (030) 314-22442

A comparison of a one-dimensional finite differences model and an analytical model is carried out considering the same
solar cell structure. Both models are based on current, continuity equations and Poisson's equation, on SRH recombination
statistics and on spectral absorption by Lambert's law. The different results are explained by the assumptions and
approximations, which are required for the analytical calculation. The changes of solar cell performance are discussed, if a
diffused emitter profile is considered instead of the abrupt junction. The maximum tolerable emitter surface recombination
velocity in that case increases from Sn '" 104 cm/s up to Sn '" 105 cm/s.

The prediction of reliable values for device geometry and 1.1 ANALYTICAL MODEL
doping concentrations is essential for solar cell
preparation. These parameters can be evaluated by The analytical model describes weak injection in the
model calculations based on the current and continuity considered solar cell. The transport equation is solved for
equations for electrons and holes and Poisson's equation. the minority charge carriers at the edges of the SCR
Analytical expressions for the current voltage neglecting the field current
characteristic and spectral response can be found
assuming simplified conditions. The optimum structure 2n(x)
can be found by parameter variation. However, an exact od 2 - Vex) + G (x) = 0 ,
dx
calculation for a real device can be done in a numerical
way only. This is not useful for the purpose of
In emitter and base a linear recombination law is the
optimization, because of the large amount of computing
consequence of SRH-theory
time required. The intention of this work is to compare the
results of an analytical model and an exact numerical
evaluation applied to an optimized pn-GaAs solar cell. v (X) = n (x) - no(x)or V (X) P (x) - P o(X)
·n "P
Emitter, SCR and base properties are considered to be
homogeneous. The maximum generation recombination
current in the SCR is described by an' approximation after
1. MODEL DESCRIPTION Sah et al. /1/, which is also derived from SRH-theory

The common base of the analytical and the numerical

model are the one-dimensional basic equations for
steady-state behavior irg
2 q mWSCR sinh
{"'n'p.!L
Nr 11:

2 .
kT (Vo - V)
dn
q fin n E + q On dx '
Detailed specification of current contributions refers to /2/.
q flP P E -
!!.£
q Op dx '
1 din 1.2 NUMERICAL MODEL
+ - - - V(X) + G(X) = 0,
q dx
For numerical calculations the method of finite differences
-1.~df -V(X)+G(X)=O,
q x is used as published in /3/. The basic equations as given
q
+ - before are solved simultaneously. Electron and hole
-dE
dx = -£O£r (No - NA +P- n). currents, recombination rate and generation rate are
calculated starting from an assumed balance distribution
Shockley/Read/Hall statistics assuming midgap states are of electrons and holes across the device. The minority
used for the description of the recombination rate carrier concentrations and currents at the front and the
back surface are determined by the surface
pn - nr recombination velocities (SRV) Sn and sp.
v (X) =
'n (P + m) + 'p ( n + m)
in (x=O) = q Sn'( n (x=O ) - no ) and
The rate of optically generated charge carriers follows the
absorption described by Lambert's law in the window Dn
LG' n ( ~ X ) + Sn no
layer, the emitter, the space-charge region (SCR) and the n(x=O)
base On
~x + Sn
G (X) = fo"" E). a GaAs().». e -aGaAs().) x - aA1o.sGao.2As(,i) fWd), .
he With these values a new distribution of charge carriers

549
 :0' ,..--.--------- ------------,
E 10~
Ala .Gao 2A$'WlI'ldOw layer I~ - l00nm
,<'U ~o"
l 810nm

..
"I E
o 3200nm u
"-
:0'
E 10 2
\ "-
>-
10'
So - 10' Cmis \ o 822 10" cm·3 f-
lO' erIc-prof I le.
'"z
w :0-' numerical

-\
Q
t:c lO-'
f-
Z
L.J 10- 3
0::
NA = e 25 10" em" \ Sp 0
0::
~
u
10-' lon,
10-'
L Ev
10-'
Emil1e, Base
r,r,--~~~,~~-----------------" • 0.3 0.5 0.7 0 9 1.1 1.3
~O ~ 0
VOL TAGE / '
Fig. 1: Energy band diagram of the considered solar cell Fig. 2: Dark IV-characteristics of the same GaAs solar cell structure
calculated with different models
can be found by solution of the continuity equations and n
Poisson's equation. This new distribution is the starting '", 10"
point of the next iteration. The calculation is finished, if the u V - l.2V
relative change in the carrier concentrations and the "-
potential is less than a limit value eRE anywhere in the z 10"

.
aM
device. f-
0.6V
....
0::
z 10'
w
u
1.3 SOLAR CELL STRUCTURE WITH INPUT PATA z
a
u
0:: 10' O.• V
The considered solar cell structure (see Fig. 1) consists of w
<r0::
an absorbing Alo.aGao.2As window layer with a thickness
'"
u o.oJ
of 100 nm, a junction depth of 810 nm and the doping 10-' -,,--,,~_-..l........ ............... __ ...L
concentrations NA= 8.25 1017 cm-3 in the emitter and
0 1000 2000 3000
No = 8.22 10 17 cm-3 in the base. They are constant for
x / nm
the calculation of the abrupt junction, respectively the
Fig. 3: Charge carrier distribution In the non-irradiated eric-profiled so-
surface concentrations for the complementary error lar cell at different voltages
function profiled emitter. The front surface passivation is
specified by a surface recombination velocity 10'
Sn = 104cm/s. The back surface is assumed to be ideal E v - O . OV
u O.SV
"-
(sp= 0 cm/s). Input data for diffusion lengths and charge >
carrier mobilities were taken from /4/, values for the "- l.OV
:I: 10'
extinction coefficients of GaAs and Alo.aGao.2As from /5/ f-
t!)
Z
and /6/. For optical generation the measured standard UJ L2V
0:
spectrum AMO /7/ is taken into account. ....
'"
Q 10'
...J
W
H
.....
2. DIOPE CHARACTERISTICS ...J
u'" 10'
The plots of the dark IV-characteristics for different a:
....
u
models and profiles are presented in Fig. 2. The U
...J
saturation current values of diffusion and recombination w
10'
current were extracted from these. The input values of the 0 1000 2000 3000
analytical calculation could be reproduced with 1.8%, • / nm
respectively 2.8% deviation. Analytical and numerical
calculations exhibit nearly the same results in the voltage Fig. 4: Field distribution in the eric-profiled solar cell at different volta-
range above 1.1V, whereas the analytical current exceeds ges
the numerical one by a factor of 1.8 at lower voltages. The
latter can be referred to the maximum estimate of
Model used lidO I mNcm 2 'raO / mNcm 2
recombination current in the Sah/Noyce/Shockley model.
Analytical:
The calculation using an error function profiled emitter,
input value 3.85 10-17 1.01 10-8
which is the most realistical profile for diffused junctions,
leads to an increase in current at any Voltage. The extraction 3.92 10-17 9.8210-9
explanation for this phenomenon is an increased Numerical:
recombination rate, because of the widening SCR at abrupt junction 3.6810- 17 5.16 10-9
voltages less than 1V and an increased diffusion erfc-profile 8.0610- 17 3.14 10-ts
saturation current above this limit. The exact values for the Tab. 1: Diffusion saturation current jdO and recombination saturation
extracted saturation currents are presented in Tab. 1. The current I,ao extracted from simulated IV-characteristics

550
 10'·
beginning of strong injection can be seen in the less
steepening IV-characteristic of the error function profiled '"
I
'e
P n
u 10'"
emitter at higher currents.
"-
The charge carrier concentration profiles across the
z
device are shown in Fig. 3. The exponential increase in ~ 10"
minority carrier concentration as a function of voltage I-

leads to the exponential diffusion current slope.

'"
a:
I-
z
IU
The distribution of the potential across the erfc-profiled u 10"
z
device reduces the field peak at the pn-junction by an u
0

order of magnitude compared to the abrupt junction. On a: I

~ 10'·
the other hand in emitter and base the field increases a:
a:
from approximately zero to significant values, as can be 0.8, 0.9, '.0 1/
u""
10·
10" V • 1. 3Y
0; 0 1000 2000 3000
I. I V xI n
'"
'e

~
.!! Fig. 8: Charge carrier distribution In the irradiated eric-profiled solar
10"
~ 0 . 9Y cell at different voltages
10'·
____ "
"-
w 0.7V
I- 10'~ Iii • 1.3V
""

~
a:
z ~ 10"
0 0.5V E
'"' ~
I-

z"" 10" ::::

....
til
O.3V "- 10~ I. I V
:£
0
U
~
w
a: 10~ O. IV ""
a:
Z 10'·
0.9V
8
0 1000 2000 I-
«
• I nm ~ 0 IV. o 7V
m 10"
Fig. 5: Distribution 01 the recombination rate in the non-irradiated erlc- 0
:>:
profiled solar cell at different voltages u
UJ
a:
10 "
o a 1000 2000
• I nm
,N erlc-profl1e. nurr.er ic al "
Fig. 9: Distribution of the recombination rate In the irradiated erlc·profl-
e
~ -10
led solar cell structure at different voltages
""E
"- Model used Jphoto / Voc/ V ,,/%
>-
::; mAlcm 2
~ -20 Analvtical: 31.34 1.051 20.66
UJ
o
...z abrupt junctlD~
- analytical"
Numerical:
abruot 31.79 1.056 21.36
w
a:
~ -30 - numerical \. " erfc 32.53 1.003 19.97
u
Tab. 2: Perlormance of the same solar cell structure calculated by
different models

o 0 .2 0.4 0.6 o.e 1.0

seen in Fig. 4. Fig. 5 shows the distribution of the
VOLTAGE I V
recombination rate increasing exponentially due to the
Fig. 6: IV·characterlstics 01 the same GaAs solar cell structure under voltage. The peak of the recombination rate in the
AMO Illumination calculated wHh different models space-charge region increases by approximately 10
orders of magnitude between V = OV and V = 1.2V,
whereas the level of recombination in emitter and base
6.0 increases by 1020 . This is due to the predominant
recombination current in the lower voltage range
expressed by a slope factor of two within the exponent
w and the predominant diffusion current at higher voltages
~ 4.0
o
0..
with a slope factor of one.
III
W
a:
...J

...u
~ 2.0
3. GENERATOR CHARACTERISTICS
W
0..
III

:n Fig. 6 the generator characteristics of the considered

o models are presented. The photo current of an abrupt
300 500 700 900 junction increases, if calculated numerically due to the
WAVELENGTH I nm small electrical field in emitter and base. An error function
Fig. 7: Spectral response of the same GaAs solar cell structure under profile leads to a further increase, for the gradient of the
AMO illumination calculated with different models doping profile provides better carrier collection, especially

551
 40
N
the neutral regions leads to worse charge carrier
<
e collection of the illuminated device. The maximum
.....u estimate of the recombination current yields larger
""
....."
30 currents in the lower voltage range. The differences in the

-
>- predicted electrical parameters are in the order of some
I- aDrupt junctlon. percents. Therefore analytical optimization should only be
( J)
n u"'.r ! c~1
:z
w 20 used for an evaluation of the optimum structure. The
a
I-
exact determination must be done numerically. If a
:z diffused erfc-profile is used, the surface passivation
w
a:
a:
=> 10 becomes less significant, because the electrical field in the
u abrupt junctIon. emitter prevents the minority carriers from diffusion to the
a analytical
I-
0 surface.
:r
a.
0

10' 10' 10· 10· 10'·

SURFACE RECOMB I AT ON VELOCITV / cm/s
Fig. 10: Photo current versus emitter surface recombination velocity 6. LITERATURE
calculated with different models for the same GaAs solar ceil structure

in the base. This can be seen from spectral response data /1/ C.T. Sah, R.N. Noyce, W. Shockley, Proc. IRE
in Fig. 7. The photo current rises mainly in the "red" range 45, (1957), p. 1228
of the spectrum. Another consequence of the built-in field /2/ H.J. Hovel, Semiconductors and Semimetais, 11:
is a steeper current increase with increasing voltage Solar Cells, Academic Press, New York,
analogous to the dark IV-characteristics, leading to worse (1978), p. 15
open-circuit voltage and efficiency for the considered /3/ M. Kurata, Numerical Analysis for Semiconductor
device (Tab. 2). Devices, Lexington Books, Massachusetts,
The variation of charge carrier concentration (1982)
corresponding to the voltage is shown in Fig. 8. There is /4/ H.C. Casey, M.B. Panish, Heterostructure
little change between V = OV and V = 0.8V, conclusively Lasers, Part B, Academic Press, New York,
the current is approximately constant in that range. The (1978), p. 170
minority carrier concentration gradient changes its sign at /5/ E.D. Palik, Handbook of Optical Constants of
higher voltages, indicating the transition from generator to Solids, Academic Press, New York, (1985), p.
diode characteristics. The recombination rate across the 429
emitter and the base of the irradiated solar cell (Fig. 9) is /6/ D.E. Aspnes, S.M. Kelso, R.A. Logan, R. Bhat, J.
approximately independent of the voltage up to 0.8V. The Appl. Phys. 60, (1986), p. 754
recombination of the generated charge carriers /7/ M.P. Thekaekara, Appl. Opt. 13, (1974), p. 518
determines the photo current. The recombination rate
nearly follows the characteristic of the non-irradiated
device at higher voltages.

4. SURFACE PASSIVATION

The quality of surface paSSivation described by the front

surface recombination velocity Sn is essential for the
device performance. In Fig. 10 one can see the
dependence of the photo current on Sn calculated with the
different models. For analytical and numerical
conSiderations of an abrupt junction the front surface
should be passivated better than Sn = 104 cm/s.
Numerical results show less losses for recombination
velocity values above this limit. The electrical field in the
emitter prevents the electrons from diffusion to the surface
and recombination via surface recombination centers. The
steeper potential gradient in an erfc-profiled emitter
amplifies this effect and rises the maximum tolerable value
of the SRV for high efficiency up to Sn = 105 cm/s, which
is within reach by epitaxial growth of the AIGaAs-window
layer.

5. CONCLUSIONS

The differences in the results of numerical and analytical

calculations of the same solar cell structure can be
explained by the assumptions and approximations made
in the analytical model. The neglected field component in

552
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOVOLTAIC ACTIVITY IN TIDN POLYCRYSTAlLINE GaAs FILMS

PRODUCED BY ION-ASSISTED DEPOSITION

D. Bonnet, S. Oelting, U.Luke

Battelle Institut e.V., D-6000 Frankfurt 90

ABSTRACT. Ion-assisted deposition has been applied to deposition of polycrystalline GaAs-films for solar cells using a conventional high
vacuum system. Intrinsic film quality approaching MOCVD-films regarding electronic properties and photovoltaic effect has been achieved.
EpUayers n-doped to 1017 cm·3 show mobilities of 2600 cm%Ns. Schottky-diodes generated on such films show photocurrents of about
15 mA/cm%. Polycrystalline films show high preferential orientation; polycrystalline Schottky-diodes show short-circuit currents of around
7 mA/cm% and give potential for higher currents.

1 Introduction increased surface flatness and polycrystalline films on neutral

substrates show increased preferential orientation in the 111-
In recent years the use of ions and ionized clusters has direction.
lead to a significant progress in thin film technology for The electronic properties, however, are not satisfactory
dedicated applications: due to radiation damage caused by high energetic ions. Epitaxial
layers show defect densities in the 10 17 cm·3 range. Doping levels
Use of Si-ions in Si MBE facilitates the incorporation Significantly below 1018 cm-3 by coevaporation of Sn are difficult
of Sb dopants into growing films, allowing doping to achieve. Nevertheless, Schottky barriers fabricated on such
superstructures to be made /1/. epitaxial films show promising photoelectric properties.

Use of accelerated ions and ionized clusters allows

single phase Rutile Ti0 2 and Zr0 2 films with preferen- 2 Film-Preparation
tial orientation to be made, which can be applied in
oxygen sensors /2,3/ Epitaxial and polycrystalline Films are deposited in
parallel by using GaAs-wafers and Mo-pre-coated 7059 glass
Optical coatings made by use of energetic ions show substrates. Substrate temperatures presently are around 550·C.
reduced optical scattering due to increased density /4,5/. Ga was first evaporated from the Mitsubishi ICB source
(which mainly ejects Ga-ions) and accelerated to energies of
TiN and other films made of refractory materials show several hundred eV. As is evaporated from a graphite crucible,
increased mechanical properties, such as hardness under bearing a catalytic cracker (using hot Ta-wire&). N-doping is
ion-assisted deposition. being achieved by coevaporating Sn from a third source.
Since the defect density of layers fabricated with the ICB
In all cases kinetic energies of 100 to 5000 eV are being source was rather high (due to radiation damage), doping levels
used. In the case of semiconductors, such as Si and GaAs below 1018 cm·a were difficult to achieve. Therefore the Ga-
energies have to be limited to the threshold for defect genera- source has been replaced, Ga is now thermally evaporated from
tion, usually lying in the region of 50 to a few 100 eV. Recently a BN crucible. Defect density decreased significantly, doping
plasma-assisted epitaxial growth of GaAs has been achieved and levels between 1016 cm-3 and 1019 cm-3 can now be achieved.
advantages have been attributed to ionized species. Electron densities of around 10 17 cm-3 are ideally suited for a
space charge region adapted to solar light absorption.
Two distinct effects of the ionized species have to be
taken into account:
3 Results
The input of increased kinetic energy leads to activated
films growth, allowing lower temperatures to be used in 3.1 Epitaxial. films
view of commercially interesting substrates. Films deposited onto GaAs wafers without using ions
show increased room-temperature mobilities of up to \2600
The low energy ion-bombardment of the growing cm%Ns, which favorably compares to the ideal value of around
surface re-sputters physisorbed impurity species con- 4000 cm%Ns. This has become possible by avoiding defects due
densing on the growing surface from the ambient, to radiation damage and systematically reducing carbon sources,
allowing a cost-efficient lower grade vacuum to be used, leading to a carbon-content below values detectable by SIMS.
compared with the classical technique of molecular The oxygen-content, as measured by SIMS-depth-profiling has
beam epitaxy. also been reduced to that of the substrate-wafers, only an
increased value persisting at the substrate/epilayer boundary.
Using the commercial Mitsubishi ICB (Ionized Cluster
Beam) evaporation source for Ga while coevaporating As, it has
been shown, that epitaxial /6,7/ and polycrystalline /8/ films of
improved structural properties can be made: Epilayers show

553
 I I mA/c:rn' I
3.2 PolyaystaJ1inc films
Polycrystalline films deposited onto 7059 glass coated
with 1 !Lm of Mo do not allow mobilities to be measured
directly. It is assumed, that the intrinsic quality of the material
is comparable to that of epilayers. They show strong preferential
orientation, as demonstrated in Fig. 1, especially if produced

,. .
under the influence of ions of around 300 eV kinetic energy.

.,It •• ",., •• , •• U [V[

: ,.
: .
E•
~
~

u
.
r:•
to'
Figure 3: I-U Curves of Schottky-Diodes on epitaxial n-
GaAs under AM1.5-111umination. Upper Curve:
Epilayer made by Ion-Assisted Deposition.
I
Lower Curve: MOCVD Reference Layer. Cur-
" rent corrected for Au-Reflection.
." Similarly, Schottky diodes have been produced on
polycrystalline films. If these films are below 4 !Lm in thickness,
the diodes suffer from strong shunting, as to be expected, as
grain boundaries tend to be highly p-doped. This p-doping can
Figure 1: X-Ray Diffraction Spectrum of polycrystalline be decreased by the n-doping used in our films. So it has been
GaAs deposited on Mo-coated Glass. possible to get some Schottky diodes with lower shunting on
somewhat thicker films (;<: 5 J.Lm). Fig 4 shows a result, indicat-
Figure 2 shows a SEM picture of such a film, illustrating the ing already a promising photocurrent of 7 mA/cm1, in spite of the
morphology of the films, characterized by a significant scatter in still present shunting. Standard grain-boundary passivation-
grain size. techniques will be applied in the near future for improving grain
boundary behavior.

U [VI
05

10

Figure 4: I-U Curves of an Au-Schottky-Diode on poly-

crystalline n-GaAs in the Dark and under
Figure 2: SEM Photograph of a polycrystalline GaAs AM1.5-111umination. Current corrected for Au-
Film on Glass Reflection.

3.3 Diodes 4 lon-Assisted Technology for GaAs Deposition

Schottky barriers as an cvaluation tool for the quality of
the films have been produced on epitaxial and polycrystalline Ionized cluster beam evaporation sources commercially
films produced in parallel. Figure 3 shows the I-V-curves of fabricated by Mitsubishi still suffer from a low overall ionization
diodes on epitaxial films, on being the film produced as de- effiCiency, high energy requirement and thus heating of the
scribed above, the other film has been produced by standard components inside the vacuum system. Commercial alternatives
MOCVD at Chemetall GmbH. Our films - having been deposit- are presently not available.
ed in a standard vacuum system - therefore do compare not High energies of the ions have to be avoided, but a
unfavorably with state-of-the-art films. certain amount of low-energy-ions will improve morphology

554
without changing electronic properties for the worse. Equipment NOTE: This work is supported by the German Federal Ministry
allowing ionized species to be efficiently generated is in develop- for Research and Technology, co-supported by
ment at Battelle /9/. This new ionization system uses focussed CHEMETALL GmbH and the CEC, and performed in
low energy electron beams and leads to exceptionally high cooperation with Chemetall GmbH (H. Jager). Thanks
ionization efficiencies. Additionall1y it allows arbitrary accelera- are due to KFA Jiilich-ISI (H. Wagner, G. Herion, R
tion energies. Therefore it is suitable for all semiconductor Carius) for SIMS- and PL-Measurements and University
materials, which can be evaporated from crucibles or by electron of Stuttgart ( W.H. Bloss, H.W. Schock, H. Dittrich) for
beam heating, such as Si and GaAs. X-Ray analyses.
The ionization device has been designed for and
installed on the As source. Deposition of GaAs layers using
As + -ions will be optimized.

5 Conclusions

It has been demonstrated, that polycrystalline GaAs

films can be made by ion-assisted deposition in a standard high
vacuum on neutral substrates showing high quality, allowing
promising photocurrents to be measured.
In the near future a significantly improved ionizer will
be used together with post-grain-boundary passivation for
optimizing the photoelectronic properties of polycrystalline
GaAs films.

References

/1/ Jorke, H., Herzog, H.-J., Kibbel, H.: Secondary implan-

tation of Sb into Si molecular beam epitaxy layers.
Appl. Phys. Lett. 47 (1985) 511 - 513

/2/ Fischer, D., Koppen, H.: Comparison of Some Proper-

ties of Ti0 2 Films Prepared by Sputtering and ICB.
Proc. 12th Symp. on ISIAT (1989) 309 - 316

/3/ Kao, A.S., Gorman, G.L.: Modification of Zirconia Film

Properties by Low-Energy Ion Bombardment during
Reactive lon-Beam Deposition. J. Appl. Phys, 67 (1990)
3826 - 3834

/4/ McNeil, J.R, Barron, A.c., Wilson, S.R, Herrmannjr,

W.C.:Ion-Assisted Deposition of Optical Films: Low
Energy vs. High Energy Bombardment. Appl. Optics 23
(1984) 552 - 559

/5/ Martin, P.J.: Review: Ion-Based Methods for Optical

Thin Film Deposition. J. Materials Science 21 (1986) 1 -
25

/6/ Haga, Y., Takaoka, G., Tsuji, H., Ishikawa, J.: Funda-
mental Properties of GaAs Epitaxial Films by the ICB
Technique. Proc. 12th Symp. on ISIAT (1989) 327 - 334

/7/ Bonnet, D., Oelting, S.: Ionized Cluster Beam Deposi-

tion of GaAs Films for Solar Cells. Proc. 9th Photovol-
taic Solar Energy Conf. 1989, 108 - 110

/8/ Bonnet, D., Oelting, S.: Ionized Cluster Beam Deposi-

tion of GaAs Thin Films for Solar Cells. Proc. 21st
Photovoltaic Specialists Conf. 1990, S25 - S28

/9/ Fischer, D.: Proprietary Information. Details will be

published separately after patent-protection has been
achieved.

555
1afH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

GRADED Ga1.)n» WINDOW ON GaAs SOLAR CELLS

GROWN BY ISOTHERMAL LIQUID PHASE EPITAXY

J.M.Olchowik*, H. Soumana, M.Gavand and L.Mayet

Laboratoire de Physique de la Matiere ( U.A.358 CNRS )

Institut National des Sciences Appliquees
20 Avenue Albert Einstein 69621 VILLEURBANNE Cedex (FRANCE)
*Permanent address: Technical University of Lublin, Department of Physics,
20·618 Lublin,ul.Nadbystrzycka 38a (POLAND)

ABSTRACT.
Isothermal contact between multi component liql.lid and a binary
substrate leads to the apparition of a transition layer. ihe variation of
composition and then of energy gap at the interface allows an inmrovement of
the photo-carriers collect at tlie p-n junction and increases the efficiency. The
good results of Gal_xAlxAs I GaAs structures with Gal_xAlx.As window grown by
ILPE allow to consider the possibility to obt~in alteI¥atIve structures .1~y thiS
method, for inst~ce Gal-xInJ:P I GaAs w~t~.a wmgow less sensItive to
oxydization. In thiS work we show the POSSibIlIty of Isothermal growth of
GaO.51In0.49P on GaAs substrates from Ga-In-P liquid phase.

INTRODUC'TION
Optical . windows enabling to transft;r the where:
maximal portIOn of sol!lr energy t~ the. regIOn of F F
spatial charge of a semIconductmg JunctIon play. a Tij and A Sij :
great role in mqreasing ~e efficiency of photoV01t!lIC are the meltmg t~mperature and entropy of the
devices. The .V{mdows WIth graded band gap p'rovIde compounds, respectlVely.
optimal condItions for such a transfer (0. An Idea of
applying so calle'd transition layers (2) is one of Gm ex andGel m:
feasible and cheap ways to obtain graded optical are excess free mixing energy and elastic energy and
windows. These layers appear spontaneousl; on a n the moles number of IJ-components in solid
surface of solid matter m the presence 0 other solution.
heterogenic phases (3) and they result from a 1 1 sl sl s*
tendency of the sy'stem to approach a ai ,aj ,ai ,aj ,aU :
thermodynamical equilibrium through mmimizing its
energy. are the interatomIc interaction parameters in liquid
(1), liquid stochiometric (s, 1) and solid (s*) phases.
THEORETICAL CONSIDERAnONS s
y ...
The contact 0'£ a multicomponent liquid ph!lse ( i;Jthe activity coefficient of binary components in the
includmg also the Isothermal contact ) With a bmary interface.
substrate forces a substitution of atoms into the
cristalline lattice of the interface (as a result of the Assuming complete homogenization of the
difference in the chemical potential of atoms in the solution during the isothermar contact of two
neighbouring phases). A Clisturbance of atomi~al heterogenic phases, the state of both the solution and
locations in the interface introduces a deformatIon the interface can be evaluated during heteroepitaxy.
potential to a heterogenic system (4) with a change in It is especially important in applying the so called
conditions of a lIquid mterface. The so called "substrate-source method" to the solution saturation
coherent mechanism of phase c~>njugation (5) t.*es leading to numerous nonequilibrium phenomena (8).
place in the case of the small thIcKness of tranSItion Good results obtained on the MvGal_xAs/GaAs
layers (lower than the critical one (6». Assuming the structures with graded optical win(fows performed
approxImation of p'hase coherence the equations b2' means of the isothermal liquid phase epitaxy
describing quasi-equilibrium of the system consisting (ILPE) show the faisability. to use this metliod to
of "multIcomponent solution-binary substrate" take produce the alternative windows having unique
the following form (7): physical properties.
1 1 The present paper reports the attempt to use the
1 F F ai a ' ~ ILPE method to groduce the relatively oxidation
- (MI" - TAS·,) + In_ _J_ = In aiJ' resistant, graded GaxInl-xP window on the GaAs
RT IJ Iy sl sl substrate (9).
ai aj
EXPERIMENTAL
s a( Gm ex _ Gm el ) Technological process was carried out on the
RT In Yij = an .. apparatus for epitaxy from liquid phase ( the
IJ horizontal sliding-boat technique) in the atmosphere
of hydrogen punfied by the Pd-filter. The synthesis
s s* of tlie wmdows was performed at T = 1073 K and
'Yij = f «(Xij ,x, y ) consisted of three successive stages (Fig: 1)

556
 Ga-In-P: Be
CONCLUSION

I
~
GaAs GaAs GaP
The presented pattern of the technological
process and the type of the substrate orientation are
Important factors m this paper. It is known that the
limits of coherence of a heterostructure for (111)
orientation are larger than for (100) direction (10).
It stops the relaxatIOn phenomena at the first stage of
II GaAs GaAs GaP the synthesis. The isothermal supersaturation oT the
solutIOn stimulated by the elastic deformation of the
III GaAs GaAs GaP interfaces at this stage has the nature as shown in
Fig.3.

Figure 1: The scheme of synthesis of the GaxIn 1-x P

wmdow.
6 T(K)

At the first stage, the Ga-In-P solution was To= 1073 K

saturated by the undQPed GaP (ll1)A substrate. At 20
the ~econd ~tage the Ga-lt}-P solution was brought in
5 mm. lastmg contact With the auxiliaryl unaoped
GaAs (111) substrate. The GaxInl_xP winaow on the
n-~ype GaAs (P l)A substrate was formed at the
third stage which lasted 4 min.. The preliminary 10-
doping of the solution with Be facilitated the
fo~atlo~ of the p-n junction in the GaAs substrate
by diffuSIOn. OL-__ ~ ____ ~ ____ ~ ____ ~ __ ~~

RESULTS o 0.2 0.4 0.6 0.8

mol. fraction x
The presented above pattern of synthesis of the
op.ti.ca~ Ga"Inl_ P .wind~w on GaAs (l1l)A
mmlmIzed me effectIVe thickness of the transition Figure 3: The isothermal supersaturation of the
layer. and also rpinimized the loss of solar ene.rgy Ga-1n-P solution in contact with the GaP (111)A
reacl).mg the region of the spatial charge. The SIMS substrate.
profl.lc: s of the compounds for the Ga Inl_xP
transItIOn layer produced by the isothermaf contact
of the Ga-In-P solution with the GaAs (111)A
substrate are depicted in Fig: 2.

CAMECA IMS4F DEPTH PROFILE 28/06/90

10 6
ECH 90

10 5

Ga
(
10 " 142
," As
~ 10 3

.>.

-;; 10 2

.c:
III

...
c:
10

0
10
9
-I
1121
0 10 20 3121 'Ie
mn

Figure 2: The SIMS profiles of the components in the transition layer of the
Gax1nl_xPIGaAs.

557
 At the second step the isothermal contact of the
Ga-In-P solution isoperiodic with respect to its
composition (fo x = 0.51) enables to preclude a
supersaturation and to achieve the thermodynamical
equilibrium by the system. The third step is the
q1!asi-equilibrium synthesis of the GaxInhxP op.tic~l
wmdow. The I1resence of the Be acceptor Impunty m
the solution helps in the formation of the p-n
junction in the n-doped GaAs substrate. The
Isothermicity of the process significantly simplifies
the technology and. enables to precisely control
parameters or syntheuzed layers.

ACKNOWLEDGEMENTS
This work was supported by PIRSEM-CNRS,
AFME a!1~ by the Commission of the European
Communities.

REFERENCES
(l)M.Gavand, L.Mayet , B.Montegu,l A.Laugier,
20th IEEE Photovoltalc Specialists Comerence , 020,
1988.
(2)M.Gavand, L.Mayet, ALaugier, 11th Int.Symp.
GaAs and Related Compounds, Biarritz 1984,
Inst.Phys.Conf. Ser. 469, 1985.
(3)Yu.B.Bolkhovit},anov, Prog.Crystal Growth and
Charact;o)9,159, 1989.
(4)V.V 1\.uznetsov, P.P.Moskvin, I.M. Olchowik and
V.S.Sorokin, Inorganic Materials 22, 11, 1554,
1986.
5)AZUnger'D.M. Wood,I.ofCryst. Growth,98,1989.
~6)G.AB. Ball",I.H. van der Merve, Dislocations in
ohds,t.6,F.K.N. Nabarro, North-Holland,
Amsterdam, 1983.
(7)P.P. Moskvin, I.M. Olchowik, V.V. Pasynkov,
W.Sadowski, Izv.LETI,350 1985.
(8)V.V Kuznetsov, I.M. Oichowik, V.V. Pasynkov,
W.Sadowski, Izv.LETI,365, 1985.
(9)Sarah R.Kurtz, I.M. Olson, A.Kibbler, 21th IEEE
Photovoltaic Specialists Conference 138, 1990.
(lO)A.S. GutaKovski, O.P. Pshelakov, S.L Stenin,
Kristalografia, 25, 4, 1980.

558
10TH EUROPEAN PHOTOVOLTA1C SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

A TTC2 DEVICE WITH A n AI.32Ga.6SAs-P GaAs HETEROJUNCTION

L.Mayet *, M.Gavand *, B.Montegu * and A. Laugier

Laboratoire de Physique de la Matiere ( U.A.3SB CNRS

Institut National des Sciences Appliquees
20 Avenue Albert Einstein 69621 VILLEURBANNE Cedex ( FRANCE
* Permanent address: Universite Lyon I - ISIDT, Bat 203
69622 Villeurbanne Cedex (FRANCE)

ABSTRAC!. A !TC2 device was described with a n AI.32Ga.6SAs-P GaAs

heteroJunct~on as lower band-gap cell. This involves a
~implification in liquid phase epitaxy procedure. It was shown the
~mportance of the AI. nGa. 6SAs layer thickness in degradation
performances. The total conversion efficiency of this device in
TTC 2 configuration was 2S.7% (40AMI.SD)

1. INTRODUCTION The results of this device can give also

some indications on the effects of
In previous papers/l,2/ we have widely possible "pinholes" in the n GaAs laye.r on
described a particular TTC2 device realized the solar cell performances of a classical
by LPE and ILPE and designed to operate homojunction TTC2 device.
under the moderate irradiance of linear
concentrators. A possibility to obtain a
conversion efficiency close to 30% under a 2. DEVICE STRUCTURE AND FABRICATION
40 suns solar concentration has been
demonstrated. 2.1 Description.
In this device the first p-n junction Figure lb shows the cross section of
was formed by two layers of GaAS, Ge and Sn the new device. The difference between this
doped, grown by LPE on a GaAs(Zn) substrate one and the former (Fig.la) was only in the
which constitute the lower band-gap cell suppression of the n GaAs layer. All the
so-called the lower cell. After which a other layers and their parameters are
thick Sn-doped AI. 32Ga. 6SAs (about 10 ~) was identical with those described elsewhere
grown and finally a thin graded band-gap /2/. A large value (> 6~) of the thickness
AIGaAs(Be) window was deposited by ILPE/3/. of the AI. 32Ga. 6SAs layer is chosen ·to
During the growth of the window Be diffuses facilitate the chemical etching process
in the AI.3ZGa.6SAs layer to form the upper necessary before the n contact deposition.
p-n junction. A schematic cross-section of
this device is shown in Fig.la.
The Isothermal Liquid Phase Epitaxy
used to complete this structure presents
many advantages:i) Simplicity and rapidity:
at BOO°C an one step process of about 3mn
is only necessary to fabricate in the same
time the window and the diffused p-n A.R. coating
junction at a depth of about 1 . S~. ii) The AuBe
window t<500nm
ultra-thin window (SOnm) obtained by this
technique shows a weak absorption
principally for the photons of high energy.
+
In addition the interface between the GaAlAs p(Be) 510
window and the upper cell presents no
reflection of photons because her gradual
composition in aluminum. iii) Especially GaAlAs n(Sn) 4 10 17
the diffused junction is not located at a
cristallographic interface; so, the possible GaAs n(So) 4 10"
"pinholes·, which are frequent in LPE thin
layers, have no effect on the junction GaAs p(Ge) 210
quality.
A diffused junction is not possible GaAs
for the bottom cell because the
impossibility to define exactly a diffusion
time. But we have considered the
possibility to replace, in the TTCZ
a b
structure, the n-p GaAs homojunction by a n
AI.32Ga.68As-P GaAs heterojunction. This to A~e
simplify the liquid phase epitaxy procedure Fig.l
and to evaluate the performances of this a) TTC2 stucture with homo junction
simplified new device in comparison with b) TTC2 stucture with heterojunction
the former.

559
 Stepl :2 hours saturation
Step2: Ge doped layer grown by

.
equilibrium cooling
Step3 : Ala 32Gao 6SAs layer growth
Step4: AI;Gal_xAS graded window layer

J
pGaAs P GaAs
growth + diffused p-n junction formation.

n GaAlAs
\ x=.32
The cell design was made by a
\'--------J photolithographic process.After evaporation
of metals, the contacts were annealed at
•
<50nm
•• •• ... .. II 450°C during 2mn.
Two series of 30 TTC2 cells (active
..
O.8·1.8~ 5~ 350~
area of each cell:0.5x2.5mm) were realized
8 or 16 ~
on two GaAs substrates. The structures of
which differs only by the thickness of the
Fig. 2 shows the schematic energy band AI.32Ga.6SAs layer i.e.: tl = 16~m (sample
diagram of the device under short circuit BIC26) and t2 = B~(sample BIC27) .
r.onditi·ons. The height of the "spike" is
about 0.4 Volt.
3. RESULTS
2.2 Realization.
Schematic growth process and 3.1 Quantum efficiency.
temperature cycle utilised to fabricate the Figure 4 shows the external quantum
structure are shown in Fig. 3. Only three efficiency of the devices before any
melts are necessary to make the whole antireflexion coating. The two samples
structure with a p GaAs buffer layer and presents, for the high band gap cell, an
even two melts if the p GaAs substrate had identical response in spite of the
good diffusion lenghts which permit to thickness difference of the AI.32Ga.6SAs
avoid a buffer layer. Our stuctures layers. This means that the absorption edge
presents a buffer layer. After two hours of the AI.32Ga.6SAs material is very sharp.
saturation at BOOoC under Pd-purified The quantum efficiency of the low band gap
hydrogen the Ge doped GaAs layer and the cell is represented in comparison of this
(Sn)Al.32Ga.6SAs layer were grown by of the former device (with homo junction).
equilibrium cooling. Then, after a weak This shows a reductio"n in the photocurrent
dissolution (dT=0.25°C), an isothermal especially for the infrared photons. So the
growth realizes in the same time window and photocurrent collected in the GaAs cell
diffused junction. before antireflection coating was 8.2
mAcm- 2 (AM1.5D, 100mWcm- 2 ) against B.97
AI AI mAcm- 2 for the former device. The current
x = 0.32 y = 0.85 leakage was B.5%. After an antireflection
coating (2nm MgF2 + Slnm ZnS + 9Bnm MgF2)
Ge Sn Be the value of photocurrents were: 10.7
mAcm- 2 for the lower cell and 14.0 mAcm- 2
..... " ....... for the upper .
""" """" "
" ...
"" "" "
" "" "" ... 3.2 Jsc versus Voc
The experimental short circuit current
(1)
(2)
versus open circuit voltage characteristics
(3) so-called the pseudo "dark current"
(4) characteristics represents the diode
quality and allows to anticipate the future
device's performances.

Po)ycrystal
100
Monocrystal ......
<;'(

>- 80
z
"" 60
,-- - - _................. '\

~
~
~

::!<
~'---fC.'-:-: f-
Z
(I) (3) (4) <
(2)

r-
x
t::

. ~ .75 5
WAVELEKGTH (~ m )

Fig3: Growth process and time- Fig.4.External quantum efficiency of

temperature cycle for the three terminal the device.Also represented Q.E. of TTC2
solar cell: device with homojunction (dashed line) .

560
 .
The upper cell presents similar
characteristics and diode parameter values
than the former device and this for the two
samples i.e.: for the high currents, the '.
o

diode factor n=1 (diffusion current) with

J01 in the range 0.6-1.3 10- 24 Acm- 2 and for .. ' 0

the low currents n=2 (recombinaison

current) with J02 =10- l3 Acm- 2 . So under one . ,
' 0

.
sun illumination SO% of the dark current
was from diffusion current and 92% under 40 100 o
,
suns. For the best cell, under 40AM1.5D, o
Voc=1400mV, FF=0.B5 and the conversion , 0
,.,
efficiency was 11 =16.6% .
, 00
Jsc-voc characteristics of the lower
::;-
cell (heterojunction) were shown in Fig.S.
Different behaviours depends to the E
u
. 0

«: o
different samples: 5 o
Sample BIC26. For the high currents the IAMI.S
behaviour was always of diffusion type ~ 10 o
(n=l) with Jo1 in the range 3-18 10- 19
(I)
Z o
Acm- 2 , but for the lower currents the diode w
Q
factor was n >2, typically in the range: o
f-
n=3 to n=4.S with high values of Jon: 10- 9 Z o
Acm- 2 to 4 10- 6 Acm- 2 . Under one sun only 2%
to 11% of current was of diffusion type but ~
U o
o

60% to 93% under 40 suns. So voc = 790 to

920mV under one sun, but 1025 to 1060mV ::.:
..... o
under 40 suns. The best efficiencies were 11 o
=6.2% (lAM1.SD) and 9.6% (40AM1.SD). o • ... aIe2' (1,.,.-1
Sample BIC27. The values of the diode
parameters were close to these of a o o "te27 (81U'1

homojunction diode: Jo1 = 3 10- 19 Acm- 2 , n = o

2-3, Jon = 10- 10 Acm- L 7 10- 12 Acm- 2 with 73%
o
to 82% of the current was diffusion current
o
under 1AM1.5D and 9B% under 40AM1.SD. The
best efficiency was 9.B% (40AM1.SD). o
The total performances of these devices o
were 11=25.5% (BIC26) and 2S.7% (BIC27) in I
TTC2 conditions of utilisation. O.S 0 .9 1 1. 1
OPEN crRCUIT VOLTAGE (Veils)

Fig. 5 Typicals experimental Jsc "dark

4. DISCUSSION. current" versus Voc for two thicknesses of
the GaAIAs layer: BIC26 (16~m) and BIC27
Behaviour difference between the two (8~) .
samples can be only explained by the
~hickness difference in AI.32Ga.6SAs layers
~.e. 16~m for the BIC26 cells and 8~m for 5. CONCLUSIONS.
the BIC27 cells. A possible explanation can
be found in the Matthews's mechanism /4/. i) Use of heterojunction as lower
The lattice of GaAs and AlAs are matched cell for the TTC2 device leads to a
around 900°C, but at room temperature the simplification in the LPE procedure.
lattice mismatch was about 1.4 10- 3 because ii) This use involves a limited
the difference in thermal expansion degradation in the photovoltaic
coefficients of this two materials. For performances under 40 suns irradiance: 11=
layers grown at 900°C the thermal strain 25 . 7% towards 26.6% for the former device.
involves a tetragonal deformation in the iii) For use under one sun the
structure but only an elastic shear stress thickness of the AI. 32Ga. 6SAs layer should
is shown. Our layers were grown between be drastically limited.
BOO°C and 780°C and a weak misfit can relax iv) A pinhole in the n GaAs layer of
the stess by generation of an interfacial a homojunction lower cell as shown in Fig.6
network of dislocation if the thickness of is in fact a local heterojunction and
the layer was sufficiently large/5/. therefore does not involves degradations.

These dislocations increase the Pinhole

recombinaison states at the heterojunction
interface and can explain the diode factor
and JOn increases. This effect was still
limited and not dramatic for the high
values of photocurrent. However for an
utilisation under one sun irradiance, the
drop in performances can be avoided only by
a drastic thickness reduction in the
AI. 32Ga. 6SAs layer.
However the photocurrent decrease in
the lower cell, as shown in Q.E.Fig.4,
cannot be avoided because the interfacial
junction does not permit obviously a
junction depth optimisation. Homojunction Heterojunction

Fig.6: The pinhole in a n-GaAs layer is

in fact a heterojunction.

561
ACKNOWLEDGEMENTS
This work was supported by PIRSEM-CNRS,
AFME and by the Commission of the European
Communities.

REFERENCES
/1/ L.Mayet, M.Gavand, B.Montegu, A.Laugier
High efficiency AI.32Ga.68As-GaAs tandem
solar cells with three terminals.
20th IEEE Photovoltaic Specialists
Conference, 597, 1988.

/2/ L.Mayet,M.Gavand,B.Montegu, J.P.Boyeaux

and A.Laugier.Monolithic tandem solar
cell based on GaAIAs-GaAs system.
21th IEEE Photovoltaic Specialists
Conference, 64, 1990.
/3/ M.Gavand, L.Mayet and A.Laugier.Graded
band gap heterostructures for high
efficiency GaAs solar cells made by a
modified isothermal liquid phase
epitaxy. 11th Int.Symp.GaAs and Related
Compounds. Inst. Phys .Conf. Ser. N°74,
469,1985.

/4/ J.W.Matthews and A.E.Blakeslee. Defects

in epi taxial layers. J. of Cryst.
Growth, 27, 118, 1974.

/5/ G.A.Rozgonyi,P.M.Petroff,M.B.Panish.
Control of lattice parameters and
dislocations in the system Gal_gAlgAs/
GaAs. J. of Cryst . Growth, 27,106,1974.

562
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Cost Perspectives of GaAs Thin-Film Solar Cells

E.A. Alsema, R.F.A. Cuelenaere, w.e. Turkenburg

Dept. of Science, Technology and Society, University of Utrecht,
Oudegracht 320, NL-35 I I PL Utrecht, The Netherlands.

ABSTRAcr. To become a viable option for grid-connected electricity supply in the Netherlands
GaAs thin-fim solar cell modules should achieve a break-even module cost of ECU lOO-170/m2 or
ECU 250-35O/m2 if social costs are accounted for. The most important bottle-neck are the costs of
the substrates for epitaxial layer deposition, which are much too high if GaAs or Ge are used as a
substrate material. The cost prospects for mono- or semicrystalline silicon substrates appear
somewhat better (minimum substrate costs of ECU 651m 2) but major technological difficulties still
have to be solved for this option. Deposition on low-cost, non-crystalline substrates by means of
the grapho-epitaxial method appears even more difficult.
The second bottle-neck requires that the utilization rate of source gases for MOCVD is improved
and that their costs are very much reduced. Thirdly equipment cost of MOCVD reactors will have
to become much lower, e.g. by scaling up reactor size. Finally raw material supplies may become a
bottle-neck for large-scale application of GaAs cells, especially if GaAs or Ge substrates are used_
Our conclusion is that meeting the cost targets is very difficult but not impossible. It requires major
break-throughs in each of the above areas.

l. INTRODUcrION items like module encapSUlation, labour and overhead.

As large-scale production is still a long way in the future it will be
In the past years research on high-efficiency solar cells has been clear that there are large uncertainties about estimates of production
one of the main subjects in the field of photovoltaic energy costs. Therefore we have laid emphasis on making estimates of the
research in the Netherlands. Because of their excellent efficiency minimum costs, defined here as the lowest conceivable costs in a
prospects thin-film solar cells based on III-V materials were given future production process. for the various aspects we have
much attention in the National Solar Energy Research Programme. examined.
In 1988, however, questions were raised concerning the long-term
cost perspectives of III-V solar cells (I). Therefore NOVEM, the
Netherlands Agency on Energy and the Environment, commissioned 2. COST TARGETS
a study on the prospects of III-V solar cells of achieving cost levels
at which application in grid-connected PV systems in the Nether- For Ill-V thin-fil.m solar cells cell efficiencies of more than 27%
lands may become feasible. already have been achieved in a stacked configuration. For the
In this paper we will summarize the results of the study. which was long-term an efficiency of between 25% and 35% for a flat-plate
concluded in 1990 (2). module may be considered as feasible.
At this module efficiency one can estimate that module costs must
At present the most extensively investigated III-V material for use not exceed ECU lOO-170/m2 to achieve a break-even price ofECU
in solar cells is GaAs. although interest for other materials, like 0.05/Kwh for electricity from grid-connected PV systems in the
InGaP and GaSb, is increasing. Almost all III-V thin-film solar Netherlands .
cells are made.by epitaxial growth of a thin film on a monocrystal- If we also account for the social costs of conventional electricity
line GaAs (or Ge) substrate. Metal Organic Chemical Vapour generation (including CO2 emissions etc.) and accordingly assume
Deposition (MOCVD) is the most commonly used deposition that the break-even price of PV electricity will be ECU O.09/kWh,
process. Because of the existing research and production experience then module costs must be no more than ECU 250-350/m2 at most.
we focussed our study on GaAs thin-film cells prepared by
MOCVD and related processes. Although corrections for raw We shall take these break-even module costs of ECU 100-170/1112
material prices will be necessary. many conclusions will be valid and ECU 250-350/m 2 as the basic criterion against which the cost
for other III-V solar cell materials as well. prospects of GaAs solar cells are to be tested.
In the next section we will discuss first the cost targets which
should be achieved for III-V thin-film cells in grid-connected 3. RAW MAlERIAL SUPPLIES AND RAW MATERIAL cosrs
systems in The Netherlands. Subsequently we will investigate the
five major bottle-necks which can be identified for large-scale If GaAs solar cells are to be applied worldwide on a large scale
application of GaAs thin-fIlm solar cells, namely: problems may arise with the raw materials supply.
>I< raw materials supplies; In table I we have therefore drawn up an inventory of the current
.• substrate costs; production levels and the extractable stocks for the raw materials
• utilization rates and costs of source gases for MOCVD; gallium, arsenic and gennanium (the latter being a candidate
* MOCVD equipment costs; substrate material).
* environmental and health aspects. It must be noted here that the existing production capacity for Ga is
In section 3 we will discuss raw material requirements and supplies estimated to be about three times higher than the actual production
for solar cell production. In the subsequent sections we will draw level. Nonetheless it will be clear that the gallium supply places
up estimates for the costs of GaAs, Ge, Si and other substrates, for severe constraints on future GaAs solar cell production. Arsenic
the cost of materials in the thin film, for equipment costs, for supply, on the other hand, will hardly be a problem.
environmental protection costs and ftnally for the remaining cost

563
 current max. prod. max. extraction raw material progress into a number of alternatives for GaAs substrates.
world with current from coal ash price (6N purity)
production e1ltr. method The first area being investigated is the multiple use of GaAs sub-
(t/year) (t/year) (t/year) (ECU/kg) strates using what is known as CLEFT technology (3). Multiple use
might be able to cut the price of substrates by a factor of 5 to 8. If
Ga 45 4800 27.000 450-480 the technology proves viable and at the same time the minimum
As 50.000 » 50.000 n.a. 100·130 price for GaAs substrates. referred to above. becomes a reality. it
Ge 100 1400 7.000 630·700
might be possible to achieve effective substrate costs of ECU 70·
120/m2 with CLEFT. However. we do not consider that the probab·
Table I: Estimated current and maximum world production and
ility of both these conditions being fulfilled is very high.
current prices of raw materials for the manufacture of
GaAs solar cells.
The second alternative is to use germanium substrates. Although Ge
is extremely well suited to the epit31lial growth of GaAs. current
The material requirements for solar cell production can be distin·
raw material prices are so high that the use of Ge substrates is
guished into the requirements for the thin ftlm and for ~e substrate.
unattractive purely from the cost angle (min. cost ECU 1300/m2).
Assuming an active cell thickness of 3 pm and assummg that the
unless there were some way of managing with Ge layer thicknesses
utilization rate for Ga in the thin·film deposition process can be
ofless than 50·100 pm.
greatly improved to 50%. the gallium consumption will be around
60 tort/GWp (see section 5 for discussion of utilization rates).
The third alternative that we have examined is the use of mono·
If we compare this with the production levels we can conclude that
crystalline silicon substrates. Epitaxy of GaAs on such substrates
the supply of raw materials for the thin films in GaAs solar cells is
runs up against a large number of difficulties due to the mismatch
unlikely to prove a major obstacle. Existing production ca~acities
of the crystal lattices and the (tltermal) e1lpansivity of tlte two
for gallium appear to be sufficient for solar cell production of
materials. The consequence is that tlte electrical properties of GaAs·
approx. 1.5 GWp per annum. . . on·Si cells often leave much to be desired.
The estimated m31limum e1ltraction capacity for gallium (and
Due in part to keen interest on the part of tlte electronics industry. a
arsenic) using present extraction methods would seem to be suf·
fair amount of research has been done in recent years into intprov·
ficient for solar cell production of at least 20 GWp per annum.
ing tlte quality of GaAs·on·Si material. This has led to some
With the latter solar cell production level it would be possible to
intportant advances. For example. GaAs·on·Si solar cells have now
generate approx. 10% of present world electricity consumption
been made with an efficiency of 20% (4). Recently. inte1'e5ting
using GaAs solar cells.
results have also been obtained witlt Ge buffer layers on a silicon
substrate (5). However. much work still needs to be done on tlte
However. the supply of raw materials for the substrates on. which
stability of the Gas-on·Si materials after deposition.
the thin·film cells are to be grown may well present some dlfficul·
Moreover. the costs of monocrystalline Si substrates are com·
ty. The problems arise if solar cell produc~on has to be ~ed on
paratively high (minimum cost appro1l. ECU 150/m2) and unfor-
the use of germanium or GaAs wafers. With a substrate thickness
tunately it looks as if the complicated processing needed to produce
of 200 ).1m for Ge or 400 ).1m for GaAs present extraction m~thods
good quality GaAs·on·Si material will also entail considerable
for germanium and gallium would permit solar cell production of
costs. Estimates of tlte ultimate costs of GaAs-on·Si material are
no more than 260 MWp/y and 900 MWp/y respectively.
titus in tlte region of ECU 800·2000/m2. This would put GaAs·on·
Whereas using alternative extraction methods. e.g. from coal ash.
Si in about tlte same price category as GaAs-on·Ge. tltough prob·
would enable a large increase in Ga and Ge output. it would prob·
ably with lower efficiency.
ably also lead to a sharp rise in raw material prices.
The conclusion to be drawn is that if GaAs thin·film solar cells are
The cost prospects for semi·crystalline Si substrates . the fourth
to be used on a large scale the use of substrate materials other than
possibility· are a lot better. witlt a minimum pricx: of ECU 65/~2.
GaAs or Ge wafers will be necessary.
The problem is. however. that the technology reqUired for grow~g
GaAs films on this material is not yet available and that tltere IS
An intportant determinant of GaAs solar cell costs is the price of
very little interest in developing it. even inside tlte solar cell in·
the raw materials.
dustry.
At the tinte of writing the price of highly refined (6N) arsenic is
Also it might be interesting to look at tlte possibilities of using
approx. ECU lOO/kg. One major determinant here is the stringent
environmental and health requirements that the refining process has metallurgical·grade silicon substrate as this would very much
reduce the silicon cost.
to meet. Cost reductions of 30% may be possible with economies
of scale.
In the fift1t place we investigated the possibilities of the grapho·
The price of gallium fell rapidly during the sixties. but in the past
epitaxy metltod. Efforts have been made using this very interes~g
ten years has remained more or less constant at ECU 450/kg.
technique to bring about epitaxial growth on a non..crysta11ine
despite a fourfold increase in production.. There .is ~ow a high
substrate onto tlte surface of which a relief structure has been
degree of overcapacity in gallium productIOn which IS probably
formed. If this works. it will open up prospects of using cheap
having the effect of depressing prices. Further price falls cannot be
substrates.
ruled out. but they will only be achieved if and when investors see
Interesting results from grapho·epitaxy have recently been reported
prospects of large sales of gallium.. ..
by Kanata et al.• who have succeeded in growing films of CdS and
To determine the effects of a possible fall tn raw matenal pnces we
Ge of fairly good crystalline structure on a specially prepared Ni
have also examined the cost prospects with a Ga price cut by 50%
substrate. So far. however. tlteir experiments witlt GaAs on such a
and an As price down 30%.
substrate have not proved successful (6). Though we recommend
that the development of tltis technology be followed closely. the
prospects for grapho·epitaxial growth of GaAs do not appear
4. SUBSTRATE COSTS
particularly favourable given tlte present state of tlte technology.
At the present time most GaAs thin·film solar cells are made by
epitaxial growth on monocrystalline GaAs wafers. The present costs
5. MATERIAL COSTS FOR THIN·FILM GaAs SOLAR CELLS
of these substrates are extremely high. viz. ECU 9.000/m2 to
45.000/m2. We estimate that with increased production volumes.
Current metltods of achieving epitaxial growth of thin films of
higher ields and scrap recycling a minimum price of ECU 750·
I OOO/m~ might be attained for GaAs substrates. In view of our cost
III·V material by Metal Organic Chemical Vapour Deposition
(MOCVD) are characterized by a low to very low rate of utilization
targets this still is prohibitively high. . .
of the raw materials (~15% for gallium. <1 % for arsenic). This low
It is partly because of these high costs that research IS currently In
utilization rate, coupled with tlte very high prices of metal organic

564
source gases, results in high material costs for GaAs thin-film solar Scaling up reactors for larger production volumes of I m2 substrate
cells. surface per run, while not entirely unproblematical, does appear
possible in principle. Here again the principle of the pulsed reactor
New, more efficient methods of deposition that would be suitable being investigated at the University of Nijmegen may offer good
for the fabrication of thin-film solar cells are so far only at an prospects.
embryonic stage. One of the methods with which better material If the investment costs involved in such scaled-up reactors can be
utilization might be achieved is the so-called pulsed MOCVD held at ECU 1.3 million and at the same time the production cycle
reactor currently being developed at the University of Nijmegen. time can be reduced by a factor 4, equipment costs might be able to
With such improved deposition methods it may be possible in the fall to approx. ECU 901m 2• In very favourable conditions (5
future to achieve Ga utilization of 50% and an arsenic utilization m2/run, investment ECU 1.7 million, 8000 runs/year and 90%
rate of 10%. A utilization rate of 70% for both Ga and As would yield), equipment costs of around as little as ECU 101m 2 might be
appear to be possible only if deposition technology develops extre- possible.
mely favourably.
On the basis of an analysis of the production costs of crystalline
Apart from raw material utilization, other determinants in materials silicon modules, we estimate other production costs, i.e. the mate-
consumption are, of course, the thickness of the solar cell and the rial costs for the module, labour costs and ovemeads, at ECU 31-
overall yield of the production process. The present generation of 391m2, ECU 52-591m 2 and ECU 121m2 respectively.
GaAs thin-film cells have a typical thickness of approx. 5 pm. One possible way of reducing labour costs might be to produce
Possibly this can be reduced to 2 pm. We estimate the ultimate integrated modules, as is now done with amorphous silicon solar
feasible overall yield for the production process at between 70% cells. However, as long as GaAs solar cells are made on the basis
and 90%. of crystalline substrates it looks extremely unlikely that integrated
module production will ever be possible.
The metal organic source gases used in the MOCVD process are Because of the relatively high current density and the use of gold
currently extremely costly - up to about ECU 6500/k:g for the Ga as the contact metal, the material costs of the contacts for GaAs
source. This is probably determined by the small market for such cells can be quite high. We estimate them at ECU 6.5lm 2•
gases. Increasing the size of the market might be expected to lead
to prices falling by a factor of 2-3. The minimum price level for the
source gases we estimate at twice the price of the elementary raw 7. ENVIRONMENTAL AND HEALTH ASPECfS
material.
The environmental and health hazards arising from the production,
However, even at prices reduced by a factor 2-3, the material costs use and decommissioning of GaAs solar cell system lie mainly in
associated with the deposition of a thin film of GaAs for a solar the fact that arsenic is used [or the production of GaAs cells. In the
cell would still be extremely high. Even assuming slightly improv- case of the MOCVD process the chief dangers lie in the use and
ed MOCVD technology, with utilization rates for gallium at 30% storage of the highly toxic gas arsine.
and for arsenic at 3%, film thickness of 5 pm and an overall yield The extraction, refining and processing of gallium, by contrast,
of 70%, the material costs would still be as high as ECU 8oo/m2 appear not to pose any particular threat to the environment or to the
cell arel\. health of the public at large or process wodcers in particular.
It will only be either by extremely favourable developments in Using dust fIlters or gas scrubbers it is possible to reduce emissions
MOCVD technology or by a very sharp reduction in the price of of arsenic during refming and cell production to the extent that they
the source gases to the minimum level together with a simultaneous will cause only a negligible increase in the annual average back-
substantial improvement in MOCVD technOlog~, that it may be ground level. The costs of emission control measures can be es-
possible to cut material costs to ECU 40-170/m . If it is also as- timated at ECU 71m 2.
sumed that the raw material price of Ga and As will fall, then it
might be possible for the material costs to fall to ECU 20-130/m 2• The transport, use and in particular storage of arsine can carry very
serious risks for workers and those living near the plant. Replacing
However, it has to be remembered here that the production volumes arsine with metal organic arsenic compounds, like trimethyl arsine
required if our minimum price for the source gases is to be achiev- or tertiary butyl arsine, would seem to be possible and would result
ed will only be reached if solar cell production reaches the very in a significant reduction in the hazards of producing GaAs solar
substantial level of over 1 GWp/year. Although the price of the cells. Despite the extra capital investment in equipment and the
gases may fall slowly as the market grows, it looks very much as if possibly higher cost of the source gas, using alternative source
the high cost qf metal organic source gases will continue to be an gases may well produce a cost benefit compared with using arsine
almost if not wholly insuperable obstacle to the introduction of because it will be possible to dispense with an expensive gas
GaAs solar cells. scrubber.
Another possible way of reducing the dangers of using arsine might
This being so, one might also wonder whether it might not be a be to make the arsine on-site, as and when it is needed.
good idea to look for other methods of deposition in which it
would be possible to use elementary or other cheap source materials
instead of metal organic gases. The investigation of such alternative
deposition processes has yet to progress beyond the most tentative 8. OBSTACLES TO REDUCING PRODUCfION COSTS
investigations.
In Table II we have summarized the various cost estimates for the
production of GaAs solar cells on the basis of epitaxially grown
6. EQUIPMENT AND OTHER PRODUCfION COSTS thin films. The cost estimates themselves are based on the present
state of the technology, a state of the technology that we regard as
One major element in the production costs of GaAs thin-film cells feasible in the future, and an optimistic view of the technology of
is .the interest and depreciation costs for the production equipment. the future. These last estimates may be regarded as what we con-
The most uncertain factor here is the equipment needed for the sider the lowest conceivable costs assuming present raw material
MOCVD process. prices: in other words, our 'minimum price '. In the fourth column,
The present generation of MOCVD reactors have only a limited finally, we give an estimate of costs assuming both favourable
processing capacity of at most 1,000 cm 2 substrate surface per run, technological development and reduced raw material prices of 50%
and call for an investment of approx. ECU 400,000. Assuming a and 30% (the reduced minimum price) for Ga and As respectively.
20% annual interest and depreciation percentage, 1000 reactor runs
a year and an overall production yield of 70%, the equipment costs The very high level of present production costs is an indication of
would be more than ECU 1,000 per m2 cell surface. the fact that at the present time there is still virtually no production

565
 ~.~- - .------
present future future optimistic + reduced
costs feasible optimistic raw material price
(ECU{m2) (ECU{m2) (ECU{m2) (ECU{m2)

substrate 9,000-45,000 170-800 a) <40 b) <40 b)

thin-film material 4,000 170-800 40-170 20-130
MOCVD equipment 1,000 90 10 10
envir. control costs 7 2 2 2
other costsc) 120-130 <90 ? <90 ? <90 ?

Total 13,000-50,000 500-1800 <180-300 <160-270

Table II: Estimates of the production costs of GaAs solar cells based on thin epitaxial films, for present technology, future feasible
teclmology, best-case teclmology (,minimum price ') and best-case teclmology combined with reduced raw material price ('reduced
minimum price').
a) crystalline Si substrate; upper limit is for factor three cost increase due to complicated processing.
b) grapho-epitaxy (non-crystalline substrate); exact substrate costs are difficult to estimate.
c) includes: labour for cell processing, contact material, labour and materials for module assem bly.

teclmology for GaAs solar cells. However, it cannot be deduced of raw materials.
from this that sufficiently low costs are an impossibility. If the price of source materials can be greatly reduced at the same
time as material utilization rates are sharply improved, material
In the past, for example, a cost reduction of more than two orders costs of less than ECU 401m2 might be possible.
of magnitude has been realized in crystalline silicon teclmology.
When considering the costs that can be regarded as feasible in the While the third obstacle, the cost of MOCYD equipment, is less
future we see that there are at least three obstacles in the way of serious than the two others just referred to, substantial advances
reducing production costs to a break-even value of ECU 100- will nonetheless be needed in the scaling up of deposition reactors
l70/m2 or ECU 250-350/m2. In order of imponance these are: if equipment costs are to be brought below ECU 170/m2.
- substrate costs;
- material costs for the thin film; Finally attention will have to be devoted to cutting 'other' costs,
- the costs of the MOCYD equipment mainly the cost of labour and materials at the module assembly
Major teclmological breakthroughs will have to be achieved in the stage. It is debatable whether the cost of assembling non-integrated
areas of all three of these obstacles, since each one individually is modules can be reduced any further.
capable of adding so much to the total module costs that they rise
above the target costs. Even if the raw material prices of Ga and As
fall below their current level, important advances in these areas will 9. CONCLUSIONS
still be necessary.
A fourth obstacle that may arise, although it does not emerge from Although III-Y solar cells materials offer the prospect of very high
the table is the availability of raw materials. As observed in 7.2, the module efficiencies, in the range of 25-35%, their production costs
reserves of Ga that can be extracted with present methods are are also very high.
probably too small for the manufacture of sufficient areas of sub- It might be possible to reduce the production cost of GaAs thin-
strate based on crystalline GaAs or crystalline Ge for large-scale film solar cells to a level of ECU 250-350/m2 or lower, but this
solar cell production. will require major teclmogical break-throughs in each of the
following three areas: deposition on cheap substrates, source mate-
There are therefore compelling reasons for trying to use substrates rial utilization and scaling up of deposition equipment.
other than GaAs or Ge wafers. In the first instance crystalline Research on III-Y solar cells as an option for grid-connected
silicon substrates look the most attractive alternative. In the longer electricity supply in the Netherlands or in countries with a similar
term, however, it is still uncertain whether the costs of these will be irradiation should therefore be viewed as a high-risk investment.
low enough. The only remaining option would then seem to be the
use of cheap, non-crystalline substrates combined with e.g. grapho-
epitaxy or a (Ge?) buffer layer on a non-crystalline substrate. ACKNOWLEDGEMENTS
Using techniques of this kind it might be possible to bring substrate This study was conducted in commission of NOVEM B.Y., the
costs below the ECU 40{m2 mark and there would be no problems Netherlands Agency for Energy and the Environment. The authors
with raw materials supplies. The likelihood of these techniques want to thank Prof.Dr. LJ. Oiling, Dr. M. Muradin, Prof. Dr. W.F.
being successfully achieved would, however, appear to be not very van der Weg and Dr.Ir. R.J.c. van Zolingen for their valuable
large. contributions.

To solve the second problem, that of the material costs involved in

the thin film, two things are needed: 1) a fall in the price of the REFERENCES:
source material (currently metal organic gases) and 2) an improve- 1) F.e. Treble, E.A. Alsema, W.C. Turkenburg, The Prospects of
ment in deposition technology as regards the utilization of sources. Amorphous Silicon and III -Y Solar Cells for the Netherlands,
There is little that can be done about the price of metal organic University of Utrecht, October 1988/ January 1990.
gases, since the determining factor is the size of the market One 2) E.A. Alsema, R.F.A. Cuelenaere, Cost Perspectives of GaAs
point that might be worth investigating is whether it would be Thin-Film Solar Cells, University of Utrecht, W 90033,
advantageous to make the metal organic gases on-site, as part of the September 1990.
whole solar cell production process. (This would also be beneficial 3) R.W.McClelland,e.O.Bozler,J.C.C.Fan, Appl.Phys.Lett. 37
from a safety point of view.) (1980), 560.
A strong effort will be needed in the field of deposition teclmology. 4) Y.Kadota, Y.Ohmachi, M.Yamaguchi, PYSEC-4, Sidney, 1989.
The MOCYD process will have to be greatly improved as regards 5) A. Leycuras et aI., 21st PY Spec. Conf., Kissimmee, 1990.
materials utilization. Depending on how the prices of metal organic 6) T.Kanata, H.Takakura, Y.Hamakawa, Appl. Phys. Lett. S4
gases develop, it might even be wondered whether we ought not to (1989), 706.
be devoting attention to the development of deposition processes in
which the source materials are used in elemental form. lmproving
utilization rates is also highly desirable in view of the availability

566
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DEVELOPMENT AND EVALUATION OF CdS/CdTe THIN FILM PV CELLS

J.Skarp, Y.Koskinen, S.Lindfors, A.Rautiainen, T.Suntola

Microchemistry Ltd
PO Box 45
SF-02l5l ESPOO, FINLAND

ABSTRACT

A new CdS/CdTe thin film cell process has been developed. Cells
processed were used for an evaluation of the performance and technical
potential of thin film CdS/CdTe cell concept. Atomic Layer Epitaxy (ALE)
method was used to grow CdS/CdTe - thin film layers.

The cells processed were characterized by electrical measurements in

light and in the dark. The highest conversion efficiency, 14 %, was
measured in a cell of 12 mm2 active area under 100mW/cm2 (ELH bulbs).

The environmental risk of cadmium in commercial cell products was

figured out by comparing the amount of cadmium involved in the cell to
the ammount of cadmium in other existing products.

1. INTRODUCTION
2. CELL FABRICATION

Polycrystalline thin film PV-cells

have a potential to become a real A CdS/CdTe thin film structure, as
challenger in PV - power applications. This shown in Fig. I, was grown on a commercial
is because of the cost advantage compared Sn02/glass-substrate. The glass was 1.1 mm
to single crystal PV- cells and because of thick and cut to 5 x 5 cm2. The sheet
an efficiency and stability advantage resistance of the Sn02 was approx. 10
compared to amorphous silicon thin film n/square.
cells. One polycrystalline thin film
candidate is CdS/CdTe.

This study has . been made for Neste Au/eu - electrode

Advanced Power systems, NAPS, to develop
polycrystalline CdS/CdTe thin film cell and CdTe (3um)
so to investigate the feasibility of using
CdS/CdTe in power applications. CdS/CdTe -
graded layer
The theoretical maximum PV efficiency (2S0nm)
of CdTe is over 27 %, and there are several
reports of thin film cells with an CdS (SO run)
efficiency above 10% /1 ... 5/. The raw
materials, Cd and Te, are relatively
inexpensive and available in reasonable TCO «SOOnm)
quantities. However, cadmium is classified
as an environmentally harmful 1 material. We
here compare the amount of cadmium involved glass
in the CdS/CdTe-TF-cell to the cadmium
amounts in existing products or processes.

CdS/CdTe layers were grown using Atomic

Layer Epitaxy (ALE) /6/. ALE is a thin
film growth method which is based on
sequential chemical reactions on the
surface. It is very suitable for II-VI Figure 1. Schematic picture of
compounds thin film deposition. Through its CdS/CdTe - structure grown on a glass/TCO-
inherent mechanism of growth, ALE features substrate.
a digital control of the process.
Characteristics of thin films produced by Cadmium sulphide and cadmium telluride
ALE are good crystallinity, excellent thin films were grown by Atomic Layer
uniformity over a large area and a low Epitaxy (ALE) using elemental Cd, Sand Te
pinhole density. as reactants. The process temperature was
420 C. The processes of CdS and CdTe layers

567
are presented in references /7,8/ in more Some attempts were made for optimizing
detail. individual cell parameters. The best open
circuit voltage achieved was 0.84 V, the
There is a 9.7 % lattice mismatch short circuit current 26 mA/cm2 and the
between CdS and CdTe. To minimize the fill factor 0.73. These results were
effects from this mismatch on the measured on separate samples.
electrical function of the heterojunction a
graded layer was processed between CdS and The high Voc and FF are signs of good
CdTe /9/. In this layer the CdS/CdTe crystallinity of CdTe and good hetero-
proportion was gradually changed from pure junction properties. These could be
CdS to pure CdTe affected by growing either the CdS or the
graded layer thicker. This has, however, a
After the ALE-process the samples were negative effect on the short circuit
given CdC12-heat treatment followed by current due to reduced short wave length
Br2/methanol-etch /10/. The back electrode transparency in the window layer.
(Cu/Au) was then evaporated through a
mechanical mask. The best Jsc was achieved either with
very thin (20 nm) CdS layer or when using
3. RESULTS AND DISCUSSION Cd(x)Zn(l-x)S instead of CdS. However, in
both cases we observed a decrease in Voc or
increase in series resistance.
Current-voltage measurements were made
in light and in the dark. The area of The fill factor could still be
measured samples was 12 mm2. improved by reducing series resistance,
which in our test cells was about 3 O/cm2
The light source used was based on ELH at best .
halogen bulbs calibrated with a Si
reference cell. Illumination level was The I-V-curve in dark is shown in
adjusted according to /11/ to suit CdTe Fig . 3.
band gap value. The total illumination
level was 100 mW/cm2. The diode characteristics of this PV-
cell shows Jo = 6.1E-l1 A/cm2 and ideality
The I-V curve of the best solar cell factor of n = 1.7 in J = Jo{exp(qV/nkT)-1}.
is shown in fig.2. The conversion The low saturation current density means
efficiency was 14.0 % with Jsc 23.8 low density of defects or impurity states
mA/cm2, Voc = 804 mV and FF = 0.73. The in the junction and/or within the depletion
thicknesses of CdS ,the graded layer and layer. The cell withstands also several
CdTe in this cell were 50 nm, 250 nm and volts negative voltage without avalanche or
3.0 ~m, respectively . increased leakage current.

j
~
5 I
9 " .I

-5
: I
-19
: /
-15
~ / ~

-28
~
/ :

-25
-
V
o 0.5 1

Fig. 2. The I-V curve in light

Eff. 14 %
Voc 804 mV
Jsc 23.8 mA/cm2
FF 0.73
active area = 12 mm2

568
 cells give no cadmium emission. Market
acceptance of CdS/CdTe PV products will
i rely on future legislation /14/ and on
general attitudes.
I. l / J 6.ACKNOWLEDGEMENTS
// !
/ !

The authors would like to thank

/ I
Ms.S.Kaipio for the help in the processing
/ of the samples.
/
REFERENCES:
~
::::: 0.5 1.5 z
u
/1/ H.Matsumoto et.al. ,Solar Cells
11,(1984),367

/2/ S.p.Albright et.al.,Proc. of the

Fig.3. The I-V-curve in the dark Polycrystalline Thin Film Prg.Meeting,
SERI,(1989),17
4. CADMIUM IN THE CdS/CdTe PV CELL
STRUCTURE /3/ P.V.Meyers, Solar Cells 27,(1989),91

/4/ G.C.Morris et.al., 21st IEEE

There are several papers about cadmium Photovoltaic Specialists Conf.,(1990),575
in PV-cells /12/. In the following we have
collected some information about cadmium /5/ A.K.Turner et.el., Technical Digest of
which would help us to view the problem of 5th International Photovoltaic Science and
cadmium in PV-cells in perspective. Engineering Conference,(1990),761
Cadmium is produced as a by-product in /6/ T.Suntola, Materials Science Reports
the purification of zinc, lead, copper and 4,(1989),261
mercury. The annual production of Cd-metal
is about 19 000 t /13/. Today cadmium is /7/ A.Rautiainen et.al.,Proc. of MRS Spring
mostly used in rechargeable batteries (Ni- Meeting -91, to be published
Cd), as a UV-stabilizer in PVC-plastics, in
steel plating against corrosion and in some /8/ A.Kytokivi et.al. ,Proc. of MRS Spring
paint pigments. Meeting -91, to be published

The total thickness of CdS/CdTe, 3 um, /9/ N.Suyama et.al.,21st IEEE

means about 10 g of cadmium in one square Photovoltaic Specialists Conf.,(1990l,498
meter of PV-panel. There is thus 100 kg of
cadmium metal in 1 MWp (10% module /10/ R.W.Birkmire et.al.,Proc. of the
efficiency, 10 000 m2) of panels. Compared Polycrystalline Thin Film Prg.Meeting,
with the annual production of cadmium this SERI, (1989) ,77
is rather small amount. There is also some
small cadmium emission from thermal coal /11/ K.Emery et.al., 20th IEEE Photovoltaic
power plants. The amount of this emmlSlon specialists Conference,(1988),1087
is the same as the amount of cadmium needed
in CdS/CdTe cells to produce the same /12/ P.D.Moskowitz et.al., Health, Safety
energy in their operating lifetime /12/. and Environmental Issues relating to
Cadmium Usage in Photovoltaic Energy
Cadmium. has a low vapOUr pressure, Systems,SERI,(1990 ,
which makes the process control easier than
with toxic gaseous substances. The dust of /13/ T.O Llewellyn, Minerals Yearbook
cadmium and its compounds, however, must be 1988,1990
taken into account when planning cell
production. /14/ The Council of the EC, Council
Resolution 88/C30jOl
The CdS/CdTe thin film structure is
encapsulated within two glass plates in the
PV-module. There is no emission of cadmium
in normal use. However, there should be
instructions for broken modules, which
cannot be treated as normal waste.

5. SUMMARY

This study, together with published

works on CdS/CdTe cells, shows that there
are several ways of producing thin film PV-
cells of 10% 14% efficiency and
therefore it is a viable alternative for
large scale power production. cadmium can
be easily handled safely in a CdS/CdTe
manufacturing plant.

In storage and operation CdS/CdTe

569
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOVOLTAIC PROPERTIES OF AN EPITAXIAL CdS/CdTe/ZnTe

n-i-p HETEROSTRUCTURE GROWN BY MOVPE

M. Y. Simmons, H. M. Al-Allak, K. Durose, A. W. Brinkman

Applied Physics Group, School of Engineering and Applied Science,
University of Durham, South Road, Durham, DHI 3LE, U.K.

Abstract

Epitaxial ZnTe/CdTe/CdS p-i-n heterostructures have been fabricated by the

MOVPE growth of CdTe and subsequently ZnTe onto n-type single crystal (0001)
CdS substrates. Analysis of the current-voltage characteristics suggests that
the electrical transport across the junctions is dominated by a multi-step
tunnelling process. Preliminary photovoltaic measurements without
optimisation of the contact geometries gave open circuit voltages of up to
0.1 V and short circuit currents of 5 rnA cm- 2 . The fill factor of devices
formed was low -0.33 mainly due to the highly resistive p-ZnTe layer.

The reaction was then quenched by rinsing the

1. Introduction substrates in deionised water for 30 minutes.
The substrates were loaded into the MOVPE
The n-CdS/i-CdTe/p-ZnTe is currently of reactor immediately after preparation and
interest as an alternative to the single n- given a heat clean at 450°C for 10 minutes
CdS/p-CdTe solar cell heterojunction. The under hydrogen.
addition of the ZnTe window layer avoids the
problem of contacting highly conducting p- Growth took place in a conventional horizontal
CdTe. The novel p-i-n structure was initially reactor using dimethyl cadmium, dimethyl zinc
proposed by Meyers [1] who later went on to and di isopropyl telluride as the precursors.
produce polycrystalline devices having The CdTe (-1 ~m) and ZnTe (-0.1 ~m) epitaxial
efficiencies of up to 9% [2]. The epitaxial layers were_grown at 325°C on both the (0001)
growth of CdTe onto (0001) CdS by MOVPE has Cd and (0001) S faces of the CdS with a total
previously been demonstrated [3]. In these flow rate of 7000 SCCM. Au and In were vacuum
studies shiny epilayers having best surface evaporated onto the ZnTe and CdS surfaces
morphology were those grown on the (OOOl)B respectively for device fabrication. The
face of CdS. RHEED patterns confirmed the structure and morphology of the resulting
epit!xial relationship was (1Il)CdTe II layers were assessed routinely using 100 kV
(OOOI)B CdS. Layers of CdTe on (OOOl)A CdS RHEED and SEM. The electrical properties of
were matt in appearance and assessment of the the CdTe/ZnTe films were measured using
films using RHEED showed them to have a random standard resistivity measurements. Current-
'polycrystalline structure. There have been, voltage (I-V) characteristics were obtained
as yet, no systematic studies of the using a Keithley electrometer. Temperature
structural and electrical properties of dependent measurements were performed using an
epita~ial CdS/CdTe/ZnTe heterostructures, Oxford Instruments cryostat and temperature
grown by MOVPE. One problem here is the controller. Photovoltaic output character-
expense and poor availability of the single istics were recorded under simulated solar
crystal n-CdS substrates. Epitaxial CdS/CdTe/ illumination of 90 mW cm- 2 • SEM EBIC studies
ZnTe structures have been fabricated on (0001) of the junction were also studied using a S600
CdS single crystal substrates in situ in the Scanning Electron Micro-scope.
MOVPE reactor. The aim of the present study
being to investigate current-transport
mechanisms and interfacial recombination in 3. Results and Discussion
the absence of polycrystalline grain
boundaries. The epitaxial ZnTe/CdTe layers on the (OOOI)B
CdS surface were characterised by a smooth,
flat morphology. The crystallinity of the
2. Experimental CdTe and ZnTe layers was excellent as
illustrated by the RHEED pattern of figure 1.
Intrinsic (0001) - oriented CdS mechanically Pairing of the spots, however, reveals that
polished layers, supplied by Eagle-Picher with the epilayer is twinned as expected from
resistivities in the order of 5-8 n em were results achieved earlier ~ith the epitaxial
used as substrate material. The polarity of growth of CdTe onto (OOOI)B CdS [3]. The
the (0001) surfaces was identified by the slight streaking of the spots along a
supplier. The oriented CdS wafers were direction perpendicular to the shadow edge
initially boiled in C2 HGl, for 10 minutes to confirms the flat surface morphology. Layers
degrease the substrate surface. They were of CdTe/ZnTe on (OOOl)A CdS were highly
polished in a mixture of 4.8 g CrOa, 9.1 g HNO a facetted, having a polycrystalline structure
(70%) and 50 m1 of deionised water at 30°C for as evidenced by RHEED. Growth on the S-face
10 minutes as prescribed by Zhuk et al [4] . was thus' found to give superior crystallinity

570
 current 10 , The deviation from exponential
behaviour at high current levels was
attributed to the high series resistance of
the CdTe and ZnTe. This became increasingly

A
A
A
A A o 0

--
'.
IT'
Figure 1 : < 110 > RHEED pattern takeo from the CdTe/ZoTe on 0 ".
o ,"",
. 'M.

and was
(OOOI)B CdS indicating twinning within the epilayer.

therefore adopted for device

• .,...
0""
A""

fabrication.
•

Typical photovoltaic output characteristics

from an epitaxial CdS/CdTe/ZnTe device are IT'~0~tOl~~0~'~OtA~~O~I~~IO~~'~1~l~,~1'6~lilll
shown in figure 2. The overall efficiency VIV'''<I
was low, due to the high series resistance and
contact resistance losses of the cells. These Figure 3 : Forward current-voltage characteristics in the temperature
led to low values of the fill factor (-0.33) range 290·400 K.
and the short circuit current (5 rnA cm- 2 ) and
were predominately, the direct result of the
high resistivity of the p-ZnTe layers. Recent
attempts to produce conductive p-type ZnTe
films in this laboratory have been successful
producing resistivities of 10-20 0 em,
although these films have yet to be in-
corporated into the p-i-n diode structures.

60 eo 100 V(mVI
0.1
0.2
0.3
II( .6.1IlAu,·Z
0.4 VO(.O IV
O.S ff ,0 Jl
,"",
. t'K1l

0.7
xl0-6 Amps
U 0' GO 01 '0 U I, U II
V !WIll

Figure 2 : Photovoltaic output characteristies under 90 m\V crn- l

illumination. Figure 4 : Reverse current~voltage characteristics in the temperature
range 200·100 K.

The forward (If) and reverse (Ir) current

voltage characteristics for a range of
temperatures between 250 and 400 K are given pronounced as the temperature was lowered.
in figures 3 and 4 respectively. It can be Rectification ratios were typically -50:1 at
seen that over a broad range of forward 0.9(V) bias. The slopes of the forward In(I f )-
current (10- 6 10-4 A) the behaviour is V characteristics in the current regime where
exponential and beyond that (If > 10-4 A) the series resistance was not dominant (10- 6 - 10-4
curve deviates from this behaviour and rises A) were found to be constant over the
less steeply. It has been well established temperature range measured. The If-V
[5] that in heterojunctions where the current characteristics are not well described by the
transport is dominated by the tunnelling usual diode relation,
process the forward current is related to the
voltage by
(2)
If - Io (T) exp(BT) exp(AV) (1) since this would predict slopes in the In(I f )-V
curves that varied with temperature. The
where A is the tunnelling constant and B gives absence of any thermal activation in the If-V
the temperature dependence of the saturation

571
 behaviour was confirmed by the failure of smaller due to the large lattice mismatch
i!n(Ifl vs 1fT plots to display any straight between CdTe/CdS and possibly as a result of
lines. Instead the forward I-V characteristics the substrate polishing techniques leaving an
were reasonably well described by the imperfect surface for epitaxy. The value of
tunnelling expression (1), as shown by the the current, I, at a distance, d is given by
straight line in the plot of In(Io) with T
(fig. 5). This confirms that the T and V
dependence of If can be separated. A value for I(d) _Ae- d / L (3 )
B, the temperature coefficient of Ie was
estimated from the slope to be 0.061 K- 1 in where L is the minority carrier diffusion
length and
d is the distance from the junction

Figure 7 shows a plot of in I(d) vs d from

which the value of L, calculated from the
gradient was found to be -0.7 pm. This is in
agreement with the values of L of between
0.6 - 0.8 pm predicted by Awan [9] for
10" CdTe/CdS heterojunction, cells but is less
than values normally observed for bulk single
crystal CdS [10]. This reduced value of L may
-' be again attributed to the large lattice
.5
mismatch between CdS and CdTe,which will give
rise to a high dislocation density at the
interface.

~~~~)oo~~lN~~~~~U~O~~~~~~~-7.
4 N~~

""",-",0.yw. .''''
T~ur. IK)

Figure 5 : Graph oC In (I.) versus T .

agreement with values obtained by other

workers on related junctions [6] .

. An SEM EBIC line trace across the junction is

given in figure 6. From the trace it can be
seen that the maximum intensity of the signal
occurs at the CdTe/ZnTe junction. This signal
indicates that there is considerable re- -2
combination at the CdS/CdTe and CdTe/ZnTe

Figure 7 : Graph oC In I (d) versus d.

4. Conclusion

It is concluded from the analysis of the I-V

characteristics that the current-transport
across the junction is dominated by a
multistep tunnelling/recombination process.
This is in general agreement with many other
Figure 6 : EOIC line scao taken acr088 a cleaved ZoTc/CdTe/CdS
junction.
studies of CdS/CdTe heterojunctions [7,8 ] and
suggests that current transport mechanisms in
epitaxial devices behave similarly to their
polycrystalline counterparts. The CdS/CdTe
interface will contain a high density of
dislocations due to lattice and structural
interfaces since the signal does not decay to mismatch which will lead to a large density of
zero over the depletion region of the CdTe. interface states. These states support such a
The decay of the signal intensity from the transport model.
CdTe/CdS interface pro.gressing into the CdS
can be used to calculate the minority carrier From the point of view of the solar cell, the
diffusion length in the CdS. However, this principal objective is now to reduce the
cannot be calculated at the ZnTe interface resistance of the p-ZnTe. This will lead,to
since the thickness of the CdTe is less than an improved fill factor and higher efficiency
the diffusion length. The intensity from the of cell. Voitage losses can be primarily
CdS/CdTe interface would be expected to be attributed to junction recombination. In

572
 order to reduce these losses, it is proposed 5. M. Arienzo and J. J. Lafersky, J. App1.
to grow graded junctions between the CdS/CdTe Phys., 51 (1980) 3393.
and CdTe/ZnTe interfaces.
6. A. M. Mancini, A. Valentine, L.
Vasanelli and A. Losacco, J. Crystal
Growth, 72 (1985) 530.

7. K. W. Mitchell, Evaluation of the

5. References CdTe/CdS Heterojunction Solar Cell
(Garland, New York, 1979).
l. P. V. Meyers, Solar Cells, 21 (1988) 59.
8. C. Ercelebi, A. W. Brinkman, T. ·S.
2. A, Nouhi, R. J. Stirn, P. V. Meyers and Furlong and J. Woods, J. Crystal Growth,
C. H. Liu, J. Vacuum Sci. Technol., A7 101 (1990) 162-166.
(1989) 833.
9. G. R. Awan, Ph.D. Thesis, University of
3. M. Y. Simmons, P. D. Brown and K. Durham (1987).
Durose, J. Crystal Growth, 107 (1991)
664. 10. W. D. Gill and R. H. Bube, J. Appl.
Phys. 41 (1970) 1694.
4. V. A. Zhuk and V. T. Sotnikov, Phys.
Chem. Mech. Surfaces, ~ (1987) 2544.

573
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THIN FILMS OF CADMIUM TEllURIDE PRODUCED USING STACKED

ELEMENTAL LAYER (SEL) PROCESSING

M.T . Bhatti, E.P. Groarke, R.W. Miles, M.J . Carter and R. Hill
Newcastle Polytechnic, Ellison Place, Newcastle upon Tyne, NEI SST, UK

ABSTRACT. The stacked elemental layer (S.E.L.) technique has been used to synthesise
thin films of cadmium telluride (CdTe). This technique consists of depositing alternate
layers of Te/Cd in the stoichiometric ratio needed to form CdTe and then annealing the
stack to re~ct the .elements to form the. compound. In this work the Cd and Te layers
were deposited usmg thermal evaporation and the layers annealed in nitrogen at
temperatures in the range 350-5500 C for 15 minutes. Transmittance and reflectance data
X-ray diffraction data and scanning electron microscopy observations of the surfac~
topology are given for the layers produced. This technique promises to be a low cost
method for producing large areas of CdTe suitable for use in solar cell structures.

1. INTRODUCTION Corning 7059 glass substrates whereas the deposition of

tellurium films of uniform thickness and without pinholes
It is generally recognised that CdTe, with its direct was found to, be straight-forward. Good films were
optical bandgap of 1.5 eV, is a promising thin film obtained using unheated substrates (at a temperature
photovoltaic material. Because of its sharp optical "'200 C) for a Te source temperature in the range
absorption edge and large absorption coefficient, solar 220-2600 C. Increasing the substrate temperature was found
radiation with energy in excess of the energy bandgap is to produce non-uniform layers with pinholes present and
essentially all absorbed within 1-2 pm of the surface, and hence an ambient substrate temperature was used in this
a relatively short minority carrier diffusion length in the work. A tellurium source temperature of 2350C (for such
material can be tolerated.

The means that material costs are minimised

(be';ause thin layers are used) and that low cost
deposition methods can be used to produce the layers 100
(because a long minority carrier diffusion length is not
required).
90
The development of CdTe solar cells has been good
in recent years with cell efficiencies >10% for the CdTe
produced using four different methods: close spaced
sublimation [1], galvanic deposition [2,3], reactive 80
deposition [4] and screen printing [5]. The work
presented here uses S.E.L. processing to produce CdTe
thin films. This technique was originally developed in our 70
research laboratories to produce thin films of copper
indium diselenide [6] . This technique is simple, and it ~
w
promises to be a low cost method to produce large areas u
z 60
of good quality thin film materials for use in solar cell c(
devices. ~
~
z 50
<f)

c(
2. EXPERIMENTAL PROCEDURE a:
I-
The evaporations were performed in an oil pumped 40
vacuum system operating in the 10- 5 to 10- 7 Torr range
of vacuum pressures and the source heaters were of the
type described by [7]. Two shutters were used, one close 30
to the substrate to prevent deposition onto the substrate
and the other shutter just above the the quartz bottles
containing the cadmium and tellurium could be adjusted
20
to cover both the sources or just one source to prevent
either cadmium or tellurium vapour or both reaching the
substrate. For all the work, heated Corning 7059 glass
slides were used as substrates. Transmittance and 10
reflectance measurements were made using a Cary 17-D
spectrophotometer and the X-ray diffraction data was
obtained using a X-ray diffractometer with a Cu KO' 0
X-ray source. 600 1000 1400 1800
WAVELENGTH (nm)

3. RESULTS Figure 1. Transmittance versus wavelength for

an unannealed stack and for the
It was found to be difficult to evaporate stack annealed in nitrogen at 3500C
homogeneous, uniform films of cadmium directly onto and 45(}oC

574
 10' ~------------------------' 270

240

'E 5
~ 210
~
Z
W
U 180
ii:
LL
W
0
U
">...
E
150
<I>
Z
0
F
...
'0
IL
N 120
.c
~
II:
0
VI HI" ~
III 90
0(

60

5
30

0 •
0.4 0.5 0.6 0.7 0.8 0.9 0.8 1.0 1.2 1.4 2.0 2.2
WAVELENGTH (microns) hI' leV)

Figure 2. Ci versus hv for a care sample

produced by annealing a CdlTe stack Figure 3. (ah.) 2 versus hv far a care
at 4500 C in nitrogen sample produced by annealing a
CdlTe stack at 45(}oC in nitrogen

a substrate temperature) was adopted so that the quartz glass substrates with further optimisation of the annealing
crystal monitor in the system could be calibrated to conditions or perhaps annealing after CdTe synthesis may
permit controlled deposition of the tellurium. The improve the grain size. X-ray diffraction data for a
deposition rate for these conditions was "'10 A/sec. synthesised CdTe layer is given in Figure 5. The data is
typical of the layers produced to date. The diffraction
Deposition of the cadmium onto such tellurium layers peaks are as expected for CdTe although a small amount
resulted in uniform films free from pinholes, for cadmium of excess Te appears to be present in the layers. It is
source temperatures in the range 120-150OC. A standard hoped that increasing the ratio of cadmium/tellurium in
cadmium source temperature of l350C was adopted and the stacks will eliminate this problem.
the quartz crystal monitor calibrated to permit controlled
deposition of the cadmium. The deposition rate of the
cadmium for these conditions was 10 A/sec. Alternate 4. CONCLUSION
tellurium/cadmium stacks were deposited using the
optimised conditions such that the relative thickness of the CdTe thin films have been produced by annealing
layers were in the stoichiometric ratio (6:5) and these repeated Cd/Te stacks in nitrogen. The initial results are
stacks annealed in nitrogen to react the elements to promising: the layers have good absorption edges and the
synthesise cadmium telluride. Results from annealing a 6 technique has proved to be simple and reproducible.
layer stack with a cadmium thickness of 700 A and a
tellurium thickness of 600 A are given in Figure 1. The
annealing conditions were optimised by annealing the stack
for 15 minutes for temperatures in the range 350-5500 C
starting with· the lower temperature and then annealing at
500 C intervals until a good absorption edge was obtained.
Annealing at too high a temperature >500OC was found
to degrade the maximum transmittance and the absorption
edge and an annealing temperature of 4500 C (for an
annealing time of 15 mins) was considered to be
optimum.

Both transmittance and reflectance measurements on

optimised samples were used to determine the absorption
coefficient, Ci, in the wavelength range 600-2000 nm. A
plot of Ci versus X is given in Figure 2 and a plot of
(ab.» versus h. in Figure 3. The latter plot gives a
straight line with an intercept of 1.5 e V on the h. axis,
i.e. the energy bandgap of the CdTe, and is consistent
with that expected for a successfully synthesised CdTe
layer.

A secondary electron micrograph of the CdTe

produced is given in Figure 4. The grain size is small
with a typical grain diameter of 0.1 pm. Using many
repea ted stacks to increase the total thickness to >1 pm
and then annealing did not produce an appreciable Figure 4. Secondary electron micrograph of a
increase in the grain size. It is hoped that CdTe synthesis CdTe layer produced by annealing a
on CdS coated glass substrates instead of Corning 7059 CdlTe stack at 45(}oC in nitrogen

575
 Cdf.
(111)

(FWHMJ
_ 1'""0.4'

edTe
(220)
edTe
(311)

_ ~0. 4·

_ ~O .25·

fe
A(l")

50 40 30 20

2-THETA (degr ..... )

Figure 5. X-ray diffraction data for care

produced by annealing a Cd/Te stack
at 45{)oC in nitrogen

Further optimisation of the technique should

eliminate the small amount of excess tellurium. in the
layer and increase the grain size. Further work will also
investigate the changes in the layers produced by
annealing in air, vacuum and in a reducing environment.
Annealing in the former environment could produce
p-type material [8] which could enable the CdTe to be
used directly in hetero junction solar cells. This technique
promises to be a low cost method to produce large areas
of CdTe for use in n-i-p and heterojunction solar cells.

S. REFERENCES

1. Y.S. Tyan and E.A. Perez-A1buerne, Proc. of the

16th PhotovoItaic Specialists Conference, San Diego,
(1982), p. 794
2. B.M. Basol, Solar Cells, 23, (1988), p. 69
3. P.Y. Meyers, Proc. of the 7th E.C. Photovoltaic
Solar Energy Conference, Seville, (1986), p. 1211
4. T .L. Chu, Solar Cells, 23, (1988), p. 31
5. S. Ikegami, Solar Cells, 23, (1988), p. 89
6. M.J. Carter, I. I'Anson, A. Knowles, H. Oumous
and R. Hill, Proc. of the 19th IEEE Photovoltaic
Specialists Conference, New Orleans, (1987), p. 1275
7. S. Arshed, PhD Thesis, Newcastle Polytechnic, UK,
1990
8. R.W. Birkmire, B.E.M. Candless and W.N.
Shafarman, Solar Cells, 23, (1988), p. 115

[This work is supported by the Commission of the

European Communities under Contract No. JOUR-0065-C
(MB)]

576
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

Screenprinted CdS-CdTe Solar Cells

I. Clemminck, M. Burgelman, A. Vervaet and J. De Poorter
Rijksuniversiteit Gent, Laboratorium voor Elektronika
St.-Pietersnieuwstraat 41, B-9000 Gent, Belgium
tel.: (32)91233821/ext.2522j fax: (32)91332219

The influence of different preparation conditions on the properties of screenprinted and sintered CdS-CdTe solar cells
was investigated. The use of CdS base layers of different thickness and microstructure and the sintering conditions
of the CdTe layer influence Voe and spectral responses. The addition of copper to the graphite electrode on CdTe
can improve V 00, J,e and FF. The interdifFusion of CdS and CdTe was studied to explain some spectral response
features. 5.5 % intrinsic efficiency was obtained on a .56 cm2 cell, higher efficiencies should be obtainable ifthe sheet
resistance of thin CdS layers could be lowered.

propanediol. An equimolar mixture of 5N Cd and Te powders is

1 INTRODUCTION
ground in an agate ball mill, under H20 for 24 hours, to obtain
The preparation of CdS-CdTe thin film cells, using a screenprint- a grain size of ~ 1 JIm. Cd+Te pastes are prepared from this
ing and sintering technology has already shown its potential as a mixture with CdCl 2 powder and 1,2-propanediol. Screenprinting
low-cost production technique for high efficiency and stable solar of the pastes is done using 80 or 200 mesh stainless steel screens.
cells. Ikegami et al. [1-3] and 1m et al. [4] reported the produc- The prints are dried prior to sintering at 150° C during 1 hour.
tion of all-screenprinted CdS-CdTe solar cells with an intrin~ic The sintering is performed in a belt type furnace, under N2
efficiency on small area of 12.8 % and 11.2 % respectively. Also atmosphere at teID.peratures ranging from 575°C to 700°C. Dur-
1m et al. reported a screenprinted CdZnS-CdTe solar cell with ing sintering the CdCh is evaporated from the film. In order to
an intrinsic efficiency of 12.5 % [5]. In this paper we present the control the CdCl 2 evaporation rate, the sintering is performed
latest developments in our research on screenprinted and sintered in an alumina case with a cover containing a variable number of
CdS-CdTe solar cells. The influence on CdS-CdTe cell proper- Imm diameter holes. Prior to Cd+Te deposition, all CdS films
ties of different preparation conditions, such as the use of CdS are resintered for 30' at 650°C, placed open on the transport belt
base layers of different thickness, Cd Te sintering conditions and in order to remove any residual CdCh (all time indications refer-
heat treatment conditions of the C:Cu electrode on CdTe were ring to cycle time in the belt furnace). A colloidal graphite paste
investigated. The interdiffusion of CdS and CdTe was studied to (Aquadag), containing a small amount of CuCh as a source of
explain ~ome spectral response features. copper impurities is painted on top of the CdTe to act as an elec-
trode and is heat treated in N2 atmosphere. Contact to the CdS
2 CELL TECHNOLOGY layer is made by an In-Ga alloy.

Cells are manufactured on 2" x 2" Corning 7059 substrates by

3 RESULTS AND DISCUSSION
consecutive deposition of CdS and CdTe layers, each deposition
step consisting of screenprinting (of CdS and Cd+ Te pastes re- S.l CdS BASE LAYERS
spectively) and subsequent sintering with the aid of CdCl2 as a
Since the CdS layer is to act as a conductive window, it should
flux. CdTe patterning is done as to obtain unit cells with a CdTe
have both high optical transmission and low sheet resistance.
area of 5 x 9 mm2 •
Also, the CdS microstructure should form a solid substrate for
CdS pastes are prepared from 5N CdS powder with an av-
the deposition of a CdTe layer on top [6].
erage grain size of ~ 1.5 JIm, anhydrous CdCl2 powder and 1,2-
We prepared thick and thin CdS layers by performing the
screenprinting with 80 mesh and 200 mesh screens respectively.
A In order to get a good microstructure, the CdS sintering condi-
tions (Le. initial flux content in the paste and flux evaporation
control) had to be adjusted to the deposited thickness. Some
relevant properties of both types of CdS layers are listed in Table
I.
SEM photographs (Figures lA and lB) show grain sizes up
to 20 JIm for both thick and thin CdS layers. A significant mi-
crostructural difference is that thick layers have multiple grains
in the thickness dimension, whereas thin layers are essentially
single-grained and have on the whole a more compact structure.
At wavelengths, for which the CdS is transparent (>. > 520nm),
B the window layer transmission will be limited by diffuse scatter-
ing from grain boundaries and porosities in the film (reflective
losses). Measured optical transmissions (diffuse + specular) at
750 nm are 40% and 53 % for thick and thin CdS layers on glass
respectively, which reflects the difference in microstructure. The
actual transmission into the absorber in the sintered CdS-Cd Te
structure will be larger than these measured values, because a
CdTe layer in contact with the CdS will significantly reduce re-
flections from the CdS backplane.

Figure 1 - SEM photographs of the cross section of 80 Mesh (A) and 200
Mesh (B) sintered CdS films.

577
 Sintering ParlLllleters I Thickness I Transmission (750nm) Sheet Resistance
11 wt% CdCb, 650·C, 4 holes, 60' 40 % 100 11 - 1kl1
16 wt% CdC!" 70g·C, 0 holes, 60' 53 % 600 0 - 10 kO

Table I-Properties of thick and thin CdS layers: optimum sintering parameters (initial CdC!, content, number of holes, sintering
time), sintered thickness, optical transmission at 750 nm (diffuse + specular), sheet resistance.

8.2 CdTe SINTERING CONDITIONS A

In order to study the influence of CdTe sintering conditions on
the properties of CdS-CdTe cells, Cd+Te pastes containing 1
wt% and 4 wt% (solid fraction) of CdCh were prepared and
screenprinted on both thick and thin CdS layers using a 200 mesh
screen. The composite structures were sintered for one hour in
an alumina case with a 9-hole cover at temperatures of 575°C,
625°C and 675°C.

CdTe Microstructure
It is seen from X-ray diffractions of sintered layers on glass that
pure CdTe of cubic modification is synthesized from the elemental B
powders during sintering, with no other phases remaining in the
film.
Figures 2A and 2B show SEM photographs of the surface of
CdTe films which contained 4 wt% and lwt% of CdCl2 before
sintering and were sintered at 675°C. The microstructure of both
types of layers is on the whole inferior to the CdS microstructure,
showing smaller grainsizes and more porosity. CdTe layers that
contained 4 wt% of CdCh are relatively dense with 5 - 10 /Lm
grains and have essentially the same microstructure for sintering
temperatures between 575°C and 675°C. The microstructure of
CdTe layers that contained lwt% of CdCh is on the whole worse, Figure 2 - SEM photographs of the surface of 200 Mesh CdTe layers
with porosity increasing and grain sizes decreasing on lowering which contained 4 wt% (A) and 1 wt% (B) of CdCI 2 before sintering and
the sintering temperature (":l 5 /Lm at 675°C, ":l 2 !lm at 575°C). were sintered at 675°C.

Global cell parameters A comparison of Voc values shows that cells prepared on thin
Table II lists the properties of CdS-CdTe cells as a result of dif- CdS layers always show higher Voc than the corresponding ones
ferent preparation conditions. on thick CdS. The highest values of Voc are to be found for cells
Apart from the cells that were prepared on thick CdS layers on thin CdS layers with CdTe layers that contained 1 wt% of
and sintered at high temperatures (cells 2,3,6), all cells suffer CdCl 2 (cells 10,11,12) and for cells sintered at 575°C (cells '4,7) .
from a large series resistance which limits J.e and FF. J. e of cells There is a tendency for high Voc to coincide with high CdS sheet
2,3,6, although not obstructed by series resistance is low as well, resistance. Typical values for the dark diode parameters mea-
as a result of low h. It can be seen that the series resistance sured on cells with good Voc are 3.10- 10 A/cm 2 < J o < 1.10-9
arises from the sheet resistance of the CdS layer. The sheet A/cm 2 and 1.6 < n < 1.9.
resistance of as deposited CdS layers was, for thin layers in any
case, for thick layers in most cases, higher than an acceptable Spectral response features
value of 100 0 (Table I). Ro is however reduced during CdTe
sintering as a result of the interaction between CdS and CdTe Normalized spectral response curves for different CdS-CdTe cells
or between CdS and the CdCh in the CdTe layer. Acceptable are ' shown in Figures 3A and 3B for CdTe sintering at 675°C
values can however only be obtained on thick CdS layers and for and 575°C respectively. Spectral response is nearly panchromatic
sintering temperatures ~ 625°C (30 010 is about as low as we between the bandgaps of window and absorber materials for the
could go).
CdS Layer CdTe Layer I Voe J .. Re CdS SR type SR plateau
CdC!, (wt'70) I Ts (DC) I (mV) (rnA/crn 2 ) (0) (%)

4%
575
625
1
2
650
600 - 630
5- 8
9 - 11
• 1k
60 - 65
A-B
C
40
40
80 MESH 675 3 600 11 - 13 70 -90 C 50
(THICK) 575 4 740 9 - 13 0 400 - 500 A-B 60
1% 625 5 650 - 670 9 - 13 0 100 A-B 55
675 6 540 - 600 12 - 13 100 B 50

4%
575
625
7
8
700
670
2- 3
13 - 15 0
• 3k-lOk
300 - 500
A
B
< 30 0
50
200 MESH 675 9 650 12 - 15 0 300 - 500 B-C 60
(THIN)
1%
575
625
10
11
760
720
2- 3
6 - 11 •
• 3k- 5k
650 - 800
A
A
55
73
0

675 12 700 5- 6 • 2k A 65

Table II- Properties of sintered CdS-CdTe solar cells: Voc and Joe are measured under 100 rnW /crn 2 Xenon arc illumination.
CdS sheet resistance is measured in situ, as the resistance of 1 square of CdS between t.wo line contacts on either side of a unit
cell, under illumination. Spectral response types refer to the shapes of the curves in Figure 3A. Spectral response plateau was
measured as the absolute quantum efficiency ..t 750 nm (values listed ..re the best values of 5 cells on the same substrate). No
bias light and zero bias voltage were used in spectral response measurements. Cells marked (0) or .(.) in the J. e column have J"
which are appreciably, resp . heavily (J. e = Voc / Rs) affected by series resistance. Cells marked (0) in the SR plateau column
may have values of absolute quantum efficiency affected by series resistance.

578
 3.3 INTERDIFFUSION OF CdS AND CdTe
In order to clarify spectral response features in the short and long
wavelength regions, we investigated the optical and crystallographi
chemical properties of window and absorber layers after sintering
of the CdTe on top of the CdS base layer. We stripped the CdTe
layer of parts of the sintered CdS-CdTe structure, by means of
etching in K2Cr207: H2S04 solution, to reveal the properties of
the underlying CdS window layer.
500 600 700 !DO Figure 4 shows the optical transmission curves of the window
wAvnENVTH (""')
layers that gave rise to the spectral responses depicted in Figure
3A. It can be seen that the optical transmission shows qualita-
..
~roo
~ tively the same short wavelength features as the spectral response
~ 40 curves, which suggests that a poor short wavelength spectral re-
sponse is at least in part due to a deterioration of the optical
::,~ 60 properties of the window layer.
~ Figure 5 shows the optical transmission of a complete CdS-
~ '0 CdTe structure (A), depicting the absorption edge of a CdTe
~ layer that was sintered on a CdS base layer, as compared to a
!':! 10
~ CdTe layer that was sintered on bare glass (B). The absorption
~ 0 edge of the CdTe that was sintered on CdS seems to shift some
30 -40 nm towards longer wavelengths. Essentially the same ab-
W.AVEl..~IICrH In,")
sorption edge was seen for all structures in Table II and this
corresponds to an extended long wavelength spectral response
which was measured in all cells.

Figure 3 - Normalized quantum efficiency of CdS-CdTe cells in which

the CdTe layer was sintered at 67SoC (A) and S75°C (B) - A : thin CdS
layer, lwt% of CdCl1 in CdTe, B : thick CdS layer, 1 wt% of CdCl2 in
CdTe, C : thick CdS layer, 4 wt% of CdCl2 in CdTe.

cells prepared on thin CdS layers with CdTe layers that contained
1 wt% (all temperatures) or 4 wt% (575°C only) of CdCh. All
other cells show an attenuation of the short-wavelength (520 nm
< >. < 700 nm) spectral response, which is more pronounced for
thick CdS layers, higher sintering temperatures and high CdCh
content in the CdTe. A poor short wavelength spectral response
in sintered CdS-CdTe cells was reported earlier in literature [3,4]
The loss in short wavelength spectral response of curve 3A- C as 100 !DO !/GO

. compared to 3B- A e.g. would introduce a factor of ~ 0.70 in J. e WA.VELENGTH (nmJ

on illumination with AM 1.5.

The long wavelength spectral response extends to about 900 Figure 4 - Optical transmission of CdS window layers from sintered CdS-
CdTe structures, after the CdTe layer was stripped. The curves refer to
nm, which is about 35 to 40 nm further than what is normally
the specimens of Figure 3A.
encountered in CdTe solar cells. This phenomenon was also
mentioned earlier in literature for screenprinted CdS-CdTe cells
IOO,-------------------~~====~==~~
[1]. The long wavelength spectral response edge is always quite
abrupt, which indicates that the minority carrier diffusion length
in sintered CdTe is sufficiently large.
The absolute values of quantum efficiency, corresponding to
the spectral response plateau are always larger for CdS-CdTe cells
prepared on thin CdS layers as compared to the corresponding
ones on thick CdS layers (apart maybe from cells 7,10 in which
10
the spectral response current itself is limited by a large CdS sheet
resistance). This is at least in part due to a higher window layer
O+-----r---~----,_----r_--~----~
transmission associated with the microstructure of thin CdS lay- 900 HOO
ers (Up to now we could observe the difference in reflectivity from
CdS-CdTe structures with thick and thin CdS layers only in a
qualitative way). Again, the highest values of absolute quantum Figure 5 - A : Optical transmission of a complete CdS-CdTe structure,
efficiency are to be found for cells prepared on thin CdS layers prepared on a thick CdS layer with a CdTe layer that contained 1 wt%
with CdTe layers that contained 1 wt% of CdCh before sintering. of CdCl 2 and was sintered at S75°C. - B : Optical transmission of a CdTe
layer sintered in the same circumstances on bare glass.
It can be concluded that both Voc and spectral response con-
siderations favour cells prepared on thin CdS layers with Cd Te The spectral response features meantioned have been associ-
layers which contained 1 wt% of CdCh (575°C sintering not ev- ated in literature [3,4] with the interdiffusion of S and Te species
idenced), although the cells sintered at 575°C (thick CdS, 1wt% during the sintering of CdTe on CdS, in the presence of CdClz.
CdCh and thin CdS, 4wt% CdCI 2 ) should not be excluded. These Indeed, the CdS-CdTe equilibrium phase diagram [7] predicts a
conditions however, were seen to be incompatible with a low se- limited intermixing of CdS and Cd Te at the sintering tempera-
ries resistance for cells prepared on CdS layers that had too high tures, and the coexistence of 2 solid solution phases CdS 1_ r Te",
a sheet resistance in the as deposited state. and CdTel_ySy. Figure 6A shows the X-ray diffractogram of
Our best cell up to now was made on thick CdS with CdTe the surface of the CdTe layer in a CdS-CdTe structure that
that contained 4.5 wt% of CdCh and has 5.5% intrinsic effficiency was prepared with a CdTelayer that contained 1 wt% of CdCl 2
on a .56 cm 2 surface (Voc = 670 mY, J. e = 16.4 mA/cm 2 , FF = and was sintered at 675°C. Figme 6E shows the X-ray diffrac-
0.5). togram of the surface of the underlying CdS layer after edTe

579
 stripping. The existence of CdS1_zTez and CdTel_ySy solid so-
lution phases is evidenced by shifts from the original CdS and .7.---------------------,
CdTe peak positions. The surface of the absorber layer is sin- .6
gle phase CdTcl_ySy, whereas both pure CdS and CdS1_zTez
solid solution phases are present in the underlying window layer
. Values of x and y derived from the peak shifts l7J (x ~ u.on) .t
, y ~ 0.15) are only slightly smaller than what is predicted by
the CdS-CdTe phase diagram at 675°C (x = 0.10, Y = 0.18). .3 '--2::0!:::O:----30
=O:----t~00
~---;5~0:;;'0--'
EDAX microanalysis evidenced the presence of S throughout the 16r---------------------,
absorber layer in this specimen, with no appreciable change in 14
concentration as a function of depth. On the other hand no Te
can be detected in the bulk of the CdS grains. These results sup-
port the picture of a homogenious CdTel_ySy absorber layer and
a window layer consisting of pure CdS and a (thin) CdS1_zTez
solution layer at the interface and/or the CdS grain boundaries
in sintered CdS-CdTe structures that were made with CdTe that
contained 4 wt% of CdCI 2 • 200 300 400 500
The absorption edge of both CdS1_zTe., and CdTel_ySy shifts
towards longer wavelengths for moderate values of x and y (a min- .61-
imum in bandgap energy of 1.41 eV is found for x = 0.75, Y =
",- .51-
0.25) [8] The absorption edge of CdSo.92Teo.os and CdTeo.ssSo.15 "'-
e.g. would be shifted 80 nm and 50 nm respectively, with re- .41-
spect to pure CdS and CdTe. The compositional picture drawn
from XRD and EDAX analysis is therefore capable to explain in . 3~~20~0----3~0~0----4~0~0----5~00
--~
a qualitative way the measured optical properties of window and HEAT TREATMENT TEMPERATURE ( OC)
absorber layers and hence the short and long wavelength spectral
response features of the cell.
Figure 7 - V00, J,e and FF of sintered CdS-CdTe solar cells in which the
graphite electrode was heat treated. for 30' in N2, as a function of heat
treatment temperature; for graphite pastes with no intentional doping
(0) and pastes that were doped with 200 jlg/g (t» and 1000 jlg/g (D)
of Cu. CdS-CdTe structures were prepared on thick CdS layers with a
CdTe layer that contained 4 wt% of CdCl 2 and was sintered at 67SoC.

of V00 are found for a heat treatment at 200°C, although there

is some loss in FF at this temperature. Joe was low for heat
treatment temperatures of 200°C and 300°C in these specimens,
but could be increased to ~ 14 mA/cm 2 (1000 j.Lg/g copper paste)
by increasing the heat treatment period, without a loss in V00 or
FF.
These results suggest that copper, which acts as a p- type
dopant to CdTe, diffusing into the CdTe during the heat treat-
ment of the carbon electrode is significantly altering the semicon-
"'" ductor properties, even in the vicinity of the junction, affecting
® such properties as collection of light generated current and Voc
I'0OI
buildup.

fo~ I"" ACKNOWLEDGEMENTS

(n1) We acknowledge the support of the Belgian Fund for Scientific
fo" Research (N.F.W.O.)(M.B., research associate). This work was

t.rrr- .In
IIo<l

supported by the Commission of the European Communities

r~. - - - ......~ (JOULE program).

./

" » ,.
REFERENCES
1. S. Ikegarni, Solar Cells, 23( 1988), 89
Figure 6 - X-ray diffractograms of a CdS-CdTe structure, prepared on
a thick CdS layer with a CdTe layer that contained 4 wt% of CdCl2 2. N. Suyama, T. Arita, Y. Nishiyama, N. Ueno, S. Kitamura
and was sintered at 675°Cj of the surface of the CdTe layer (A)j of the
and M. Murozono, 21st IEEE PVSC, Kissimee(1990), 498
surface of the underlying CdS layer, after CdTe stripping. Diffraction
peaks of pure CdS and CdTe are shown for reference. Peaks marked (.) 3. H. Uda, H. Matsumoto, Y. Komatsu, A. Nakano and S.
correspond to solid solution phases. Ikegarni ,16th IEEE PVSC, San Diego(1982), 801

4. Y. K. Yun and H. B. 1m, J. Electrochem. Soc., 135 (1988),

3.4 C:Cu ELECTRODE 1658
Figure 7 depicts V00, Joe and FF of sintered CdS-CdTe solar cells 5. Y. S. Seol and H. B. 1m, 9th E.C. Photovoltaic Solar Energy
as a function of the heat treatment temperature of the graphite Conference, Freiburg(1989), 294
electrode on CdTe, for graphite pastes with different amounts of
copper impurity (in the form of CuCh) added. It is seen that 6. I. Clemminck, A. Vervaet, M. Burgelman and A. Demeyere
for all heat treatment temperatures the presence of copper in the ,20th IEEE PVSC, Las Vegas(1988), 1585
graphite paste significantly enhances both VDC and J 'c, and in 7. K. Ohata, J. Saraie and T. Tanaka, Jap. J. Appl. Phys.,
most cases also FF, with the largest effect found for the graphite 1.2(1973), 1198
paste that contained 1000 j.Lg/g of copper. The highest values
8. K. Ohata, J. Saraie and T. Tanaka, Jap. J. Appl. Phys.,
12(1973), 1641
580
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOVOLTAlC PROPERTIES OF n-CdTe/p-Te JUNCTIONS

I. YOUMab, M. CADENEb,D. LlNCOTc and D. LAPLAZEa

a. Laboratoire des Semiconducteurs et d'Energie Solaire. Faculte des

Sci.ences. Unive~sitl! Cheikh Anta Diop. Dakar (Sl!nl!gal)
b. Laboratolre des Matl!naux Photoconducteurs. Universitl! de Montpellier
11- U.S.T.L. 34095 Montpellier Cedex (France)
c. Laboratoire d'Electrochimie Analytique et Appliqul!e. E.N.S.C.P. 75005
Paris (France)

ABSTRACT

We describe the fabrication by photo electrochemical reaction of CdTe

junc:t~on~. Elec~ri~al and photovoltaic properties of this devices are also reported. The
rectifl~atlon ratio IS of the order of 10 4 and the ideality factor in the range 1.5 - 1.9. The
~apa~lta~ce measurements show that the barrier height is around 1.2 eV. Under white
Illumination (65 mw/cm2) the open circuit voltage is close to 400mV and short circuit
current is 0.5 mAlcrn2 . Spectral response indicates than the thin layer of tellurium acts
only as an optical filter and that active material is CdTe.

I-INTRODUCTION

Cadmium telluride is one of the best is first determined by anodic scanning. For correct
semiconductor compound for photovoltaic solar value of this potential, and under illumination, the last
energy conversion on account of its energy gap reaction produces a thin layer of polycrystalline
(Eg= 1.47 eV) which is well matched with solar tellurium onto the surface of CdTe crystal. The
spectrum. Photoconversion devices can be made, thickness (0,2-0,8 pm) of this layer is dependant of
either from single cristalline CdTe samples obtained the reaction time, we have chose between 3 and 8
by Bridgman method or from thin films, deposited by minuts. Ponctual top contact are made with silver
electrochemical process. This semiconductor can be epoxy or electrodag paste.
oxydised by photo-electrochemical reaction,
producing a thin layer of tellurium which is a small
gap p-type material (Eg= 0,32 eV). Using such III-EXPERIMENTAL
reaction, its seems easy to elaborate CdTe/Te
junctions. Fabrication and new results about 1) Dark I-V characteristics
photovoltaic properties of such devices are reported
Analysis of I-V characteristics of these devices
in this paper.
in the dark indicates a very low saturation current
density in the range of 10-8 Alcm 2 . The rectification
ratio at 1 V is of the order of 103 to 10 4 • A typical
diode characteristic is shown in figure 1. The forward
current varies exponentially with voltage over four
11- DEVICE FABRICATION order of magnitude according to the well known
relation:
All the CdTe/Te junctions, used in t his
investigation, are fabricated from n type single J = Js Exp (9 (V - Rs J) - 1)
cristals. Ohmic back contacts are obtained using nkT
indium alloy on which external wire are attached with
silver epoxy. Samples are then placed in a one were Js is the saturation current density, Rs the serie
compartment electrochemical cell, with standard resistance and n the ideality factor, which is obtained
calomel electrode, and as counter electrode, either from the slope of the curve Log(J) versus V . Its values
platinium wire or grid. Electrolytic solution is made vary between 1.5 to 1.9 for the studied junctions.
with propylene carbonate and tetrabutyl ammonium Deviations from exponential behaviour for higher
hexafluorophosphate. The value of electrode voltage than 0.56 V are due to serie resistance effect,
potential giving rise to the following decomposition the calculated values of Rs been around few ten
reaction: ohms by cm2 . We can notice also than reverse
current do not saturate at Js value indicating that
CdT e + 2 h+ --. Cd 2++ Te (Solid) complex current mechanisms occur in the junction.

581
 2) Capacitance measurements

The C-V measurements in the dark have been

J (A) made in the frequency range 10-100 KHz. A set of
-4
10
1/C2 -V plot obtained for a typical diode is shown in
figure 2, which exhibit a linear variation . The slopes
of these curves are frequency dependant and their
gradual increase can be explained by excitation of
traps in the bulk of CdTe crystal as they are a function
of Nd + Nt, Nt been the trap density whose energy is
equal to Fermi level in the space charge region (1).
From these . measurements, the uncompensated
-5
10 donor concentration can be estimated around
8.1016/ em 3 in good agreement with values found by
electrochemical measurements (2). The intercept of
these curves with V axis gives a barrier height equal
to 1.2 eV, which is very close to the values obtained
for CdTe homojunction (3)

3) Photovoltaic properties

USing frontwall illumination, devices show an

open circuit voltage Veo = 400 mV, a short circuit
current density J se= 0.5 -0.8 mAlcrr12- and a fill factor
FF= 0.54 for a light flux of 65 mW/crrI2-· The low value
of J se can be attributed to the weak collection
efficiency of carriers by the ponctual top contact and
the strong absorption of the Te layer. Classical
600 800 WmV) variation of J se and Veo are observed with incident
light intensity.
The spectral response have been obtained using
Jobin-Yvon monochromator supplied with chopped
Figure 1: Dark I-V characteristics at
iodine lamp and a typical curve is shown in figure 4.
different temperature.
All spectra exhibit a gradual increase from short
wavelenghts to a cutt off wavelenght near 0.83 ttm
which corresponds to the value of CdTe energy gap.
These spectra show that the active part of the cell is
CdTe material and the tellurium layer acts only as an
optical filter. On other hand, the lowering of spectral
response for short wavelenghts indicate that
collection efficiency in space charge region is
reduced (4,5).

c- 2 ( 10 1O F- z )

.
s..
III 1000

· , '"
• iQl.Hi
J mA/cm 2 .
1.0
• 'oow
30 '

'. '. ., 0,5

-4-
20
DD

ClOQ:~: ::-.
10 ' 0 • ••
D.

0
100 300 mV
-1 .5 -1 - 0.5 0 0.5 V

-0.5
Figure 2 : Variation of junction
capaCitance with applied voltage

Figure 3: I-V characteristics under

illumination

582
 modelised by an homojunction. Such transformation
10 r---------------------------~ has been found by electro reflectance measurements
R 5 (u a) (8).

0,8 We have shown that it is possible to realise easily

CdTe/Te junctions with very interesting photovoltaic
0,6 properties. Actual photovoltaic performances are
affected by the low value of Jsc and efforts are now in
progress to improve this value by optimization of Te
0,4 thickness and realisation of better front contact. On
other hand, studies of junctions made with thermal
evaporation of tellurium on CdTe crystal can help us
0,2 to confirm our model.

0,4 0,6 0,8

V- REFERENCES

Figure 4: Spectral response. (1) E.H. RHODERICK, Proc. Inst. Electr. Eng.,
Part1, 129, (1982) 1.
(2) D. LlNCOT, Private communication.
(3) G. COHEN SOLAL, D. LlNCOT, M. BARBE,
4th E.C. Photovoltaic Solar Energy
Conference
(4) M.A. GREEN, J. Applied Physic 50 ,(1979),
IV- CONCLUSION 1116
(5) M. KANE, G.w. COHEN SOLAL, D. LAPLAZE
All these results, and particularly the values of G. COHEN SOLAL, Solar Cells 15 (1985)
barrier height obtained from C-V measurements 171
indicate that these devices cannot be modelised by (6) S.M. SZE Physics of Semiconductor Devices
Schottky diode. Using this last model and taking into (Wiley) New York, 1981.
account electronic affinities of the two materials (7) E.H. RHODERICK Metal Semiconductor
(XCdTe = 4.28 eV; XTe = 4.76 eV) we must have a built Contacts Clarendon, Oxford (1978).
in potential around 0.5 eV which is very different from (8) F. VASQUEZ-LOPEZ, J.S. SANCHEZ
measured values (6,7). We must consider that during SINENCIO, HELMAN J.L., PENAA.
Te layer formation, migration of Cd2... ions, and donor LASTRAS-MARTINEZ T.M., RACCAH M.
impurity, lead to the formation of a interfacial p or p+ and TRIBOULET R.J.
CdTe layer, so that the junction can be best J. Applied Phys. 50 (1979) 53, 91 .

583
10TH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

VARIATIONS IN THE RESISTIVITY OF SPRAYED FILMS OF CADMIUM SULPHIDE

R W Buckley B Cotier K Dixon T Gallagher N Goddard

Twyford Church of England High School

Twyford Crescent Acton
London W3 9PP UK

ABSTRACT. The way in which the resistivity of cadmium sulphide varies

with thickness has been investigated. The films were deposited on glass
substrates using a modified paint spray, the spraying solution being a
mixture of cadmium chloride and thiourea. This is an important step in
the.fabr~cation of thin film ~dS - CdTe solar cells. Variations in dark
reSIstIvIty over the range 10 _10 7 ncm were observed and this variation
is interpreted in terms of an increase in the excess cadmium content of
the film with time.

1. Introduction 3. Experimental results

The electrical properties of evaporated Figure 1 shows the resistivities of depos-
films of cadmium sulphide have been investigated ited films as a function of thickness. Each
extensively in the past. One of the first point plotted represents the mean of at least
suggestions that the resistivity of an evaporat- six films of the same thickness. The curve
ed film of CdS might vary with thickness was shows that the resistivity of the films varied
made by Berger et al (I). Following this, from 2 x 10ncm at a thickness of 0.15pm to 10Gcm
Wilson and Woods (2) in an extensive series of at 1.47pm.
experiments using an open evaporation system
observed a dependence of resistivity on thick-
ness which was explained in terms of an in-
creasing deviation from stoichiometry of the
source material as the evaporation proceeded.
They proposed a tentative model to explain this
process which was based on considerations of
dynamic equilibrium. This theory made use of
the general shape of the CdS phase diagram as
proposed for example by Shiozawa et al (3) and
predicts that the film will become richer in 10'
cadmium during the course of the deposition. \
In the light of this situation we have been
concerned to investigate the electrical resis-
10'
\
tivity and Hall mobility of films which would be '5
.
10'
suitable for use in the preparation of solar 9 \ ", A
cells. Although we used an entirely different
deposition method similar results have been
obtained. "
~
::
[0'
~---- '~.
Q
~ [0'
\
2. Experimental details. [0'
CdS films have been prepared by spray
pyrolysis. Despite the great potential of this 10
method for producing cheap cells for terrestrial
applications very few attempts have been made by
other workers to make cells in this way. 05 to
The apparatus we used for this process film tI'uckn t n ~)
conSIsted of a modified paint spray available Figure I. The resistivity or films or CdS u a runClioa or their thickness.
fro m mas t laboratory suppliers. Tin oxide
coated glass was used as a substrate, as this
forms an ohmic contact to cadmium sulphide and
it is transparent, The substrate was heated to
220'C. This imp£oved the adhesion and structur-
al quality of the films. The spray solution was
a mixture of 5g of thiourea (which provides
sulphur ions ) and 5g of cadmium chloride ~pro­
vIding cadmium ions) dissolved in 250cm of
distilled water. These compounds react on the The way in which the Hall co-efficient of
substrate to form cadmium sulphide. The spray- the different films varied with thickness is
ing was carried out in 5 to 10 second intervals shown in figure 2. The Hall coefficient of the
to allow the substrate to re-attain the desired films changed from 3.6 x 10 to 5.5 x IOcm 3 C-1
spraying temperature of 220'C. After spraying, as their thicknesses were increased. The simi-
the grain size was increased by baking the films larity in shape of the curves in figures I and 2
in an inert atmosphere at 400'c for )0 minutes. is strikingly obvious.

584
 10'
I fact the carrier c~~centration is increased from

\
z_tO· 8 x 10 to 8 x 10cm as the thickness increased
10
to 1.5\lm.
1<1)" Decreases in the resistivity of II-IV
:- 10' compound films with thickness have been reported
...
~
ldo" previously and attributed to a variety of ef-
l 10' fects. For example, Berger et al (1) explained
6>10"
the effect in CdSe films in terms of an in-
& creased mobility. Vergunas (5) suggested that
:: 10'
!
~ the resistivity of CdS films decreased because
I \ j of increased fibre-axis orientation. However,
10 g the results reported here prove that the de-
Ii ... crease in resistivity is associated with an

1a1'
10' increase in carrier concentration. It fo llows
that a film changes composition during growth.
10
u
Wilson and Woods suggested three ways in which
this might happen:
(i) The impurity content of the source
O~ 10 IS might increase with time, particularly with
f'4to Ihockn... <OO_1
resistively heated sources.
' .... Z. The Hall cocft\ctcnll of CdS films. (ii) A gettering process might occur,
particularly in an open vacuum system.
( iii) The excess cadmium content of the
film might incl'ease with time.
The variation of the Hall mobility with Wilson and Woods discounted the first two
thickness is shown in figure 3, where for con~ possibilities and our results demonstrate that
venience of comparison with the results of this was justifiable, supporting the idea that
Wilson and Woods (2) and Buckley and Woods (4 ) the excess cadmium content of the film does
reciprocal mobilities are plotted against thick- indeed increas e with time.
ness. The mobility increased by a factor of 2.5
over the range of thicknesses investigated and
was clearly inadequate to explain the thousand- 5. Conclusion
fold increase in electrical conductivity ob- For optimum photovoltaic efficiency the
served. With the thickest films t~e _To~Fity resistivity of the CdS film should be of the
tended to a constant value of 3.7 em V S • order of 100Qcm. It is also necessary to employ
a rather thick film (4\lm - 20\lm) in order to
avoid the incidenc~ of pin tide short circuits.
This method of CdS thin film preparation meets
these criteria.
07

06
References

1. Berger H et al 1969
Phys Stat Solid 33 , 417-24
2. Wilson JIB and Woods J 1973
J Phys Chern Solids 34, 171-191
3. Shiozawa et al 1969 Aerospace Research Labs
Report ARL 69-0155
4. Buckley RW and Woods J 1973
J Phys D 6 1084-89
5. Vergunas FE 1967
IS
Sov Phys Christal log 11 420-1
FIaurt 3. Reciprocal Hall mobililies or CdS tilmL

It, is interesting to note that all films

exhibited an efficient green edge emission when
excited with ultra violet radiation at 77k.
The y also displayed substantial photoconductive
effects. These properties demonstrate that the
films closely approximate to the bulk material.

4. Discussion
The Hall measurements allow a distinction
to be made between the contributions of the
carrier concentration and the mobility to
changes in the resistivities of the films. The
obvious similarities between the changes in the
Hall coefficient and the resistivity of the
films demonstrate that the change in resistivity
is almost entirely due to an increase in the
carrier concentration with thickness. In

585
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THE DEVELOPMENT OF CUINS2 SINGLE CRYSTALS AND THIN FILMS FOR PHOTOVOLTAIC
APPLICATIONS

M.L. Fearheiley, N. Dietz, ' R. Scheer, M. Kanis, S. Fiechter, M. BriiJHer*, C. Dzionk*,

H. Metzner*, H. Migget, and H.J. Lewerenz

Bereiche Photochemische Energieumwandlung, Schwerionenphysik*, Strukturforschungt,

Hahn-Meitner-Institut,
Glienicker Strasse 100, 1000 Berlin 39, Federal Republic of Germany
(49) 30 8009 2081
FAX (49) 30 8009 2434

ABSTRACT. CuInS2 is being developed both as single crystals and as thin

films for the use in photovoltaic systems. Good quality single crystals are
required in systems utilizing solar concentrators. To assist in this, the
phase relations in this ternary have been investigated by DTA, X-ray
diffraction and Perturbed Angular Correlation (PAC) of ,,{-rays from l11In
tracers. The results from the PAC investigation are compared to published
thermodynamic values and found to be in agreement. CuInS2 single crystals
are currently being grown by a modified gradient freeze method and by high
pressure Bridgman. Future work includes growth from Cu/ln solutions using
one form of the traveling heater method and by liquid phase epitaxy.
Initial success has been observed in material grown under elevated argon
pressures by the gradient freeze method. The second part of our work is
concentrated in the area of thin films for large area applications. Ttlin
films are produced by thermal evaporation from three sources and/or by
reactive evaporation in an H2S atmosphere. A closed looped system is used
to monitor and adjust the component fluxes during film formation, Here, PAC
will be used to investigate interfacial regions.

1. INTRODUCTION 2.1 The Cu, In, S system

The chalcopyrite, CulnS2, possesses a The binary boundaries, Cu-In/8/, Cu-
room temperature band gap of 1.55 eV. This S/9/ and In-S/10 / , have previously been
lies near the optimum for a homojunction investigated. However, in the ternary only
device/l / having a theoretical maximum the CU2S-In2S3/3/ pseudobinary and t h e
efficiency of 25 %. Cuo . SlnO.s-S/ll/ liquidus have previously
At present, the best single crystal been investigated. With DTA and ,x -ray
device is a photoelectrochemical cell diffraction we have investigated the
utilizing a 1-/13- redox couple/2/ pseudobinaries: In-CulnS2 / 12/, Cu-
exhibiting a power conversion efficiency of CulnS2/13/, and other compositions in the
9.7 %. The development of CuInS2 single metal rich portion of the ternary (Fig. 1).
crystals is hindered by the presence of two As can be seen on the Cu-CulnS2
solid-state phase transformations/3 / . pseudobinary, the liquidus shows a broad
These transformations would promote the region of liquid immiscibility (Fig. 2) .
formation of voids and cracks in material This could make the growth of this material
g ro wn from the melt. One way to overcome from solution more difficult. Thiel/ll/
these difficulties would be to grow the
ma~erial below these temperatures or vary
the growth conditions during processing. s
with regard to thin film devices,
Mi t chell et al. /4/ have developed a 7.3 %
thin film solar cell based on this
material. The properites of these devices
are inherently dependent on the interface
and grain bou ndary states. Their r o le in
these devices will have to be understood in
order to optimize CulnS2 thin film solar
c e lls.

2. SINGLE CRYSTAL GROWTH

To utilize solar concentration
techniques, good quality single crystals
will have to be produced. The growth of
single crystals have been previously
reported/s,6,7 / . However, it is felt that
too little was understood of the phase
relations in this system. Without this
understanding the development of good
Cu L -____~~------~-- __~~------~----~
quality CuInS2 single crystals will be 1n
greatly hindered, We are currently
investigating the phase relations in this
ternary system, and with this information Figure 1 : The Cu-In-S ternary phase field
we hope to better control the crystal showing the pseudobinary regions
growth process. investigated.

586
indicated that this two liquid region
occurs along the CUO. sInO. s-S pseudobinary,
and i t was also found along the In-CuInS2
pseudobinary/12/. The liquidus isothermal
-0.08
projection has been constructed and shows
this two liquid region to dominate the
metal-rich portion of this ternary system
(Fig. 3).

:.::L------~J
IU •

.... .~~
.....
J \'
~,,~
...
,/
o ,00 200 300 4 00

...
l ime Ins)
• ... ~,LI.\.I IL. · ~

...
\ \ • Cftl ~
\'
. \ ___ . _ _ _ ~:ti:~=t=
~ Figure 4: PAC spectrum of stoichiometric
...
~ -~--_~ . _.~I~.~

f ........ _ -. -.- - . --
CulnS2. The spectrum shows cubic symmetry

... /
/
,- for all lllIn-lllCd probes and gives no
indication for tetragonal lattice
"
r "------ ---··-·--·- distortion/14/ .

... I
I
/
/

between CulnS2 and InS to be drawn. The

:0. .'
method was used to identify all In
". containing phases by means of. their PAC

c.
.. .. .. .. .A_,_c.
". "
parameters and to complete the

ternary system (Fig. 6).

room
temperature tie-line picture for this

Figure 2: The Cu-CuInS2 pseudobinary

showing a broad region of liquid
immiscibility/13/. CY ln5 , with 11'\ . .. C; • •• (2:0 . JO"':.)

-002

- 0 .0 4

10) S - O . OS

'"

-0 . 12

-0 . , "'0:----:-:'0:":0, - - -2"'0:":0, - - - : -)0"'0, - - - -40:.0

l im e (n,)

'?
c,
f (1) . -_ _ _1'T2l _ _
2.

:~
(b) ;
E
ii
E
o

c. I,

Figure 3: The liquidus isothermal

o 200 .00 6 00 8 00
projection for the Cu-In-S system. A broad 20/l" ( loU-' I )
two liquid region dominates the metal
portion and is crossed by two
eutectics/13/ . Figure 5: PAC spectrum (a) and its Fourier
transform (b) of a sample prepared from
stoichiometric CulnS2 and metallic In. The
PAC was used to analyze the room Fourier spectrum domonstrates the
temperature In-containing phases/14,15/. appearance of two non-cubic components
Phases containing more than about 1 % of all which are identified as metallic In (1) and
In atoms will exhibit a well defined InS (2) /14/ .
electric field gradient. Several In-
containing phases can be identified
simultaniously at this accuracy, making PAC Calculations on the room temperature
superior to x-ray diffraction of mixed ternary system 1 16/ substantiate the tie
phase samples. Results for CuInS2 and a lines found in the P.AC study and show
mixture containing CUlnS2, In and InS are compatability with existing thermodynamic
shown in Figures 4 and 5, respectively . data.
The presence of these three phas es in
equilibrium allows the tie-line between
CulnS2 and In as well as the tie-line

587
 After metallization of the substrate, a
s quadrupol mass spectrometer, which is
coupled to the temperature programmer, will
monitor the component fluxes ensuring well
defined conditions over the substrate.
Evaporated CdZnS windows can be deposited
in-situ as well as the front contact. A
transfer chamber is available to transfer
the substrate after any process step to the
surface analysis lab.

I
CuS . · .....
.. '
. ., .'
.:.-::.,.. " ::-',
3.2 Interface investigation
As the thickness of the active layer
("JI ~S , ' ...... . decreases so increases the importance of
{u~ ,.5 • . , .,. ...",
Cu:S "" . E . the interfaces and grain boundaries in
these active layers. These planar defects
must be taken into consideration when
optimizing these thin film devices. Their
investigation, by nature, is difficult,
comprising a fraction of the device's
volume requires very sensitive techniques.
Cu n By depositing the lllIn probe at the
interface between the CulnS2-metal or
CulnS2-window layers, it is possible to
Figure 6: The room temperature phase
investiqate the chemistrv of these
relations as determined by PAC/15/ and
corraborated by thermodynamic interfaces. This method is part icularly
calculations/16/. sensitive for this type of investigation.

REFERENCES

2.2 Growth at elevated Ar pressure 1. Loferski, J. J., Proc. IEEE. 51: 677,
The use of an inert gas, such as Ar, 1963.
changes the activity of S in the liquid and
solid phases. In altering the system in 2. Lewerenz, H. J., H. Goslowsky, K.-D.
this way, large single crystals have been Husemann and S. Fiechter, Efficient solar
produced/7/. The material was shown by x- energy conversion with CulnS2' nature.
ray diffraction to be single phase CulnS2, 321(6071): 687, 1986.
RHEED showed the cleavage plane to be
(112), however rocking curves showed an 3. Binsma, J. J. M., L. J. Giling and J.
unusually high degree of dislocations. By Bloem, Phase Relations in the system Cu2S-
EBIC analysis, the material showed minority In2S3, Journal of Crystal Growth. 50: 429-
carrier diffusion lengths up to 1.5 and 2 436, 1980.
!lm, for n- and p-type crystals,
respectively. 4. Mitchell, K. W., G. A. Pollock and A.
V. Mason, 7.3% Efficient CUlnS2 Solar Cell,
2.3 Hiqh pressure qrowth
To improve the quality of these Proceedings of the 20th IEEE Photovoltaics
crystals, a high pressure system is being Specialist Conference, Las Vegas, 1988,
set up so that CulnS2 can be grown by the 1542.
Bridgman method or the gradient freeze
method. The latter allows for the control 5. Hsu, H. J., M. H. Yang, R. S. Tang, T.
of the S partial pressure during M. Hsu and H. L. Hwang, A novel method to
processing. grow.large CulnS2 single crystals, J.
Crystal Growth. 70: 427-432, 1984.
2.4 Traveling heater method
Future plans include the growth of 6. Goslowsky, H., S. Fiechter, R.
CU lnS 2 by the traveling heater method using K6nenkamp and H. J. Lewerenz, Chemical
In and Cu/ln zones. A double ellipsoid vapour transport of CulnS2: correlation of
mirror furnace will supply steep growth, Sol. Energy Mater. 13: 221, 1986 .
temperature gradient along the melt/crystal
interfaces. 7. Dietz, K., M. L. Fearheiley, R.
Scheer, H. Hoffmann and H. J. Lewerenz,
3. THIN FILM GROWTH Lamellar Metastable Phase in the Cu-In-S
For large area applications, the System: Structural and Electronic
development of thin films is necessary. Properties, in preparation. : 1991.
The growth of CulnS2 thin films from one,
two/17/ and three/18/ sources have been 8. 'Subramanian, P. R. and D. E. Laughlin,
reported. Success depends on both the The Cu-In System, Bull. Alloy Phase
independent control of the component flux Diagrams. 10 (5): 554-568, 1989.
rates to insure desired composition,
electrical and optical properties of the 9. Chakrabarti, D. J. and D. E. Laughlin,
films; and the control of interface Cu-S, Bull. Alloy Phase Diagrams. 4(3):
properties. Two parallel methods are used 254-271, 1983.
to address these requirements.
10. G6decke, T. and K. Schubert, On the
Phase Diagram InSM' Z. Metallkde. 76 (5) :
3.1 Flux control
A quasi-UHV chamber (10- 10 Torr) is 358-364, 198;;.
designed to produce a complete solar cell.

588
11. Thiel, F. A., The Phase Relations in
the Cu, In, S System and the Growth of
CUInS2 Crystals from the Melt, J.
Electrochem. Soc. 129(7):·1570-1, 1982.

12. Fearheiley, M. L., N. Dietz, M.

Birkholz and C. Hapfner, Phase Relations in
the system In-CuInS2' J. Electronic Mat.
20(2): 175-177, 1991.

13. Fearheiley, M. L. and A. Hartmann,

Phase Relations in the Cu, In, S System, in
preparation. : 1991.

14. Brtissler, M., H. Metzner, K.-D.

Husemann and H. J. Lewerenz, Phase
identification in the Cu-In-S system by y-y
perturbed angular correlation, Phys.Rev.B.
38 (13): 9268, 1988.

15. Metzner, H., M. Brtissler, K.-D.

Husemann and H. J. Lewerenz,
Characterization of Phases and
Determination of Phase Relations in the Cu-
In-S System by y-y Perturbed Angular
Correlations, in preparation. : 1991.

16. Migge, H., Thermochemistry in the

system Cu-In-S at 298K, J. Mat. Research.
submitted: 1991.

17. Kazmerski, L. L., M. S. Ayyagari and

G. A. Sanborn, CuInS2 thin films:
Preparation and properties, J. Appl. Phys.
46(11): 4865-9, 1975.

18. Wu, Y. L., H. Y. Lin, C. Y. Sun, M. H.

Yang and H.L.Hwang, On the growth of CuInS 2
thin films by three-source evaporation,
Thin solid films. 168: 113, 1989.

589
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PREP ARATION OF CdSiAs2 - HOMOJUNCTION SOLAR CELLS AND ITS

ANISOTROPIC OPTICAL PROPERTIES

F.P. Baumgartner, H-P. Schweikardt and E. Bucher

Universitiit Konstanz, Fak. fur Physik,
Postfach 5560, D-7750 Konstanz, F.R.G.

ABSTRACT
Cd~iAs2 homojunct~on~ h~ve been fa~ricated on p-t~pe single crystals by a closed tube
dOplllg procedure wIth llldIUm. As denved from electncal transport measurements indium
on a.cadmium site acts as a deep donor. Using p-type starting material with a net carrier
densIt:y of 4.0.10 15 cm-a, a resistivit:y of 20 ncm Cl;nd a Hall mobility of 90 cm2·(Vs)-I, the
bes.t dIOde structures coul~ be. obtalll~d ~t a doplllg temperature of 410 'C / 15 minutes.
It IS ?-ssumed that ~he dIffuSIOn ?f llldIUm. and the escape of the volatile components
cadmIUm and arsenIC are the malll mechanIsm of the junction formation. Open circuit
voltages up to 635 m V have been reached in combination with a short circuit current
density of 0.5 mA·cm- 2. From deep level transient spectroscopy (DLTS) measurements an
acceptor level at 350 me V and a donor level were detected. Fits of the dark I-V
characteristics r~sulted in ~ hig~ series resistance of > 200 n·cm 2 • Spectral response
measurements wIth un,Polanzed light show.ed an onset at 800 n~ (A-t~ansition) a~d a
decrease at 700 nm. It IS speculated that thIS large value of the senes resIstance is caused
by. the high resistance n-type layer which is due to the deep donor level. Optical
anIsotropy has been demonstrated by spectral response measurements with linearly
polar~z~d li~ht. For cells prep~red o~ the (101) crystal face the photocurrent at the A-
tranSItion 1.55 eV) was 3.1 times hIgher for a polarization parallel to the c-axis than
perpendicu ar to it. This indicates photoconductivity of the n-type layer.

INTRODUCTION Measurements with polarized light might

be applicable to other chalcopyrite solar cells
The ternary semiconductor CdSiAs2 is the like CuInSe2 in order to determine the ratio of
closest chalcopyrite-analogon of GaAs. With a bulk to surface recombination.
first direct gap at 1.55 eV large spectral
absorptivity and good electronic properties, the
material meets the basic requirements for solar EXPERIMENTAL
cell application /1/. A controlled wide variation
of the carrier concentration has been obtained CdSiAs 2 single crystals with a size of
by doping in combination with large Hall (6x3x2 mm3) were used as bulk material. They
mobilities. Both p- and n-type conductivity has were doped with silver during the chemical
vapor growth procedure. A chemical analysis
been achieved /2/. revealed a doping concentration of 3.1018 cm·3 •
Little information on the properties of
CdSiAs 2-diode structures are given in the These p-type crystals have the following
literature. The first investigation of CdSiAs2 electronic properties; a carrier concentration of
pn-junctions revealed an open circuit voltage 1:1Q16cm'a, a Hall mobility of 90 cm 2 .(Vs)·t, a
V DC of about 600 m V and a practically constant resistance of 20 ncm and an acceptor level of
photosensitivity up to 3 eV /3/. ZnO-CdSiAs2 210 meV, as determined by temperature
heterojunctions exhibited open-circuit voltages dependent Hall measurements/1/.
of 250 m V and a short circuit current density of Prior to the fabrication process the crystals
12 mA ·cm·2 /4/. Spectral response were etched in 1% Br-Methanol for 3 minutes .
measurements on In203-CdSiAs2 structures A thin layer of indium (80-600 A) was
(V oc=600 m V) showed a large anisotropy of the evaporated onto the (112) or (011) faces.
photo current with linearly polarized light /5/. Together with 100 mg indIUm these crystals
Polycrystalline CdSiAs2 with a top layer of were filled into a silica ampoule, which was
cadmium snlfide has been studied as a possible evacuated and sealed at a residual pressure of
candidate for large area thin film solar cells /6/. 10- 5 mbar. The ampoule was heat treated at
CdSiAs 2 was also used successfully as a about 400 C for between 15 and 30 minutes.
0

source of spin polarized electrons and as a By diffusion of indium into the p-type
detector of linear polarized light }1 / / 5/. crystal a n-type layer was formed. On that part
Information on CdSiAs 2-dIOde structures of the crystal surface which was not covered by
dealing with different characterization the indium layer the volatile Cd and As
measurements are still lacking. In this paper we produced also a thin n-type layer. Therefore the
report on the fabrication and the properties of crystal was mechanically polished in order to
CdSiAs2-homojunction diodes. The dark and remove this layer before making the back
illuminated current voltage characteristics, the contact. Front and back contact were obtained
anisotropic spectral response and DLTS by using silver paste on 1000 A evaporated
measurements will be discussed. silver.

590
1-V MEASUREMENTS Table 1: Parameters of the dark and the illuminated I-V
characteristics compared with different annealing
conditions; the series resistance Rs, the shunt resistance
Some electrical properties of three diodes RSH , the diode quality factor n and the saturation current
produced with only small differences in the 10 are listed.
fabrication process are compared in table 1.
All I-V meaurements were carried out with
a Keithley 238 source-measure unit where the Sample H41A H41B H41C
sensitivity of the complete experimental setup
was 10 picoamperes. The dark current In-layer thickness [A] 80 200 600
characteristic of the three p-n junctions are Temperature [C 0] 418 410 400
shown in fig 1. Annealing- time [minutes] 10 10 30
The measured points for sample H41C and
H41B were fitted with a single diode model Dark characteristic:
with the parameters listed in table 1.
The small rectification ratio of sample Rs [!t.cm 2 ] 1050 120 240
H41B is attributed to a low shunt resistance. RSH [M!t.cm 2 ] 20 7.2.10- 4 1.5
That is correlated with a narrow range of a n 1.7 2.6 1.9
exponential rise of the current between 0.5 and 10 [mA·cm- 2 ] 2.5.10- 7 8.3.10-6 5.2.10-6
0.7 V. This leads to some inaccuracy in Rectification ratio at IV 1600 110 2100
determining n. In order to explain the reverse
branch of H41A in addition to the RSH a second Illuminated at 85 mW ·cm'2:
diode was assumed. The parameters of the first 591 634 530
Voe [mV]
diode which is dominant near 0.5 V are also [mA.cm'2] 0.65 0.49 0.53
listed in table l. Ise
Fillfactor 0.36 0.37 0.46
Rectification ratios up to 16000 at 1 V
could be obtained by etching off the IIi front
contact, apart from the area under the silver
front contact. It should be mentioned that by
removing the naturally formed n-type layer, the
shunt resistance could be increased. .......,e 10 -,

The dark current measurements on sample 0

10 -.
H41A were carried out at temperatures between "-
«
170 K and 270 K as shown in fig.2 . The fit E
....... 10 ...
parameters to the single diode model for each >-
temperature are given in table 2. Because of f-<
experimental limitations of the measuring Cii 10
~
circuit it was not possible to collect enough r.:l
0 10 -,
accurate data for determining the shunt f-<
resistance RSH below 230 K. Z
r.:l
The logarithmic decrease of the series Ik: 10 ..
resistance with rising temperature can be Ik:
;:,
described by an activation energy of 210 me V U 10 -,
1.0
0.0 0.2 0.4 0 .6 0.6
as shown in fig. 3. VOLTAGE [V]

Figure 2: Dark 1- V caracteristic of the sample H41A

at different temperatures between 170 K and 270 K
N' 10 . . . - - - - - - - - - - - - - - - - . 10 (stepsize 10 K).
Ie
()

<e 10 .,
.......
Table 2: Fit parameters for the temperature dependence
10 -, of the dark I-V characteristic of sample H41A. They are
based on the single diode model.

Temperature Rs n 10 RSH
[K] [!t·cm 2 ] [mA·cm- 2] [MQ·cm 2 ]

170 3.5.106 1.7 1.3.10-13 -

-0.5 0 .0 0.5 180 1.1.106 1.9 4.6.10. 12 -
VOLTAGE [v) 190 6.8.10 5 1.7 5.1-10- 12 -
200 3.7.10 5 1.8 4.0.10- 11 -
Figure 1: Dark current characteristic at room 210 2.3.10 5 1.4 3.4.10. 12 134
temperature for the three samples listed in table l. 220 1.4.10 5 1.5 6.4.10- 11 176
The best fits with parameters from table 1 are shown 230 8.7.10 4 1.4 3.5·10'11 97
by the solid lines. 240 5.2.10 4 1.5 5.9·10'10 64
250 3.2.10 4 1.6 3.9.10-9 63
260 1.9.104 1.5 3.7.10-9 34

591
 10' ~--------------------------'
4 CdSiAs. H41A
......,
.E
~
:i 2

~ Or---------~T-------~----~
.E
u
10 •

.r: ~
Z
-2
£ l.'J
.....
10 • CIJ -
I
a! CIJ
5o
10 5 3. 4.0 4.5 5 .0 5.5 6 .0
1000 ! TEMPERATURE [1!K] -lf60 180 200 220 240 260 280 300
TEMPERATURE (K]
Figure 3: Series resistance Rs of sample H41A as a
function of temperature. The best iit leads to an
activation energy of 210 me V. Figure 5: DLTS - signals of sample H41A at different
temperatures measured at 2332 Hz and 1249 Hz
Lock-in frequency.

The illuminated 1-V characteristics are Using two overlapping gaussian curves for
shown in figA and the data in table 1. It can be the acceptor and donor peaks the donor level
seen that sample H41C with a low series was calculated to be between 150 meV and
resistance and a relatively high shunt resistance 300 me V. Indications of an onset of a second
exhibits the best fill factor. Without an donor peak with a maximum above 300 K is
antireflection coating an efficiency of about also observed. With the present experimental
0.15 % was observed. setup, DLTS-peaks for T > 300 K cannot be
evaluated.

VOLTAGE [mV]
..
~

I 0.0
0 100 200 300 400 500 600 700 800

CdSiAs, H41A
E
u
H41B -2.0
-0. 1 E, : 335-380 meV
*
] -0.2
...-.. -3.0
~ -0.3 (II
E-<
en
Z "-
0
-0.4 Q)
W -4.0
.s
0 '-'

<- -0.5
Z
w H41A
0: -0.6 -5.0
0:
::>
u -0.7

-6·~.L.
6 ----5-'.0-----5-'-.2-----5.L..
4 ----5-'.6-----="5.8
Figure 4: muminated 1- V characteristic of the
samples from table 1 at 85 m W·cm- 2 lOOO/T [11K]

Figure 6: Arrhenius-plot of the acceptor level of

DLTS-MEASUREMENTS sample H41A for two sets of similar measurements.

The DLTS-system DLS 82 E used in this

work is based on the lock-in technique. The ANISOTROPIC SPECTRAL RESPONSE
capacitance measurements were carried out at a
frequency of 1 MHz. Lock-in frequencies The tetragonal crystal field and the spin
between 10 Hz and 2500 Hz resulted in an orbit interaction split the upper valence band
emission rate in the range of 22.3 S-1 to 5575 S-I. into three levels. In order to excite electrons
In fig. 5 the DLTS-signals of sample H41A are from these levels into the conduction band
plotted versus temperature. The minimum at minimum three different energies at k = 0 are
about 200 K indicates an acceptor and the required (A = 1.55 eV, B = 1.72 eV and
maximum at about 230-240 K a donor level. C = 1.98 eV) . With linearly polarized light the
The Arrhenius plot of the emission rate selection rules result in different transition
revealed a deep acceptor level of 350 me V probabilities for the A, Band C transitions.
(fig.6). An Arrhenius plot of the donor level The spectral response of the diodes were
could not be achieved because of the influence measured as . a function of wavelength for
of the stronger nearby acceptor peak: An different angles between the direction of the
estimate was obtained from the maxlmum linearly polarized light and the crystal c-axis.
halfwidth of the donor peak in fig. 5.

592
Fig.7 shows the curves for sample H41A. It can DLTS measurements clearly showed an
be seen that at the A-transition the response acceptor level (at 350 me V) and a donor level
for Eli hi parallel to the c-axis (q; = 0 is 3.1 0
) within the space charge region. These levels
times targer than for Elighl perpendicular to the correspond to the p- and n-regions of the
c-axis (q; = 90 This. value is v~ry close t~ the
0 ). homojunction. The donor level is estimated to
ratio of the arusotropIc absorptIOn coefficIents lie between 150 and 300 meV. More DLTS-
011/ 01 = 2.9 at 1.55 eV /7 f. T~~ supports t~e measurements have to be carried out in order
assumption t~at. photoconductivIty .of a thi? to clarify whether this donor level can be
and highly reSIstIve n-type layer dommates this correlated with the series resistance activation
anisotropic spectra. EBIC (electron beam energy.
induced current) measurements revealed .an From etching experiments it is obvious
onset of the induced current at an acceleratIOn that the value of the shunt resistance correlates
voltage above 17 kV, indicating the existence of with the existence of a naturally formed n-type
a top dead layer. .. layer (outdiffusion of Cd and As) at the edge of
Fig. 7 also shows that for the B-transItIOn the In front layer. To increase the rectification
the response for Elighl II to the c-axis is larger ratio (i.e. up to 16000 at IV) a soft etching
than for E lighl 1 to tlie c-axis. This corresponds procedure should be developed in order to
also to the results of electroreflectance spectra reduce the natural n-type layer without
reported elsewhere /7/. . damaging the In front layer.
Spectral response measurements WIth Spectral response measurements showed an
unpolarized light reaches a maximum of ~nlr 5 increase at the first direct transition and a
% quantum efficiency. These spectra are SImIlar decrease at about 700 nm, attributed to high
in shape to the curves in fig. 7 for q; = 45 The 0 • values of surface recombination. The same
decrease at 700 nm indicates a large value of measurements carried out with linearly
the surface recombination rate. polarized light confirme the assumption that
the spectra are dominated by
photoconductivity of the front (n-type) layer.
In order to increase the short circuit
;::J 8 A current density other n-type dopants will be
iii
.0 investigated. To impede the decomposition of
the CdSiAs2-compound heattreatment
>- 0.8 experiments will be carried out in the presence
U
Z of cadmium and arsenic vapor.
~
U 0.6
1: ACKNOWLEDGMENTS
e:..
fil
0.4 The authors would like to thank
~..... K.M. Nigge for assistance with the preparation
0.2 of the samples and G. Willeke for reading the
~ manuscript.
~
(Y
0.0
600 650 700 750 800 850
WAVELE GTH [run]
REFERENCE
/1/ L.C. Burton, C.F. Smith, A.F. Carroll and
Figure 7: Anisotropic spectral response with linear
L.H. Slack; Proc. Southeast, IEEE Region
polarized light for sample H8-~1. The numbers 3 Conf., 1979, pp. 25-28.
indicate the angle between the E-light vector and the /2/ F.P. Baumgartner, M. Lux-Steiner, G. Doll
optical crystal c-axis. The junction was formed on the and E. Bucher; Journal of Crystal Growth
(10 1) crystal plane. 109 (1991) 318.
/3/ Ch. Dovletmuradov, K. Ovezov, V.D.
Prochukhan, Yu.V. Rud and M. Serginov;
Sov ..Semicond. 10 (1976) 986.
CONCLUSIONS /4/ M. KImmel, M. Lux-Stemer, M. Vogt and
CdSiAsz-homojunctions with an open E. Bucher; J. Appl. Phys. 64 (3) 1988.
circuit voltage of up to 635 mV have been /5/ V. Yu. Rud, Yu.V. Rud and M. Serginov;
demonstrated on single crystals by diffusion of Phys. Stat. Sol. (a) 121 (1990) K81.
indium at 410 C / ~O minutes...
0
. /6/ DOE Contract No. DE-AC04-79 ET 23007
Experiments WIth In by m-sItu dopmg Exhibit A, Statement of Work, 9 April
during the crystal growth process confirm the 1979
assumption that In on a Cd-site acts as a deep /7/ J.L Shay and E. Buehler; Phys. Rev. B 3
donor -/1/. Therefore it is proposed that .In a.cts (1971) 2598.
twofold· it forms the n-type layer by dIffuSIon
and it ~cts as a front contact by the remaining
In on the crystal surface. .
1-V fit characteristics revealed a senes
resistance of Rs > 200 !I·cm Z • This value is over
100 times larger than the resistance of the p-
type base. From temperature dependent I-V
measurements Rs is activated by 210 meV.

593
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Thin Film Preparation of FeS2 (Pyrite) by Argon Sputtering, Plasma

Reaction, MOCVD and Spray Pyrolysis

C. Hopfner, A. Ennaoui, D. Lichtenberger, M. Birkholz, G. Smestad, S. Fiechter and H.

Tributsch

Hahn-Meitner-Institut Berlin,
Abteilungen "Materialforschung" und "Solare Energetik"
Glienicker StraBe 100, D-1000 Berlin 39, Federal Republic of Germany
(49) 30 8009 2081
FAX (49) 30 8009 2434

Abstract: Pyrite seems to be an interesting material for solar cell

applications because of its band gap of 0.95 eV and its very high
absorption coefficient. The preparation of pyrite thin films by four
different methods is reported. The differences concerning the
stoichiometry and the crystalline phases are discussed.

1. INTRODUCTION

Pyrite, with a band gap of Eg = 0.95 eV,

- AI

attracts interest because of its environ-

mentally sound components and its extraor-
dinarily high optical absorption coeffi-
cient (a > 105 cm- 1 ) which makes the mate-
rial a promising candidate for thin film
solar cell applications /1/.

Chemical analysis of natural and synthetic Cold Will R•• ctor

pyrite crystals points to a strong thermal
dependence on composition /2/. Since with GI ... Sub.t .. t,
increasing growth temperature the sulfur
deficiency increases up to 5 atom% the app-
lication of synthesis temperatures as low
as possible is required.
-......-[".:..... 1. SuscoptOl
2. METHODS

To realize this demand we are working with o.uattzrod

several thin film techniques to produce py-
rite thin films. They are described below:

• Metal organic chemical vapor deposition

(MOCVD)

We used an atmospheric pressure MOCVD-reac-

tor (Fig. 1) /3/ with iron pentacarbonyl
and elemental sulfur as precursors and
Argon as carrier gas_ The substrate is in
contact with an RF-heated graphite suscep-
tor and has a temperature of 350 0 C during
the deposition. The Fe(CO)5 flux controlled Figure 1: Diagram of the atmospheric pres-
the growth rate which was between 0.7 to 5 sure MOCVD reactor
A / s.
X-ray diffraction (Fig. 2) showed that the
layers produced consisted of the FeS2 After annealing at 500 0 C in sulfur atmo-
phases pyrite and marcasite. The relative sphere the marcasite phase disappeared, the
amounts of these two phases could be X-ray diffraction showed only the pyrite
influenced by the growth rate, however we peaks (Fig. 2). The stoichiometry of the
had no success in complete suppression of pyrite layer was FeS1.95 ± 0.1'
the marcasite phase. The stoichiometry was
measured by Rutherford backscattering and
had the value of FeS1.95 ± 0.1 (Fig. 3).

After annealing at 500 0 C in an Ar-flux

much of the layer decomposed to monosulfide
phases and had a average composition of
S/Fe = 1.2.

594
 1 IO~-GVH

.'
M I REACT I UE GAS
~ H,S

~
~
it' ,
I

~~';i
I'!~ i
I
" ~ . ) btl," . _ .....
M. Py. M.
... r,
;j P,
... M.
M•

;j
:;\. ,... ». il .• ilo. ." Zo . .\.I' .1..
tWD - n.:;u IU£Q1f.t.SJ

Figure 2: X-ray diffraction of a thin film

prepared with MOCVD a) before annealing and
b) after annealing. Py = Pyrite peaks, Ma =
Marcasite peaks. punp IHG
SY$TEn

Energy (MeV) Figure 4: Diagram of the sputtering system.

2SO r-______~03~------~lr·O~------~1~3~----~ The target is pressed from pyrite powder .

200

:2
•
):150 f.
• Spray pyrolysis

•
"0
A s o lution of iron III chloride hexahy-

i~
.!:! drate and thiourea was sprayed onto a hot
D
~ 100 substrate using a spray nozzle Fig. 6
o shows the geometry of the configuration
Z
/4 / . Typical experiments were performed
so using a substrate temperature of -350 0 C
---~- and a growth rate of -5 A / s.
or-~--,_-----.----~~-A~~----~
a 200 400 800 800 1000
Channel
3. COMPARISON
Figure 3: 2 MeV Rutherford back scattering The common feature of these techniques is
spectra of a pyrite film prepared by MOCVD. that the film formation takes place far
The comparison of the integral intensities from thermodynamic equilibrium. In this way
gives a composition of FeS1.95' the temperature controlled sulfur content
in the layers could be balanced by the
• Reactive Ion Beam Sputtering sulfur activity in the deposition regime.
Using ' the MOCVD technique the sulfur
Fig. 4 shows the sputtering system . A H2S activity is controlled directly by the ele-
ion beam is directed at a 10 cm diameter mental, gaseous sulfur, whereas the concen-
target which was pressed from pyrite pow- trations of HiS as wel l as thiourea control
der. The deposition occurs at -75 0 C with a the sulfur activity in the other techni-
growth rate of -0.3 A / s. After annealing ques.
in H2S atmosphere at 250 0 C the layers had
a s~oichiometry of Fes1.95 ± 0.1' ITO glass, boron silicate glass, fused si-
Experlments with an Ar-beam and H2S as a lica, Kapton foil and single crystalline
reactive gas yielded layers with a substrates, e.g. si or GaP which offer the
composition of S/Fe <1 .5 . opportunity for epitaxial growth of the py-
rite film due to suited lattice match, were
used as substrates .
• Plasma-assisted sulfurization of iron
The layer thicknesses produced varied from
Fe layers were exposed to a H2S plasma 700 to 10,000 A. Best films were deposited
(10- 1 mbar) in a rf plasma reactor on glass. They exhibited columnar growth
(Fig. 5). At temperatures of 200 0 C a and their XRD-pattern showed exclusively
mixture of pyrite, marcasite and pyrrhoti- pyrite lines. The layers have p-type con-
tes, with an average composition of S/Fe ductivity. The electrical properties are
<1.5 was obtained. currently being investigated /5/.

595
 REFERENCES

1. Ennaoui, A. and H. Tributsch, Iron

sulphide solar cells, Sol. Cells. 13: 197-
200, 1984.

2. Luck, J., A. Hartmann and S. Fiechter,

Stoichiometry and Impurity Concentration in
Synthetically Grown Iron Pyrite Crystals
and Their Constituents, Fresenius Z. Anal.
Chern. 334: 441-446,1989.

3. Chatzitheodorou, G., S. Fiechter, G.

Fleming, D. Herm, R. K6nenkamp, M. Kunst,
J. Richard and H. Wetzel, Photoactive Fes 2
Figure 5: Diagram of the plasma reactor (Pyrite) Films, Proceedings of the Intern.
Symposium on Trends and New Applications in
Thin Films, StraBburg/France 16.-
20.3.1987, 1987, 85-89.

4. Smestad, G., A. D. Silva, H.

Tributsch, S. Fiechter, M. Kunst, N.
Meziani and M. Birkholz, Formation of
Semiconducting Iron Pyrite by Spray
pyrolysis, Sol. Energ. Mat. 18: 299-313,
spray nozzle entrance
1989.

@~
5. Schieck, R., A. Hartmann, S. Fiechter,
R. K6nenkamp and H. Wetzel, Electrical pro-
perties of natural and synthetic pyrite
(FeS2) crystals, J. Mat. Res. 5(7): 1567-
1572, 1990.

bell jar

exhaust
hole

Figure 6: Diagram of the spray pyrolysis

apparatus

596
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

THIN FILMS OF MoSe2 AND WSe2 PREPARED BY SOFT SELENIZATION AS BULK

MATERIAL FOR SOLAR CELLS

A. Jager-Waldau, M. Lux-Steiner and E.Bucher

Universit§.t Konstanz, Fakultat fur Physik, Postfach 5560, 7750 Konstanz, FRG

ABSTRACT. MoSe2 and WSe2 thin films have been prepared by the method of soft selenization of r.f.
sputtered metal films of molybdenum and tungsten respectively in a closed tube system. Depending on the
selenium partial pressure two types of textures could be obtained for both films. Type A with the c-axis
preferentially parallel and type B with the c-axis preferentially perpendicular to the substrate. The lowest
direct and indirect band gaps were found to lie between 1.29 and 1.40 eV (E gd) and 1.05 and 1.12 eV (Egi)
respectively for MoSe2. For WSe2 the corresponding values are 1.45 ± 0.05 eV for Egd and 1.25 ± 0.05 eV for
E{ The "as grown" films had room temperature resistivities in the range of 10- 3 to 10 4 ncm for MoSe2 and
1 to 102 ncm for WSe2. These values are in good agreement with those of single crystals. For the first time it
was possible to fabricate photo-sensitive heterojunctions with p-MoSe2 and p- WSe2 thin films as bulk material
and n-ZnO as window material. For MoSe2 diodes Isc was up to 18 mA/cm 2 but Uoc = 16 mV only whereas
the best WSe2 diodes achieved Ise = 16 J1.A/cm 2 and Uoe = 92.5 mY.

The dependence of the absorption coefficient a on

1. INTRODUCTION photon energy for a MoSe2 and a WSe2 film prepared by
soft selenization is shown in Fig. 1. By extrapolating plots
The interest in the application of transition metal of a 2 or a 1/ 2 versus E to zero, the direct and indirect
chalcogenides such as MoSe2 and WSe2 as photoelectrodes
energy gaps of the films were estimated. The lowest
in photoelectrochemical cells or as bulk material in
indirect band gap was found to lie between 1.05 and
heterojunction solar cells has increased in the last few
1.12 eV for MoSe2 and 1.20 and 1.30 eV for WSe2. The
year.s. These materials have properties such as
lowest direct energy gap was found to be between 1.29
exceptionally high solar-optical absorption and chemical
and 1.40 eV for MoSe2 and 1.40 and 1.50 eV for WSe2.
stability [1-5], which make them attractive as solar cells.
These values (Table I) for the thin films were in good
Although single crystals of these semiconductors have
agreement with the results obtained from single crystals
shown high efficiencies in PEC cells [6-8] and somewhat
[4,15]. However, the optical measurements revealed no
lower efficiencies in heterojunction solar cells [9,10], they
difference in the optical behaviour of films with type A or
are typically of small area, making their use for large scale
type B.
conversion of solar energy impractical. Polycrystalline
films, ·on the other hand, can be grown as large area films
with simple experimental procedures. Table I: Estimated values of the energy gaps from
The MoSe2 and WSe2 films have been prepared by optical studies and photoyield for thin films (1)
the method of soft selenization of d. sputtered metal and single crystals (2)
films in a closed tube system as previously reported
[11,12]. In this work, optical and electronic properties of energy gap [e V] material method reference
these films are presented and results of first experiments Egd E gi
to prepare li'ght sensitive devices for photovoltaic use are
given.
1.29-1.40 1.05-1.12 MoSe2 (1) opt. absorpt. present work,[16]
1.19±0.01 1.09±O.01 MoSe2 (2) opt. absorpt. Goldberg [4]
2. EXPERIMENTAL RESULTS 1.35-1.38 1.09-1.12 MoSe2 (2) photoyield Kam [15]

2.1 Optical Studies

1.40-1.50 1.20-1.30 WSe2 (1) opt. absorpt. present work,[12]
The thickness of the MoSe2 and WSe2 thin films
used for the optical and electrical transport measurements 1.35-1.39 1.20-1.22 WSe2 (2) photoyield Kam [15]
varied between 0.2 and 8 J1.m for MoSe2 and between 0.4
and 1 JLm for WSe2.
The optical absorption spectra of type A films with 2.2 Electrical transport properties
the c-axis preferentially parallel and type B films with the With the van der Pauw geometry (VDP) the
c-axis preferentially perpendicular to the substrate [11,12] resistivity of the MoSe2 and WSe2 films was measured
were measured at room temperature in the spectral range between 300 and 50 K. The Hall measurements were
between 0.7 and 1.8 eV. The transmission spectra were performed in a magnetic field of 0.65 T. The experimental
obtained by a conventional optical setup. The absorption setup is described elsewhere [17].
coefficient a has been calculated either from the In order to obtain ohmic contacts, gold was
transmission and reflection spectrum of the films as thermally evaporated onto the MoSe2 and WSe2 films.
described by Pankove [13] or from the transmission Gold wires (d = 0.05 mm) were attached to the gold dots
spectra Tl and T2 of two different samples 1 and 2 with (d = 1 mm) using silver paste. At room temperature the
different thickness tl and t2 respectively as described by ohmic behaviou,r was checked within ± 5 V. The type of
Bichsel and Levy [14] for MoSe2 films. conductivity of the films was determined by thermoprobe.

597
 2.5 2.5

,.......,2.0 ...........,... ,.......,2.0

S
C)
."
S
C)
"
~1.5 ~1.5
o o

. ctl
1.0

." ,.q
Q.,
:;:0.5

0.0 0.00 . 7 09 3 5 7
0 .7 0.9 1.1 1.3 1.5 1.7 . 1.1 1. ,le' V]1.
Enerey [eV] Energy L
a) MoSe2 film b) WSe2 film

Fig. 1: Energy dependence of the absorption coefficient

10 • 10 •
A
10 •

,...., 10 • B ,...., 10 •
§ 10' (5) B
E
()
c: c:
'-' I....J 10 '
10 •
Q. Q.
» 10 »
......., ....+l
...,>
..... . ~ 10'
+l
rJl
'iii 10 -.
....rJlrJl
Ql
$..,
$.., 10 -I 10
(3) A
10 -t

1 0~+--------r-------'--------r-------4
3 4 5 6 7 5 10 15 20
temperature lOOO / T [1 1K] Temperature 1000 / T [1 1K]

a) MoSe2 films b) p-WSe2 films

n-type: # 1, 2, 4, 5, 6 type A: # 3, 4, 5, 6
p-type: # 3 type B: # 1, 2

Fig. 2: Temperature dependence of the resistivity for different MoS~ and WS~ films of type A and B as indicated

The MoSe2 films were p- or n-type whereas all WSe2 carrier concentration (NCD). The electrical transport
films were p-type. At room temperature the resistivity properties of the MoSe2 and WSe2 films were comparable
was between 10 to 10 4 .acm for n-MoSe2 films, 10- 3 to with those of single crystal data as listed in Table II.
10 .acm for p-MoSe2 films and 2.2 .acm to 2.9.10 2 .acm
for the WSe2 films. The temperature dependence of the 2.3 Characterization of the diodes
resistivity for some of the MoSe2 and WSe2 films is shown The heterojunctions were fabricated by the following
in Fig. 2. procedure: gold contacts from the Hall measurements
Temperature dependent Hall measurements were were used as ohmic contacts. In order to produce the
made and the measured Hall voltage UR was in the range heterojunction, 720 A n-ZnO was reactively sputtered
between 10- 2 to 10-4 V. Grain boundaries give rise to onto the masked films from a 99.9% pure ZnO target in a
voltage fluctuations and therefore Neff and the Hall H2/ Ar atmosphere. The distance between the ohmic
mobility were only estimated. A single-carrier model was contacts and the ZnO was about 2 mm. Details are
applied to calculate the Hall mobility JLR and the net described in earlier papers [22,23).

598
Table II.: Electrical properties of MoSe2 and WSe2 single crystals and thin films

type NeD activation Hall a resistivity material reference

(RT) energy mobility from: (RT)
[cm- 3] [meV] [cm 2/Vs] JLH - T-a [ncm]

p 10 18 _10 21 67 2-10 2.1 10- 3-10 MoSe2 thin film this work
n 10 12 _10 13 18-210 0.5-70 2.09-7.97 102_10 4 MoSe2 thin film this work
p 3.10 16 --- 83 2.4 2.5 MoSe2 single crystal Agarwal [18]
n 10 13 _10 17 140 55 1.5 4.0-10 MoSe2 single crystal EI Mahalawy [19]
n 4.10 16 100 70-100 2.3 1.0-3.0 MoSe2 single crystal Spah [20]
p 10 15 _10 16 50-80 100-150 0.9-1.8 1-10 MoSe2 single crystal Jager-Waldau [10]
n 10 12 _10 16 50-450 10-70 2.1-2.5 4-10 3 MoSe2 single crystal Jager-Waldau [21]

p 10 15 _10 20 84-301 0.1-100 2.03-7.37 2.2-10 2 WSe2 thin film this work
p 170 135 2.6 1.0-2.0 WSe2 single crystal Agarwal [18]
p 1017 _10 18 85 110-120 2.7 0.3-2.0 WSe2 single crystal EI Mahalawy [19]
n 10 16 110 100-120 2.1 0.8-2.0 WSe2 single crystal Spah [20]
p 10 16 80 200-330 1.2 0.5-2.0 WSe2 single crystal Spah [20]

These n-ZnOjp-MoSe2 and n-ZnOjp-WSe2 hetero-

'O.-------~r--------------------------.
junctions are the first photosensitive all solid state devices
using MoSe2 and WSe2 thin films.
The different contacts from the Hall measurement
were used as ohmic contacts. Depending on the
ohmic contact the short circuit current density Isc of the
n-ZnO/p-MoSe2-diode varied between 11 and 18 mA/cm 2.
The conversion efficiency 1/ = 0.12 % was very low due to
the small open circuit voltage (Doc = 12 - 16 m V) and a
low fill factor. Fig. 3 shows the I-V characteristic of the diode A B
n-ZnO/p-MoSe2 device under halogen lamp illumination ... I ..... /em·] 8 .9 1~ . 9
(70 mW/cm 2).
U.. [mY] 92.5 208
IT Ill] 2~2 2~8

10r--------,---------------------------,
,......,
"S
u
~ 0+--------4~------------~~--------~ Fig. 4: Illuminated (85 mWjcm 2 ) l-V-characteristics of
E
'-' the ZnOjWSe2 device

COY-measurements were made in the dark. The

shunt resistance RSh and series resistance Rs were
I.., - 18.33 rnA/em'
...., extracted from the impedance plots. The real part of the
U"" m 10.4 mV
c: - 20 impedance of the MoSe2 diode exhibited the same
<I) FF - 26.84 %
;..
;.. 'I - 0.12 % behaviour for two different ohmic contacts. The shunt
::l resistance of this diode was small with RSh = 250 r.?cm 2 ,
o
- 30·~~~~~~~~~~~~~~~~~~~ which might be responsible for the low open circuit
- 10 o 10 20 30 voltage. The series resistance RS = 1.57 r.?cm 2 is also low
Voltage [mV) and makes the high short circuit current understandable.
The shunt resistance of the WSe2 diodes was about two or
three orders of magnitude larger as well as the series
Fig. 3: Illuminated (70 m Wjcm 2) 1- V-characteristics of resistance.
the ZnOjMoSe2 device For the MoSe2 diode the Mott Schottky plot at
10 kHz exhibited a non-linear dependence for low reverse
The n-ZnO/p-WSe2 devices depicted in Fig. 4 bias. This indicates an increase in the acceptor
showed much lower short circuit currents but higher open concentration at the ZnO-MoSe2 interface. The net carrier
circuit voltages. Dnder simulated sun illumination concentration calculated from the slope of the Mott-
(AM 1.5; 85 mW /cm 2), diode A exhibited an open circuit Schottky plot at reverse bias was 7.2.10 17 cm- 3 and is in
voltage Doc = 92.5 mV but only an ISC = 8.9 /LA/cm 2 good agreement with the values of the electrical transport
whereas diode B had an Isc = 16 /LA/cm 2 and an measurements. The extrapolated diffusion voltage VD was
Doc = 20.9 mY. The fill factors were rather similar with about 750 m V and the width of the space charge region
FF = 25.2 % (A) and FF =25.8 % (B). was about 0.2 /Lm.
At present it is not yet understood why the WSe2
devices have a much lower short circuit current than the
MoSe2 devices but better open circuit voltages.

599
 [4] A.M. Goldberg, A.R Beal, F.A. Levy
r: 100 and E.A. Davis,
~ Phil.Mag., 32 (1975) 367
.xl
........
[5] W. Jl!.germann and D. Schmeisser,
>. 80 Surf. Sci., 165 (1986) 143
0
Q
.....
Q) [6J R Tenne and A. Wold,
60 Appl. Phys. Lett., 47 (1985) 707
.~
.....
.....Q)
[7J G. Campet, C. Azaiez, F. Levy, H. Bourezc and
40 J. Claverie,
S Active and Passive Elec. Comp., 13 (1988) 33
...,;::l
s:: [8] R Srivastava and V.M. Pathak,
al 20
;::l Journal of Materials Science Letters, ~ (1990) 294
0"
[9] M. Vogt, M.Ch. Lux-Steiner, H.P. Schweikardt,
400 500 600 700 800 900 1000 P. Dolatzoglou and E. Bucher,
Int'l PVSEC-4, Sydney, Australia, (1989) 79
wavelength [nm]
[10J A. Jl!.ger-Waldau, M. Lux-Steiner, P. Dolatzoglou
and E. Bucher,
Int'l PVSEC-5, Kyoto, Japan, (1990), 287
Fig. 5: Quantum efficiency of the n-ZnO/p-MoSe'l device
[l1J A.Jager-Waldau, M. Lux-Steiner, R.Jager-Waldau,
(normalized to maximum quantum efficiency)
RBurckardt and E.Bucher,
Thin Solid Films, 189 (1990) 339
The external quantum efficiency of the ZnO jMoSe2
[12J A. Jager-Waldau and E. Bucher,
device as shown in Fig. 5 exhibits a high infra-red
Thin Solid Films, 200 (1991)
sensitivity which is due to the strong absorption of MoSe2
(a = 1.1-104 cm- l at 850 nm [4]) above the lowest direct [13J J.I. Pankove,
band gap. Because of the different distances of the ohmic "Optical Properties in Semiconductors",
contacts from the ZnO the two plots have different Dover Publ. Inc. (1975) ISBN 0-486-60275-3
intesities. However, the dependence of the plots from the
[14J R Bichsel and F. Levy,
wavelength is the same. The decrease of the intensity on Thin Solid Films, 124 (1985) 75
both sides of the maximum at 545 nm indicates the
existence of crystallographic defects. This is very [15J K.K. Kam, C.L. Chang and D.W. Lynch,
reasonable if the lamellar structure of the polycrystalline J. Phys. C., 17 (1984) 4031
films is taken into account.· K.K. Kam and B. Parkinson,
J. Phys. Chern., 86 (1982) 463
[16J A.Jl!.ger-Waldau, M. Lux-Steiner, R Jl!.ger-Waldau
4. CONCLUSION and E. Bucher
Proceedings of the POLYSE 90 (1990)
Optical and electronic properties of MoSe2 and
WSe2 films prepared by soft selenization were investigated [17J M. Kimmel,
with respect to photovoltaic application. It was found, Konstanzer Dissertationen Vo1.197 (1988),
that these film properties were in good agreement with Hartung & Gorre Verlag, Konstanz,
those of single crystals. For the first time it was possible ISBN 3-89191-160-2
to fabricate photosensitive n-ZnOjp-MoSe2 and [18J M.K. Agarwal, P.D. Patel and O. Vijayan,
n-ZnOjp-WSe2 heterojunctions with MoSe2 and WSe2 phys. stat. sol. (a), 78 (1983) 469
thin films as bulk material.
From the obtained data it can be concluded that [19J S.H. El Mahalawy and B.L. Evans,
further dev~lopment might result in thin film solar cells phys. stat. sol. (b), 79 (1977) 713
with efficiencies comparable to those of the single crystal [20J R. Spah, Konstanzer Dissertationen Vo1.130 (1986),
solar cells. Hartung & Gorre Verlag, Konstanz,
ISBN 3-89191-060-6
ACKNOWLEDGEMENT [21J A. Jager-Waldau, M. Lux-Steiner, P. Dolatzoglou
This work was supported by the German and E. Bucher,
Bundesministerium fiir Forschung und Technologie, Verh. Dtsch. Phys. Ges., (VI)25 (1990) p. HL 27.4
BMFT contract No. 0328453-B, which is gratfully [22J M. Lux-Steiner, M. Vogt, P. Dolatzoglou,
acknowledged. A. Jl!.ger-Waldau and E. Bucher
Int'l PVSEC-3, Tokyo, Japan, (1987) 687
REFERENCES [23J M. Vogt, M. Lux-Steiner, P. Dolatzoglou
and E. Bucher,
[1] A.R. Beal, W.Y. Liang and H.P. Hughes, 18th Int. Photovoltaic Solar Energy Conf.,
J. Phys. C., Solid State Phys. ~ (1976) 2449 Florence, (1988) 1112
R Beal, W.Y. Liang and H.P. Hughes,
J. Phys. C., Solid State Phys. ~ (1976) 2459
[2J RF. Frindt, J. Phys. Chern. Solids 24 (1963) 1107
[3J A.R Beal, J.C. Knights and W.Y. Liang,
J. Phys. C., Solid State Phys., £ (1972) 3540

600
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

INTERFACE RECOMBINATION VELOCITY OF AI/WSe2 AND ZnO/WSe2 SOLAR CELLS

DETERMINED BY EBIC AND LIGHT INDUCED CURRENT MEASUREMENTS

M. Vogt, K. Friemelt, M.Ch. Lux-Steiner, M. Keil, W. Reetz and E. Bucher

Universitiit Konstanz, Fak. ftlr Physik, Postfach 5560,
W-7750 Konstanz, F.R.G., Fax: 07531-883895 , Phone: 07531-883446

ABSTRACT. Measurements of the dark and light I-V characteristics indicate the invalidity of the superposition
principle in single crystal WSe2 heterojunction solar cells. This is caused by a strong voltage dependence of the
photocurrent I L. The EBIC signal depends on the bias voltage VB in a similar way. This difference between the
ideal photocurrent ho and h can be explained neither by series resistance effects nor by the variation of the
depletion layer width. Good agreement between theory and measurements is achieved by taking the interface
recombination into account. To relate the interface recombination velocity Sr with I L, a model was used previously
developed by Eron for CulnSe2/Cd(Zn)S p-i-n diodes. We modified this model for Schottky diodes and applied it to
the evaluation of EBIC measurements for the first time. The influence of different emitter materials
(AI, Ag, In, ZnO) and their deposition methods on Sr was systematically evaluated. Interface recombin~ti?n
velocities between 9.10 2 cmls and 5.105 cmls have been measured. Photocurrent-voltage and EBIC characterIstics
show excellent agreement between experimental data and the interface recombination model.

1. INTRODUCTION 2.2 EBIC

In the configuration with the electron beam
In the group of layered compound semiconductors, perpendicular to the collecting barrier the EBIC signal can
WSe2 is a promising material for photovoltaic energy be expressed as [6J:
conversion. The net carrier concentration of single crystals
can be varied between 1014 cm- 3 and 1019 cm- 3 by doping. IEBrc = 'T/EaIB-(l-f)/Eeh (3)
Hall mobilities up to 330 cm 2/Vs are typical. Accompanied where EB is the energy of the impinging electron beam, IB
by a high absorption coefficient, WSe2 is a possible the beam current, f is a correction factor taking losses due
candidate for thin film applications. With the lowest to backscattering of electrons from the sample into
indirect energy gaps between 1.16 and 1.22 eV and the account and Eeh is the energy to create an electron-hole
lowest direct energy gap of 1.3 eV, open circuit voltages pair (Eeh oe3.Eg). The charge collection efficiency 'T/(V) is
up to 750 mV can be expected theoretically. Conversion the fraction of carriers collected by the junction. The
efficiencies over 20 % should be reached under ideal voltage dependence arises from the variation of the
conditions for a single junction cell [lJ. Because of the depletion layer width with applied voltage. There are two
layered structure with Van der Waals bonds between the contributions to "I. The first results from the generation in
chemically saturated layers no surface states are expected. the space-charge layer and is assumed to be unity. The
This may provide the possibility to fabricate highly lattice second is governed by the diffusion of generated carriers in
mismatched heterojunctions without interface states. the quasi neutral region to the boundary of the
However for the best ZnO/WSe2 heterojunction solar cells space-charge region and therefore depends on LD. By
open circuit voltages up to 540 m V and conversion analogy to the quantum efficiency it is assumed that there
efficiencies less than 10 % have been reported [2J. This led is no contribution from the emitter. This method is well
to the conclusion that interface states cannot be neglected. established and is a valuable tool to determine diffusion
Results reported by Mahalu [3J support this assumption. lengths of minority carriers.
In this paper the influence of interface recombination on
the quantum efficiency of WSe2 solar cells is reported. 2.3 Interface recombination
Starting with a model by Eron [4J a quantitative analysis If interface states are present, eq. 2 or eq. 3 is a poor
is performed. This model is also applied to EBIC approximation and the interface recombination velocity Sr
measurements of the collection efficieny of induced carriers has to be taken into account. Rothwarf gave a detailed
under bias conditions. analysis [4,7J of this case. He showed that the
photocurrent depends on the voltage dependent electric
field Fr(V) at the interface. This leads to an invalidity of
2. THEORY the superposition principle as demonstrated in fig.!.
2.1 Quantum efficiency 8r---~----------,
As reported by Phillips [5], quantum efficiency
measurements (QE) are strongly influenced by the series NE 4
resistance R. of the solar cell and the load resistance RL of u
"- 0
«
the electronic circuit. Therefore the quantum efficiency Q
has to be calculated from the measured quantum efficiency .s -4
Qex by: ]:;-
'GO -8
Q = Qex-[HID-(q/nkT)(Rs+RL )}, (1) "(l)

..., -12
"0
where n is the diode quality factor and ID the dark
current. For n+p- and Schottky-barrier solar cells Q
depends on the applied voltage V, the minority carrier
t" -16
::0
u -20
diffusion length LD as well as the absorption coefficient a
of the base, the transmission T of the emitter, the -200 0 200 400 600
reflection R and can be expressed by
Voltage [mVl
Q = T(l-R){l-Kexp{-aW}} (2)
where W denotes the depletion width Figure 1: Experimental and calculated 1- V-curves of
W = [2coe:/qNA·(VD-V)J, K = (HaL D)-l and the diffusion an illuminated Zn O/WSe2 solar cell.
voltage YD.

601
Consequently the quantum efficiency QS can be written as: 3. EXPERIMENTAL RESULTS
Q[1+5J(V) / JLFJ(V)jl
3.1 Crystal growth and device fabrication
Qf(sr(V)) (4) WSe2 single crystals were grown by vapor phase
transport in a closed quartz ampoule as reported earlier
where JL is the carrier mobility of the emitter. Under the [8J. The fabrication of ZnO /WSe2 solar cells has been
condition that the drift velocity JLFJ(V) is smaller than the described elsewhere [2J. The thickness of the RF
thermal velocity Vth, the influence of empty interface magnetron sputtered ZnO layer was about 70 nm.
states of density NJ on SJ is given by SJ = UNJpVth. Schottky diodes with sputtered Al emitter have been
According to eq. 5, N';p is a function of the applied voltage. prepared in the same way. In order to investigate the
influence of the deposition process on surface
(5) recombination, AI, In and Ag have been deposited by
thermal evaporation. The metal thickness of all emitters
was about 60 A. Before the evaporation the system was
where Un / p are the electron/hole capture cross evacuated to a pressure of less than 4.10- 6 mbar. All
sections, NI is the total density of interface states and devices had Mo back contacts. The characteristic data of
ex = NJp/(Nr-NJp). For n+p junctions and Schottky diodes the illuminated solar cells (85mW/cm 2) are given in
almost the entire depletion layer can be located in the table 2.
lower doped base. Therefore the density of electrons nr(V)
at the n side of the interface is equal to the electron 3.2 Measurement principle
density no of the emitter and independent of voltage. At To eliminate the influence of the load resistance on
the p side of the interface the density of holes is the QE and EBIC measurement the bias voltage of the
PJ(V) = po·exp(q(V-Vo)/kT) where Po is the bulk hole device has been sensed. It was kept constant by a feed-
concentration. Introducing quasifermilevels, this expression back mechanism. Therefore the voltage variation of the
for Pl(V) no longer holds. As depicted in fig. 2 we assume source output was proportional to the induced current .
that the difference Ll. V will be smaller than V due to The schematic arrangement is shown in fig 3.
emptying of the donor like interface states with applied
voltage. For this reason we change the expression for pj(V)
into Pr(V) = poexp(qB'(V-Vo)/kT). Depending on the
density of interface states and their distribution within the
energy band gap, B is a function of V.

Bose
l
Emit er

Figure 3: Schematical illustration of the experimental

o emp y stotes arrangement.

.. empty states (EBle) 3.3 Spectral quantum efficiency

EV-------' From spectral quantum efficiency measurements
• OCCUpied sla es under various bias conditions it is evident that interface
recombination takes place. This results from the
Figure 2: Schematic band diagram of a n+p wavelength dependence of Q(bias)/Q(OV) [9J. Especially at
het.erojunction. ~,p,l: Quasi fermi level under QE short wavelengths this ratio is dominated by interface
condition, ~,p,2: Quasi fermi level under EBIC recombination because the variation of depletion layer
condition. width has no influence on quantum efficiency (fig. 4).

Assuming a constant density of interface states, B is 1.4

1~0
independent of voltage. With no :> Po, Njp(V) can be

~
expressed by: ~

120 1.2
NjiV) = NrfA '·exp(qB'(V- Vo)/kT) + D'] (6) ~
Using this expression, a parametric description of SI >-
<.>
c:
100
- - - - - -;~ I-O(~o:vf-::~- 1.0
~
is obtained:
:§
-
Q)
80 0.8 0
o( - o.sV)
SJ(V) = Aexp(BV) +D (7) 0.6
"-
Q) 60 .-...
>
Combining eq. 4 and eq. 7 results in a function used to fit E 0.4
0
:> 40
the corrected (eq. 1) experimental data. C 0
0
This discussion shows that in theory the influence of :J 20 0.2
0
interface recombination can be taken into account by
multiplying the expression for the ideal photocurrent with L.-_<----'--'~.-J.~-'-~_:_"_:~~~ 0.0
400 500 600 700 800 900 1000
the function f(SJ(V)), The model used is independent of
the generation process of the minority carriers. Therefore Wavelength [nm]
the influence of interface recombination on the EBIC
signal can be described by the same model.
Figure 4: Spectral quantum Q(V) efficiencies of a
PEBIC(V) = IEBIC·f(SJ(V)) (8) ZnO/WSe2 solar cell under different bias conditions.
For bias voltages V = 0.4 V and -0.8 V the ratlOs
Q(D V)/Q(V) are given.

602
3.4 Measurements and modelling 3.5 Influence of the emitter material on surface
As depicted in fig. 5 and fig. 6 Q(V) and .,,(V) recombination
exhibit a strong dependence on bias voltage which cannot The origin of the interface states in ZnO/WSe2
be explained by a theory which ignores interface diodes remains still unknown but two possible explanations
recombination. Even with the correction of the can be given: (a) sputter-induced defects and (b) chemical
experimental data due to eq. 1, the deviation between reactions at the interface. More investigations with
experiment and theory still remains large. Good agreement different emitter materials were carried out, using
between experiment and theory can be achieved by taking sputtered ZnO and Al as well as evaporated AI, In, and
interface recombination into account. This was done by Ag. The solar cell data are given in table 2. The results of
fitting eq. 4 and eq. 7 as described. Vo was determined by QE and EBIC measurements are shown in fig. 7 and fig. 8
C-V measurements. and summarized in table 3.

0.3 VodmVJ hdmA/cm 2J FF[%J .,,[%J

ZnO (1) 532.7 16.5 51.4 5.3
Al (1) 207 5.4 56.4 0.7
Al (2) 505 9.5 66.6 3.7
In (2) 316 2.0 57.2 0.4
Ag (2) 172 10.2 38.6 0.8

,, Table 2: Characteristic data of WSe2 solar cells with

E 0 .1
o a
~
000 meosurements
.. .. . .. corrected meosurements
, different emitter materials under AM 1.5. (1)
::0
C - - theo. cu....... with S, sputtered; (2) evaporated.
o - - - theo. cu ...... e without S,
::0
o
0.0 1.4
-1.5 -1.0 -0.5 0 .0 0.5 1.0 ......
Bios voltage [ V] E
..... 1.2
0
Figure 5: Bias dependent quantum efficiency of a ~
c:
1.0
ZnOjWSe2 solar cell at .\ = 850 nm. (*) measured >-
data, (0) corrected data, ( - - ) litted data
..
u
c: 0 .8

0.8
-..
:~

E
0.6
••••• ZnO spuHered
00000 AI evaporated
0.4 •• ••• AI spuHered
::0
••••• Ag
---------------- -- --- , C
0 ••••• In
evaporated
evapora ted
::0
0.2
~ 0 .6 0
>- 0.0
u
c: - 1.5 - 1.0 - 0 .5 0 .0 0 .5 1.0

- [V]
.~ Bios vo ltage
.~ 0 .4
IV
c: .0000 measurements Figure 7: Experimental and calculated data of t.he
.2 ...... .... correc·ted measuremenls
0.2 - - theo . curve with S. quantum efficiencies of WSe2 solar cells WIth
U - - - theo. Cu""'" without S, different emitter materials.
~
o
U
0 .0
- 1.0 - 0 .5 0 .0 0 .5 1.0 1. 4
Bias volfage [V] ......
E 1.2
'-
0
Figure 6: Bias dependent collection efliciency of a c:
ZnOjWSe2 solar cell. The electron beam acceleration 1.0
>-
voltage was 9 ke V. (*) measured data, (0) corrected u
c:
data, ( - - ) lit ted data .,!! O.S

From both measurements it can be concluded that SI

increases with bias voltage (see table 1). The bias
-.,
~

c:
0 .6
•• • .• ZnO spul!ered
independent part D of eq. 7 can be neglected in the case of .2 0 .4 ••••• AI evaporated
quantum efficiency measurements. U II III 0 II 0 AI sputte red
••••• In evaporoted
In the case of EBIC measurements SI is dominated by D. ~ 0.2
We attribute this effect to a stronger activation of '0
U
interface states due to inelastic scattering of electrons. For 0.0
bias voltages larger than 400 m V int~rfac~ recombination -1.5 -1.0 -0.5 0 .0 0 .5 1.0
velocities of about 5.10 5 cm/s are obtamed m both cases. Bios voltage [V]

Parameter QE EBIC F'Igure 8: Experimental and calculated data of the

collection efliciencies of WSe2 solar cells with
NA cm· 3 J 3.8.10 15 5.0.10 15 different emitter materials.
Lo JLmJ 0.19 0.4
B l/VJ 12.5 15.3
SI~OV) cm/sj 4.2.10 2 2.6.10 5
SI 0.5V) cm/s 1.7.105 4.0.10 5

Table 1: Parameters of the theoretical curves in

figure 5 an d figure 6.

603
 sputtered sputtered evaporated evaporated evaporated
ZnO Al Al In Ag
NA [em- 3 ] 1.0.0 17 3.8010 16 7.7.10 16 3.5.10 17 7.7.10 16
LD [p,m] 0.145 0.19 0.2 (0.25) 0.12
Q B [ l/V] 13.9 14.2 9.0 6.0 24.9
E Sr( OV) [em/s] 1.1.10 3 6.6010 3 3.9.10 3 3.8.10 5 1. 2 .10 3
Sr(0.2V) [em/s] 1.8.10 4 1.1.10 5 2.4.10 4 1 . 3.10 6 1.7.10 5
Sr( O. 4V) [em/s] 2.8.10 5 2.0.10 6 1. 5.10 5 - -
VD [V] 1.0 0.9 1.0 0.8 0.6

NA [ em- 3 ] 4.5.10 16 9.0.10 16 1.1.10 17 6.9.10 16

E LD [p,m] 0.11 0.14 0.16 0.13
B B [l/V] 17.4 16.8 11.0 16.9
I D=Sr( OV) [em/s] 3.0.10 5 6.4.10 5 1.1.10 6 3.7.10 5
C Sr( O. 2V) [em/s] 3.1.10 5 8.1.10 5 1.2.10 6 5.4.10 5
Sr( O. 4V) [em/s] 5.1.10 5 - 1.6.10 6 -
Table 3: Parameters from QE and EBIC measurements of WSe2 solar cells with different emitter materials.

From QE measurements it can be seen that the References

sputtered Al diode exhibits stronger interface
recombination than the evaporated one_ This supports the [1] E.Bucher: Physics and chemistry of materials
explanation of sputter-induced defects_ On the other hand with low dimensional structure, Kluver
Sr is comparable for devices with a sputtered ZnO and an Academic Publishers, Doredrecht NL, to be
evaporated Al emitter_ This is contrary to explanation (a). published
Additiona.lly the experiments on solar cells with an
evaporated In and Ag emitter indicate the influence of the [2] M.Ch.Lux-Steiner, M.Vogt, P.DolatzQglou,
emitter material on Sr. For example at a bias voltage of A.Jager-Waldau and E.Bucher,
+200 m V the In/WSe2 and Ag/WSe2 samples exhibit 3rd Int. Photovoltaic Science and Engineering
interface recombination velocities of 1.3.106 cmls Conference, Tokyo, Japan, 687 (1987)
and1.7-10 5 cmls respectively, while Sr of sputtered ZnO
samples is about 1.8.104 cm/s. This is a hint that chemical [3] D.Mahalu, A.Jakubowicz, A.Wold, R.Tenne,
reactions could take place at the interface. In a.ll cases Sr is Phys.Rev.B, 38, 1533 (1988)
strongly voltage dependent and the constant term D can
be neglected. [4] M.Eron, A.Rothwarf,
In contrast to this, EBIC measurements show a J.AppLPhys., 57, 2275 (1985)
sma.ller voltage dependence of Sr, whereas D is the
dominant parameter. Genera.lly, Sr determined by this [5] J.E.Phillips, M.Roy
method is larger than determined by QE. This fact can be 20th IEEE Photovoltaic Specialists Conf., Las
explained by an activation of interface states. EF ,p,1 is Vegas, 1614 (1988)
strongly influenced by this process and nearly independent
of bias voltage. With a Sr of about 1.1-106 cmls and [6] C.J.Wu, D.B.Wittry,
6.4.10 5 cmls the AI/WSe2 diodes showed the largest J.AppLPhys., 49, 2827 (1978)
Sr-values of a.ll samples. Because of the large V oc of the
evaporated Al/WSe2 cell and the sma.ll Sr of this sample [7] A.Rothwarf
obtained from QE measurements we concluded that Al 13th IEEE Photovoltaic Specialists Conf.,
creates deeper impurity levels caused by chemical Washington D.C., 1312 (1978)
reactions. Sputtering possibly changes these reactions and
therefore the nature of the impurity levels can be changed. [8] R.Spah, Dissertation University of Konstanz,
From these data it cannot be concluded, that sputtering (1986)
has a destructive influence on solar cell performance.
[9] K.W.Mitchell, A.L.Fahrenbruch and
R.H.Bube,
J.AppLPhys., 48, 4365 (1977)
5. CONCLUSIONS

In this work the invalidity of the superposition

principle in WSez diodes was demonstrated. This is caused
by a bias dependence of the photocurrent. From the
results of voltage dependent spectral quantum efficiency
measurements, we attribute this to recombination effects
at the interface. A model was introduced in order to get
quantitative results. Good agreement between
measurement and theory was achieved. Comparing
samples with sputtered and evaporated emitter led us to
the conclusion that sputter induced defects are not the
single reason for interface recombination. The interface
properties were even more affected by the choice of the
emitters material. Experiments with sputtered and
evaporated emitter gave a hint that the chemical nature of
the Al/WSez interface can be altered by the sputtering
process.

604
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ZnO TRANSPARENT CONDUCTING THIN FILMS DEPOSITED BY ULTRASONIC SPRAY PYROLYSIS,

FOR SOLAR ENERGY CONVERSION.

A. TIBURCIO-SILVER (0) and J. C. JOUBERT.

Laboratoire des Materiaux et du Genie Physique, ENSPG-INPG.

U.R.A. 1109 du CNRS. B.P. 46, 38402 st Martin d'Heres, FRANCE.
(O)CINVESTAV-IPN, Depto. de Ingenieria Electrica-SEES.
Apdo. postal 14-740, 07000 Mexico, D.F., MEXICO.

ABSTRACT
Undoped and indium-doped zinc oxide thin films were deposited by
ultrasonic spray pyrolysis (pyrosol process) onto soda-lime glass substrates.
Several precursors and solvents were tested for undoped zinc oxide thin film
deposition. Only a solution of zinc acetate on methyl alcohol allowed us to
obtain good quality ZnO thin films; indium doping was achieved by adding a few
drops of a saturated solution of In(CsH102)3 on acetylacetone. Resistivities
as low as 2'10-3 n-cm, Hall mobilities as hi~h as 21 cm2/v-s and effective
carrier concentrations as high as 1. 3 '10 20 cm- have been obtained. Electron
concentrations are always lower than indium contents on films. Average optical
transmission on the whole visible range as high as 85% for the best conductive
films has been obtained. Refractive index of layers is modified by ~he ~rowth
temperature and indium doping. Haacke's figure of merit up to 5'10- n- in a
500 nm. thick films was obtained. Indium doping improves stability of the
electrical properties of ZnO conducting films. This films are suitable for
solar energy conversion, particularly in photovoltaic applications (ZnO/CdTe
and ZnO/Si solar cells).

I. INTRODUCTION. monly used in the laboratory were not homo-

geneous in thickness. In order to improve
Transparent conducting thin films have the uniformity in film thickness, we have
been investigated over several years be- developed a new experimental set-up (10),
cause of their electro-optical properties, showed in fig.-1.
that made them sui table for solar energy
applications, such as heterojunction solar
cells, heat mirror coatings and photother-
mal conversion systems. Sn02 (Undoped and
doped) and ITO (tin-doped indium oxide)
have been the most widely used materials.
Recently, ZnO has prompted a great interest
as transparent conductor, owing to its pro-
mising properties. ZnO is an n-type wide
band-gap semiconductor (Eq=3.3 eV.); its
conductivity can be controlled over several
decades by hydrogen thermal treatement or
by an adequate doping. Transparent con-
ducting films of ZnO have been prepared by
several techniques such as sputtering
(1,2), reactive evaporation (3), chemical
vapour deposition (4) and pneumatic spray
pyrolysis (5,6). In the present work we re-
port on the electrical and optical proper-
ties of undoped and indium-doped ZnO thin
films deposited by ultrasonic spray pyro-
lysis (pyrosol process). 1-
A) ULTRASONIC SPRAY GE ERATOO
B) SOLUTION CONTAINER
II. EXPERIMENTAL DETAILS. G I ULTRASONIC GENERATOR
H) SUBSTRATE HEATER ( MOLTEN T IN BATH)
2.1. Film deposition. S) SUBSTRATE
Undoped and indium-doped ZnO films
were deposited by ultrasonic spray pyro- Fig . l.-New experimental set-up for Pyrosol deposition
lysis (patented in France as "Pyrosol
process"). The Pyrosol process is a kind of
spray pyrolysis where the aerosol is produ-
ced by ultrasonic vibrations, instead of Several precursors, dopants and sol-
pneumatic means (7). In this way it is vents were tested for undoped and indium-
possible to have a better control on the doped ZnO thin film deposition (10). After
mean size of aerosol droplets, and hence a number of trials, only a solution of
better homogenity of the films (8). A dihydrated zinc acetate in methyl alcohol
complete description on the Pyrosol process (C=0.2 M.) allowed us to obtain good qua-
can be find in (9). In spite of their lity undoped ZnO thin films. It is also po-
interesting characteristics, the films ob- sible to obtain undoped ZnO thin films by
tained in the experimental apparatus com- the Pyrosol process using C2HsOH and a mix-

605
ture of C2HsOH + H20, but the final aspect Preferential orientation along the (002)
of films is hazed. Indium doping was achie- axis is observed, changing from parallel to
ved by adding a few drops of a 'saturated perpendicular to the substrate surface as
solution of indium acetYlacetonate in ace- Ts increases. Indium doping modifies this
tylacetone (C=0.3 M.) to the alcoholic so- behavior ,becoming strongly oriented along
lution of zinc acetate. The films were the (110) axis parallel to the substrate
deposited onto silicon and soda-lime glass with doping.
substrates simultaneously, cleaned prior to The mean grain size increases linearly
deposition following the procedure descri- whith T. for undoped ZnO films. For indium-
bed in (11). Substrate temperatures ranged doped films' the mean grain size tends to
between 350 and 575°C. Purified air was saturate for high substrate temperatures.
used as carrier gas. Indium contents in films increases
with indium contents in the solution, and
2.2. Characterization of films. decreases with T (fig.-3).
Structural studies of films were
20
performed with a computerized SIEMENS
KRISTALLOFLEX 9/29 Diffractometer (eu ano- iii 550'C
de, Akal=1.5441 A). Quantitative volumetric • 500'C
analysis of zinc and indium contents in o 45O'C
films were made by the atomic absorption
technique. Depth profiles of representative
films were obtained by SIMS and ESCA tech-
niques using a MIG 300 Ionic nanoprobe
(Vacuum Generators Instruments) for the
first and a SSX-100-06 Spectrophotometer
(Surface Science Labs. Inc.) for the se-
cond one. Sheet resistance was measured by
the four point probe technique. Resistivi-
ties, Hall mobilities and effective carrier
concentrations were obtained from Hall-Van
der Pauw measurements. Optical band gap of o 2 3 • 5 6
films was calculated after transmission InlZllsol (%)
measurements in the uv-visible range Fig.3.-Indium contents in films as a function of
(190-800 nm) made with a VARIAN SuperScan 3 indium contents in the solution and subs-
spectrophotometer. Total transmittance and trate temperature.
reflectance spectra in the uv-visible-near
IR range (0.3-2.5 ~m.) were obtained with a
BECKMANN Acta M IV double beam spectropho- As we can see, indium contents in films are
tometer equiped with an integrating sphere. higher than indium contents in the solution
Refractive index, absorption coefficient for low Ts. Similar efect of Sn accumula-
and film thickness were calculated using a tion in Im03: Sn spray deposited films on
method developed by Manifacier et al. (12). borosilicate glass substrates has been ob-
Calculated film thicknesses were verified served by MANIFACIER et al. (13). The ori-
by measuring the steps in SIMS analized gin of this effect remains unexplained.
films, using a Dektak Sloan profilometer. SIMS profiles show that Zn, 0 and In
are distributed uniformly in films (fig.-4).
This is an outstanding feature of pyrosol
III. RESULTS AND DISCUSSION. deposition.
3.1. Structural properties.
Films obtained are strongly texturi- to· 10'
zed. Preferential orientation of grains de-
105
pends on the substrate temperature (Ts) and In In I In
indium doping in the solution (%Inso\), as L
-,
2. z~ z"
10 1
we can see in fig.-2. III
0 0 ~iO> No

1'\0 .... j ~

~ ~
K .J SIO'
10' 10 1

~ fi 1m (500nm l -l film(500nml~
Zn06'J.!n . 550"C T =SSOoc T =S2S oc
5 5
67.ln 501 27.ln 501
ZnO 2'!.i n .55O"C

Fig.4.-SIMS profiles of indium-doped 2nO films.

ZnO. 550"C

We may note an indium accumUlation at the

surface of films. In spite of this, low
angle X-ray analysis does not reveal any
30' 40' 50' 60' 70' spurious phase other than ZnO.
29
3.2. Electrical properties.
Fig.2.-Influence of the substrate temperature and Electrical properties of films are
indium doping in the solution on the prefe- strongly influenced by substrate tempera-
rential orientation of grains. ture and indium doping (fig.-5).

606
 This can be explained by the fact that ca-
e \ll-cml jJ{crrftV- 51 rrier concentrations as measured by Hall
t0 2 -r-----------------------------. effect in polycrystalline materials are not
30 the overall carrier concentrations but only
10 t _
effective carrier concentrations taking
part in electrical conduction across the
20 films (15, 16). I t is also possible that
excess of indium segregates to grain boun-
dary, producing there mixed compounds wich
can be either amorphous or microcrystalline
10 or both.
10-2 ---2 0 / 0 1n. o l
.... .... 6 % [nsol
3.3. Optical properties •
Total transmittance and reflectance of
some representative films are shown in
400 550 600 fig.-7 (transmission loss due to substrate
included).

Fig.Sa.-Influence of substrate temperature on the

electrical properties of 2nD films.
-
'... ....'''~-'':''':- .. ....: - - -.....
"

,
~ . . .... --.-
.... - '-.,- - "-
~ ________________________ ~ 101
....

10 '
----- ----- --
u
.
.
I
,I

.
,_ ,,
'.
'.J
I '.
1.0 1.5 2.0 2.5
I """
""41 ... P
... ...... .... 0-.... .. ...... ..• ..... ... .

Fig.7.-0ptical transmittance and reflectance of ZnD

o 2 4 6 8
films.
18
(---) TS=4S0oC. In/2n=Z%. N=1. 7xl0 19
( - ) TS=47SoC. InlZn=Z%. N=S.Sxl0 19
(-.-) Ts=47SoC. InlZn=6%. N=7.4xl0 Z0
Fig.Sb.-Influence of indium doping on the electrical ( ... ) TS=SSOoC. InlZn=6%. N=1.3xl0
properties of 2nD films.
From fig.-7 it is clear that selectivity of
Behavior of resistivity with Ts can be as- ZnO pyrosol deposited thin films is not
cribed to an excess of Zn (or sUbstitutio- high enough (as compared with that of
nal atoms, in the case of indium doping) in Sn02:F and ITO films). This could be
the films (14). Mobility increases with Ts expected, considering the higher carrier
as grain size does. concentrations obtained in our films. Cal-
As we can see in fig.-6, effective culating the plasma cut-off wavelength
carrier concentrations are always lower
than indium contents in films (represented A =21lC[(C
poL
C m')/(e 2N)fl2
by the straight line in fig.-6).
where c is the speed of light in vacuum, C
the permitivity of free space, c the d.c. o
10 22
dielectric constant of ZnO, m' L the effec-
tive mass of electrons in ZnO, e the elec-

/
NltO'~ tron charge and N the effective carrier
he r-1m concentration in film, one obtain, for the
best case (N=1. 3 '10 20 cm- 3 )
" 500'C
• 550'C 4.67 j,Lm for m*=0.3 m [
see ]
E A= { or ° ref. 17
u
5.26 j,Lm for m*=o.38 m see ]
i
p [
ref. 18
10 20 °
optical spectra correspondjng to this case
seems to indicate that m =0.3 m is the
0.5 2 • 6 10 20 . 0 100
most probably value for the effec~ive mass
0.1
in our films.
WZn IiIm (%) Refractive index of films obtained at
higher Ts are about n=2.0 on the whole vi-
sible range and is not very dependent on
Fig.6.-Evolution of effective carrier concentration indium doping. On the other hand, refrac-
with indium doping in films. tive index 'reduces sensitively for lower

607
Ts, being about n=1.87 and more dependent 8. G. BLANDENET, M. COURT and Y. LAGARDE,
on doping. Thin Solid Films, 77 (1981) 81.
9. G. BLANDENET, Y. LAGARDE, J. ERNOULT
and J. P. TARDIF, "Le procede Pyrosol",
%In n (Ts=550°C) n (Ts=425°C) Compte Rendu DMG N° 153/79, CEA-CENG,
sol
DMG-LEMM, 14 dec. 1979. Grenoble,
0 2.06 - 1.98 1.90 - 1.85 France.
2 2.10 - 1.95 1.92 - 1.84 10. A. TIBURCIO-SILVER, These
Nouveau Regime, ENSPG-INPG, Grenoble,
Docteur
6 2.10 - 1.95 1. 79 - 1.80
France, 1989.
TABLE I.-Dependence of refractive index on indium 11. J. MIMI LA-ARROYO , J. A. REYNOSO, E.
doping and substrate temperature. (visible SAUCEDO and J. C. BOURGOIN, J. Cryst.
range) . Growth, 68 (1984) 671.
12. J. C. MANIFACIER, J. GAS lOT and J. P.
FILLARD, J. Phys. E, 9 (1976) 1102.
3.5. Figure of merit and time-stability. 13. J. C. MAN I FACIER, L. SZEPESSY, J. F.
The best Haacke's figure of merit (19) BRESSE, M. PEROTIN and R. STUCK, Mat.
for ZnO P~foSO_t dep~sited thin films is Res. Bull., 19 (1979) 109.
about 5·10 Q. Th~s value is smaller 14. J. C. MAN I FACIER, M. DE MURCIA, J. P.
than those obtained for spray deposited FILLARD and E. VICARIO, Thin Solid
SnOa:F (20) and ITO (21) thin films. Films, 41 (1977) 127.
Some experiments under illumination 15. J. Y. W. SETO, J. Appl. Phys., 46,
and darkness have been carried out with our (1975) 5247.
films, showing that indium doping improves 16. J. W. ORTON, B. J. GOLDSMITH, J. A.
the time-dependence of the stability of the CHAPMAN and M. J. POWELL, J. Appl.
electrical properties of ZnO thin films. Phys., 53 (1982) 1602.
This can be explained by assuming that 17. E. ZIEGLER, A. HEINRICH, H, OPPERMANN,
interstitial zinc concentration reduces Phys. Status Solidi (a), 66 (1981) 635.
with indium doping (22). 18. A. P. ROTH and D. F. WILLIAMS; J. Appl.
Phys., 52 (1981) 6685.
19. G. HAACKE, J. Appl. Phys., 47 (1976)
IV. CONCLUSIONS. 4086.
20. A. TIBURCIO-SILVER, M. Sc. Thesis,
ZnO thin films have been deposited by CINVESTAV-IPN, Electrical Engineering
the pyrosol process (french patent). Dept., Mexico, D.F., 1985.
An extensive study on their structu- 21. A. L. DAWAR, J. C. JOSHI, J. Mater. Sci
ral, electrical and optical properties has 19 (1984) 1.
been made. 22. R. A. SWALIN, "Thermodynamics of Solids
ZnO transparent conducting Pyrosol John Wiley & Sons Inc., New York-London
deposited thin films match requirements for 1962, chapter 16, p. 310.
photovoltaic applications such as ZnO/CdTe 23. J. A. ARANOVICH, A. L. FAHRENBRUCH and
(23), ZnO/WSea (24), ZnO/p-MoSea (25), R. H. BUBE, Proc. 14th IEEE Photovolt
zno/a-si (26) and ZnO/mono-si solar cells. specialists' conf., San Diego, CA, jan
In the latter case, some preliminary 7-10, 1980, IEEE, New York, p. 633.
studies have been carried out (27), showing 24. M. CH. LUX-STEINER, M. VOGT, P.
that ZnO can sUbstitute favorably the well DOLATZOGLOU. A. JAGER-WALDAU and E.
known SnOa:F and ITO films. BUCHNER, 3rd. Int' 1 Photovol taic Sci.
and Eng. Conf., Tokyo, Japan, 1987,
p. 687.
V. ACKNOWLEDGEMENTS. 25. A. JAGER-WALDAU, M. LUX-STEINER, P.
DOLATZOGLOU and E. BUCHNER, 5th Int'l
One of the authors (A. T. S.) is gra- Photovol taic Sci. and Eng. Conf. ,
teful to mexican gouvernement (COSNET- SEP) Kyoto, Japan, 1990, p. 287.
for their finnancial support while working 26. A. SANCHEZ, A. ORTIZ and J. CAMPOS, Me-
at LMGP-ENSPG-INPG. morias XII Reunion Na1. de Energia
Solar, Puebla, PUE, Mexico, octubre 5-7
REFERENCES. 1988, ANES, Mexico, p. 77.
27. J. H. MONDRAGON, M. de la L. OLVERA, A.
1. J. B. WEBB, D. F. WILLIAMS and M. MALDONADO and R. ASOMOZA, Memorias XIV
BUCHANAN, Appl. Phys. Lett., 39 (1981) Reunion Nal. de Energia Solar, La Paz,
640. BCS, MeXico, octubre 1-5, 1990, ANES,
2. o. CAPORALETTI, Solar Energy Mater., 7 Mexico, p. 276.
(1982) 65.
3. D. E. BRODIE, R. SINGH, J. H. MORGAN,
J. D. LESLIE, L. J. MOORE and A. E.
DIXON, Proc. 14th IEEE Photovoltaic
specialists' Conf., San Diego, CA,
january 7-10, 1980, IEEE, New York,
p. 468.
4. A. P. ROTH and D. F. WILLIAMS, J. Appl.
Phys., 52 (1981) 6685.
5. J. ARANOVICH, A. ORTIZ and R. H.BUBE,
J. Vac. Sci. Technol., 16 (1979) 994.
6. K. L. CHOPRA, S. MAJOR and D. K.
PANDYA, Thin Solid Films, 102 (1983) 1.
7. G. BLANDENET, Y. LAGARDE and J. SPITZ,
5th Int'l. Conf. on Chemical Vapor
Deposition, London, 1975. Published by
The Electrochemical Society, Princeton,
NJ, USA, 1975.

608
WfH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

SOLVENT EFFECT ON ZoO THIN FILMS PREPARED

BY SPRAY PYROLYSIS

I.-F. GVILLEMOLES, D. LINCOT, P. COW ACHE and I. VEDEL,

Laboratoire d'Electrochimie Analytique et Appliquee, V.A. CNRS 216.
ENSCP, 11 rue P. et M. Curie, 75005 Paris FRANCE.
Tel: (331) 43 54 5385 ; Fax: (331) 43 25 7975

ABSTRACT; Solvent (water and alcohols), doping and temperature effects on the characteristics of ZnO thin
films prepared by spray pyrolysis have been studied. Improvement of transparency and conductivity of the layers by
the use of alcohols is shown. The best results are obtained at 400°C using a solution of zinc acetate in methanol
containing 1% at. of an indium salt (average transparency: 92%, resistivity: 5.10-3 ncm). It is also shown that the
presence of indium in solution lowers the growth rate and improves transparency and conductivity of the films.
These results are briefly discussed according to the properties of the solvents and the growth mechanisms
specific to spray pyrolysis.

INTRODJJCTION noozzle, improving by the way the lateral

homogeneity of the layers.
Zinc oxide is a n-type' semiconductor of Main parameters of deposition are the distance
increasing importance, especially in photovoltaic between the nozzle and the substrate (20 cm), the air
applications where it has proven to be very effective flow rate (7 l/min), the solution flow rate (5ml/min),
(1) because of its high transparency in the visible the solution concentration (O,IM Zn and eventually
range (over 90%, the gap value is 3,3 eV) combined O,OOlM In for an extrinsic doping), the spray duration
with an attainable low resistivity (below 1O- 3o.cm). (typically 20 minutes), the nature of the solvent (see
Thin layers can be deposited by various methods: below), the nature of substrate (soda-lime glass Imm
sputtering (2), reactive evaporation (3), MOCVD (4) thick) and the regulation temperature (400°C and
and spray pyrolysis (5-9). This work aimed to 500°C) corresponding to a substrate surface
optimize the properties of ZnO layers obtained by temperature lower by about 100°C due to gas flow
spray pyrolysis (at temperatures low enough to be cooling. The nature of ambient atmosphere (air or
compatible with the deposition on other photovoltaic nitrogen) was not found to induce significatives
materials) by a study of the specific influence of differences between the films formed.
precursors and solvents. Little attention has been paid by the authors
Spray pyrolysis is a technique in wich zinc to the solvent despite its importance in the deposition
oxide precursors are dissolved in a solvent which is steps. The solvent has to solubilize precursors, to
then finely dispersed in small droplets with a volatilize easily and to allow a good pulverisation
pneumatic noozle (size range: 10 to 30 11m). These (narrow distribution in size) of small droplets. Some
droplets are then transported to a heated substrate solvents fulfill those conditions and were studied:
where they vaporise and react depending on the water (as in (5,9», mixtures of water (25%) and
reaction conditions (temperature, presence of water) alcohol (75% vol.), where alcohol stands either for
according to (when precursor is zinc acetate), (5,11): methanol or ethanol (as in (7» or isopropanol (as in
(6,8», and methanol alone.when necessary, acetic acid
a) Zn(CH3C02>2 ---> ZnO+C02+CO(CH3)2 was added to prevent formation of hydroxides as in
or solutions with water content (which causes
b) Zn(CH3C02)2+H20 ---> ZnO+2CH3C02H precipitation of zinc and indium hydroxide), but not
for methanol solutions where no precipitation occured.
One of the advantages of the method consists
in its great adaptability to industrial processing which
means the possibility to produce large surfaces at low RESULTS AND DISCUSSION
cost.
1) Comparison between nrecursors
We found that zinc acetate gave better results
EXPERIMENTAL on all aspects (optical, electrical and morphological
properties) than zinc chloride, which is attributed to
The apparatus designed for the deposition by vapor phase transport of zinc acetate (sublimes at
spray pyrolysis is of classical conception and is 180°C (11» and absence of pollution with CI (5).
equipped so as to allow a sweeping movement of the Consequently, we used it systematically afterwards.

609
 Water Alcohol and Water Methanol

0% at. In 1% at. In 0% at. In 1% at. In 0% at. In 1% at. In

Growth rate(nm/min) 100 85 55 35 45 45

Resistivity (Ohm-em) 17 18 0,35 0,04 0,4 0,005

Average transmission (%),

·n the 450-850 nm range 40 50 89 92 82 92

Tab. 1: Typical values of the characteristic parameters for layers prepared at 400°C with various so/vents and
various indium concentration after 20 min. of deposition.

2) Comparison between solvents the deposition temperature. SEM photographs showed

The main discrepancy between solvents a decreasing surface roughness from layers prepared
appears in the comparison between water and other with H20 to mixtures water-alcohol and from
alcoholic solutions in the range of deposition mixtures to methanol. The grains boundaries of the
temperature studied (Tab. 1): water gives granular layers are more revealed with methanol than with
layers of poor transparency and poor conductivity, mixtures water alcohol (see photographs a·and b).
with X-Ray spectra similar to ZnO powder spectra Methanol was found to gi ve the highest
(Fig.I). Oppositely, alcoholic solutions give layers of conductivity, due to a better electronic mobility as was
good transparency (around 90%) with interference shown by Hall effect measurements. It indicates that
fringes (Fig. 2) and a more abrupt absorption edge at less scattering defects are present in these layers.
the gap value wich indicates a better crystallinity and a Carrier density was similar for all the doped samples
lower defect density near the band edge. obtained with alcoholic solutions (about 3.10 19cm-3)
irrespective of the deposition temperature. The
difference in conductivity originates from higher
100 mobilities (30 y- I cm 2s- 1) with methanolic solutions
002
than for water-alcohol mixtures (3 y- 1cm 2s- I). Sheet
resistances as low as 15 n have been obtain~d with
80 longer deposition times (one hour).
For the water-alcohol solutions, there was
some difference between the various alcohols and it
r-..
;3 was noticed that variations were exactly in opposite
cd 60
'-"
orders in doped and undoped samples. In-doped samples
>. were in this order for increasing transmission and
.....
~

til conductivity: methanol>ethanobisopropanol. The

I::: 110 best average transmission (450-850 nm), 95%, was
~
~
I:::
40
/ obtained at 400°C with a methanol-water solution.
All this was discussed in terms of solvents
properties related to the various possible growth
20 mechanisms: methanol seems more likely to allow a
gas phase transport of the precursors onto the substrate
and so diminishes the defect density.

3) Effect of temperature
15 20 25 An increase of deposition temperature from
8 (degree) 400 to 500°C induces an improvement of the optical
and electrical characteristics of the layers obtained from
Fig 1: R-X spectra of In-doped ZnO layers prepared at water solutions and a deterioration in those obtained
4000(; with: methanol (upper) and water (lower). from alcoholic ones due to the formation of a white
haze on the surface wich was attributed to ZnO powder
Conductivity is improved of 2 to 4 orders of formed in the vapor phase (13). This did not occur
magnitude, and X-Ray spectra show a marked with methanolic solutions, the layer transmission
preferential orientation with the c axis perpendicular to becing quite indepcndant of the deposition temperature.
the substrate, increasing with the layer thickness and

610
SEM photographs and RX spectra showed an should be increased, because mass transport to the
increasing grain size (about a,s Ilm at 500°C against substrate diminishes as temperature increases due to
O,21lm at 400°C) and a better crystallinity and gas convection and thermophoretic forces that repell
orientation along the c axis. . droplets and reactants.
The growth rate was lower with higher
temperature, despite the fact that the reaction kinetics

100

90

80

70
~
c:
0
"(i)
60
CI)
.§
50
'"
~ 40
..-
30 ./
;""
20 f
;
i
;
10 i

0
350 450 550 650 750 850
Wavelenght (nm)

Fig. 2: Comparison of the specular transmission of In-doped layers prepared at 400°C from different solutions.

(mainly with methanol, see tab. I) and a decrease of the

0.4 0.04 growth rate with water containing solvents but not
with methanol. Crystallite morphology is modified by
. .•.. / .,A
indium addition: they are smoother and seem scarcely
,-..
0.3 :;0 smaller in undoped samples (220 nm against 250 nm
~ 0.03 in doped samples as determined by RX mesurements)
/ ' ~
SQ)
/ .
..
(ii '
E c. and shown on SEM photographs. Grains size and
0 ./ < resistivity vary with the thickness (Fig. 3).
~ .
..... 0.2
u
0.02 ---.
601)
0
:::T 5) General discussion
N 3 To interpret the effects of solvent and
'Vi
0.1 0.01 '"n temperature, one has to distinguish between the three
.~ 3
.... '--' growth mechanisms involved in the deposition process
0 -t- (12):
O 0 -1- When a droplet reaches the substrate, the
0 0.5 1.5 2 remaining of solvent vaporizes rapidly while the
precursors react violently, conducting to a disordered
Film thickness (Micrometers) growth.
Fig.3: Evolution of resistivity and grain size with -2- If all the solvent is vaporized, the precursors
film thickness in a doped sample prepared at 400°C sublimate and diffuse towards the substrate where they
with methanol. react: this is similar to CVD.
-3- If the gaseous precursors react in the vapour
4) Effect of indium phase, fine powder of ZnO is formed. This powder can
stick on to the substrate.
Addition of indium induces an improvement
of the conductivity and transparency of the films
611
The process shifts from the conditions of 1 to those of SEM micrographs a and b : in-doped layers prepared at
3 by increasing the substrate temperature or solvent 400°C. Thickness is about 0,5 ~ m for each.
volatility or with smaller droplet size or kinetic a (upper): methanol solution
energy. b (lower): ethanol-water solution
The conditions of 2 are the more favourable
to the formation of good layers. Because of its
properties, methanol allows the smallest dispersion in
droplet size and kinetic energy and is also volatile
enough: with this solvent the conditions of 2 are
dominant. A more extensive discussion of these
phenomena is given in (11).

CONCLUSION
The influence of the solvent on the growth
mechanisms (and hence on the quality of the layers)
was underlined and tentatively interpreted. Zinc acetate
in solution in methanol was found to give the best
results (P= 5.1O- 30cm, Taverage=92%) at 400°C
without annealing, which compares to the best results
obtained by this method after annealing. Methanol
appears so to be a promising solvent for further
investigations. The effect of indium on conductivity,
transparency and on growth rate was shown, the last
two findings beeing new for ZnO as far as we know.

ACKNOWLEDGEMENTS: This work is supported

by a PIRSEM-AFME contract on thin films solar
cells (N° 90N80/0078). L. Robiola has also to be
warmly thanked for the SEM studies.

References
(1) D. Pier, K. Mitchell, 9th E.C. PhotovoltaYc Solar
Energy Conf., 1989, p 488
(2) R. Schropp, J. Appl. Phys., QQ (1989) P 2027
(3) J.H. Morgan, D.E. Brodie, Can. J. Phys., @
(1982) P 1387
(4) Chi Kwan Law, S.K. Tiku, K.M. Lakin, J.
Electrochem. Soc., 127 (1980) P 1843
(5) J. Aranovitch, A. Ortiz, R.H. Bube, J. Vac. Sci.
Technol.,.lQ (1979) p 994
(6) S. Major, A. Banerjee, K.L. Chopra, Thin Sol.
Films, J.Q8. (1983) P 341
(7) D. Cossement, J.M. Streydio, J. Cryst. Growth,
11. (1985) p 57
(8) H.R Paes, L.MOC. Pinho, W. Losch, 9th E.C.
PhotovoltaYc Solar Energy Conf., 1989, p 63
(9) C. Ebespacher, A.L. Farenbruch, RH. Bube, Thin
Solid films, l3.6. (1986) P 1
(10) J.F. Guillemoles, D. Lincot, P. Cowache, 1.
Vedel, C. R Acad. Sci, in press.
(11) H.G. Mc Adie, J. Inorg. Nucl. Chern., 2li (1966)
p2801
(12) J.C. Viguie, 1. Spitz, J. Electrochem. Soc., ill
(1985) p 1394
(13) A. Tiburcio-Silver, J. C Joubert, M. Labeau,
Thin Sol. Films, 121 (1991) P 195-214

612
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

SPRA YEO ZnO THIN FILMS AS OPTICAL WINDOW IN CuInSe 2 BASED SOLAR CELLS

K. Belghit, M.A. Subhan, U. Riilhe*, S. Duchemin, J. Bougnot

Centre d'Electronique de Montpellier, Universite de Montpellier II, Sciences et Techniques du Languedoc

Place E. Bataillon - 34095 Montpellier Cedex 5 - France
*I.P.E., Universitat Stuttgart - Germany

ABSTRACT. The purpose of this paper is the study of un doped and (AI or In) doped ZnO layers as electrically conduc-
ting and optically transparent window in a photocell. ZnO thin films are obtained by chemical spray pyrolysis using
two different Zn source materials. The properties of these metal oxide films are analysed and related to the nature
of the Zn source and dopa,,~ts. We ~~ow that it is possible to obtain ZnO layers with a transmission at about 80%
and a resistivity between 10 and 10 .'l..cm.

1. INTRODUCTION. 3. CHARACTERIZATION.

In the field of energy conversion a promising 3.1. ZnCI2 as Zn source:

device is constituted by CuInSe2 based solar cells. The Zn source was ZnCl2 and the nozzle displa-
Our studies (l) reported until now on CdS-CuInSe2 cement an elliptical sweeping.
sprayed photocells were successfull in the high 3.1.1. Undoped films: .
short circuit value (up to 30 mA/cm') in spite of For Ts = IflO°C and C = O.IM/1 (standard values)
a low photovoltage (~ 300mV) leading to efficiencies a minimum in the resistivity (3.n.cm) is observed
between 4 and 5% for !em' area. In a previous with a solution flow rate f = 6cc/min and a carrier
paper (2) we showed that the poor quality of the gas flow N2 : 71/min, which are the selected values
CdS films was largely responsible for the photovol- afterwards.
tage failure in the sprayed photocells. Arco Solar '" Ts variable, C and f fixed (standard values):
achieved 14% efficiency with some cells by stacking When Ts is decreased from 1f10°C to ;50°C a little
ZnO on CdS (very thin)/CulnGaSe2/Mo/glass. So improvement of the preferential orientation along
our studies concern the metal oxide deposition and the 002 axis is observed, the grain size kll1m) being
properties variation related to the fabrication parame- quite unchanged. The optical transmission is also
ters and the doping content. increased as shown in Fig.2, the optical gap and
the electrical properties f7~ma!~ unaffected (Eg
2. EXPERIMENT. = 3.24eV, p =3..ncm, n = 10 cm ).
ZnO layers with thicknesses between 0.5 and
211m are sprayed onto heated substrates (pyrex
glasses) from aqueous solutions containing the consti-
tuent materials and with the help of a carrier gas
(N2 : 71/min) as previously described for CdS or
I.
C.I.S. samples (3). Two different Zn sources with
different molar concentrations were used : ZnCl2 00
dissolved in deionized water and a small amount
of HCI (2 vol%) or Zn(CH3C02)2 dissolved in a 60

mixture of methanol (75 vol%) and deionized water

(25 vol%) or isopropyl alcohol (75 vol%) and deionized ••
water (25 vol%).
To deposit homogeneous, conducti ve and transparent
films some fabrication parameters, especially the 2.
substrate temperature Ts, the solution flow rate
f and the molar concentration C of the aqueous
solution must be taken into account. Standard values ), ""'
were fixed to obtain samples with an hexagonal
structure and a preferential orientation along the Fig.2. Optical transmission versus Ts for a 211m
002 axis (Fig. J). The structural and physical proper- thick film.
ties are studied in relation with the fabrication
parameters variation around the standard values. 'ItC variable, Ts = 41 DoC and fixed (standard
value) :
When C is being decreased the crystallographical
properties strongly change: the preferential orienta-
tion decreases (Fig.3) and the hexagonal shape of
the crystallites becomes triangular with a reduction
in the grain size. The resistivity which decreases
slowly, is always in the range I to lO.Q.cm (Fig.lf)
whereas the optical properties are unchanged.
" Heat treatments:
Generally the resisti vi ties of as sprayed ZnO films
mentionned in the literature (If) (5) (6) are higher
than those obtained on our samples. However, the
values obtained, although low, are too high to use
the ZnO films as conducti ve electrode in place
of I.T .0. or 5n02. So, to increase the conductivity,
heat treatments under H2 atmosphere (360°C, 16min)
Fig.1. Typical X-rays diffraction for standard ZnO.

613
 were performed leadi ng to decrease the resi sti vi ty
by two orders of magnitude and to increase the
free carrier density by three orders of °magni tude
as shown in Fig.lf, the mobility decreasing by one
order. . ........ @
~: : : ~~~: ~

Fig.5. Typical X-rays diffraction pattern for:

ZnO(In) wi th In = 2at%

Fig.3. X rays diffraction patterns for C

Similar change in the orientation and in the shape
of the grain was also observed by D. Cossement
et al. (6) in In-doped samples.
10' r -_ _ _ _ _ _ _ _ _ _ -,10"..-----------1 JI< Electrical properties:
In spite of that expected the resistivity increases
10"
with the dopant concentration, more especially
JO= in the case of In-doped films (Fig.6). At the same
..
10' 1

IOU ti me rhe free carrier densi ty and the mobility decrea-

10' se. If heat treatments are performed, the same

(r
lOI'
effects are observed than on non-doped samples
10" decreases by two orders of magnitude).
1011' OW

10 I , AJ dtpO:i,ud
10' r------------,
p(O.CIII)
• na'"C
.. tla"C

10'

. . ..
~
1

Fig.4. Resistivity and free carrier density as molar

10'
concentration of the spray solution ZnCI 2 •
• .. ,o"C

~11"'iIo)
3.1.2. Doped films:
In the Ii teratua:e by Al or In doping (6) (7) (8) the 10 : 0' 10 1 10 I 10 1 10 1
conductivity and t~ free carrier I~ens~ty respectively
increase up to 10 :n:\-:m-' and 10 cm '. The In and
Al dopants are introduced in the spray aqueous Fig.6. Resistivity as a function of doping concentra-
solution in lhe form of InCl3 or AIC13 with different tion for ZnO films obtained from ZnCI 2 •
molar concentrations.
iE: Crystallographical and morphological properties:
The preferential orientation which slowly decreases -t Optical properties:
when the films are Al doped, entirely disappears The optical transmission remains unchanged what-
in favour of the 101 orientation when In doping ever the dopant but a decrease in the absorption
is used (Table l)·(f, (7 ~) edge is observed when In content increases (Fig.~).
3,5
~ _____ %at 0.01 0.1 1 2 Gap (eV)
1002/110L I
ZnO(AI) 16.5 13.5 12.5
3,3
I
ZnO(!n) 25.6 0.3 0.17
I
3,1
Table I
Preferential orientation variation versus [In] and
[AI] 2,9

The crystalli te shape is also affected : the hexa-

2,7
gonal structure not obvious for as-deposited films
can be restored by heat treatment under N2 atmos- in (at % )
phere (350°C, 45min) for AI-doped films whereas
the crystallite shape of In-doped films is always 2~0·3 10·' 10. 1 10· 10 1 10 2
triangular even after annealing Fig.7. Gap as dopIng concentration (Zinc acetate
as Zn sourCe).

614
3.2. "l n{ CH3C02)2 as Zn source : at a temperature of 360°C for 10 min. and 45 min.
To avoid the effect of chlorine, as discussed respecti vel y does not change the resisti vity value.
later, on the electrical and optical properties of This is an indication that the chlorine present in
ZnO films [undoped and dope(jJ zinc acetate a small quantity in the AICl3 or InCl3 spray solution
Zn(CH3C02)2 has been chosen as the source mate- is insignificant in the samples.
rials like other investigators (4) (7). The films with Yet ZnO films so obtained are comparable with
a thickness about I~m are now produced by a linear I.T.O. only in respect of optical transmission in
sweep of the nozzle (X-direction). The growth parame- the visible range but its electrical conductivity
ters chosen to deposit ZnO films are very close is still one order less than that of I.T.O. So, instead
to the growth parameters for the preparation of of using ZnO alone as a window layer, ZnO(In) depo-
I.T.O. which uses also the linear sweep of the spray- sited on I.T.O. has been used to realize CulnSe2
ing nozzle (X-Y) : substrate temperature Ts = 350°C backwall solar cell. e.1.S. (R2/Ri)/ZnO/I.T.0. photo-
for a flow rate of spraying solution f = 9cc/min, cell with an area of Icm' have yielded efficiency
a carrier gas flow N2 : 2.5/min and no post deposi- from 0.4% to 0.6% with an open circuit voltage
tion treatment done. about Voc 150mV and a short circuit current Jsc
near 16mA whereas, when a very thin CdS layer
3.2.1. Undoped films. is sprayed between ZnO and C.I.S., an enhancement
Standard films undoped have the same characl~ist~'3s of Voc is obtained (270mV) but Jsc is lower (lOrnA)
as ZnO produced from ZnCI2( f#1a.cm, n = 10 cm , leading to an efficiency (0.6%) still far below the
~ = 40cm'/v.s.), and an optical transmission T for efficiency value about 5% (I) obtained on classical
wavelength in the visible range about 80%. I.T.O./Cd¥C.I.S. sprayed cells. To enhance the
photo voltaic parameters, a new optimization of
3.2.2. Doped films: the C.I.S. growth is certainly necessary.
Doping by In and Al decreases the resisti vi ty of
ZnO films. However the decrease is more marked
for In ~3ping as shown in Fig.S where a resistivity REFERENCES.
of 5.10 -fL .cm. is o?ltained for 2at% (0.002M/J)
In doping whereas 1.10 .fL .cm is only obtained for (I) S. Duchemin, J. Bougnot, A. EI Ghizizal, K.
Al doped films. The transmission of I~m thick film Belghit, 9th E. Photo Sol. En. Conf. (Sept 89)
is about 80% for Al doping and 85% for In doping. Friburg (Germany)
(2) S. Duchemin, J. Bougnot, M. Kaka, M. Cadene
Thin Solid Films, 136 (1986) 289-298
10 1 (3) S. Duchemin, V. Chen, J.e. Yoyotte, J. Bougnot
p(n .em) M. Savelli, 8th E. Photo Sol. En. Conf. (May
88) Firenze (Italy)
(4) J. Aranovich, A. Ortiz, R.H. Bube, J. Vac.
10° ~ Sci. Technol., 16 (1979) 994

• • (AI)
•
(5) M.S. Tomar, F.J. Garcia, Thin Solid Films,
90 (1982) 419
10. 1
• (6) D. Cossement, J.M. Streydio, J. of Crystal
Growth, 72 (1985) 57-60

.. (7) S. Major, A. Banerjee and K.L. Chopra, Thin

Solid Films, 108 (1983) 333-340
10. 2
. .. .. .. . [In] (8) H.R. Paes, L.M.O.C. Pinho and W. Losch,
9th E. Photo Sol. En. Conf. (sept 89) Friburg-
(Germany).
(at%)
10' 3
0,00 0,01 0,02 0,03 0,04

Fig.8. Resistivity versus doping concentration for

ZnO films obtained from zinc acetate.

The change in the preferential orientation is identical

to those Al and In doped ZnO films prepared by
spraying ZnCI2 aqueous solutions.

4. DISCUSSION.

Whatever the Zn source, undoped films (as

deposi ted) are not high conducti ve to act as an
electrode in place of I.T.O. To obtain conductive
films post deposition treatments or doping is neces-
sary.
However when the films are sprayed from ZnCl2
solutions, doping is ineffecti ve, probably of a compen-
sation effect of the 02 vacancies by the chlorine
(4) occuring in the samples during the fabrication
process. In this case heat treatments only decrease
the resi sti vi ty by one or two orders of magnitude.
The choice of Zn(CH3C02)2 as Zn source is to
avoid the chlorine effect since as deposited films
exhibit resistivities of ,I.ctcm. decreasing by as
much of two orders of magnitude when the films
are doped. Post deposition heat treatment of doped
films in an atmosphere of hydrogen and nitrogen

615
 POSTER SESSION 3B
High Efficiency Crystalline Silicon Solar Cells
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ANALYSIS OF THE OPERATION

OF LASER GROOVED SILICON SOLAR CELL
Antonio G.M. STROLLO and Gianfranco VITALE
Department of Electronic Engineering, University of Napoli
via Claudio 21, 80125 Napoli, Italy.

ABSTRACT. A significant improvement in the performance of Silicon solar cells has been
achieved in recent years with the so-called laser grooved cell /1/ . In these cells, the top contact is
formed within deep grooves, formed on the cell surface by laser ablation. This cell structure has
demonstrated to achieve very high efficiencies exceeding, in some cases, 20% /21-
This paper provides an insight in the origin of the very good performance of the laser grooved
cell, with the help of a closed-form solution for the two-dimensional minority carrier distribution
within the cell.
Results of the analysis demonstrate that the improvements in the efficiency of the laser grooved
cell, as compared with the classical, planar junction, cell, is due to a combination of factors that
include, among others: an improvement in the Spectral Response of the cell, particularly in the
long wavelenght range, and a substantial current collection that takes place in the part of the
junction along the grooves, where a low resistance path is encountered.

with the corresponding quantities, calculated for a Cell

with plane junction allows to evaluate the improvements
I. INTRODUCTION.
that are to be expected for the Laser grooved Ceil, as
A two dimensional model of the laser grooved Solar shown in Section 3.
Cell has been developed, with the purpose of providing a
means to understand the operation of the cell and to pro- II. ANALYTI C AL MODEL.

vide a design tool. The model is based on a closed-form The analysis of the laser grooved cell, is based on
solution of the fundamental transport equations within a closed-form solution for the two-dimensional minority-
the 2D domain limited by the grooves. This result has carriers distribution in the P base of the cell, accounting
been achieved by suitably approximating the problem, as for the optical generation and the Shockley Read Hall
discussed in detail in the following Section. The use of a recombination terms.
closed - form solution allows to carry out a large paramet- With reference to Fig.1 and by assuming low-
ric analysis of the cell, with a reasonable computer effort. injection regime and uniformly doped P base, the con-
Furthermore, because of the nature of the solution, the tinuity equation for electrons may be written as:
contributions of the various regions of the cell are easily
separated each other, thus allowing to understand many n(x , y) aF -a(z+w)
details of its operation.
..,2
v
(
nx,y - -V--D
) _
e (1)
n n

where n(x, y) is the excess minority-carriers concen-

tration, w is the depth of the N region and of the space-
w y charge region on the top of the cell (see Fig. 1) the other
symbols have their standard meanings.
:5. Contact The solution of eqn. (1) is accomplished by us-
ing the superposition principle. The function n(x, y) is
written as the sum of a 1-D term g(x) which takes into
a P Base account the optical generation, and a 2-D term f(x, y)
c which accounts for the two-dimensional distribution of the
minority-carriers:

n(x , y) = g(x) + !(x, y) (2)

with:

Fig. 1 - Schematic of the laser grooved Solar Cell d2 g _ .!L _ aF - ar(z + w)

dx2 - £2n Dn e (3)

Examining the geometry of the cell, Fig. 1, it is seen \1 2f =L

L2 (4)
that the 2D effects are likely to take place mainly into n

the base region of the cell, namely the region limited by: The boundary conditions for the function g(x) are
the grooves, the (top) p-n junction and the back contact. choosen as:
This part of the overall solution allows to calculate the
contributions of the (P-type) base region to the Spectral g(O) = 0 (5a)
Response and to the short - circuit current. A comparison
dg
This work was supported by: ENEA, Italy Sg(e) = - Dn-(c) (5b)
dx

6 19
 the boundary conditions (10)-(11) which can be rewritten
In this way the term g(x) becomes the standard one- as:
dimensional distribution for minority-carriers in the base,
that is, the distribution that would result if there were
al
no vertical grooves on the top of the cell. In eqn.(5b) u(b - x)/(x, a) + u(x - b) ay (x, a) =
S is the minority-carrier surface-recombination velocity
which characterizes the back contact of the cell. From = u(b - x)g(x) 0~ x ~ c (14)
eqns.(3),(5) one obtains:
where u(x) is the step function (u(x) = 1 for x > 0
and u(x) = 0 for x < 0). By substituting the eqn (12) in
aFT (14), one obtains:
g(x) L2 n e -ow [ cosh (x/ Ln ) -
2
a n-1
(aD n - S)e- OC + £: sinh(c/ Ln) + S cosh(c/ Ln) N

Ssinh(c/Ln) + £:cosh(c/Ln) L Ij [ cosh(hja/ Ln)u(b - x) sin(kjx/ Ln)+

j=1

sinh(x/ Ln) - e- o ",] (6) (h j / Ln) sinh(hja/ Ln)u(x - b) sin(kjx/ Ln)] =

= u(b - x)g(x) (15)
The boundary conditions for the function I(x, y) are
specified as follows:
By expanding in series of eigenfunctions sin(kix/ Ln)
both the right-hand-side and the left-hand-side of eqn.
al (15) and equating the two series one obtains finally:
ay (x,o) =0 O~x~c (7)

f
al
SI(e,y) = -Dnax(c,J) = 0 O~y~a (8) N b

L Ij [cosh(hja/ Ln) sin(kjx/Ln) sin(kix/ Ln)dx+

,=1 0
g(O) + 1(0, y) =0 (9)
j
O~y~a

(hj / Ln) sinh(hja/ Ln) sin(kjx/ Ln) sin(kix/ Ln)dX]

g(x) + I(x,a) =0 O~x~b (10)
b

al
ay (x,a) =0 b~x~c (11) fb

= g(x) sin(kix/ Ln)dx lor i = 1...N (16)

o

Eqn. (7) reflects the symmetry of the structure with

Equations (16) are a set of N linear equations which
respect to the x-axis (see Fig. 1.), the condition (8) is
related to the surface-recombination velocity at the back supply the unknowns Ij.
contact of the cell (as in eqn. (5b)), while the minority-
III. THE SPECTRAL RESPONSE
carrier distribution is imposed to be zero along the space-
charge region of the top junction through the eqns. (9)- The part of the spectral response of the cell due to
(10). the minority-carrier generation within the base layer is
Strictly speaking, eqn. (11) holds exactly only if obtained by observing that the photogenerated carriers
d = 0, that is if the structure is symmetric with respect are collected not only along the top contact of the cell
to the line y = a (see Fig. 1.); however, assuming the (that is for x = 0 in Fig. 1.), but also along the vertical
boundary condition (11) also when d "f 0 greatly simpli- groove:
fies the development of the two-dimensional model, and
it is expected to be still reasonably accurate if a ~ d.
The diffusion equation (4) is solved by using the tech-
nique of separation of variables. By imposing the bound- SR ~ 2q~. U. J_(O, y)dy + ' / J.,(x, 0)"'1 (17)
ary conditions (7),(8),(9), the function I(x, y) is expanded
in Fourier series as:
From the eqns. (2) ,(17) one obtains:

N
f(x,y) = L!isin(kjx/Ln)cosh(hjY/Ln)
j=1
(12)
S R = Dn dg (0) + Dn
F dx aF
fa

aa l (0, y)dy + DFn

x a
f b

aal (x, a)dx

y
if the series is truncated to the first N terms, hj = o 0

VI + kJ and the eigenvalues kj are obtained as the solu- (18)

From the last equation it can be seen that the spec-
tion of the equation: tral response of the cell is the sum of three components.
The first one is the standard one-dimensional term; the
SLn second one takes into account the reduction of the x-
kj = - - tan(kjc/ Ln) (13)
Dn component of the gra.dient of the electron distribution due
The coefficients Ij in (12) are obtained by imposing to the carriers collection along the vertical grooves. This
term is, therefore, negative and it is given by:

620
 IV. RESULTS.

As a first application of the model discussed in the

SR" = ~; L fi(kj/hj)sinh(hja/L n ) (19) preceding Section, some parametric study has been car-
} ried out, with the purpose of understanding the basic op-
eration of the Laser Grooved Cell. Results are compared
The third term in eqn. (18) accounts for the pho- with those obtained for a Cell with plane junction.
tocurrent collected along the vertical grooves; this is a Fig. 2 displays the electron distributions, over the
positive contribution to the overall spectral response: Base domain, for two different photon energies. In this
Figure, the front p-n junction lies along the Y axis. The
cell thickness, C = 4S0",m, the grooves are along the X
SR" = ~; L fj(h j/kj )sinh(hja/L n }[l- cos(kjb/L n ) axis and are 200 micron deep, namely less than one -
}
half the Cell thickness. Other geometrical parameters
(20) are: d = 40JLm, a = 150",m, the diffusion length is 120
micron.
A first - approximation value of the electron com- At a photon energy of 1.2 eV, the carrier distribution
ponent of the short circuit current has been obtained by displays a maximum at a significant depth inside the Cell,
modeling the solar spectrum as the Black - Body radia- resulting in a strong lateral collection along the grooves.
tion at 6000 K, which approximates the extraterrestrial As the photon energy is increased to 1.5 eV, the maximum
(AMO) Solar radiation. Because of the way the Spectral in electron distribution is very close to the front junction
Response has been obtained, the contributions of the front and the carrier collection is substantially the same as that
junction and the lateral grooves are easily separated each
other.

.
Q/
.8
.
~
0
Q,
Q/ .6

.....
EI::
Cj .4
0
Cl
Q, .2
(/)

0
1 1.5 2 2.5 3
Pholon Energy, Ev

Fig. 3 - Component of the Spectral Response due to the

base layer of the cell, compared with the same component
for the plane junction cell.

-a -
of a Cell with plane junction. Thus the Spectral Response
improves essentially in the range of low photon energies,
as shown in Fig.3.
Fig. 4 shows the short circuit (electron) current den-
sity as a function of the distance between grooves. In these
calculations, the width of the groove has been taken into
account as a shadowing degree of the Cell. Calculation of
the 1D curve assumes that a top contact, of the same size
as the groove, partly shadows the cell.
Results shows that, obviously, the short circuit cur-
rent improves for both cells as the distance between
grooves is increased. For the same geometry, the 2D struc-
ture displays an improvement of about 7 % with respect
to the 1D cell, an improvement that is to be added to the
reduction in contact resistance of this type of cell.
Subdivision of the 2D current into its components, as
shown in Fig.4b, indicates that grooves collect only 1/6 of
the maximum current of the 2D Cell thus indicating that
most of the current is still collected through the front
junction. It is worth to note that the current, collected
by the front junction in the 2D model, is different from
- b - the (total) current in the 1D model due to the different
carrier distributions.
Fig. 2 - Two dimensional electron distributions
If the depth, b, of the grooves is reduced from 200
within the base of the cell for (aJ 1.2 eV and (bJ 1.5 eV
photons. to 50 microns, Fig.5, the contribution of the grooves to

621
 .E
30
..E 30

u
u 20 ~ 20
"-
<C g
....
....8 10 0
N
10 groove

0 0
0 100 200 0 100 200 300
Cell Size ",m Cell Size ",m

-a- Fig. 6 - Components of the 2D electron current if the

width of the grooves, b, is reduced from 40 to 20 microns.

current that is collected laterally is about the same.

Further developments of the model, by including the
.E 30
two - dimensional effects on the dark current, and the ef-
u fects of the other layers, will provide other elements for
~ 20 the understanding of this device and a guideline for opti-
mum design.
..,t
0
N
10 groove

0
0 100 200
Cell Size ",m
REFERENCES
- b-
/1/ M.A. Green, C.M. Chong, F. Zhang, A. Sproul,
Fig. 4 - The component of the short - circuit current J. Zolper and S.R. Wenham: "20% Efficient Laser
due to the base layer. (aJ Comparison between the lD Grooved, Buried Contact Silicon Solar Cells" Proc.
and the 2D models. (bJ Breakdown of the 2D solution 20th IEEE Photovoltaic Specialist Conference, Las
into its components. Vegas, USA, 1988, pp. 411-414
/2/ M.A. Green, A.W. Blakers and S.R. Wenham: "Im-
provements in Silicon Cel and Module Performance"
current collection reduces substantially and the 2D cell
Proc. 9th E.C. Photovoltaic Solar Energy Confer-
shows no intrinsic improvements with respect to the ID
ence, Freiburg, FRG, 1989, pp. 301-304
structure.
If instead the width, d, of the groove is reduced from /3/ G. A. M. Hurkx: " On the sidewall effects in submi-
40 to 20 microns, Fig. 6, the total short circuit current crometer bipolar transistors", IEEE Trans. Electr.
increases, due to the reduced shadowing effect but the Dev., vol. ED-34, n. 9, pp. 1939-1946, sept. 1987.

.E 30

u
~ 20
....8
0
N
10
groove

0
0 100 200
Cell Size ",m

Fig. 5 - Components of the 2D electron current if the

depth of the grooves, b, is reduced from 200 to 50 microns.

622
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PROGRESS ON HIGH EFFICIENCY SILICON SOLAR CELLS

J.Alonso, A.Luque, I. Tobias, G.Sala and A.Cuevas
Instituto de Energia Solar - Universidad Politecnica de Madrid
E.T.S.I. Telecomunicaci6n - Ciudad Universitaria
E28040 MADRID

ABSTRACT
An approach to obtain high efficiency 5i solar cells, based on model-assisted technology is
presented. Using PC-1D program, the Deep Emitter cell, developed in our Institute, has been accurately
modeled. The theoretical results obtained in this way are in good agreement with experimental ones. It
allows us to investigate the Deep Emitter cell behavior. Then it is used to propose and develop a novel
structure to improve the efficiency of this cell. We call it the Deep Emitter Rear Locally-diffused cell.
Results on these cells for p and n substrates are presented here.

INTRODUCTION. measured sheet resistance (once the base doping is

Our effort for improving the cell efficiency given). This is a simple calculation, well known,
is largely based on model-assisted technology also done automatically by the PC-lD. A check is
development. For it fast modeling is' required and the diffusion depth, easily measured by staining.
we use extensively the PC-1D code [1], that allows In some wafers [2] Auger and incremental Hall and
for accurate and sound simulation of the cell sheet resistivity has been done showing the
behavior and, what is more important, do it very identity of doping density and free majority
fast, so that extensive model effort can be done in carriers and the accuracy of the method, at least
reasonable time. It is true that PC-ID, being for deep emitter cells (this simplified method is
one-dimensional has some limitations, but yet it is certainly not applicable to common emitters)
useful, if used with some skill, to give good The surface concentration goes from
information even on two dimensional problems. 1-2xl019 cm- 3 for sheet resistivities of about 100-50
Our baseline is the deep emitter cell, O. The junction depth is of 1. 3-1. 8 ~m for base
developed at ~ur/tnsti tute [2]. This cell is based resistivities of 0.3-20 Ocm, being this the reason
on a two steps diffusion emitter and an for the name of deep emitter given to these cells.
alloyed-diffused Al B5F. Our first task is to
accurately model the cells based on this 2.2 Back doped-region profile.
technology. Then the main limitations of this cell The back region is doped with AI. For it the
are investigated and a novel structure to suppress Al is deposited on the rear and alloyed at high
these drawbacks is proposed and developed. We call temperature so that a euthectic is formed, and
it Deep Emitter Rear Locally-diffused cell. simultaneously a deep Al diffusion is produced.
Results on this cell are presented also. When the cell is cooled again the euthectic
recrystalizes and segregates in a AI-rich 5i phase
2 STRUCTURE DEFINITION. and a Si-rich Al phase (probably with abundant
The structure, of n+pp+ type, is defined by nitrogen). The depth of the diffused junction, well
the base resistivity (linked to the doping by the away from the front of recrystalization, makes this
well known Irving curves) the emitter profile and junction much more perfect than the more common
the p+ profi Ie. alloyed AI-junctions, so that it can act perfectly
The base resistivity is measured with the well not as a high low junction but also as a pn
known four points probe. FZ from Wacker are used junction, if desired.
through this work. This diffused region starts at the interface
The cell thickness is measured physically. of the liquid phase and the solid phase during the
alloying-diffusion process. The liquid phase acts
2.1 Emitter profile. as an infinite source so impOSing, rather
The common way to define the emitter profile perfectly, the thermodynamic equilibrium value of
is based on the knowledge of the technology used in the solid solubility of Al in 5i at the 5i surface.
its formation. It consists first on a When the recyrstallization is produced this surface
pre-deposition step, where the amount of P is remains away from the real surface in a distance
controlled by the pre-deposition temperature and that is given by [3]:
the gas mixture entering into the open tube. The
theoretical study of this process is difficult, and Xrec = WA!(pAllpSi)[F/(l-F) - Feut/(l-Feutll (2)
for our purpose unnecessary. The final results is
the introduction in the wafer of a certain charge where WAI is the thickness of the deposited Al
Qem of P atoms situated in a very shallow layer layer, the pAl and PSi are the densities of the
(less than 0.1 ~m). AI-layer and of the 5i (note that WAiPAI is the
The pre-deposition is followed of a drive-in weight of the Al layer so that information of the
step after which the doping profile C(x) takes, layer density is not actually needed) and F and
rather accurately [1], the shape of a Gaussian: Feut is the liquidus mole fraction at the process
temperature and at the euthectic temperature,
C(x) = (QP/vnDt)exp-(x/IP)2; with lp = 2/Dt (1) respectively.
Feut 0.117; for the conditions of our
D is a well known value that depends on the process F = 0.62, and WAI = 2~m , so that Xrec =
drive-in temperature (for P, with anomalous 1.7f.lm.
diffusion, the low concentration value of P, below The doping profile in the conditions explained
1019 cm 2, must be taken); t is the drive-in time. We here follows a erfc profile, starting at this
usually work at condi tions (those reported here) depth. From it to the surface the profile is
with lp = 0.544 ~m. In this curve only the value of complex as it takes the solid solubility of Al in
Q, or if preferred the concentration in surface, is 5i at the temperature at which the recrystalization
undetermined but it can be adjusted by the the of each layer takes place, if the cooling is slow

623
enough as to allow for equilibrium, or at the
Refl clance
temperature of the process is the cooling is very 0 . 5 ~~~~~-------------------------,
fast. In general i t will be intermediate between
these two si tuations, decreasing when approaching - - Grid shadow -I- 0 grid
to the surface, but this is only a guess. Actually
these details are not very important for us.
For simplicity we have simulated the doping
profile as a gaussian, peak concentration that of
the All~0Ii~3solubility at the process temperature,
1.9x10 cm and IAI = 2VDt = 1.732 ILm, and a
maximum located at 1.7 ILm. We believe that this
reflects rather well, in a semi-quantitative way,
the actual doping profile, and is easy to use in
the PC-lD.
0.4 0.5 0.6 0.7 0 .8 0.9 L1 1.2
3 CHOICE OF THE PHYSICAL PARAMETERS. Th ousand
The physical parameters resident in the PC-1D
have been used here with good results: These Fig 1 Cell reflectance and fitting. Note that the
include, among others, mobilities (minority carrier visually measured grid shading is added to this
mobilities are the same as the majority carrier value, done in samples without grid processed in
ones), bandgap narrowing parameters, intrinsic the same cell wafer.
carrier concentration at standard temperature and
temperature variation law, density of states in the
conduction band to the valence band ratio, Auger
recombination coefficients in low and high Current density (rnA)
50
injection, band to band recombination coefficient
(Auger trap assisted), structure of the absorption 40
bands and temperature dependence (including direct
transitions, and phonon assisted indirect 30
transitions) and index of refraction. -- 21 Ohms ' em (Theory)
All these values are based on recent 20
literature and well selected. In any case the -- 21 Ohms ' cm (hp)
accuracy is sufficient to insure good agreement 10
with our results without fitting. Some aspects of
this parameter selection might be the object of o
minor refinements with negligible influence in our -10 ~----~------~------~------~----~
work. -0.2 o O.::! 0.4 0.6 0.8
The SRH and the surface recombination are
introduced the programme through single level Vollage (vol t )
traps. They have been set as zero in base and at (a)
the passivated surfaces, and 107 cmls at the metal
covered surfaces. Alternatively the have been set
to give 1 ms in base and reasonable values (close Current density (rnA)
to 100cm/s) at th11su~faces, with a surface charge
of around +10 cm. The differences are
40~~============~==~----~
negl igible. 30
The solar cell is not totally one dimensional.
The region under the metal fingers has a strong 20
Ohms' cm (Th eory)
recombination at the front. This effect is included
in the cell model by means of an external diode 10 Ohms' em (Exp)
connected in parallel. The saturation current of
this diode is calculated by using the cell model O r---------------------------;---~
modified by setting the front surface recombination
velocity as 107 cmls. A low injection cell has been
used for this calculation, and only the hole
current, at 600 mV, injected in the emitter has
-10 ~----~------~------~------L-----~
-O.::! o 0.2
Voltage
0.4
volt
0.6 0.8
been taken into account. This current density is
found to be 51.6 fA (b)
The programme GRID also by Paul Basore, has
been used to calculate the cell metal grid, and to
calculate the cell lateral resistance that is
included in the model as a series resistance of
61 mO.
30
4 RESULTS.
To adjust the back reflectance of DE cells we 20
0.3 Ohms ' cm (Theory)
fitted the sub-bandgap reflectance, is of about
0.19, admitting that the escaping reflectance of 10 0.3 Oms ' em (Exp)
the textured face is 0.92 ~ 1-1/n2 , the theoretical
one. It might be lower and Simultaneously the back 0~------------------------~--_1
reflectance also lower. This point has not yet been
made clear. The differences, for the DE cell are -10L-----~----~~----~----~----~
rather small (it might become more substantial for -0 .2 o O.::! 0.4 0.6 o.a
better cells) volt
The measured typical refelctivity represented
Voltage
in Fig 1 is introduced in the PC-1D for (c)
calculations.
We present now, in Fig 2 the theoretical JV Fig 2. JV curves for different .base resistivities
curves together with some typical experimental ones (0.3, 1 and 21 Ocm).
for several base resistivities (0.3, 1 and 21 Ocm).

624
 Finally we present in Fig 3 the theoretical velocity at the base-oxide interface at the cell
and experimental internal quantum efficiency of the back caused by depletion conditions at the back MOS
same cells . structure. To avoid it n-base cells have been
fabricated that should present accumulation
120lnlernal Quanlum Efficiency conditions there , and thus lower surface
120 re combination velocity. We call this structure the
100
Back- DERL cell . Results, as given in Fig 6, are as
100
expected but attributed to furnace contamination
80
~
0.30hms· cm 80 (by tube cleaning), all the batch has given low
voltage, so record efficiency, as sought, has not
21 Ohm · cm (Theory)

----
60 60 been achieved.
21 Ohms· em (Exp)
40 40
0.3 Ohm · cm (Exp)
20 9.8 ohm · cm (Exp) 1/ 20 50 Current
o 0
300 400 500 600 700 800 9001000110012001300 40~-------------------- ____
Wav length (nm)
30
#Back-DE (see the text) DE (21 Ohms· em)
20
DERL (21 Ohms· em)
10
Fig 3 Calculated and measured Internal Quantum
Efficiency for the cells of different base O ~--------------------------;r--;
resistivity in our work .. _10L-__- i_ _ _ _ ~ ____ ~ __ ~ ____ ~ _ __ _ ~L_~

As shown in Fig 3, values of the Internal o 0. 1 0.2 0.3 0.4 0.5 0.6 0.7
Quantum Efficiency over 100% are reported. We know Voltage (volt )
that these values are not realistic, but we report (a)
them to compare wi th our reference cell (measured
at Sandia). that we suspect it must be
slightly overestimated . If the measurement were
fitted according with 100% for the maximum, more Current densit
realistic values would be found (approximately SY. 40 ~~~~~~~~~~~~-------------'

less) . On the other hand, in IR and UV regions ,

experimental measurements are under the theoretical 30
ones . This must be due to the poor signal and a not
well measurement of reflectivity in these regions 20
of the spectrum. DE (0.3 Ohms· cm)

10 DERL (0 .3 Ohms· em)

5 DISCUSSION .
The preceding results proves that our modeling
and parameter selection is operationally accurate . O ~----------------------------~~
Now it must be used to predict how to improve the
DE cell - 10 L---~--~~--~--~----~--~--~
o 0. 1 0. 2 0.3 0.4 0.5 0.6 0.7
Voltage (volt)
2.5E +17 Generation & Re c ombination
(b)

2.0E+ 17
Fig 5 JV curves of DERL cells as compared to DE
1.5E+ 17 21 Ohms· cm
cells in the same batch, for several base
0.30hms· em resistivities .
1.0E+ 17
1 Ohms· em

5.0E+16~=::::::==========~~
0.0t:
It is of the highest interest to note that no
reduction of the Jse is observed in the Back
o 50 100 150 200 250 300 illuminated cells . On the contrary, our Jse record
Di lan (mi c ron
is found with these cells (see Table1l. This is
very promising as it makes our simple basic DE cell
technol ogy. with simple modifications, suitable for
Fig 4 Location of the GR along the cell. eff i ciencies above 20% .

6 CONCLUSIONS.
The most significant results for every type of
For it let us plot in Fig 4 the location of ce ll of this work are shown in Table 1.
the generations and the recombinations in the From it we conclude that our simple DE
cells . It is apparent that most of the t ec hnol ogy is sui table to obtain efficiencies in
recombination takes place at the back metal th e order of 19-20y'. This efficiency can be
contact . If this metal contact is reduced to dots exceeded by some simple modifications of this
we should observe an increase of the open-circui t technol ogy, as can be seen from the promising
voltage, and thus of the efficiency . This cell re s ults of Back-DERL cells. These improvements come
structure is called Deep Emitter Rear Locally from an exhaustive investigation of the DE behavior
diffused (OERLl cell . The results have not been by modeling . In this way. we have available a
pos i tive in this respect , as shown in Fig S. It is flexible and promiSing method to obtain silicon
believed to be so because the high recombination solar cells above 20% efficiency .

625
 Curr ent
50r---------------X-~---L------------~
TABLE 1. Experimental results for different cells
Measured at IES
40 I=~==~::=~!i.§~ CELL TYPE
1
PBASE J 2 V3 F.F. l}(%)
30 DE (0.3) SC DC

DE (21)
20 IPEl DE 21 41. 2 620 0.791 20.1
Back - DE (9.6) 7PEl DE 1 37.5 636 0.817 19.4
10 8PEl DE 0.3 35.1 654 0.829 18.9
Back-BERL (9.8)
4PE2 DERL 21 39.5 615 0.787 19.0
o f - - - - - - - - - - - - - - -H--I---i 8PE3 DERL 0.3 35.2 646 0.800 18.1
_IO~ __ ~ ____ ~ ____ ~ __ ~ ____- L_ _ _ _ ~ __ ~
9PE7 • Back-DE 9.8 40.0 588 0.755 17.7
9PE9 • Back-DE >1000 41. 8 595 0.763 18.9
o 0.1 0 .2 0.3 0.4 0.5 0.6 0.7
Back-DERL 9.8 38.7 599 0.732 16 7
9PE8 •
o ltag (vol t ) 9PEI0· Back-DERL >1000 39.9 604 0.715 17.2
(a)
3 mV
lQcm; 2mA/cm 2 ;

r-----
'Batch with difussion furnace contaminated
50
ulT e nl d e n .:::..:..cc.'----'..:..;:.:.;...:.!.--_
il (rnA) _ _- - - ,
during cleaning step

40f=~==~====~~~~~
30 DE 10.3)
DE 121) REFERENCES
20
[ 1Jh DT Rover, P Basore and GM Thorson, in Proc.
Back-DE ( > 1000)
18 PVSC, IEEE, New York, (1985), 703-9.
10
B c k -OERL 1> 1000) [2] A Cuevas and MA Balbuena, in Proc. 8th EC
o PVSEC, Kluwer Academic, Dodrecht, (1988), 1186-91.
[3] J del Alamo, J Eguren and A Luque, Solid State
-IO ~--~----~----~--~----~----~---J Electronics 24, (1981), 415-20.
o o. I 0.2 0.3 0.4 0.5 0.6 0.7
ollage ( oiL)
Ib)

Fig 6 JV curves Back-DERL cells, of Back DE cells

and of DE cells of several resitivities in the same
batch.

626
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

THE PHOTOVOLTAIC EYE : A HIGH EFFICIENCY CONVERTER BASED ON LIGHT TRAPPING

AND SPECTRUM SPLITTING

A Luque, G Sala, J C Minano, P Davies, I Tobias, J Alonso,

C Algora, G L Araujo, J M Rulz, A Cuevas and J Olivan
Instituto de Energia Solar, ETSI Telecomunicaci6n
Universidad Politecnica de Madrid, E 28040 Madrid
phone (341) 5441060 - fax (341) 5446341
P Dunn and GRice
University of Reading
J Knobloch and B Voss
Fraunhofer Institute for Solar Energy Systems
C Flores
CISE

ABSTRACT The concept of light confinement by cavities is discussed with

particular reference to silicon and gallium arsenide cells in concentration.
Light confinement increases phonon-assisted absorption in silicon, permitting
thinner cells of lower base resistance, and ameliorates the problem of grid
shadowing in gall i um arsenide. Integrating sphere, single ellipsoid and double
ellipsoid cavities are compared making reference to their performance as
predicted from mathematical models. Efficiencies in the range 30 to 34.5%
appear possible using silicon and gallium arsenide in the same cavity. A
double ellipsoid cavity termed "photovoltaic eye", which is be i ng constructed,
is presented. The use of closely arranged cells in a cavity introduces special
cooling problems in relation to which heat pipes are of interest . Progress in
an experimental heat pipe rig is reported.

INTRODUCTION

The possibility of using light confining

cavities with solar cells has recently been
attracting our attention. A consortium has been
established under the sponsorship of the
Commission of the European Communi ties (Contract
JOUR40) to demonstrate some of the concepts behind
this idea . After some discussion to adapt the
available funding to the wide goals of the
projects, two final specific goals have been
identified. This paper describes these goals and
progress in their achievement .
These two goals are the achievement of
efficiencies as high as possible using jointly
celiS/
GaAs and Si cells inside a cavity and the
development of a dense photovoltaic receiver, with area Ac
cooling by a heat pipe. The use of cavities,
especially if they are large, presents problems of
cooling which make heat pipes attractive. ~
2 THE INTEGRATING SPHERE CAVITY

2.1 Cavities for a single type of cells

The concept is presented in Fig 1. The light, Fig 1. Spherical light confining cavity.
from a concentrator, enters through a small entry
aperture and then falls onto a set of solar cells.
The light reflected by these cells, by its AR
surface or by the metal fingers, has a high Obviously the advantage for a cell having very low
probability of falling on the cells again so that p is small (the efficiency is proportional to
the apparent reflectance at the cavity entry [I-pI), but it can improve the performance of a
aperture is reduced. cell with p not so low.
If an ideal angle transformer is situated at The cavity allows for a substantial coverage
the entry aperture (area Ae), so that it converts of the cell with metal fingers and thus operation
the rays from a primary into an isotropic bundle at higher concentration. This is of direct
at the entry aperture (a CPC does this rather application to GaAs cell where the series
accurately), then a cavity of spherical shape has resistance, at very high concentration, is limited
the property of becoming isotropically illuminated by grid coverage. We present in Fig 2 the effect
by the aperture. Furthermore, i f the reflectance using a cavity i nternally covered with GaAs cells.
of the cavity internal walls (area Ac) is We can observe that the efficiency of a cell of
homogeneous, of value p, then the spherical cavity our present technology (of about 22% at one sun)
illuminates, also homogeneously, any point of might give only 24.7% at some 100 suns and 24% at
itself . 1000 suns. With a cavity in which the entry
The value of the apparent reflectance is aperture is 1/10 of the cells area it might give
27.1% at 1000 suns , which is a nearly optimal
p' = pA./[pAe+(l-p)Acl =p(A./Ac) situation. This implies having some 10000 suns at
the entry aperture; this value can be achieved
so that, for small cell reflectance the cavity with a mirror (1). We observe both the improvement
reflectance is approximately reduced by the ratio in maximum efficiency and the increase in the
of the entry aperture to the inner wall area. optimal concentration at which it occurs .

627
 escape absorption. This reflectance is effectively
40 reduced by the cavity .
The possibility of making cells very thin
also allows for an increase of the concentration
at which a cell can operate, as in silicon it is
1 the base resistance, proportional to the cell
( b) 133-7 thickness, which is the limiting factor at high
concentrations.
0~
2.2 Cavities for spectrum splitting
~30 It is possible to fill the cavity with cells
c 1 of different materials, for instance 5i and GaAs.
QI
u (a) 127·1 In principle the cells should be covered by an
appropriate filter so that the 5i cells reflect
if
~ 1 the photons with energies above 1.42 eV (the GaAs
24·0 band gap) and the GaAs cells reflect those with
(cl energies below 1.42 eV.
This allows for assembling methods that are
essentially similar to those for single cells. We
present in Fig 4 a schematic of two ways of
situating the cell in the cavity. The one to the
10 100 1000 10000 right, in which the 5i cells are illuminated only
concentration on GaAs by the light reflected from the GaAs cell is the
one giving better results (2).
Fig 2. Efficiency of (a) GaAs, and (b) Si and GaAs
cells together, in a spherical cavity. (c) applies
u;c'dsn/ "Qht
to GaAs cells without any cavity.

In addition in cells like those of silicon,

the cavity allows for an enhancement of the
internal light confinement in the sil icon wafer.
The cell must have a good back mirror, and the
front surface must be textured so as to produce
Lambertian illumination in the light entering the
cell. When located in a cavity the average path
length of the rays in the wafer becomes

1 = 4n~(Ac/Ae)

with W the cell thickness. The enhancement for

being inside the cavity is the factor (Ae/Ac), 10
in the example. This is important for the
absorption of the'sub-band photons, whi.ch is weak,
assisted by one phonon . To evaluate the importance Fig 4. Spherical cavity for GaAs and Si cells.
of this effect we present in Fig 3 the photo Left, the GaAs and the 5i cells are intercalated.
generated current in absence of losses, for cells Right, the GaAs cells are illuminated directly and
of different thickness, when the cell is located the 5i cells indirectly by the cavity.
in a cavity of (Ae/Ae) = 80.
With both configurations, the level of
In a back-mirrored silicon cell high illumination of each type of cell can be adjusted
reflectance appears at low energies as photons independently, e.g. increased by reducing the area
of the concerned cell. In this way the main
drawback of the tandem cells, situated one on top
of the other, that the cells have to be of the
same size, so that the little power retrieved from
Generated current dens (mAlcrn21 the low bandgap cell (significantly called

:::
"booster cell") may not justify its inclusion, can
45 be overcome. In this case we can achieve the
a e a efficiency boosting with a smaller booster cell

42
~g and thus retrieving high specific power from it .
An efficiency of 33.7% is achieved with a
concentration of 10000 at the entry aperture, 1000
-+- F on the GaAs cell and an area of 5i cells of only
39 ....... T
0.3 of the GaAs cells area. The 5i cell are
actually operating at a current level as if the
were at 100 suns with full spectrum and their
T8M efficiency, referred to the irradiance they
36
-a- TM8C receive, is 23.6%.

2.3 Summary of advantages and drawbacks

100 200 300. 400 In summary the use of the cavity allows for
Cell thickness (micronsl (a) reduction of the apparent reflectance, (b) an
increase of the light confinement and (c) an
increase of the concentration at which the cell
can operate.
Fig 3. Generation by a cell with light In the case of using 5i in addition to GaAs
confinement, vs. the cell thickness. F: flat cells we obtain higher efficiencies . Operation of
(untextured) cell. T: pyramidally textured. T & H: the 5i cells at a flexible concentration level
textured and with back mirror. T,H & c: textured, becomes possible. Best efficiency are obtained
with back mirror and light confining cavity. wi th current levels at the 5i cell as when they

628
operate at 100 suns with full spectrum. It is unlikely that such an small area, usually in
Three drawbacks are: (a) the high the range of 10% of the cell area, can be achieved
concentration at the cavity entry aperture; (b) in the spherical cavity.
the impossibility of covering the inner surface of In general we find that the ellipsoidal cavity
the sphere with solar cells, some regions must be is theoretically equivalent to the spherical
left uncovered and losses are produced there; cavity but has some practical advantages: (a) it
finally (c) the difficulty of removing the power does not require any secondary; (b) it does not
caused by the high concentrations involved. subject any surface to higher irradiance than the
To solve these drawbacks new cavities and cells are subject; (c) parasitic optical losses
means of removing power are being considered. can be kept low; (d) it is particularly good for a
single cell. With these arguments the results for
single cells in Fig 2 (provided we keep the cell
3 THE ELL! PSO IDAL CAVITV surface reflection diffusive) can also be applied
to this case.
3.1 Cavity for a single kind of cell
This cavity is intended to avoid the high 3.2 Cavity for two kinds of cell
concentration required at the spherical cavity The ellipsoidal cavity can also be used for
entry aperture. It is based on confinement of two kinds of cells. In this case the position of
light whereby only rays within an angularly the cells is as represented in Fig 6. The entry
limited bundle are allowed to leave the cavity. aperture is now the circle between the foci, and
Hence less energy escapes from the cavity, per the GaAs cell has an arbitrary size. The light
unity of entry aperture area, and larger areas are impinges directly on the GaAs cell and then is
allowed for the same level of light confinement. reflected, we assume diffusely. Rays in this
The ellipsoidal cavity is formed by the cavity are of four classes. (i) Rays linking (in
revolution of the outline in Fig 5. A mirrored both directions) the GaAs cell and the entry
ellipse has the foci F and F' at the cell edges. A aperture; they may escape. (i i) Rays linking the
hole EE' of arbitrary size forms the entry GaAs cell and the top cover (filled with 5i cells
aperture. Any ray issuing from the cell is or with a white paint). (iii) Rays linking the top
reflected back into the cell again. Only rays cover with itself; no ray is allowed to link the
issuing directly from the cell can escape through top cover with the entry aperture. (iv) Rays
the entry aperture. In consequence the rays I inking the entry aperture with itself are
escaping through this aperture, in contrast to the geometrically possible but are not actually
spherical cavity, are not isotropic. illuminated. Thus the escaping rays are angularly
limited.
This cavity allows for a rather simplified
Ii¢,lt theoretical analysis. We have done it (3) and
< found that, with very small losses in the white

~!
paint, the optimal situation, for an entry
aperture of 1 cm of diameter, is to use a GaAs
cell of' 0.33 cm of diameter and an annular 5i
I cell, surrounding the entry 1. 34 cm of diameter.
I .'
:': -( "::':- The concentration at the GaAs cell is of 755 suns
I and at the 5i cell of 100 suns The total
efficiency is 34.1%.
In this analysis we have used the same
radiance as before (corresponding to an isotropic
concentration of 10000 suns) and the same ce 11 s .
We have not used any dichroic f11 ter on the 5i
cells (so that almost no light turns back to the
GaAs cell). lf realistic values are set for the
different optical parameters we find an efficiency
of only 33%.

Fig 5. Ellipsoidal cavity for a single cell Incoming light

bun.,dle
The reflectivity of this cavity, assuming escaping
loss less reflection at the mirror and Lambertian
:: :
I

[Jf ray
reflection in the cell (well achieved in textured
cells), is given by
, , ,.
white paint (-' /?vhfte paint
p' = pAe/[pAe+(l-p)At] ~ p(Ae/At)

where At is the area of the ellipsoid equatorial

section. The cell area does not appear explicitly. A'
ISi~ F' , )/~"Ae/~
, '" -, , A
The geometry is determined by the cell area and
the area of the equatorial section. The entry f I r
, '" ,
,('"

aperture is arbitrary, but must be big enough as ,~ (

to allow the bundle of incoming light to fall in , ,, , . , , con t'me d" ,

the cell. The bigger it is the higher the ~-------~.--~-~--------------
concentration on the cell can be. ,;' : ray
,.
/" . l
Let us compare the losses in the two
cavities. In the spherical cavity losses are
-'/
,
' (

produced in the white regions intercalated among

the cells, let us assume that their area is Aw and
has an absorbance (one minus the reflectance) ~w.
Let be the mirror absorbance ~m. The area of white
regions giving the same losses as the mirror is

Aw = (~m/~w)pAc[l-(Ae/At)l
Fig 6. Ellipsoidal cavity for two types of cell.

629
4 DOUBLE ELLIPSOID CAVITY FOR TWO KINDS OF CELL cells and cavity walls. This illustrates the
relevance of heat pipe coolers to cavities. An
The design of Fig 6 has some drawbacks. (il important parameter as regards the power output
It requires diffuse reflection from the GaAs to
from the Si cells is the reflectivity of the GaAs
illuminate the Si cells, whereas GaAs cells to light below it bandgap. Preliminary
actually give specular reflection. Oil It does calculations show that a dichroic filter might be
not enhance significantly the illumination of the used over the GaAs to increase the efficiency to
GaAs cell. (iii) It can be seriously affected by 33.5%.
the slightest absorption in the white paint Simplifications of Fig 7, made from one
surrounding the Si cells, because for optimal instead of two pieces of quartz, are being
considered.
performance this paint should cover a large area.
As an improvement, the double ellipsoid, 5 THE HEAT PIPE COOLER
dielectric-filled cavity of Fig 7 is proposed. We
take of advantage of specular reflection from the The first stage in applying heat pipes to
GaAs cell by tilting it, reflecting light directly solar cells is to set up an array, measuring
onto one of the two Si cells. The top cover is 16x16cm, of 64 silicon cells, under a
replaced by an ellipsoidal mirror which returns to concentration of 300. An evaporator will be formed
the GaAs cell light reflected diffusely from its of a copper block covered in a mesh wick for
metal grid. The Si cells have now to be moved from uniform distribution of the working fluid which
their previously ideal positions of Fig.6 in order has yet to be selected. The fluid will condense in
to be incorporated into the top dome, but the 48, 1m-long finned tubes extending perpendicularly
cavity still gives efficient confinement of light to the plane of the photovoltaic cells. Enclosure
reflected from them. Rays which reach the entrance of these tubes in a duct, through which air flow
aperture from inside the cavity are usually will be forced by a fan, will assist cooling. The
internally reflected and few escape. energy expended by the fan will be compensated by
A prototype based on Fig 7 is being made from the increased power output due the lower cell
quartz. To ease manufacture the elliptical temperature, and the approach will be to find the
surfaces are to be approximated by spheres. optimal operating point.
Detailed calculations of the efficiency which A test rig has been constructed to establish
we can realistically expect from this device have the performance of the heat pipe, and provide
been carried out. These required 3-dimensional ray guidance on its design and operation. Electrical
tracing for different cases of the cell heating is used to simulate the heat dissipated by
reflectivities which vary according to the the solar cells. The rig will enable different
working fluids, wick structures and geometries of
incident light extended tube surface to be compared. The aim is
to demonstrate successful operation at a
sustainable evaporator heat flux of 24 W/cm2
keeping the temperature of the cells as close as
possible to the ambient temperature.

6 CONCLUSIONS

The use of light confining cavities should

make efficiencies of over 30% possible using
combinations of existing silicon and gallium
arsenide cells. The apparent reflectivity of the
cells is greatly reduced, i.e. little light
escapes from the cavity, so that greater
proportions of the light are diSSipated as
electricity and heat. With regard to the increased
heating, heat pipes are of interest to assist
mirror cooling. The concepts presented are aimed at
maXImizIng efficiency, not necessarily at
producing the most economic solution.

Fig 7. Double ellipsoid cavi"ty.

REFERENCES
wavelength of the light. The GaAs cells were
modelled according parameters from (4). For the Si (ll Luque A: Solar Cells and Optics for
cells the computer program PC1D was used with Photovoltaic Concentration, Chapt. 13, Adam
parameters which represent our own cells (5), Hilger, 1989.
except that we have supposed an improved (2) Araujo G L, Marti A and Luque A: Efficiency of
reflectivity (p=0.97) at the back surface of the Si and GaAs concentrator solar cells operated
Si cell, in line with results reported by Green et inside integrating cavity receivers, Solar Cells,
al. (6). The surface of the quartz is to be in press.
surrounded by a thin air layer, gIvIng total (3) Tobias I, Luque A and Mifiano J C: Light
internal reflection of oblique rays, and outside confining cavities for GaAs and Si solar cells,
this a silvered mirror. The primary concentrator, Int. J. Solar Energy, in press.
for the purpose of ray tracing, has been taken as (4) Algora C: Desarrollo de celulas solares de
a spherical mirror, 12 cm in dia~eter. This gives GaAs de alto rendimiento crecidas por epitaxia de
a concentration of 720 on the GaAs cell. To avoid fase liquida, Doctoral thesis, Univ.Compl.Hadrid,
shadowing, lenses may equally well be used. 1990.
The efficiency predicted is 31.1 % using (5) Cuevas A and Balbuena M: Thick-emitter silicon
present GaAs substrate, 32.5% with material having cells, 21st IEEE PV specialists conf. (Las Vegas),
lower substrate absorbance. The 32.5% consists of 429 (1988).
contributions of 25.5, 5.5 and 1.6% from the GaAs (6) Green M A, Blakers A W, Zhao J, Milne A M,
and two silicon cells respectively. Only about 1% Wang A, Dai X: Characterization of 23% efficient
of the light energy escapes from the cavity, the silicon solar cells, IEEE trans. electron. dev.
remaining 66.5% being dissipated thermally in the 37, 331 (1990).

630
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFEREl',CE S-12 APRIL 1991 LISBON, PORTUGAL

Si0 2-passivated High Efficiency Silicon Solar Cells:

Process Dependence of Si-SiO 2 Interface Recombination

A. Aberle, S. Glunz, W. Warta, J. Kopp, J. Knobloch

Fraunhofer-Institut fUr Solare Energiesysteme, Oltmannsstr. 22, D-7800 Freiburg, Germany

Abstract In comparison of simultaneously processed MOS test structures and solar cells the
effect of varying postmetallization annealing time and temperature, oxidation temperature and
ambient as well as the surface morphology is investigated. More than 20 min at 400°C are
found as the optimum postmetallization annealing conditions. No significant effect of oxidation
temperature variations between 950 and 1100°C is observed, For all oxidation temperatures
an equally strong increase of the recombination velocity at the passivated rear surface with
excess carrier concentration is found. Using measured MOS data and an extended SRH
recombination formalism in an elaborate description of the surface space charge layer this
dependence can be understood quantitatively.

(a) For the oxide quality assessment quasistatic

1. Introduction
and high frequency C(V)-curves are simultaneously
Due to enhanced quality of the starting material measured /1/ and evaluated in the usual way. Oxide
and processing sequences capable of maintaining thickness, band bending, interface state density (mea-
extremely high volume lifetimes, the efficiency of surement limit for Dit about 1010 cm- 2eV-l) and densi-
today's most efficient silicon solar cells is limited by ty of fixed oxide charges are obtained. In addition a
surface recombination. The highest AM1.5-efficien- more detailed DLTS assessment of MOS test struc-
cies are presently obtained on n + p cells with a full- tures is performed. From capacitance transients in-
area emitter at the illuminated surface. Optimum back duced by small filling pulses 12/ the distribution of in-
surface passivation is realized by covering the rear terface states with densities below 1010 cm,2eV-l as
surface almost completely with a high quality thermal well as the energy dependence of the capture cross
oxide, restricting the metallized area (aluminum) to a sections are derived. A commercial DL TS spectro-
few percent. Optimization of solar cell processing meter (Polaron S4600) with box car integrator and a
thus requires a thorough control of the effect of pro- Boonton 1 MHz capacitance bridge (type 72B) was
cessing changes on the properties of the passivating modified by adding a HP8116A pulse generator. Thus
oxide. We therefore started a research task aimed at small voltage pulses with well defined edges and of
clarifying the influence of a series of processing steps arbitrary polarity were available. The temperature
on oxide quality and, subsequently, surface recom- scans revealed clear DL TS peaks. The appropriate
bination velocity. In this paper the following process energy interval was determined from the band ben-
influences are covered: ding profile at peak temperature. Capture cross
- variation of temperature and duration of the sections for majority carriers are calculated from the
postmetallization forming gas anneal SRH-theory detailed balance equations (e.g. /31).
- variation of oxidation parameters
(b) The high injection surface recombination velo-
(temperature, gas atmosphere)
influence of surface morphology. city Shli on Si0 2-passivated test wafers without
metallic contacts is determined from photoconduc-
Measurements on identically processed solar cells tance transients /4/. A RF-bridge is balanced with the
and test structures are reported revealing the effect of wafer in dark condition. Application of an appropriate
different processing steps on both the oxide para- light pulse drives the high resistivity wafer (typically
meters and the surface recombination velocity S. By 100 Ocm) into high injection, the generated excess
applying an extended Shockley-Read-Hall (SRH) carriers unbalance the bridge and the recovery tran-
recombination formalism it is demonstrated that sient is recorded. Values for the surface recombina-
experimental dependencies of S on excess carrier tion velocity Shli are derived using volume lifetimes
density at the surface can be explained by measured determined on the wafer immersed in HF /5/.
values of interface state densities, electron and hole
(c) The back surface recombination velocity of solar
capture cross sections and fixed oxide charges.
cells affects the spectral response in the long wave-
2. Characterization techniques length region. Thus, Sback can be extracted by a
parameter fit /6/ to the measured spectral response.
For investigating the correlation between oxide However, for cells with point contacted rear surface
quality and surface recombination velocity, three lines the value obtained is an average between the recom-
of measurements are combined: (a) MOS charac- bination at the oxide passivated and the metal
terization, (b) contactless photoconductance decay covered fraction of the rear surface. For an accurate
and (c) determination of the back surface recombi- determination of Sback we illuminate bifacial solar cells
nation velocity Sback of finished solar cells. from the rear /7,8/. From these measurements a

631
strong dependence of Sback on light intensity was
found and reported in an earlier paper 18/.
a 3SO 'C • 500 'C

3. Sample preparation r 6 400 'C

x @'C
• 5!iO 'C

~
Starting material was high quality boron doped FZ- 'I' 10"
silicon from Wacker-Chemitronik. Thickness of the .5,
(100)-orientated 3" wafers was about 250 microns.
Prior to all high temperature steps, the wafers i
received a RCA-cleaning. Resistivity of the wafers 0 ""
was 0.5 Ocm (solar cells). 10 Ocm (MOS capacitors) rJ'
and 100 Ocm (contactless PCD), respectively. Oxida-
tion was performed at 950-1100oC with 2 Vol. %
e.rJ 0 5 ~ ~ ~ ~ ~ ~ ~ ~ ffi
TCA being added to the main oxygen stream. After Amedng TITe ~J
oxidation, gas atmosphere was changed to argon and
the wafers received a postoxidation anneal. Sub- Fig. 1 C(V)-measurements of interface state
sequently, the tube was cooled down to 850 0 C in density Dit.midgap versus duration of the post-
metallization forming gas anneal for 5 different
argon and the wafers vyere extracted. After metalli-
anneal temperatures
zation (electron beam deposition) the wafers received
a postmetallization anneal in forming gas (5% H2 in The differences in Dit are clearly mirrored in the
N2) at 400-500 DC. surface recombination velocity underneath the oxide
passivated rear surface of solar cells. The measured
4. Process influences
spectral response in the long wavelength regime is
4.1 Influence of postmetallization anneal plotted in fig. 2. Anneal time was chosen to 25 min,
A final anneal in forming gas after metallizing an safely within the Dit saturation regime.
oxide-silicon structure with aluminum is common
M r------.-------r------~----~
practice in MOS-technology. The density of interface
states is greatly reduced by this process /9/. The
improvement is explained by the aluminum metal layer ~
reducing residual water in the silica film. The hydro- 5l 60
gen atoms released passivate interface defects /10/.
This reaction is slower at lower temperatures. On the j [oc]
T s"'II DJt
['0 3 ary.] [~1O aTl-2....-~
•
other hand, at hiQher temperatures increasing thermal
!~ 3SO

·
o ~.O 3.0
decomposition of the annealed centers will dominate. 17 l!i1 • I
• 400
With this model the minimum in interface state
•
2.7 t8
~
density around 400°C found experimentally /11/ is • 500
4.5 2.5 before ared
• 5!iO 'D.C 303
explained. According to this model Dit will eventually
~
become independent of annealing time at any 850 900 950 me ~

temperature. For a complete overview of the Wavelength rmJ

annealing behaviour and a test of the model
predictions we chose the following procedure: Fig. 2 Measured' spectral response of n+p
batches of identically processed MOS samples and silicon solar cells annealed at different tempera-
ture.s reflecting the observed differences in Dit
n + p solar cells (with the rear side oxide passivated
(polished (1 ~Ol-surfaces, ohmic rear point
and point contacted) were oxidized at 1050 0 C (2 % contacts, oxide thickness 105 nm).
TCA) with a 60 min postoxidation anneal in argon.
The resulting oxide thickness was 105 nm. The In the insert average surface recombination velo-
batches were split before the final annealing step. cities (metallized fraction included) extracted from the
Each group received a postmetallization anneal spectral response curves by a parameter fit are
(forming gas) in the same annealing tube with a compared with measured Difvalues. Annealing at
different combination of annealing time and tempera- 400 a C leads to the lowest recombination velocity, in
ture. Temperature was varied between 350 and coincidence with the lowest interface state densities.
600°C, annealing times ranged from 5 to 40 min. The
measured dependence of interface state density on 4.2Effects of oxidation temperature and atmosphere
annealing time and temperature is plotted in fig. 1. Passivating oxides were grown at temperatures
The 600 DC anneal was dropped as the solar cells between 950 and 1100 0 C in dry oxygen with 2 %
were destroyed at this high anneal temperature. Our TCA added to the gas stream followed by a 1 h argon
experimental results are compatible in every detail anneal at oxidation temperature. Typical oxide thick-
with the model proposed by Deal et al.: at lower ness was 105 nm.
temperature Dit initially decreases, until passivation is Measured values of interface trap density
completed. At higher temperatures the initially Dit.midgap (from CIVIl and surface recombination velo-
achieved complete passivation is reduced due to city Shli (from contactless PCD) are plotted in fig. 3.
thermal decomposition. After about 20 min Dit No significant dependence on oxidation temperature is
reaches a saturation value which increases with tem- observed. Measured values for the fixed oxide charge
perature. The lowest Difvalues (9xl0 9 cm- 2eV-l) are densities fall in the range (1-2)xl 0 11 cm- 2 and exhibit
found for 400°C after an anneal of at least 20 min. no correlation with oxidation temperature.

632
 5. Correlation of surface recombination velocity and
aw .-----.-----.-----.-----.---~ ~g oxide parameters
MO.o Modelling the solar cell, the parameter "effective
surface recombination velocity" 5eft appears in the
boundary condition at a virtual rear surface (x =d, see
fig. 5) for the diffusion equation of a neutral base:

qD,,~An(x)l = qU.(d) = qS.gAn(d) (1)

40.0 poot~
ax =d
~~
2Il.O
~500'C Seft describes the recombination current flowing
0.0 L -__ ---'===---.!r-_ _'>_-==r.....:::EO"'-"'C----'-r! into a physically quite complicated surface layer. It is
900 9!50 '000 '050 1'00 1150
only constant for given illumination level, otherwise
0xKla00n ~ f'Cl
dependent on An(d). For excess carrier concentrations
below 1017 cm- 3 the dominant loss process at a
Fig. 3 Dependence of Shli and Dit.midgap on
silicon surface is SRH-recombination via interface
oxidation temperature. The lower Shli-curves
were obtained after deposition of about 50 A of traps forming a continuum of states. Therefore, the
aluminum on both surfaces and subsequent standard SRH-expression for a single level trap has to
anneal in forming gas. be integrated over the bandgap:

In addition, we tested three different postoxidation

annealing schedules: 0, 1, and 4 h in argon ambient dE (2)
at oxidation temperature. Measured interface state
densities clearly revealed the lowest values for 1 h,
the highest for no postoxidation anneal.
On simultaneously processed bifacial solar cells Us depends on the detailed variation of the inter-
with the passivating oxide grown at different face state density Dit and the capture cross sections
temperatures the effective surface recombination (Tn/p with energy and on the carrier concentrations ns
velocity Seff was determined for different illumination
and Ps taken at the physical Si-SiO z interface. ns and
levels. A clear correlation of Seft with oxidation Ps cannot be deduced in a simple way from the mino-
temperature was not observed. rity carrier concentration An(d) at the edge of the sur-
face layer or, equivalently, the separation of the
r:J quasi-Fermi levels. Their values are determined by the
tI> 0
o band bending III across the surface space charge
region caused by charges in the oxide and the work
function difference between the semiconductor and
the gate metal. Fig. 5 gives a picture of the charge
and potential distribution at the interface.
O.mtic:n:
• 950 'C
QOtt <hor~. [,,,,- ' )
r:f v 1J()()'C eIect,ode axid'. IIKon
o n;o'C
o IOJ 'C

'0'" ~ft '0'" -0" -OM 'O~

Exress 01age CaTier DenSty M Pn-1

Fig. 4 Measured dependence of effective sur-
face recombination velocity on minority carrier
excess concentration at the illuminated rear sur-
face of n+p bifacial solar cells (0.5 Oem,
Ln =500pm)

The measured values plotted in fig. 4 increase

steeply from very low recombination velocities around
50 cm/s at about 1 sun illumination intensity to more Fig. 5 Charge distribution and band diagram at
than 5x 104 cm/s for very low light level. This remark- the Si-SiO 2 interface
ably strong dependence of surface recombination on
light intensity is well suited for a quantitative test of For solar cell surfaces the treatment of the well-
the correlation between MOS and solar cell based known situation at a dark MOS structure is not suffi-
oxide assessment using SRH-theory. cient as the bend bending is strongly affected by the
excess charge carriers generated by the illumination.
An exact analytical solution of this problem is not
possible. An approximate solution has been
introduced by Girisch et al. /12/ for describing gate

633
controlled point junction diodes. The basic approxi- In fig. 8 a representative example of the experi-
mation is the assumption of flat electron and hole mentally observed dependence of Seff on excess
quasi-Fermi levels in the space-charge region. The ' charge carrier density is compared with the calculated
positions of the quasi-Fermi levels lP n and lPp and the dependence.
electrostatic potential at the surface "'s are calculated
by iteratively solving the charge neutrality condition in
the surface region: 0Si + 0it + Of + Og = O.
As can be seen from equation 2, the knowledge of
the energy dependence of the capture cross sections
of electrons and holes is an essential prerequisite for
the calculation of recombination rates at a silicon
surface. In fig. 6, the results of energy resolved Ul~ ri
DLTS-measurements of an and a p are given.

'0'"'
"()-a E;

'1
"()-Q

"()-- 'i!I
~ '0'"-. an ,-
/
/
~ Fig. 8 Correlation between experimental SeW
values determined by spectral response (SR) and
~
"()-

"
~"
contactless photoconductance decay (PCD) and
"
i
"()-.
\ theoretically calculated curves
"()-.
" ap \
/ I
10-- \ Using capture cross sections and fixed oxide
/ /
10-" charge density as determined in the MOS measure-
0.0 0.1 Q.2 Q.J 0.. 0.& O:J QB 0.9 10 \1
~
ments, a perfect correlation between the expected
Ey iq>-Enetgy ~ E"
and the observed dependence is found. The theo-
Fig. 6 un and up of a thermal oxide versus retical curves are plotted for a range of interface state
energy from small pulse MOS-DLTS (oxidation densities at midgap. The value of 4x10 10 providing
at 1050 o C, 2% TCA, lh argon anneal!. the best fit is slightly higher than the measured
average Dit of about 2x10 10 on simultaneously
With the experimentally determined parameters processed MOS capacitors. This is quite reasonable
antE). ap(E) and Dit taken at midgap the recombina- since the final anneal in forming gas was different for
tion rate Us and the effective surface recombination the MOS structures and the bifacial solar cells: the
velocity at the edge of the depletion layer was oxide at the solar cell surface was not covered by an
calculated using the method described above. aluminum layer. Thus, the additional improvement by
The calculated dependence of Seff on the light the aluminum (process discussed in section 4) did not
generated excess concentration of minority carriers is occur. Atomic hydrogen was only provided in very
plotted in fig. 7 . Parameter is the fixed oxide charge low concentration by the forming gas.
density Of' Only positive 0rvalues were taken into The results of fig. 7 and fig . 8 also apply for high
consideration, as the oxide charges at thermally efficiency cells with the rear surface point contacted
oxidized silicon surfaces are generally positive. and the passivating oxide capped by aluminum. For
d ~~~~~~~~~nmrm~~_
105 nm thick oxides, the work function difference of
aluminum and p-type silicon (about 0.95 V) has an
rJ effect equivalent to an additional Of of about
+ 2x 10 11 cm- 2. We consider the typical excess carrier
concentration of about 3x10 13 cm-3 at the rear
surface of a high efficiency silicon cell at maximum
power point and AM1.5 illumination. For this situation
it is inferred from fig. 7 that Seft-values of about 10
cm/s are achieved. Due to the effect of aluminum on
Dit actual values may even be 2 - 3 times smaller. We
conclude that the dominant loss mechanism at well
n-ILw~~~Lwmuma~~yu~~~~~~ passivated (D it S 1010 cm- 2 eV-l) rear surfaces of
d rJ d rJ rf' rJ' n" n° U' ne n· rf' n· n + p silicon solar cells is recombination at the rear
Uu::ess Quge Carier Dmsity M ~ -1 metal contacts. In a previous paper /13/ we proposed
as a solution to this problem a local back surface
Fig. 7 Calculated dependence of effective field . In theoretical investigations we find that the
surface recombination velocity on minority optimum rear surface structure could be formed by a
carrier concentration at the edge of the surface moderate and deep diffusion underneath the metal
space charge region . Parameter is the fixed contacts (N s = 1x 10 19 cm- 3, depth at least 10 pm).
oxide charge density Of (0.5 Ocm p-Si, 300 K;
eventually combined with an additional light diffusion
no gate metal: work function difference = 0)
(1-3xl0 18 cm- 3) underneath the oxide. Detailed
results on this topic will be reported elsewhere.

634
6. Influence of surface morphology 7. Conclusion
Rear surface passivation of today's most efficient Several process parameters are investigated for
silicon solar cells /14/ is accomplished on polished their influence on oxide quality. The temperature of
surfaces. The necessity for doublesided polished the postmetallization anneal was varied between
starting wafers is an enormous cost factor. On the 350 0 C and 550 o C. Optimum oxide quality was
other hand, we found no information in the MOS achieved for 400 0 C and more than 20 min annealing
literature about oxide quality on other than polished time. The annealing behaviour is in accordance with a
surfaces. The influence of a rough surface morpho- model assuming interface defect passivation by the
logy on the C(V) characteristics appears to be unclear hydrogen atoms introduced by the dissociation of
/15/. We therefore investigated the influence of water at the aluminum layer. The variation of the
different surface morphology on bifacial solar cells. oxidation temperature between 950°C and 1100°C
Three types of rear surface preparatimn are compared: did not show a significant influence on interface state
(i) polished, (ii) polished and slightly KOH-etched, (iii) densities, fixed oxide charges nor surface recom-
photolithographically defined inverted pyramids. bination velocities. The optimum duration of the post-
Bright-etched surfaces as a further possibility are oxidation anneal in argon ambient was found to be
under investigation. Sback-values are determined from about 1 h. Structuring the rear surface of a solar cell
the spectral response illuminating the cells from the by inverted pyramids was found to have a detrimental
rear. The results are given in table 1. effect on the quality of the oxide passivation. A
Table 1 Influence of surface morphology on effective shallow boron diffusion is suggested in this case.
We were able to quantitatively model the observed
surface recom b'Inatlon ve oClty
I .
strong dependence of the effective surface recombi-
Surface type Self (cm/s) nation velocity on the light induced excess carrier
polished 80 concentration . The energy dependent capture cross
KOH'etched 100 sections, interface state and fixed oxide charge densi-
inverted pyramids 3050 ties were determined experimentally with small pulse
MOS-DLTS on MOS structures processed in parallel to
The massive increase of surface recombination at bifacial solar cells. Using these experimental values in
the surface with inverted pyramids is attributed to an elaborate model description of the surface space
two sources: (i) mid gap interface state density at the charge layer, the dependence of the surface recombi-
(111 I-surfaces formed at the side walls of the nation velocity on excess carrier density is predicted
pyramids is 3-4 times higher than at a (1 OO)-surface theoretically. An excellent agreement with surface
/9/; (ii) surface area is enhanced by a factor of 1.7. recombination velocities measured with spectral
Attempts of rear surface texturizing for light response on bifacial solar cells illuminated from the
trapping or enhancing efficiency of bifacial cells will rear is found . These results prove the suitability of a
be hampered by very poor surface recombination. An SRH recombination formalism for describing the
additional boron diffusion at the rear surface might recombination at the surface of a solar cell under
resolve this problem. illumination.

Acknowlegements The authors wish to thank W. Wettling for encouraging discussions. The assistance of A. Noel, U.
Schubert, S. KluBmann, and T. Grille in sample preparation is greatly appreciated. We are indebted to E. Schaffer for preparing
the drawings and her untiring measurement assistance.
This work was supported by the German Bundesministerium fOr Forschung und Technologie.

References

E.H. Nicollian, J.R. Brews. "MOS Physics and Technology". John Wiley & Sons. New York (19821. p. 331
2 T. Katsube, K. Kakimoto. T. Ikoma, J. Appl. Phys. 52. 3504. 1981
3 E .Klausmann. W.R. Fahrner. D. Breunig in "Instabilities in Silicon Devices"
ed. by G. Barbottin. A. Vapaille. Vol. 2. North-Holland. Amsterdam (1989). p. 196
4 D.E. Kane. R.M. Swanson. Proc. 18 th IEEE PVSC. Las Vegas. 578. 1985.
5 E. Yablonovitch. D.L. Allara. C.C. Chang. T. Gmitter. T.B. Bright, Phys. Rev. Lett. 57 (2) , 249. 1986
6 A. Aberle. W. Warta. J. Knobloch. B. Vor... Proc . 9 th EC PVSEC. Freiburg. 765. 1989
7 D.R. Lillington. G.F.J. Garlick. Proc . 18 th IEEE PVSC. Las Vegas. 1677. 1985
8 J. Knobloch, A. Aberle, W. Warta. B. VoB, Proc . 8 th EC PVSEC, Florence. 1165. 1988
9 R.R. Razouk. B.E. Deal. J. Electrochem. Soc. 126. 1573. 1979
10 B.E. Deal. E.L. MacKenna. P.L. Castro, J. Electrochem. Soc .• 116.997.1969
11 see 11 I. p. 784
12 R.B.M. Girisch, R.P. Mertens, R.F. de Keersmaecker, IEEE Trans. Electron Devices 35 (2). 203. 1988
13 J. Knobloch. A. Aberle. B. VoB. Proc. 9 th EC PVSEC. Freiburg , 777.1989
14 A. Wang. J. Zhao, M.A. Green. Appl. Phys . Lett. 57 (61. 602. 1990
15 A. Goetzberger. private communication. 1991

635
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

The use of plasma silicon nitrides and

hydrogen plasma on polycrystalline silicon
solar cells

J.Coppye, M.Ghannam, P.De Schepper, J.Nijs and R.Mertens

Interuniversity Micro-Electronics Center, IMEC
Kapeldreef 75, B-3030 Leuven
Belgium

In order to improve the efficiency of polycrystalline silicon solar cells,

the surface, the emitter and the bulk of the solar cell need to be
passivated, In this paper, the combined action of surface passivation by
means of a thin thermal TCA oxide, the use of hydrogen containing plasma
silicon nitrides as anti-reflective coating and Si02-Si interface-, emitter-
and bulk passivation by means of a hydrogen plasma are investigated.

LlNTRODUCTION surface substrates.

In order to obtain solar cells with relatively II. APPARATUS

high efficiencies, the minority carrier diffusion
length should be quite large and the surface The hydrogen plasma treatment and Si3 N 4
recombination velocity quite small. To achieve deposition are carried out in a PD80 Plasma
these requirements, different gettering and Technology system. The cylindrical chamber has
passivation schemes have been proposed [1-6]. a diameter of 24cm. Two parallel plates with
Clean oxides are frequently used for high quality the same diameter act as the electrodes. The
surface passivation. These clean oxides are very substrates are put on the lower electrode which
thin and usually grown in dry 02 including TCA. is heated up to the desired temperature. The
Atomic hydrogen seems to play an substrate temperature is measured by a chromel-
important role in the passivation of the emitter alumel thermocouple situated under the
and the bulk of the solar cell, but also in the substrates inside the lower. electrode. The
deactivation of recombination centers at the plasma is created by means of a rf-generator
interface between the silicon and the silicon with a constant frequency of 13.56 MHz.
dioxide for oxide passivated surfaces. These Introducing SiH4 and NH3, silicon nitride layers
recombination centers are caused by the can be deposited in the same reactor with a
dangling bonds which, when saturated by uniformity within +/- 5% on 4 inch substrates.
hydrogen atoms, are deactivated. Their density III. PREPARATION OF THE SAMPLES
is therefore strongly reduced resulting in a
smaller surface recombination velocity. At T ~ All the tests are carried out on Wacker-SILSO
500°C, the Si-H bonds tend to dissociate so that 2x2 cm 2 solar cells with p-type substrates
the interfacial trap density increases again.
having a thickness of 350 11m, a resistance of
The grain boundaries in polycrystalline
0.1-10 Clcm, a diffusion length of 70 11m [7] and
silicon are structurally and chemically
disordered interfaces. Therefore, severe an average grain size of 6-10 mm 2. After
recombination occurs at the grain boundaries etching back 50 11m of the substrate to remove
and the diffusion length of the minority carriers the saw damage in a soiution of HN03 (70%),
is strongly affected. Dangling bonds in the grain CH3COOH (22%) and HF (8%), the back side of the
boundary are the main carrier trap sites and in wafers is covered with a soooA thick low
principle, can be passivated by hydrogen atoms. temperature deposited oxide (APCVD PYROX) to
Injection of atomic hydrogen in silicon can prevent phosphorus diffusion from the back side
be accomplished by: corona discharge, glow during the emitter diffusion step. The emitter
discharge, ion implantation and post-metal is created at the front side by phosphorus
annealing. In this paper, we only consider the diffusion in a clean tube furnace using P-solid
glow discharge process for treatment with source wafers (P20 5) at 900°C during 30
atomic hydrogen. Care must be taken with this minutes. After diffusion, an emitter sheet
plasma generation of atomic hydrogen species resistance of 20Cl/square is obtained. The
since radiation, generated in the plasma, can emitter is etched back (to remove the top dead
also introduce defects or structural damage in layer) to a sheet resistance of 45Q/square and
Si02 and at the Si-Si02 interface of oxidized the oxide at the back side is removed. This is

636
followed by growing a thin thermal TCA dry Efficiency i.f.o. H2-plasma passivation time
oxide on some wafers to passivate the surface.
13.2-r-----=------=-----------,
The back side metallization is carried out in 13.0

vacuum by electron beam evaporation of 21J,m of 118

.
~

aluminium followed by a sintering step at 420°C u

" 12.6
during 20 minutes in forming gas (90% N2, 10% .;;;
;: 12.4 - 0 - 110 thermal oxide
H2). For the cells that will receive a hydrogen f;l -a- lIMA Ibermal oxide
12.2
plasma treatment from the back side, the back
12.0
side metallization is grid-patterned such that
the metal covers only part of the surface. Front 1I.8+----,..-----,r----"""'T-~-~
D.O 1.0 2.' 3.' 4.'
side metallization of 500A Ti, 500A Pd and 3 IJ,m bours passi vatioD

Ag is carried out by lift-off for all cells after Fig 1. Efficiency vs. H2-plasma passivation time
the hydrogen plasma treatment.
It is clear from these results that the hydrogen
plasma treatment improves the Jsc and the Voc
IV EXPERIMENTAL PROCEDURE AND RESULTS of the cells with non-oxidized surface. On the
other hand, the efficiency of the oxide
a) Front side hydrQ.gm.passivation paSSivated cells decreases with increasing the
time of the plasma treatment. No indication
The front side of the cells with oxidized and about the passivation of the Si02-Si interface or
with non-oxidized surface is subjected to a of the grain boundaries by hydrogen plasma is
hydrogen plasma at a pressure of 0.8 Torr and a
observed for these cells. The thin thermal oxide
power of 660 mW/cm 2 . This treatment is seems to be damaged or partially destroyed by
performed for different times up to 4 hours. the plasma. This can be due to the relatively
Immediately after the hydrogen plasma high power (660 mW/cm2) used to create the
treatment (without taking the substrates out of plasma.
the reactor chamber) an anti reflective layer of
silicon nitride with a thickness of 900A and a b)Optjmjsatjon of the plasma
refractive index of 1.98 is deposited. The
optimum conditions (98 sccm NH3, 40 sccm Two experiments are carried out aiming at the
SiH4, 177 mW/cm2, 98 mtorr) are used for the determination of the optimum plasma conditions
deposition of this silicon nitride layer which (power and pressure) for bot~ types of cells
contains 23 atomic % of Si-H bonds as measured (oxide passivated and non-oxide passivated). In
by infrared spectrofotometry. these experiments, the time is fixed to 3 hours
The results of this experiment are plotted in and the temperature to 350°C. Three different
Fig 1 and listed in Table 1.
power densities (660, 300 and 150 mW/cm2) and
three different pressures (0.4, 0.6 and 0.8 Torr)
are investigated. As in the previous experiment,
Pass. ARC Surfac. Jsc Voc FF Eft.
only the front side is treated by the hydrogen
time oxide mAlcm2 mV % %
plasma. On all the cells, an optimized Si3N4 ARC
is deposited after the treatment by H2 plasma.
o hrs yes no 27,1 567 78 12
1 hrs yes no 28,1 577 77,6 12,6 i) cells with non-oxidized surface.
3 hrs yes no 28,4 582 77,6 12,87
4 hrs yes no 28,8 581 77,8 13
Power Press Jsc Voc FF Eft
1 hrs yes yes 29,2 581 77,4 13,1 mWlcm2 Torr mNc;m2 mV % %
2 hrs yes yes 28,8 582 77 12,85 ref ref 29.2 580 77.8 13
3 hrs yes yes 28,7 581 76 12,65 660 0.4 29.9 586 79.3 13.93
660 0.6 30.3 589 78.6 14.06
Table 1. cell parameters Lf.o. Hydrogen passivation time 660 0.8 30.0 584 74.9 12.9
(with ARC)
300 0.4 30.4 583 77.9 13.8
300 0.6 30.2 584 78.4 13.8
300 0.8 30.7 588 78.3 14.16

150 0.4 30.6 582 78.3 14.03

150 0.6 30.7 587 78.3 14.10
150 0.8 30.5 583 78.7 14.03

Table 2: Parameters of cells with non-oxidized surface for

different plasma conditions. (with ARC)

637
 (mV)
570
30.8 .,
I>Il
os
30.6
..
:::0
•t::
560
N' 30.4
5 =
...
'-'
~ .y
< 550
g 30.2 .,=
Co
'-' 0
....'" 30.0 540 +--.--r----.---'r--T--.--.....---.---'.....--!
0.4 0.5 0.6 0.7 0.8 0.9
29.8 +-...,......-r---.--.-~--.-.,..---.--,r--\ Pressure
0.4 0.5 0.6 0.7 0.8 0.9
Pressure (Torr) Fig 4: Voc for different H2-plasma conditions on oxide
Fig 2: Short circuit current of cells with non-oxidized surface passivated substrates
for different plasma conditions
(mA/cm2)
24.0.,.--------------,
Efficiency vs. Power, Pressure
=
14.5
150 mW/cm2
~
e
...
23.0

ii 14.0 E::::::::::::=:::lt-::::::::~::::.; ·S
i::
300 mW/cm2 ~ 22.0
jj
r"---_.-- 300 mW/cm2
~

21.0 +--...,.--r---r--.....,.-~
0.4 0.5 0.6 0.7 0.8 0.9
Pressure (torr)
12.5 +--.--Y-~--.-....---,..--..--...--.--I Fig 5: Jsc for different H2-plasma conditions on oxide
0.4 0.5 0.6 0.7 0.8 0.9 passivated substrates
Pressure (Torr) (%)
10.0.,--------------,
Fig 3:Efficiencies of cells with non-oxidized surface for
different plasma conditions 9.8
....'-' 9.6
From Table 2, Fig. 2 and 3, it can be concluded .,= 9.4
.Cj
that, within the limits of the experiments
performed here, the hydrogen plasma treatment ...<:
f;I<1
9.2

at a power of 300 mW/cm2 and a pressure of 0.8 9.0 mW/cm2

Torr results in the best post treatment 8.8
efficiency on cells with a non-oxidized surface. 0.4 O.S 0.6 0.7 0.8 0.9
A maximum efficiency of 14.4% and an average Pressure (torr)
efficiency of 14.16% is reached on these cells.
Fig 6: Efficiencies for different H2-plasma conditions on oxide
passivated substrates
ii) Cells with oxidized surface
A similar test is carried out on cells with an
Power Press Jsc Voc FF E ff oxidized surface. For these cells, the best
mW/cm2 Torr mAlcm2 mV % %
plasma conditions seem to be associated with a
300 0.4 21.9 546 77.3 9.26
power density of 150 mW/cm2 at a pressure of
300 0.6 21.8 541 75.4 8.9
0.5 Torr as depicted in Fig 4, 5 and 6 and Table
300 0.8 22.2 544 77.2 9.32
3. In fact, the results given in the next section
indicate that front side treatment by H2 plasma
150 0.4 22.2 559 77.6 9.61
always degrades the performance of cells with
150 0.6 23.4 565 75 9.92
oxide passivated surfaces.
150 0.8 23.1 550 74.4 9.43
c) Comparison between front- and back side
Table 3: Parameters of cells with oxidized surface for hydrogen passivation
different plasma conditions. (without ARC)
Hydrogen plasma treatments at the optimum
conditions determined earlier are carried out on
cells with oxidized and non-oxidized surface.
The results are compared for front side, back

638
side and front + back side treatments (3 hours As expected, both the short circuit current and
each side). For cells which are treated from the the open circuit voltage of the cells with a non-
back side, the back side metallization consists oxidized surface are maximum when the
of a grid with 70% coverage. The results are hydrogen plasma treatment is done from the
shown in Table 4 , Fig. 7 and Fig.B. front and back sides of the cell. On the other
hand. back side treatment only is effective in
cells with non-oxidized surface: passivating the grain boundaries on the oxide
300 mW/cm2. 0.8 Torr, 3 hours passivated cells. The result of such a treatment
is an enhanced short circuit current and open
Passivation side Jsc Voc FF Eft circuit voltage compared to the non-treated cell.
The front side treatment partially destroys the
Front Back mA/cm2 mV % % passivating Si02 layer which results in an
increased surface recombination velocity,
X 21.6 572 79.3 9.8 thereby reducing both Voc and Jsc.
x 21.8 571 73.5 9.14
x x 22.5 573 76.1 9.77 V. SUMMARY AND CONCLUSION

cells with oxidized surface: In this paper. hydrogen plasma passivation

150 mW/cm2, 0.6 Torr. 3 hours as a last step in a polycrystalline silicon solar
cell process is investigated. It is confirmed that
Passivation side Jsc Voc FF Eff this step is an effective means for passivating
the grain boundaries. Cells with no oxide
mA/cm2 mV % %
Front Back passivated surface are significantly improved
when both sides of the cell (front and back) are
X 22.3 544 78 9.49 subjected to the hydrogen treatment.
x 23.8 567 74 9.93 On the other hand, the cells with an oxide
x x 22.7 544 75 9.25 passivated surface are improved after a
ref ref 23 564 76.9 9.96 hydrogen plasma treatment from the back side
only. Any front side treatment for these cells
Table 4: Parameters of cells with hydrogen plasma passivation deteriorates the oxide passivation effect.
on front. back and front+back side (without ARC)
ACKNOWLEDGEMENTS
(mV)
580.0
: no oxide The authors greatly acknowledge W. Laureys for
570.0 • his help during the processing of the cells.
This work is supported by the Multichess
'"
>= 560.0 projects of the Joule program of the EEC under
contract N JOUR -0036-C (GOF).
550.0

540.0 1Michael L Reed 1989, semicond. Sci. TechnoL 4 980-985

2 C.H. Seager and D.S. Ginley 1981. J.AppLPhys. 52(2) 1050-
1055
3 A.Chari. P. de Mierry. A.Menikh and M.Aucouturier 1987.
Revue Phys.AppL 22 655-662
Fig 7: Voc of hydrogen passivated wafers on front. back and
4 C.Zurletto. D.Sarti. M Pasquilelli and S Martinuzzi 1989.
front+back side.
9th ECPSEC Freiburg 684-686
(mA/cm2) 5 L.A.Verhoef. S.Roorde. R. van Zolingen and W.Sinke 1988.
24.0,.-;.------------, 21th IEEE PSC 1551-1556
6 T.Saitoch 1987. Proc. 19th ConI. IEEE PVSC 1518
23.0 7 M.Le Quang Nam. C.N.R.S .• private communication
...
....
<Il

22.0

21.0 "'!==E=::::::;-r=::ll=::;r::===::!....,

Fig 8: Jsc of hydrogen passivated wafers on front. back and

front+back side.

639
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EFFECT OF TEMPERATURE AND LIGHT INTENSITY

ON THE PERFORMANCE OF MIS INVERSION LAYER SOLAR CELLS

K. Jaeger and W. Hoffmann

NUKEM GmbH, P.O.B. 1100S0, 0-6450 Hanau II, FRG
R. Hezel
Institut fur Werkstoffwissenschaften 6, Universitat
Erlangen-Nurnberg, D-S520 Erlangen, FRG

ABSTRACT. It is the purpose of this work to demonstrate how

t~e performance of MIS (~etal-Insulator-semiconductor) inver-
S10n layer solar cells 1S affected by changes in the cell
temperature and the light intensity. It was found that the
temperature behaviour of MIS-IL solar cells is quite similar
to, that, of n+p c,,:lls. This similar,ity is demonstrated by de-
ta1led 1nvest1gat10ns of the carr1er transport mechanism in
the MIS contacts, which indicate, that minority carriers domi-
nate the I-V characteristics as in n+p junctions. As an impor-
tant result, measurements of the illumination dependence of
the cell parameter reveal, that the fill factor of inversion
layer solar cells increases by abQut 3 - 4% if the irradiance
is lowered from 1000W/m 2 to 560W/m2.
I~ conclusio~, the,results clearly demonstrate the high poten-
t1a~ of MIS 1nverS10n layer solar cells for terrestrial appli-
cat10n particularly u~der normal insolation which is most of
the time below 1000W/m .

1. INTRODUCTION controlled copper block, were examined at

temperatures ranging from 20-S0'C. The
The advantages of metal-insulator- operating cell temperature was determined
semiconductor inversion layer (KIS-IL) so- with a PT-100 element fixed on the illumi-
lar cells with plasma silicon nitride as a nated surface of the solar cells.
charged dielectric are their simple low The properties of the MIS (aluminum/thin
temperature fabrication process as well as SiOx/p-Si) front contacts of the inversion
their potential for high efficiencies /1/. layer cells at elevated temperatures were
Recently high conversion efficiencies up to investigated by current-volt~ge (I-V) and
15% obtained for monocrystalline silicon on capacitance-voltage (C-V, l/C vs V) meas-
an industrially oriented wafer size (100mm urements , respectively. The measurements
x 100mm) were presented, together with were performed on MIS diodes (top contact
first experimental results concerning the area 1.smm 2 ) coprocessed on the solar sili-
module fabrication of MIS-IL cells /2/. A con material.
pilot plant of about lMW annual capacity
is currently built for p~oduction of this
promising novel solar cell type. 3. RESULTS AND DISCUSSION
For a large scale application it is neces-
sa!'y to study how the performance of the 3.1 Temperature Dependence of the MIS
inversion layer solar cells is affected by Contact Properties
changes in the cell temperature and the Dark I-V measurements revealed, that an
light intensity, since in real application increase of the dark saturation current I 1
the operation tem-perature of the solar of about 10 3 A/cm 2 occurs, if the substra~e
cells as well as the irradiance can vary temperature is raised from 25'C to SO'C.
over a wide range. For an "ideal" MIS diode the saturation
current 101 results eXClusively from car-
rier recomoination in the diffusion region
3. EXPERIMENTAL DETAILS and the p-silicon surface beneath the MIS
diode must be inverted /3/. In order to
MIS inversion layer solar cells (100mm check the MIS surface potential at higher
x 100mm) were fabricated on mono- (0.5- temperatures capacitance-voltage curves
0.70hmcm, CZ, Wacker SM1/Q, 0.35mm thick) (see Fig.l) were measured on the MIS con-
and multicrystalline (Wacker silso SM2/Q, tacts at temperatures ranging from 29SK to
0.35mm thick) p-type silicon wafers, 353K. The C-V curves (100kHz) could only be
respectively, according to the low tempera- recorded from inversion (negative bias to
ture processing as outlined in /2/. semiconductor) to depletion (small positive
The temperature and the illumination forward bias to semiconductor), since due
dependence of the solar cell data were to the rapid increase in current in the MIS
measured with a HMI lamp (Osram Metallo- diode at forward bias conditions the ca-
gen), whose spectral distribution cor- pacitance bridge went out of balance. From
responds well to the AMl. 5 spectrum. Dif- the corresponding 1/C 2 versus V plots, also
ferent metallic gauzes were used to var¥ depicted in Fig.l, the values "fo for the
the irradiance over the range 250-1250W/m surface potential at zero bias were deter-
without changing the spectral distribution mined by linear extrapolation to the volt-
and the spatial uniformity in the test age axis /3 /. The experimental 'Y'o-values
area. The cells, mounted on a temperature as a function of the temperature are shown

640
in Fig. 2, together with the calculated
surface potential 1jI inv required to drive 0.78. CI ~---Ec

the surface into strong inversion /4/. (The j

doping concentration NA was derived from
the slope of the 1/C2 vs V curves.) It can v
be seen, that the experimental surface po-
tential decreases with increasing tempera-
0.76
,,
ture from ~0=0.77V at 298K to Q=0.68V at Y ,, SiOx
3s3K, but, as an important result, is ex-
ceeding the calculated ~inv-values over the
,,
whole temperature range. Thus it can be ~ 0.74 ,,
concluded, that the silicon surface beneath ~ ,,
the MIS contact is still inverted at the
3 ,,
high temperatures of interest. In addition,
it has to be pointed out, that according to
the theory /3/, the experimentally obtained
i 0.72
,,
,,
intercept with the voltage axis, Yo' is b ,,
=-
even smaller than the real surface poten-
tial. The results demonstrate, that the
,,
current flow in the MIS inversion layer so-
lar cells is semiconductor-limited, which
",
means that the dark' saturation current 101 ""
mainly results from minority carrier recom-
= ""
",
bination in the neutral p-base region. - - calculated 'Pin. 2'PB
0.68
o experimental % (C-V plot)
"",
~
I

lie' (ARB . CALE) ~1 .4nm

lOx
300 320 J40
p-SI TEMPERATURE [K]

FREQ. IOOkliz
Fig.2 Surface potential of MIS contacts
298K (see insert) as a function of the tempera-
ture. Th~ experimental values were obtained
from l/C vs V plots. The dashed line re-
presents the calculated values of the sur-
~ace p~tential)Pinv at the onset of strong
1nverS10n.

0.5
majority carrier Schotty contact.

, , In Table 1 the temperature coefficient of

the open-circuit voltage Voc, the short-
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.41 circuit current Isc, the fill factor FF and
HlA TO E 11COND OR IVI
the efficiency for MIS-IL cells on mono-
and multicrystalline silicon, respectively,
Fig. 1 Experimental capacitance-voltage are presented. For comparison corresponding
characteristics for MIS (aluminum/ SiOx values are given for conventional n+p sili-
(1.4nm)/ p-Si) contacts at temperatures con solar cells measured under the same
298K and 3s3K, respectively. Also shown in test conditions. The results for the n+p
arbitrary scale are the corresponding 1/C2 solar cells agree well with those obtained
vs V plots extrapolated to the voltage axis in /6/. As can be seen, the temperature
in order to determine the surface potential coefficients of MIS-IL and n+p solar cells
Ylo(T) . are quite similar, which means, that the
current transport mechanism is dominated by
minority carriers in both cell types.
3.2 Temperature Dependence of the Solar For both MIS-IL and n+p solar cells no re-
Cell Data markable difference in the temperature be-
In Fig.3 the dependence (log scale) of haviour between the mono- and multicrystal-
the open-circuit voltage on the short-cir- line silicon material was observed. This
cuit current is shown, measured on large- result is confirmed by theoretical evalua-
area MIS inversion layer solar cells under tions which reveal, that the expected de-
different illumination intensities. The crease of the temperature sensitivity of
Isc-Voc plots, taken at five cell tempera- Voc (by 0.07mV/'C) for monocrystalline so-
tures spanning the range 22-7s'C, are shif- lar cells due to their higher (by about
ted to lower voltages with increasing tem- 20mV) open-circuit voltages is within the
perature. As a consequence, the dark sat- experimental error /7/.
uration current 101 (determined for each
temperature Ti by averaging the values of First outdoor measurements on an inversion
I Ii calculated from IOli=Is~*ex~(­ layer solar cell module have been performed
qijoC/kTi) for each Isc-Voc pa1r) 1S 1n- at the RWE test site in Kobern-Gondorf. The
creased. Following the procedure outlined module consists of 10 cells connected in
in /5/, from a semilog Pl~t of the IOl-val~ series and sandwiched in glass (2mm thick)
ues (corrected for the T- .4-factor) versus using a castable resin as encapsulating
the reciprocal temperature an activation material /2/. The field data received
energy of EA=(1.20+-0.00l)eV was evaluated. during a bright day with light wind (March
The experimentally obtained EA-value, which 90) were used to determine the nominal
corresponds well to the silicon band gap, operating cell temperature (NOCT) /8/. As
clearly indicates the presence of a minor- illustrated in Fig.4, the cell temperature
ity carrier MIS contact rather than a is roughly proportional to the insolation

641
 where Rs=(V*-VOC)/ISC with V* as the volt-
age from the dark I-V curve at the point
T: 75 56 43 31 22 where I=Isc. It was found that the Rs-value
10
for the large-area inversion layer cells is

----- ~~~
530 52.5 7.65 1.21 0.25 ' 10" A
about 10% higher than that obtained for the
~ diffused n+p cells.
As a consequence of the increase in fill
factor with decreasing irradiance the effi-
ciency of the inversion layer solar cells
(see Fig.5) remains almost constant down to
an intensity of about 600w/m2. This is in
contrast to the typical behaviour of con-
ventional solar cells showing a reduction
in the cell efficiency if the incident
light intensity is lowered.

EL AREA 98 m'
0. 1 L-IL-l-1---'----'--'---'--'--'---'--'----'--'--l....-.JL...J----'----L-'---'-_ 4. CONCLUSION
0.40 0.45 0.50 O.sS O.6()
Vue I I
In conclusion, it was demonstrated
that the p-Si-surface beneath the MIS con-
Fig. 3 Experimental Isc-Voc character- tact is still inverted at elevated tempera-
istics of MIS inversion layer solar cells tures up to 80 C relevant for the normal
0

at different temperatures. The values of operating conditions of solar cells. This

the dark saturation current 1 01 , deduced is an important prerequisite for the cur-
from the semilog Isc-Voc plots, are given. rent transport in MIS inversion layer solar
cells to be dominated by minority carriers
throughout the temperature range of inter-
est as in n+p junction cells. Consequently
the temperature coefficients of the MIS-IL
cells closel¥ resemble those of convention-
E
to!
A IIJIENT TEMPERAT RE 10' C
=
() r 39' AT 800IV/III' al silicon n p solar cells.
Due to the significant increase in fill
""
~40 F-__- - - -__- - - -- -__- - - - - -- -____----~
factor with decreasing irradiance the effi-
<{
ciency of the MIS inversion layer solar
""
t.I
Co cells remains constant or even reaches
::;r 30 higher values at light intensities below
t.I
~
.....J ~ _.or 1000w/m 2 . These insolations are by far pre-
vailing during normal terrestrial operation
~ 20 " O.02J°C1\Vm- 2 and hence the overall yearly efficiency of
a photovol taic plant increases quite sub-
"
<{
..J stantially .
~lOOL---~--~--~--~=---~--~--~--~v.t_.--~
200 400 600 Mru
IRIIAOIi\ • I.; IW/m2 )

Fig. 4 Determination of the nominal

operating cell temperature (NOCT) for a MIS
inversion layer solar cell module. CELL AREA 10 x 10cm2
1.04 • MI-IL
D N+P

level. From the straight line best fitting ~ 1.02

the experimental data (indicated by the
dashed line in Fig.4) a NOCT of 39'C can be ~ 1.00
deduced, a value, which correlates well
with 4l'C given in /9/ for the typical NOCT 0.98
of a flat-plate photovoltaic module with a
clear glass substrate. 0.96
0.2 0.4 0.6 0.8 1.0 1.2
3.3 Illumination Dependence of the Solar
Cell Data 1.02
Fig.5 presents the effect of light
intensity on the fill factor and the effi- 1.00
ciency normalized to their values at
1000w/m2, FF lOOO and ?'JlOQO respectively, ~
for large-area MIS invers~on layer solar ~ 0.98
cells. For comparison corresponding data
measured on n+p solar cells are also de- 0.96
picted in Fig. 5 . It can be seen, that in
contrast to the n+p cells the fill factor 0 .9~
of the MIS-IL cells increases by about 3-4~ 0.2 0.6 0.8 1.0 1.2
if the irradiance is lowered from 1000W/m IRRAOIA T E [10 J . W /m 2 J
to 560w/m2. This effect can be attributed
to the series resistance of the cells which Fig. 5 Effect of light intensity on the
due to the relatively high inversion layer fill factor and the efficiency (normalized
sheet resistance strongly depends on the to their values at lOOOw/m2) of MIS inver-
illumination intensity and thus the cell sion layer solar cells. For comparison cor-
current. The series resistance Rs was de- responding values obtained for diffused n+p
termined by the method described in /10/ solar cells are given.

642
ACKNOWLEDGEMENT /3/ J. Shewchun, M.A. Green and
The authors gratefully acknowledge the en- F.D. King, Solid-State Electron.,
gaged experimental work of G. Bende, W. Vol. 17, p.563 (1974)
Bergmann, P. Roth and B. Schum. /4/ S.M. Sze, Physics of Semiconductor
We are also gratefull to U. Beyer and J. Devices, Wiley, New York (1969)
Steeger of the Rheinisch-Westfalisches /5/ N.G. Tarr and D.L. Pulfrey, Appl.
Elektrizitatswerk AG for providing the mod- Phys. Lett., Vol 34(4),p. 295 (1979)
ule data. /6/ M. A. Green, Solar Cells, Vol. 7,
This work was supported by the Bundesminis- p. 337 (1982)
terium fur Forschung und Technologie /7/ M. A. Green, "Solar Cells",Prentice-
(Contract No. 0328772 A) of the FRG. Hall, New Jersey, p. 92, 1982
/8/ specification 501, Ispra, EU 7545 EN,
REFERENCES p. 16 (1981)
/1/ R. Heze1 and R. Schorner, J. of Appl. /9/ R. G. Ross, Proc. of the 15th IEEE
Phys., Vol. 52(4), p. 3076 (1981) Photovoltaic Specialists Conf.,
/2/ W. Hoffmann, K. Jaeger, G. Luthardt Kissimmee, p. 811 (1981)
and R. Hezel, Proc. of the 9th /10/ K. Rajkanan and J. Shewchun, Solid-
European Photovoltaic Solar Energy state Electron., Vol. 22,p.193 (1979)
Conf., Freiburg, p. 677 (1989)

lOx 10cm 2 monocryst. Si multicryst. Si

(0.5 -1.7 ilcm, CZ) (0.5 - 5ilcm, Silso)

cell type n+p MIS-IL n+p MIS-IL

disc/dT -2 -2
(2.1 ± 0.7) x 10-2 (2.7 ± 0.8) x 10 (2.8 ± 0.5) x 10 (3.6 ± 1.3) x 10-2
[mAlcm 2 K]

dV oc IdT ·2.4 ±0.2

-2.6 ±0.2 -2.4±0.3 ·2.6±0.1
[mV/K]

dFF/dT
(-13 ± 1) x 10-2 (.13 ± 4) x 10-2 (·14 ± 2) x 10-2 (·13 ± 1) x 10-2
[%/K]

dll/dT -2
(·6.2 ± 0.2) x 10-2 (·5.3 ± 0.2) x 10-2
(·7 ± 0.9) x 10-2 (·6.5 ± 0.9) x 10
[%/K]

Table 1 Temperature coefficients of the solar cell data

(isc, Voc, FF and r;) for MIS inversion layer (MIS-ILl and
conventional n+p solar cells on mono- and multicrystalline
silicon, respectively.

643
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

LASER SCANNING ANALYSIS OF SILICON SOLAR CELLS

A E Hughes, B M Neville, K C Heasman and M P Lesniak

BP Research, Sunbury Research Centre,

Chertsey Road, Sunbury on Thames,
Middlesex, TW16 7LN, ENGLAND

ABSTRACT

We describe the application of a multiwavelength (0.442, 0.632 and 1.06 microns) laser
scanner in conjunction with a wide range of analytical techniques to the identification of
defects in silicon solar cells. The laser scanner can be run manually or under computer
control, producing line scans, intensity modulated (OBIC) and isometric images under
electrical or dc white light bias. Electrical characterisation of a defective solar cell
enables the distinction of lifetime related problems from resistive effects. Laser
scanning allows the determination of the spatial distribution/locations of any defects.
Regions of low lifetime, low -shunt resistance or high series resistance can be identified.
We describe examples of the location of point shunts and top-contact shunting, of poor top-
contact formation and of the identification, using supporting analytical techniques, of
lifetime killing impurities and wafer defects. The use of short and long wavelength
excitation allows the distinction of emitter from bulk defects. Computer generated
photoresponse maps can be converted to diffusion length maps following a calibration
procedure with a solar cell whose diffusion length has been measured by an alternative
technique.

1. INTRODUCTION Light (focussed to a spot size of 40

microns) from one of three lasers, (HeCd, HeNe
Electrical characterisation of a defective and Nd YAG) selected using mirrors M1, M2 and
solar cell does not give the spatial location of M3 (not shown) is scanned by two galvonometer
defects responsible for the poor performance driven mirrors Mx and ~ across a solar cell. ~
unless the solar cell is cut up into pieces. is controlled by a sawtooth waveform to give a
Laser scanning allows us to locate and identify line scan across the solar cell and ~ by a
a wide range of solar cell defects. We describe staircase waveform to give a raster scan in
the application.of a multiwave1ength (0.442, combination with Mx. The waveforms which are fed
0.632 and 1.06 microns) laser scanner in to the galvo driver are either generated by a
conjunction with a wide range of analytioal personal computer (with a plug in multichannel
techniques to the identification of defects in 10 card) or by two scan generators. The output
silicon solar cells. from the solar cell is displayed either on an
oscilloscope scanned in synchronism with the
beam or is processed by an A to D converter and
2. EQUIPMENT stored on disc. The data is stored in a 240 x
320 point array with 8 bit resolution for
The laser scanner can be run manually, as in the subsequent analysis. Individual line scans can
schematic of fig. 1, or under computer control, be extracted and diffusion length maps generated
producing line scans, or intensity modulated from scans at 1.06 microns. For computer display
(OBIC) and isometric area maps. Scans of whole a colour coded image is produced with different
cells can be obtained under short circuit intensities represented by different colours.
conditions or with electrical or de white light
bias.
3. APPLICATIONS

3.1 Shunting
When a solar cell is found to have a low

M1
MY) shunt resistance this could be due to isolated
point shunts, distributed shunting or to top
I M,
contact shunting. Laser scanning can distinguish
I
I between the different shunts.

+I (a) Location of point shunts

I A point shunt leads to a spatially dependent

~
i; loss of collected current which increases as the
j shunt is approached. Theory describing the
appearance of a point shunt in a laser scan has
z~ ;"! been developed by Lehovec and Fedotowsky [2]. A
r" ~ line scan through a point shunt has a
characteristic cusp shape as shown in fig. 2a,
which enables accurate location of the shunt.
The shunt results in a diffuse black spot in an
y intensity modulated area scan fig. 2b and has a
whirlpool appearance in an isometric mode area
scan, fig. 2. Subsequent Energy Dispersive X-Ray
Fig. 1: Schematic of laser scanner showing analysis (EDX) in an electron microscope allows
manual control option. chemical identification of the shunting material
(e.g. AI, Ag, SiC).

644
b.
o.
a. a
1M:,."

Fig. 2: Laser scans of a point shunt (a) line

scan (b) intensity modulated area scan
and (c) isometric area scan.

(b) Shunting beneath top-contact

When the top contact shunts directly to the
base (as can occur for example if too high a top
contact firing temperature is used) this results
in a decrease in laser scan signal which is
independent of position. In fig. 3a we show a ,~
laser scan of a solar cell with a high shunt
resistance. A line scan from a similar cell with
a low shunt resistance (due to a shunted top b nol ..,.
contact) gives a lower collected current, fig.
3b (and a poorer signal to noise ratio as the
inputs of the amplifier circuit are shunted).
However, the origin of the shunting can be
demonstrated by contacting to the emitter away Fig. 4: Laser scans (with and without light
from the top contact using a fine probe which bias) from solar cells where the fine
shows the loss of collected current as the top grid lines are making (a) good and (b)
contact is approached, as can be seen from fig. poor contact to the emitter .
3c. Whether the shunting of the top contact is
localised or distributed can only be determined 3.3 Identification of lifetime killing impurities
by dividing the cell into pieces . Lifetime killing defects result in regions
of low response in laser scan images. We give an
example where the cause of the poor performance
of single crystal silicon solar cells was
identified using laser scanning in combination
c. with a number of analytical techniques . Figure
Sa-c show laser scans of a defective solar cell.
There are three circular regions where the
response is normal whereas over the remainder of
b. the cell the response is reduced by the effect
of lifetime killing point defects . Examination
in the SEM in EBIC mode, fig. 5d, show that each

a. "black spot" seen in the laser scan is a defect

cluster.

a. b.

Fig. 3: Line scan of (a) a cell with a high

shunt resistance (b) a similar cell with
a low shunt resistance (due to top
contact shunting) and (c) same as (b)
but with contact made to the emitter
using a fine probe.

3.2 Poor top-contact formation/high emitter

sheet resistance
Sawyer and Kessler [3] and others [4] have
shown that laser scanning of a forward biassed
cell (using electrical or a dc white light bias)
allows the detection and quantification of
regions of high emitter sheet resistance or poor
top-contact formation. Forward biassing a solar
cell introduces distributed shunting and the
resulting position dependent loss in short
c ircuit current can be related to the emitter Fig. 5: (a) and (b) laser scans of a defective
sheet resistance. Fig . 4 compares laser scans solar cell with different contrast
(with and without light bias) from solar cells settings (c) magnified laser scan of
whe r e the fine grid lines are making (a) good part of the same cell (d) EBIC image of
and (b) poor contact to the emitter. one of the "black spot" defects in (c).

645
 A highly defective area was selected for length maps. In fig. 8 we show a laser scan of a
cross sectional TEM which allowed isolation of a polycrystalline silicon solar cell (a) together
single defect, fig. 6 and EDX to identify the with the corresponding diffusion length map (b)
impurity as copper. Knowledge of the spatial where L has been averaged over 1em x 1em areas.
distribution and composition of the defects
allowed the process stage where the impurities

. ~"\,",~,~.It · 1
~~~:~~\~.-: '~ ~~"I ~ t .
were introduced to be identified.
,

~
,
~
~
," ,
....::-
1>00.'

.~,
"
:,". '
~,.

, ~',.'
"
.
I
'. ".',.
I. .·048

~ ...... ~. . ~ .,,\, . ' .,'1

- ... _ ... ~ " i 'I, ' ~ ' ''',~
~ . '~ 1·~04, .. ' ,
• ~ to ... , '-~"'''' '' '' ·"".r· '
- .. t"~~ji'~:' ..~ ~ , ~. :::;
-'
.~:'-
-=-;r.:..--~~
~·.... ~~.
"·~l' .,-' ·-'
.

59 88 130118130115107 94 122 100

107129 130 134 147 144 118 135 140 122

89 126113109121 121 107113109115

113142140113122119107113141 135

lOa 144 140 139 114 115 106 119 132 127

lOa 138 109 122 119 119 123 127 139 124
Fig. 6: TEM lattice image (x 5 million) of
98 143 129 143 109 126 116 127 121 123
single defect identified by EDX to be a
copper inclusion. 95 115135128132128122 128 109 113
~1~1~~41~1221~g41~g8

3.4 Identification of wafer defects

57 130112149149148 130 135142112
Defective wafers can give distinctive laser
scan images. This is illustrated in fig. 7 which
shows the effect of intrinsic wafer defects
(swirl defects in Czochralski grown silicon Fig. 8: (a) Laser scan of a polycrystalline
wafers) in a defective solar cell. This has been silicon solar cell with 1.06 micron
confirmed by mapping the lifetime of both light and (b) the corresponding
unprocessed and processed wafers using the time diffusion length map.
resolved photoconductivity technique.

4. CONCLUSIONS

Laser scanning in conjunction with a range

of analytical techniques is demonstrated to
provide an invaluable method of identifying
defects in silicon solar cells.

ACKNOWLEDGEMENTS
The authors would like to thank Dr D W White
and Mr K J Meade for the electron microscopy
measurements. This work has been funded by BP
Solar Limited.

REFERENCES
(1) K Lehovec and A Fedotowsky, "Scanning light
spot analysis of faulty solar cells" Solid
State Electronics, 23, 565-576, 1980.

Fig. 7: Laser scan of a defective silicon solar (2) D E Sawyer and H K Kessler, "Laser scanning
cell showing swirl defects. of solar cells for display of solar cell
operating characteristics and detection of
cell defects" IEEE Trans. Electron Devices
3.5 Diffusion length mapping ED27, 864-872,1980.
The response of a solar cell at 1.06 microns
can be related to the base diffusion length. (3) P Kowalski, W F Lankford and H A Schaff,
Stokes and Chu have shown that the long "Non-destructive measurement of solar cell
wavelength response of a solar cell R is given sheet resistance using a laser scanner. IEEE
by Trans. Electron Devices ED31, 566-570, 1984.

R ~ kaL/(1 + aLl (4) E D Stokes and T L Chu, "Diffusion lengths

in solar cells from short circuit currents"
where k is a constant, a is the absorption
Appl. Phys. Lett. 30, 425-426,1977.
coefficient and L is the diffusion length. The
laser scanner can be calibrated by measuring a
solar cell of known diffusion length (measured
by an alternative technique such as the Stokes
and Chu photocurrent technique [4]) under the
same conditions. This allows computer generated
response maps to be converted to diffusion

646
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DEFECTS INDUCED IN BULK SILICON BY RAPID THERMAL

ANNEALING

E.Susi, R.Fabbri, APoggi

CNR - Istituto LAMEL, Via Castagnoli 1, 40126 Bologna, Italy

L.Passari, M.C.Carotta, M.Merli

Dipartimento di Fisica dell'Universita, Via del Paradiso 12, Ferrara, Italy

ABSTRACT. The influence of Rapid Thermal Annealing (RTA), on silicon

electrical and structural characteristics is studied. A degradation directly
connected to the technique is observed in electron beam anneal~d samples,
while the lamp annealed ones do not change crystal propertIes for low
temperature anneals.

INTRODUCTION depending only on the position in the ~afer c~ ~e

explained by supposing that the gettenng actIOn IS
The control of defect introduction in silicon performed by some oxygen aggregates non
during the thermal treatments needed by the devi~e homogeneously distributed in the non treated crystals
processing is an old research aim. The use of the RapId as a consequence of the cooling process after the
Thermal Annealing (RTA) techniques has added new crystal growth and/or the thermal donor stabilization
interest to the topic, both from the fundamental and pre anneal. The absence of correlation with thermal
technological points of view (1,2,3). donor concentration suggests, also, that only neutral
The purpose of our work is to understand the clusters can give this gettering effect.
interaction between oxygen, carbon and metal At temperatures hig~er tha~ 709·~. a
impurities, point defects an~ t~~ir aggreg~tes degradation of lifetime and an mcrease m reSIstiVIty,
responsible of the changes m sIhcon electnc~l strongly dependent on temperature, was obsel."'ed. in
properties during a thermal process. The problem IS FZ crystals. An increase of surface recombm~tlOn
very complex, owing to the presence of different defect velocity was present too. In CZ crystals at 750 C a
reactions and to the dependence on many parameters. behaviour similar to the 400'C one, with an mcrease
In this paper we report the development of a as well a decrease oflifetime was present. At 1050'C a
study, carried by us since some years. (4),. about ~he strong degradation of lifetime, an increase of
effects of thermal treatments on mmonty carner resistivity and of metal contamination (detected by
lifetime and crystal disorder in lightly and mo~erately DLTS measurements) have been observed in FZ as
doped silicon as a function of dopant concentratIon and well as CZ samples (7).
oxygen and carbon contents. In order to test our hypothesis about the
We have begun by studying the effects on different mechanisms interacting at low temperature,
minority carrier lifetime and on crystal order of we have compared the furnace ~OO'C is~therm on .CZ
conventional furnace thermal treatments in p-type FZ crystals with isotherms at 400 C for tImes rangmg
silicon in the temperature ranges 350-450'C and 750- from 5 to 40 s obtained using RTA techniques (lamp
1050'C (5). At low temperature a degradation of and electron beam).
lifetime and an increase of crystal disorder, as detected In fact, the processes we hypothesize, namely
by X-ray double crystal diffractometry, with a
maximum effect centred around 400'C were detected. Fe-B pair formation and dissociation;
Both phenomena depend on boron content and diffusion of intersti tial Fe;
disappear in moderately doped samples (N a =1017 cm- rearrangement of oxygen aggregates associate
3).
with a gettering activity;
For CZ crystals, in the same temperat~re range, formation of thermal donor;
we have observed two different mechamsms, one
resulting in a lifetime enhancement. and the oth~r in.a are likely to have different velocities, being slower the
lifetime decrease (6). The degrading mechamsm IS ones involving oxygen movement. The use of RTA,
similar to the mechanism observed in FZ crystal, it do allowing very short heating times, can help to ~eparate
not depend on oxygen content and on thermal donor the overlapping mechanisms. and to ~~amblguously
presence, while the enhancement is oxygen dependent clarify the role of each one m detenmnmg the final
but no correlation with thermal donors has been material characteristics.
observed. The occurrence of one of different behaviours Annealings at 750'and 1050'C were also
depends only on the position of the sample in the performed,in order to have informations about the
wafer. A clear correlation between lifetime and crystal processes induced by high temperature treatments.
disorder variations exists and the variations of these
parameters are strongly dependent on ~ime and EXPERIMENTAL
temperature. It is possible to change drastIcally the
400'C behaviour by a preanneal at hig~ temperatu:e. We have used FZ and CZ wafers, boron doped at
We have explained these expenmental findings concentration varying from 8xl0 14 to 2xl015 cm-.3. The
by supposing that Fe-B pairs, present in the as-grown initial oxygen content of the CZ wafer were m the
crystal as a consequence of room temperature range 6xl0 17 - lx1018 cm- 3. Initial lifetime values
rearrangement of Fe atoms, dis~ociate. during ranged from 10 to 40 ~s. The CZ wafers had been
annealing at low temperat~re, produCl?-g ~n mcrease subjected to donor stabilization anneal. Before any
in interstitial Fe, an effiCIent recombmatIon centre, treatment wafers were tested for lifetime, resistivity,
which is responsible for lifetime diminutio~. The oxygen content and cryst~l p~rfec~io~ homogeneity.
dependence of lifetime increase, observed m CZ Appreciable non-homogeneIty m hfetIme has been
crystals, on oxygen content suggests a gettering action observed with a maximum difference of 50% between
of oxygen on free interstitial Fe ..The occurre~~e of the cent;e and the periphery of the wafer. To avoid
either behaviour in the same expenmental condItIons, experimental uncertainties due to this non-

647
 homogeneity, samples to be compared were taken from electron beam annealed specimens show a minority
nearby zones of the same wafer, so that lifetime carrier lifetime degradation already for annealing time
variations of more than 20% have been considered as of 5 s; no correlation with lattice order variation, no
significative. variation in resistivity and only a weak time and
The resistivity and the oxygen variations were temperature dependence are observed. The extremely
not higher than 10%. weak dependence on time and temperature annealing
The lamp annealings were carried out with a of the lifetime decrease points to a very quick process,
commercial Heatpulse 410 (Ag Ass.) in nitrogen probably connected to the electron beam in itself.
.ambient. The annealing chamber is heated by means At 750'C (Fig.2) we observe a diminution of
of 13 tungsten - halogen 1.5 kW lamps. A lifetime in FZ specimens annealed by lamp. The time
thermocouple feedback system is provided for the dependence is very weak, while no variation in crystal
temperature control and measurement. The order and surface recombination velocity is observed.
irradiations were performed with heating and cooling
rates similar to the furnace and electron beam
systems, so that the two RTAs had similar thermal
histories, while the only difference between the
furnace and the RTA treatments was the dwell time at 0.5 . - - - - - - - - - - - - - - - ,
the chosen temperature.
Minority carrier lifetime, resistivity, lattice r--- ------------- -----.
perfection, oxygen and carbon content were measured 0.0 FZ 750°C
·l~ ~---------------------.
before and after the annealing. For the details of the
measuring techniques see ref.(8). ~ --0.5 -
The 400'C isotherms by the two RTA systems <I
(lamp and electron beam) give different results (Fig. 1). -1.0- 1 . ' • • , •
The data obtained by the lamp are well compatible
with the furnace results (8). On the contrary the ~ 1. 5 '--'--'--'-.l.-l--'--'--'-'--'--'--,-/ /--'-..L...J--L-1

2.0 , - - - - - - - - - - - - - - - - ,
1.5
1.0
CZ 400 °C 0.5
•• •
10
LAMP
0.0 ~i:J=-=.=
-0.5 .
====:= ====== ==::t= ='
-
05
1- • - - ELECTRON BEAM
'-.... 00 . ~ - .; - . ---- ~---------- - 1. 0 L..L-'--'--L...L-'--'--L...L-'--'--L..j j-J........l.-'--.-L-J
I-
<l ---------- r-- - -- ----- o
o 3.--------------,
.,. ...
- 0 .5
••
1.0 '" ...
1.0

0.5 -
~
Q...
<l
00 • -=! -==.==!: :
05

o
o
.
x
Figure 2: Minority carrier lifetime, resistivity and
n::E 0 Rmax variations with respect to the initial values as a
>< - ( function of annealing time at 750'C in FZ silicon.
" 2 Lamp RTA and 2 h furnace annealing results are
E reported.
n:: "-'-'..Ll • w..U.u.u,..........L'-LL"-'-'" ......_L.LL'-'-'-'L4Ju..L.I............u.~
..., 0 20 40 0 20 40
li lli e (s) Lime (s)

Figure 1: Minority carrier lifetime (t), ~e~istivity (p) In CZ crystals (Fig.3), the lifetime degradation
and maximum of the X-ray reflectIVIty (Rmax) is lower for annealing times higher than 20 seconds;
variations with respect to the initial values as a the crystal order is higher after the anneal, with a
function of RTA time at 400'C in CZ silicon. The slight time dependence. The resistivity variations are
dashed lines, drawn in 6th and 6p/p graphics, point negligible.
out the spread of the exper!mental data. 'fh:e bars, At I050'C we observe a strong degradation of
drawn in 6RmaxIRmax graphic, are the expenmental lifetime and an increase of the resistivity in FZ as well
errors on the measure of this parameter (sensitive to as in CZ crystals (Fig.4, 5). Crystal order shows an
very weak crystal distortion) evaluated by double- increase, higher in CZ samples, more appreciable at
crystal X-ray diffractometry. longer annealing times.

648
 1.5 . - - - - - - - -- - - - -----, o.~ ,------------------------------,
1. 0
-- 0 .0
~---------------------
CZ 1050 °C

·
.t.
I-
0 .5
0
.
0 J:-------------C-Z- 7- 5-0-oC---
1=.-------- -----------..:
~
\_ - O.S -
~ ---------------------

<] - 0 .5 :. - - • : <1
- l. 0 - . • •• • • -
-1. 0 - I • • •• •
- 1. 5 L.....-'----'-.J.........o---L.-L.....L................--'-...Lj /.-'---'-'--'--'
2. 0 , - - - - - - - - - -- - - - , S r---------------------------~-, •
1.5 - I] -
1.0 - 3 • •
0 .5 ~ 2 - • •
Q.. • • •
0 .0 - !:~ _=_= ! =
- 0 .5
======= =:!== <]

o ~ ~ =======.== == =======~
- 1. 0 ,,--,---,--,-,--,--,-..;..-l--'--'-->--1-{ /--'----'--'----'--"'
I . _,--,--,-L . . . I • , • I I

o 3 ,--::----- - - - - - - - - - - , o JI--------~----------------,
o o
......
x 2-
«I

J1
"-.
~ 0 --------- --------------------
E
0:: - 1 I 7
<] 0 20 4.0 s 2 11 o 20 40 s 2h
Lim Lil C

Figure 3: Minority carrier lifetime, resistivity and Figure 5: Minority carrier lifetime, resistivity and
Rmax variations with respect to the initial values as a Rmax variations with respect to the initial values as a
function of annealing time at 750·C in CZ silicon. function of annealing time at 1050·C in CZ silicon.
Lamp RTA and 2 h furnace annealing results are Lamp RTA and 2 h furnace annealing results are
reported. reported.

0 .5 , - - - - - - - - - - - - - - ,

0 .0 FZ 1050 °C

.
I- --------- --------- ---
~ - 0 .5 - DISCUSSION
<I
-1. 0 • - •• • •
- The results up to now obtained can be so
summarized:
- 1. 5 L-...--'---'-.J.........o---L.-L.....L....~__'_~u/_'_-'-.l-L-
• electron beam heating produces always a
2. 0 .----------------~ degradation of minority carrier lifetime, nearly
1.5
independent of time and temperature;
•
Q.. 1. 0 • lamp heating induces a behaviour similar to the
........... 0 .5 • • •• i furnace induced one, with the suppression of all
Q..
<I
- 0 .5 -
O . O ~ ~ ======= __ === = ==== ~ ==
low temperature mechanisms inducing minority
carrier lifetime variations.
In 750·C RTA annealed CZ crystals the increase
- 1. 0 t.........-'-.&-l-'-............--L....L.-!--'--..Lf~
1_ - L -L......L....l
~flifetime observed in some furnace annealed samples
o 3 ..-------------___,
o
IS not present.
To analyze these data we have to take into
2 account that during the rapid thermal processes a
large number of carriers are generated by photon-
matter and electron-matter interactions in the lamp
and electron beam systems, respectively. This presence
--- -- -------------- could influence the defect equilibria giving different
results for rapid annealings with respect to the
furnace one. However, we have observed similar trend
in lamp and furnace annealed samples, while the
electron beam treated specimens behave differently.
Therefore, we can conclude that the mechanism
responsible for the reduction of lifetime in the electron
Figure 4: Minority carrier lifetime, resistivity and beam annealed samples is directly connected with the
Rmax variations with respect to the initial values as a electron irradiation, although the energy of our
function of annealing time at l050·C in FZ silicon. electron beam annealer is well below .the threshold
Lamp RTA and 2 h furnace annealing results are energy for electron induced damage.
reported.

649
By comparison between lamp and furnace results we
can deduce that the effects of anneaIings performed by
the two systems are similar for the lamp longer
annealing times. At 400·C the use of the lamp with
anneal times up to 40 seconds allows to avoid all the
defect movements induced at the same temperature by
the furnace. At high temperature the strong
degradation observed even for the shortest RTA
annealing times can be explained by dissolution orland
dissociation of complexes and subsequent diffusion of
quick metal contaminations, as pointed out by
resistivity increase due to partial compensation of
dopant. The trend towards a reduction of this effect for
longer anneal times at 750·C can be explained by the
presence of a slower mechanism, connected to the
oxygen presence. More investigations are now
performed on this point.

REFERENCES
(1) D.Mathiot, Quenced-in defect removal through
silicide formation by rapid thermal processing,
Appl.Phys.Lett. 58, 131 (1991)
(2) W.Eichhammer, Vu-Thuong-Quat and P.Siffert,
On the origin of rapid thermal process induced
recombination centers in silicon, J.Appl.Phys
66, 3857 (1989)
(3) A.Barhdadi, H.Amzil, J.C.Muller and P.Siffert,
Thermal annealing effects on grain boundary
recombination activity in silicon, Appl.Phys. A
49, 233 (1989)
(4) E.Susi and A.Poggi, Effect of rapid thermal
annealing on electrical and structural properties
of silicon, J. Electrochem. Soc., in press
(5) P.Negrini, L.Passari, A.Poggi, M.Servidori and
E.Susi, Lattice disorder and recombination
centers in heat-treated FZ silicon,
phys.stat.so1.(a) 92, 167 (1985)
(6) E.Susi, G.Lulli and L.Passari, Minority carrier
lifetime in furnace and e-beam annealed CZ
silicon, J.Electrochem.8oc. 134, 1239 (1987)
(7) E.8usi, A.Poggi, G.Lulli, C.Monteverdi and
M.Merli, Lifetime and crystal order in annealed
CZ silicon, phys.stat.so1.(a) 108, 503 (1988)
(8) E.Susi, A.Poggi, R.Fabbri, M.Merli, M.C.Carotta
and L.Passari, RTA-induced defects: a
comparison between lamp and electron beam
techniques, Materials Science and Engineering
B4, 231 (1989)

650
 8-12 APRIL 1991 LISBON, PORTUGAL
10111 EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE

EVALUATION OF SILICON NITRIDE SURFACE PASSIVATION OF CRYSTALLINE

SILICON BY MEANS OF TIME RESOLVED MICROWAVE CONDUCTIVITY

T. Bickl (University of Hamburg) * **

U. Creutzburg, D. Silber (University of Bremen)
W. Ebner, M. Eyckmans, G. Wandel (Telefunken Systemtechnik) ***

* Presently with ** FBl Elektrotechnik *** Telefunken Systemtechnik

University of WUrzburg Kufsteiner StraBe Space, New Technologies
Postfach 33 04 40 Industriestr. 23-33
2800 Bremen 33 2000 Wedel
Germany Germany

ABSTRACT
Silicon solar cells must have an effective surface passivation
in order to reach high efficiencies. Silicon dioxide (Si02) provides
extremely low interface state densities resulting in record efficien-
cies for terrestrial sil icon solar cells. Unfortunately the refrac-
tive index of Si02 is so low that Si02 cannot act as an optimal AR-
coating for encapsulated cells. For the use of commercial solar cells
one has to look to other alternatives.

INTRODUCTION In the second process silane (SiH4) and

NH3 are believed to form an SiNH-film at rela-
In this work silicon nitride (Si3N4) has tively low temperatures (300·C-500·C). Hydrogen
been investigated in order to achieve surface concentrations of up to 33% can be incorporated in
recombination velocities similar to Si02 and hence the films. With this process the production of
get the full benefit of the superior optical homogeneous layers is more critical as with the
adaption. Hereby both the LPCVD and PECVD deposi- previous method but it has much less problems with
tion processes have been considered. In order to "waste" gases, has a higher layer growth rate
measure the passivating properties of these films, (100-250 A/min) and range of refractive index
the effective minority carrier lifetime has been variation (n=l ,9-2,5). Figure 2 shows the influ-
determined by the Time Resolved Microwave Conduc- ence of the SiH4/NH3 ratio on the refractive index.
tivity Method (TRMC).
Samples have been prepared by varying the THE TRMC-MEASURING TECHNIQUE
following parameters of the nitride deposition
condi tions: temperature, plasma power, underlying This contactless method, described else-
oxides, and annealing procedures. Furthermore where [1], avoids any interference of contact pads
reference samples have been oxidized under wet and with the surface under investigation. Original
dry conditions. TRMC-method only allows an assessment of the com-
bined influence of bulk and surface properties.
Thus a way has been developed to separate the two
SILICON NITRIDE LAYERS signals and take the presence of the emitter layer
into account.
Silicon nitride has an optimal refractive
index to act as an AR-coating on encapsulated On an undoped Si-wafer the effective minori-
silicon solar cells and has at the same time the ty carrier lifetime is defined by a combination
potential to passivate a silicon surface. The of 'Bulk and the surface recombination.
characteristics of an Si3N4 layer depend on the
deposition processes and parameters. Here two l/teff. = l/TB + 1/(d/2S + d2/0rr2)
di fferent processes have been considered: the Low
Pressure Chemical Vapor Deposition (LPCVO) process Where d is the waferthickness and 0 the
and the Plasma Enhanced Chemical Vapor Deposition diffusioncoefficient for minority carriers. If one
(PECVO) process. In the first process makes two measurements, one with a known surface
dichlorasilane (SiH2CI2, DCS) and ammonia (NH3) recombination velocity (S=~) and one with the
are thermally activated to react to Si3N4 at tem- surface recombination in question. The last one
peratures between 700' C and 900' C. The advantages can be calculated from both Teff values.
of these methods are the homogeneity and stability
of the depOSited layers. Disadvantages for the If one wants to measure the surface recombi-
contrary are the small margin of refractive index nation on a doped emitter region one first has to
variation (n-2) , the relatively low growing rates make a measurement on an undoped wafer with a
« 100 A/min) and the high reactivity of the known surface recombination (s=~) to define
waste gases with the vacuum pumps and pump oils. tBulk' In a second and third measurement on a
In figure 1 the variation of the index of refrac- doped (n+ - p - n+) wafer S' and S" have to be
tion with the NH3/DCS ratio is shown. Here one can measured, where S' is the combined influence of
see that only small variations of n are possible. the emitter and surface recombination for S =""
and S" for emitter and the surface recombination
in question. S can then be calculated from the
following relationship [2]:
S = Dh / W(S'/S"-1)

651
Experiments Measurements on wafers with pn-junction
To evaluate the TRMC-method for measuring
of surface recombination velocities, a comparison
with the CV-technique has been done on wafers LPCVO-Nitride
without pn-junction. Hereby a fairly good agree-
ment could be established (table I).
720' (1.0 ± 0.5) * 10 3 cmls
For further measurements the Si3N4-films
have been deposited on identical npnstructures on 780' (2.8 ± 1.3) * 10 3 cmls
SSOlum thick CZ-wafers. In table II the S-val\les
for LPCVD-Nltride, deposited at different tempera- 840' (2.7 ± 1.1) * 10 3 cmls
tures, are given together with the measurements of
a wet Si02-layer. 720' + 100A TCA-Oxid (1.1 ± 0.7) * 10 3 cmls
Also the influence of an underlying TCA- 780' + 300A TCA-Oxld (5.5 ± 4.3) * 10 2 cm/s
oxid and high temperature sintering treatment are
shown. 780, 12h 1010' (5.8 ± 2.9) * 10 4 cmls
Table III gives the results for PECVO. Here Wet Oxid (1.0 ± 0.8) * 104 cmls
the influence of the reaction temperatures, anneal-
ing conditions and plasma power has been investi- Table II: S-Values for LPCVD-Nitrides
gated. The influence of etching off the native
oxide in a pn-etch step in a C2F6-Plasma is also
listed.
Measurements on wafers with pn-junction
DISCUSSION AND CONCLUSION
In conclusion we have shown that both PECVD 350' , 500 W (1.8 ± 0.6) * 104 cmis
and LPCVD S1-Nitrides can passivate doped Si-sur-
faces. The best surface recombination velocities 350' , 1000 W (1.8 ± 0.6) * 10 4 cmls
(3.10E3 cm/s) are achieved with LPCVO.
450' , 500 W (2.8 ± 1.2) * 10 3 cm/s
Both nitrides loose the passivating capabilities
after a extensive high temperature treatment, 450' , 1000 W (3.6 ± 1.2) * 10 3 cmis
which indicates that hydrogen is the passivating
species. 450' , 500 W, C2 F6 (6.0 ± 2.0) * 10 3 cmis
Both nitrides need the thin native oxide on 450', 500 W,
top of sil icon in order to achieve low surface 12h 1000'C (1.0 ± 0.5) * 105 cmis
recombination velocities.
Table III: S-Values for PECVO~Nitrides
A high power density in the plasma process
will damage the surface region and give higher
S-values.
r::
REFERENCES § .0- ............ .::.'
....... : ..................:........ :........... .
.....
2.%0
.. . .., '
111 U. Creuzburg, H. Paradies, D. Silber, t: . .
R. Ruckteschler and D. Gringel, 'Minority
carrier lifetime topography for polycryst- '"s..
't
a • .lG ~
..
. .
•• • • • • • • • • • • • • • • • • • • • • •
...
Of •••••••••••••••••• :•••••••• ;••••••••••••

alline silicon wafers obtained by advanced ··· ..

·
0::

time-resolved microwave-conductivity', · .
'0 2 •••
Proc. 9th PVSEC, P. g. Freiburg, BRO, 1989
~
'0
121 T. Bickl, 'Thesis at the University of r:: .a..~o'-- _ _ _- - '_ _. l . -_ _ _ _ ~_~ _ _- - '

Hamburg', 1990 0 ••

Gas ratio NH3/0CS

Fig. 1: Index of Refraction of LPCVD-Nitride
TRMC-Measurement on bare p-Si wafers
TCA Oxid s = 1,8 * 102 cmls ± 15%
Wet Oxid s = 1,3 * 10 3 cmls ± 15%

CV-Measurement
TCA Oxid 1200'C s = 2,5 * 10 2 cmls
Table I

2.a 3.0 ".D ~.o ".0 7.0 a.a '.0 .10.0 J..L.O .L2.a l:1.r

Gas ratio NH3/SiH4

Fig. 2: Index of Refraction of PECVD-Nitride

652
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOTHERMAL MEASUREMENT OF MINORITY CARRIER DIFFUSION IN DEVICES

Bettine Buchner*, Norberto Cella +, and David Cahen

Department of Structural Chemistry, The Weizmann Institute of Science

IL-76100 Rehovot, Israel
*present address: Institut fUr Physikalische Elektronik, Universitiit Stuttgart
D-7000 Stuttgart 80, FRG
+Inst.Politecnico do Rio de Janeiro, Novo Friburgo, RJ Brasil

ABSTRACT. By refining a photothermal method developped earlier for analyzing electronic carrier- and thermal
fluxes in photovoltaic cells, photothermal data on single crystal Si solar cells are obtained that show, already at rather
low modulation frequencies, the effects of electronic carrier diffusion in the base. Thus, these methods are able to
examine actual transport properties, in addition to their use for characterizing devices, (by separation of photogenerated
and injected carrier-mediated processes). Thermal diffusion and diffusion of minority carriers influence the experimental
results of measurements, obtained with spatial separation by varying the distance between the area of photo-excitation
and detector area. The data are analyzed by generalization of the model developped for electronic carrier and energy
flow in solar cells. This is done by solving the coupled thermal diffusion/carrier diffusion equation.

I. INTRODUCTION
The basic argumentation is as follows: Modulated mono-
The photoacoustic technique has been used in the past to chromatic light with energy h/l greater than the bandgap energy
investigate optical and thermal properties in semiconductors and E~ of the absorber material is absorbed in the solar cell and
semiconductor devices (1). Already early in its development electronic carriers are generated. The excess energy, h/l-E ,
it was applied tophotovoltaic devices (2,3). Photoacoustics (PA) will be transferred to the lattice as heat, by non-radiati~e
is one of a number of so called photothermal (PT) techniques, deexcitations. This occurs instantaneously on the time scale of
which have as their basis the detection of modulated heat that the experiment given by the modulation frequency. The carriers
is created in the sample following absorption of modulated light, with potential energy Eg, will diffuse towards the junction's
by non-radiative deexcitation. In the case of photoacoustics this space-charge region, freemg all their energy as heat by recom-
heat will produce a modulation of the surface temperature and bination before reaching the junction, or part of it as heat and
thus a modulation of pressure of the gas that is contained in part as electrical energy by crossing the junction. Furthermore,
a gas-tight chamber around the part of the sample that is excited carrier injection will occur at finite forward bias voltage under
and inside the chamber. The pressure modulation is sensed by illumination. These injected carriers take energy from the lattice
a microphone connected to the closed chamber. Particularly, to cross the junction and release Eg upon recombination. Collecting
in photovoltaic cells the heat that is generated is complementary all the terms of the energy changing processes we can express
to the electrical energy that is generated. Therefore, this technique the PA-signal in terms of the currents IL and Id of photogenerated
can be used to analyze power dissipation mechanisms in solar and injected carrier flows, respectively, the voltage across the
cells (4). terminals, V, Eg, the differences in energy between Fermi level
and valence and conduction band edges, and quantum and collection
In addition to photoacoustics other photothermal techniques, efficiencies. The explicit expressions are given in Ref. (4,6, 7).
such as photothermal beam deflection and photopyroelectric We will restrict ourselves here to unity quantum and current
detection, have been used to analyze photovoltaic cells (5-7). collection efficiencies, which simplifies the formula for the voltage
In all these techniques the signal that is detected is proportional dependence of the photothermal signal (PT-V) to:
to the amplitude of the temperature modulation on the surface
of the photovoltaic cell. Therefore, the signal will be related hv
PT-Voc ( - - V)SlIL + VSld (1)
to the amount of energy that is dissipated in the sample and q
that reaches the surface. The question, which we will adress
in this work, is, in how far, and under which circumstances, where S, is the fraction of the illuminated area of the solar cell
can the signal be influenced by electronic, in addition to thermal in the photoacoustic chamber, and S is the ratio of the area of
parameters, i.e. is there a possibility to obtain quantitative the photoacoustic chamber and the total area of the solar cell.
information on electronic parameters, specifically the minority S, expresses the fact that photogeneration occurs only in
carrier diffusion length and surface recombination velocity, from the illuminated area of the solar cell and that only the part of
PT measurements on complete solar cells? Our experimental power, dissipated inside the PA-chamber, is sensed by the
approach is based on our earlier finding that when measurements microphone. The argumentation for S is that injection occurs
are done under conditions that not all of the solar cell' s area in the 1Q!& solar cell area, but also only part of the respective
is illuminated or not all of the cell is enclosed in the microphone power dissipation will be sensed.
chamber, so called area factors need to be considered in order In this model, the area factors do not contain effects of either
to understand the photothermal signal (4). thermal or electronic diffusion.

2.2 Contribution of Thermal Diffusion

2. MODELING AND EXPERIMENTAL SETUP One of the consequences of the different areas of photo-
generation and injection is the possibility to systematically eliminate
2.1 Stationary Photothermal Signal the contribution of the photogeneration by illuminating less and
Earlier we have shown that the photoacoustic/photothermal less inside the PA-chamber. The relative contribution of the
signal of a photovoltaic cell can be analyzed quantitatively by injection to the PT-signal becomes larger and thus it becomes
using equations describing the power generation in it (4,6,7). possible to analyze the diode behaviour of a photovoltaic cell

653
under illumination. Such an analysis is not possible by electrical 3. EXPERIMENTAL RESULTS AND DISCUSSION
measurements.
In Figure 1 we show the experimental set-up to carry out For all the measurements we used a large area (8cm2) c-Si
such an experiment. The results of experiments of this type are solar cell of ASEC. The IV-and electrical power curve for ca.
not described properly by equ. (I). Data measured under conditions I-sun illumination is shown in Figure 2.
that the thermal diffusion length is of the order of the mean
distance between the excitation beam and the measuring chamber 3.1 PT versus voltage
can not be explained by a geometrically calculated area factor. For the first series of voltage dependent measurements with
The numbers determined for SI are not in accordance with the changing area factors we fixed a ring of perspex with an inner
experimental ones, as we will see below. In addition, the shape and an outer radius of 5mm and 6mm, respectively. Figure 3
of the curves changes and the PT-V data show a pronounced shows the arrangement. The PA chamber is closed by a cover
frequency dependence. This is true not only for the PT amplitude glass pressed on top of the ring. The miniature microphone (Knowles
but also for the shift in phase angle, which becomes now of BT#1754) is stuck with its nose in a small hole in the ring. A
interest. 6mW He-Ne laser, expanded to a diameter of ca. 8mm, modulated
In a first approach we describe the signal as a sum of an by a mechanical chopper, provided the excitation. In one run
inner and an outer heat source. The inner source, caused by the PT signal as well as the (ac) IV-curves are recorded simulta-
injection is equal to the third term in equ.(l) and thus does not neously. The measurements are carried out at different modulation
depend on position of illumination. Oscillating heat sources at frequencies and different distances: center ofillumination -center
the position of modulated illumination cause temperature gradients of detection, xo. A typical result is plotted in Figure 4 for Xo
laterally and in direction of illumination in the solar cell. This = O. Data collected at 20Hz and at 80Hz modulation frequency
gives rise to oscillating heating/cooling of the sample surface show a linear decrease of dissipated heat with voltage corresponding
by thermal diffusion in the area of the illumination and in a to the increase in electrical power output, cf. Fig.2. On this
ring-shaped area around it, given by one wavelength of an scale the influence of the exponentially increasing contribution
overdamped thermal wave. We take into account this contribution of the injected carriers near open circuit voltage is hardly resolved.
to the PT signal as a thermal wave, spreading in one dimension. This is due to the small area of detection compared to the whole
These waves, solutions of a heat diffusion equation, are not only photovoltaic area.
strongly attenuated but also delayed, resulting in a phase angle A different behavior of the data collected at 20Hz and at
shift between inner and outer heat sources. The real and imaginary 80Hz can be seen from Figure 5. These data are measured at
parts of the PT-V signal are given by a distance of 9.5mm, i.e. the illumination is completely outside
the PA-chamber. For 80Hz there is no contribution of photo-
PT_Voc(SI+S/)CO -S/C, +i[S/Co+(S'+S/}C,l (2) generation left, but the increase in injection is clearly to be seen.
The 20Hz curve has a finite constant contribution of photo-
where Co' C1 depend on thermal diffusio!1 perpendicular to the generation, even if the illumination is completely outside. The
sample surface, the inner (ER) source S1 = (2r/Qsc)IdV, with large thermal diffusion length for 20Hz gives rise to a PT signal
re, chamber radius , and ~sc' length of solar cell. The outer source, caused byphotogeneration. Curves, calculated according to equ. (2),
Ee, is are shown in Fig.6. The fit with Fig.5 is readily observed.

3.2 PT versus distance

Measurements under conditions like the one in the previous
with xe = Xo - rL - re + Itthln(2V2re/Itth) and chapter can be explained by solutions of the thermal diffusion
xf = Xo - rL - re - 3'l/'/4Itth, equation, only. To follow also the carrier diffusion we have
to shrink the distances in the arrangement further and to go to
Xo is the distance: center of illumination - center of detection higher modulation frequencies. Thermal and carrier diffusion
(PA chamber), rL is the radius of illumination and Itth is the lengths have to be of the same order of magnitude.
thermal diffusion length, given by modulation frequency, f, and The PA-chamber that is now fixed on the PV cell has a
thermal diffusivity, a, as Itth = Va/ ri. The In-term in xe originates diameterof2.4mm. AnAr+ laser beam, which is slightly focussed
from averaging over the detection area. For areas of illumination to O.9-lmm diameter, is impinging on the sample surface. For
and detection with radii smaller than one thermal diffusion length fixed (reverse bias) voltage and fixed frequency the distance
their axial symmetry should be taken into account by solving PA-chamber -laser beam is scanned laterally between 400J,lm,
a heat diffusion problem for cylinderical symmetry. the smallest possible arrangement for outside illumination and
For the demonstration of the influence of thermal diffusion ca. 3-4mm. While the "zero distance" is not well defined
the radius of the area of illumination was 4mm and that of the experimentally, the relative change in distance is controlled by
detection area 5mm. a micrometer. Figure 7 shows results of such an experiment.
The PT-xo curves for different frequencies exhibit characteristic
2.3 Contribution of Carrier Diffusion behaviour. For the smallest frequency of 25Hz the exponential
In semiconductor wafers the carrier diffusion is taken into decrease with distance can be explained by thermal diffusion
account by a coupling term in the thermal diffusion equation. alone, but already for 200Hz the curves show a kink, that is
Weuseaslightly modified formulaofFoumieretai (8), including shifted to smaller distances for higher frequencies. For comparison
the application of zero or reverse bias voltage to eliminate the curves simulated according to equ.(3) are plotted in Figure 8.
influence of injection. Along the surface, the PT signal is then Here we did not average over the detector area, and consequently
given by the absolute values of the distance do not correspond so well.
But the kink caused by the fact that the first 2 terms in equ.(3)
1 _~ _(+~e-.to(H) sID hv-qV-E nearly cancel can clearly be seen.
PT-x oc-e .to [ + __ + 8)
o ~ tD(sID+()((2-(~ sID+( Eg We examined the frequency range 215Hz to 300Hz in more
detail. The experimental and the calculated results are shown
(3) in Figures 9 and 10, respectively. We also show results for the
where.\ = 1IlteJ = VIIDT ; ~ = I/ltth; shift in phase angle. The sinus of the phase angle changes from
D is the minority carrier diffusion coefficient, T their lifetime -I to + I in the very small range of distance where also the
and s the surface recombination velocity. amplitude shows the kink. We did the simulation for different

654
 values of thermal and electronic diffusion, carrier lifetime and ACKNOWLEDGEMENT
surface recombination velocities. The best results we received
The work was funded in part by the Israel Ministry of Energy
were a = O.77cm 2/s, D = 14cm2/s, 1 = SOILS, s = 1400m/s. and Infrastructure and the MINERVA Foundation. B.B. likes
This gives a minority carrier diffusion length of 266J.'m.
to thank A. and R.Lempp for their hospitality and possibility
We want to point out that the simulation should be extended to use their computer. N.C. thanks CAPES for a fellowship.
from a one-dimensional model to a higher dimensional one, taking
into account the two axial symmetric systems of illumination REFERENCES
and detection. (I) A.Mandelis, ed. Photoacoustic and Thermal Wave Phenomena in
Semiconductors, North-Holland, Amsterdam, 1987.
CONCLUSION (2) D.Cahen, Appl.Phys.Lett., 33, 810 (1978).
In our effort tocharacterizephotovoitaiccells we have shown (3)S.N.M.Mello, C.C.Ghizoni, L.C.M.Miranda, H. Vargas, J.Appl.Phys. ,
61, 5176 (1987).
how a relatively simple photoacoustic set up used before to follow (4) H.F1aisher, M.Wolf, D.Cahen, J.Appl.Phys., 66, 1832 (1989).
heat dissipation can be modified to provide information on thermal (5) D .Cahen, P.E.Norda!, S.O.Kanslad, Appl.Phys.Lett. ,49, 1351 (1986).
and electronic diffusion. (6) B.Biichner, D.Cahen, M.Wolf, Proceedings of 9th ECPV Solar Energy
As we can see from comparison ofexperiment and calculations, Conf., W. Palz e\ al. eds., Kluwer, Academic Publishers, Dordrecht
one-dimensional models can explain qualitatively the amplitude 1989, p. 145.
and phase angle shift of photothermal data collected under (7) D.Cahen, B.Biichner, F.Decker, M.Wolf, IEEE Trans.Elec.Dev., ~
498 (1990).
conditions that thermal and electronic diffusion lengths are (8) D.Fournier, A.C.Boccara, A.Skumanich, N.M.Amer, J.App!.Phys. ,
comparable to the typical dimensions of the experiment. 59, 1013 (1985)
Quantitative results for diffusion governing parameters can (9) F.A.McDonald, J.Appl.Phys., 52, 381 (1981)
beobtained, but this procedure should beimpro~ed by dev~lo~ing
a more sophisticated model for photothermal signals taking mto
account three-dimensional spreading of thermal waves. 50 20

< 40
e ~---7~a 15

..
'&
"",0
c
..
t
30
/ x.~: 10 ::
~

.
::I
~ 20 3
:;:
/
'0
-="- D 5
10

0.0 0 .2 0.4 0 .6
.oltag~ 1

Fig.I: Geometry ofillumination, radius rL and remote detection, Fig.2:ASEC crystalline Silicon cell ~&m2). IV- andpower curves
radius rc' at ca. I sun. Isc =37.5mAlcm . Voc=574mV; Max. Power
Voltage 470m V, Max. Power 16. 6m W;' FF=O. 77; '1 = 0.17

Fig.3: Schematic arrangementfor PT-Vand PT-~ measurements.

The sollJr cell is mounted on translation stage to change
the relative distance between excitation laser beam and
PA chamber.

Fig.S: Experimental results of normalized PT amplitude versus

voltage/or 20Hz and 80Hz. Illumination outside the PA
chamber xo= 9.5mm.

~ 140 .• 20 H.
30
. 80 Hz I
.
8

."
- - 20 Hz
0

80 Hz ~ 25 20 Hz
Z " ---- 80 Hz
::::0 ..
. .
~120 ::I .~ 6
.: )C:Ie • 0 if 20 Xo = 9.5 mm ! Q.
e calculated, Xo 9.5
f':.o
~
..
x x

t"
)I

~100 "'">O,p,p~ :'

.".§
15 4
.; = experiment
Q.
e 80
Xo 0 Q. , 0 0° •
"
.~
~ 0
.J <i
"
'-
~ t 5
,.",x x Xx x
Ill: 'It X o. 0. E 2
• • • IIb")( .,/' 0
c
60 ,po. e. 0
III xx.x
0 --------------- - , '
0 .1 0.2 0.3 0.4 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4
vo113ge ' 0113g. 1
voJlag~ ' V

Fig.4: Experimental results of normalized PT amplitude versus

voltage for 20Hz and 80Hz. Illumination inside the PA Fig.6: Calculated curves for normalized PT amplitude versus
chamber, i.e. xo=O. voltage for 20Hz and 80Hz. Illumination like in Fig. 5.

655
 3

<II ~ -1.0
-.::I :J
::I
:::c. 2 .........
.u
Co
13cQ E
ID -2.0
~ ~
1:1. a..
c.o
..2
1
E
.
25Hz III 300 t.
+ 215 "450 g -3.0 calculated
0 tn
0 1 2 3 .....o
distance I mm
-4.0+-~-+~--r-;-~~~

Fig.7: Experimental results of normalized PT amplitude versus 1.0 1.5 2.0 2.5 3.0
distance for various frequencies. distance / mm
Fig.S: Calculated curves for normalized PT amplitude versus
distance for various frequencies.
100
+ a
..... .1]
...
c: + I:! + 270Hz
<II
::I
q'
<II
o •
+ I] ° 300Hz ."
J:.

+" 0
0
....
J..
" 240Hz
•<II 0 " c 215 Hz
{!. O. B

+ "I:!
... 50
-.::I
::I 0
+ ,,0
c. 0+"1]
0

13cQ

·~;~1iiiiij
0 .6

~
1:1.

1,~ * 0.•

b
~ 0,6
c.o 0 .2
C " I:!
~
+ 270Hz
<II
(I)
COl 0,0
" ° 300Hz
..c: " 240Hz
o.o+--.....--------=""~__j
0
c. "0 !!
I:! 215 Hz ~ 1.0
~
•
1:1. -{l,6 ••
•co
.c

·1 ,2 +-.,.........,........-....""T""-...-...-.-r--~...-j ·
oS

0,4 0,9 1,4 1,9

distance I mm

Fig.9a: Experimental results ofnormalized PT amplitude versus

distance for frequencies between 215Hz and 300Hz.
Fig.9b: Results for sin of phase angle

Fig.lOa: Calculated curvesfor normalized PT amplitude versus

distance for frequencies between 200HZ and 300Hz
1.85 2 . 3'5 2.85 3.3'5
Fig.lOb: Results for sin of phase angle. dl.t.nce I _

656
HJI'H EUROPEAN PHOmVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON. PORTUGAL

NON-CONVENTIONAL EMITTERS FOR POLYCRYSTALLINE SILICON SOLAR CELLS.

J.Coppye*, E.Demesmaeker*, H.E.EIgamel*, J.szlUfcik*, M.Ghannam*,

J.Nijs*, R.Mertens*,
Le Quang Nam+, M.Rodot+, J.M.Laroche+ ,
D.Sarti**, P.Loubly**.
* IMEC, Kapeldreef 75, B-3001-Leuven
+ CNRS/LPSB, 1 place A. Briand, F-92195-Meudon
** Photowatt, 6 rue de la Girafe, F-14043-Caen

ABSTRACT. The fabrication of emitters may improve the

quality of the base of a solar cell by a gettering pro-
cess; this has been studied, using LBIC measurements,
for three different phosphorous sources. The ability of
these emitters to be passivated depends on their surface
doping concentration; it has been possible to improve it
by using etched-back emitters and also selective emit-
ters where only the regions under the grid contact have
a high doping level.

1. INTRODUCTION
CHRS- MEUOOH OPTH' PROfl.E (P)
________________________________
The efficiency of conventional solar IOZI ~~~~lF ~

cells depends both on the quality of the

base - expressed by the electron diffusion
length 4 - and on that of the emitter -
which depends on its thickness, doping and
surface recombination velocity -. These two
issues are not independent: it is well
known that some of the processes used to
fabricate the emitter produce also an
increase of 4, due to some gettering z P
effect (see e.g. [1). Apart from this )1

physical effect, a second relevant issue is

the ability to passivate the emitter. This
depends on its surface doping level Ns [2].
Whereas the usual emitter formation
processes lead to Ns 2: 10 20 cm- 3 , a decrease .2 . " . G . 0 L' 1.1; 1. 0

of Ns by an order of magnitude would result MICRONS

in better conditions for oxide passivation.

The goal of this paper is to give some Fig. 1. SIMS profiles of P after various
new results on the gettering action processes for emitter formation:
brought about by three different types of 1. POCl3 (850'C, 12 min)
emitters and to present various attempts to 2. Screenprinted paste (940'C, 8 min)
decrease the emitter surface doping level, 3. Si02: P (940'C, 8 min)
including the use of selective emitters in
which Ns is reduced under the oxide but
remains very high under the contact grid.
2. EMITTER FORMATION PROCESSES; GETTERING CNftS · MCUOON
DtPTH
______________ ____PROFII.E
____ IP_J_ _ _ _ _ _ _ _ _ _ _ _ _- ,
EFFECT OF P DIFFUSION. IMSlF
10Z0.~

The current industrial process of I) SK>z,P 1I0SO-C .1l .... 1

Photowatt uses a POCl 3 source. An important 13 11 5IO,... P (,40 ' C . 12 f'IWI,
gettering effect in the base is obtained )) SlOz.P ,910 ' ( . " .... )
[l][3J. We have verified this point by LBIC ' 1510>-. IISO· C . II ....1
measurements of 4 made before and after
the standard POCl 3 diffusion process (12
min. at 850'C): cf. Fig. 3. A mean conver-
sion efficiency of 13% is reached in
industrial production (10xlocm2 cells) .
More sophisticated gettering treatments may z•
raise 4 to 180~m [4 J, which is likely to "
help raise the efficiency to 14%.
IMEC has a large experience with
screenprinting pastes as diffusion sources.
Using the "integral screenprinting techno- MtCRON$
logy" a best efficiency of 12.4% (for a
2x2cm~ cell) could be obtained with high
quality Italsolar wafers (4 120 to Fig. 2. Diffusion profiles of P from a
160~m). phosphorus doped silica source.

657
 In both technologies, we can expect 3. PASSIVATED EMITTERS: a) CONVENTIONAL.
only a slight (if any) passi vating effect
of a thermal oxide on the emitter surface, Removing the superficial part of the
because the latter is too heavily doped emitter is another evident way to reduce
(Fig. 1). We also studied emitters formed its doping level. This etchback step was
by diffusing P from a P-doped Si02 film. used long ago for a combined optimization
This film was deposited by CVD from SiH4 + of the series resistance and the open
O2 + PH3 at around 400'C. The Si02: Player circuit voltage [5]. We have studied this
was 2500A thick and covered by a 2500A process and its influence on the possi-
thick undoped protecting Si02 layer. The P bility of emitter passivation, using a
profiles for various annealing temperatures standard laboratory technique of fabri-
are given in Fig. 2: the surface P doping cation of high efficiency cells.
is limited to 4 .10 19cm- 3, which will be used 5x5cm2 wafers from Wacker (L" ~ 60~m)
( § 4.1) to make selective emitters. or Italsolar (L" ~ 140~m) were used. The
emitter on all wafers was formed by P
130
diffusion from a solid source (P20 5) at
120 DIFFUSION LENGTH OF AS GROWN WAFER = 68 ~m 900'C for 15 minutes. An emitter sheet
'iii' resistance of 20 O/D was obtained.
~
'10 ~ Si02+P (10 ~m REMOVED)
L:J Afterwards the emitter was etched back
2
~ 100 ~m
~
POC L3 (I REMOVED)
in a HN0 3 -HF solution, to two different
!
;:;
U
90 ~ POCl3 (I0~m REMOVED) values: 45 O/D al1d 75 O/D This was
:I: 80
followed by the growth of a thin (80 A)
t;
2 ~ 70 layer of Si02 in a dry 02 atmosphere; for
~
2 ~ 60 comparison other cells were kept free of
o
~ ~ 50
oxide. Ti-Pd-Ag contacts were evaporated
B ~
"- (9% coverage) and finally an AR coating of
40
Ta205 ( of refractive index 2.05), 690 A
~ 30

20
;:; thick, was evaforated. Results of measure-
ments on 2x2cm cells are shown by Table 2.
'0 ~
/.
The high conversion efficiency, up to
14%, obtained by this process, even when
850 9'0 940 '050 DIFFUSION TEMPERATURE 'C the initial wafer had relatively low L"
(.=12 min) (Wacker), suggests that the P20 5 process
allows an efficient gettering effect of the
impurities in the base, like the POCl 3
Fig. 3. Change of diffusion length in Polix process; however this has not been directly
wafers after P diffusion from 2 different
sources: 8J 00
POCl 3 (standard conditions),
verified by LBIC measurements.
We find that the oxide surface
~ Si0 2: P (as a function of temperature).
passivation significantly improves, by 6 to
The measurements are made after acid
15% , the short circui t current; this
etching of the most superficial layer (l~m
improvement tends to increase with sheet
in general).
resistance, which illustrates the role of
surface doping. The comparison of Voe for
the 45 and 75 O/D cases, and for passivated
The gettering action of P introduced
vs. non passivated emitters, does not give
by these three processes was tested by LBIC
clearcut results; the same is true for the
measurements of the base diffusion length fill factor.
L". It was found that the gettering effect
was small in the Si02: P process (fig. 3)
and also in the screenprinting process
(Table 1).

MATERIAL DIFFUSION DIFFUSION L" Ln

SOURCE CONDITIONS BEFORE GETTERING AFTER GETTERING
SILSO P-PASTE 925'C, 45min 64~m 56~m

SILSO P-PASTE 900'C, 45min

SILSO P-PASTE 897'C, 45min 58~m 63~m

TABLE 1. Change of diffusion length in Wacker-Silso

wafers after diffusing P from screenprinting paste.

658
 Wafer Emitter J se Vee FF Best
origine sheet (rnA/cmZ) (mV) (%) efficiency ARC
resistance vs. total
area (%)

NON OXIDE PASSIVATED EMITTER

Wacker 45 20.5 561 77.7 8.97 no

Wacker 75 19.5 552 77.1 8.3 no
Italsolar 45 21.8 547 75 8.96 no
Italsolar 75 21.9 560 74.9 9.17 no
30.5 571 74.1 12.9 yes

OXIDE PASSIVATED EMITTER

Wacker 45 22.1 565 78.5 9.6 no

30.2 574 76 13.0 yes
Wacker 75 22.8 570 77.4 10.1 no
30.7 577 78.8 14.1 yes
Italsolar 45 23.8 561 74.9 9.97 no
30.7 575 76.1 13 .5 yes
Italsolar 75 23.3 555 77.7 9.93 no
30 579 75.9 13.1 yes

Table 2. Performances of conventional cells with etched back emmitters.

4. PASSIVATED EMITTERS: b) SELECTIVE EMIT· With this technique, we could already

TERS. reach a conversion efficiency of 13.1%.
Selective emitters have been made in 3 This was made possible by the use of very
different ways. good initial wafers: the Polix wafers were
first submitted to the gettering action of
4.1. The SiOl:P - screenprinting process. POCl 3 (850'C, 12 min), then thinned to
In this process (Fig.4), (i) a n layer 170J,Lm, then back-coated with SiO z: B to
was obtained from a SiO l : P source; (ii) a obtain a BSF effect [6]. More details are
n H layer, with the pattern of the front shown by Fig.5. The BSF action was not
contact grid, was obtained from a screen- fully obtained because the sheet resistance
printed paste; both layers were annealed of the p+ was quite high (170 OlD ) and no
together at 940'C during 8 min. The P backside passivation was used (direct
profile is shown in Fig. 1. metal contact).
The sheet resistance for the n layer
was varied from 45 to 250 0/ D by changing 4.3. The screenprinting-etchback process.
the PH3 content during CVD of the Si02: P; The two above processes to build
for the n+ layer the sheet resistance selective emitters include a masking step
equals 21 0/0. The formation of a thermal with a photoresist. The grid contact has to
oxide did improve the short wavelength be made afterwards on the same pattern;
spectral response. But, as mentioned above, these two steps require a perfect alignment
the P gettering action in the base was of their respective masks. In order to
weak, so the best efficiency obtained was simplify the procedure, a third process is
ll.7%. proposed [7]. In this process the grid
contact is made before etching back the
4.2. The solid source-etchback process. emitter, and protected by means of an acid
This process derives from that of §3: resistant polymeric paste. The emitter
the etchback step is realized on only a between the fingers is then etched back in
part of the surface, by masking with a an HF-HN03 solution. The advantage of this
photoresist the pattern corresponding to process lies in the easiness and accuracy
the front contact grid which will be of the alignment of the protecting paste to
deposited afterwards. Several batches of the metal grid. using this new process
cells were made, using Polix wafers, preliminary results show an efficiency of
evaporated contacts and plasma deposited 12.9% on 2x2cml cells using Italsolar
silicon nitride AR coatings. material: this represents an average of
0.6% absolute improvement in efficiency,
when compared to the conventional emitter.

659
 FRONT CONTACT

ARC SO ..1

.
Sltp'

..... -
THERMAl S;02 (80 A' I
n

3102 4' IcY/1.."."".)

SIN!' 2 ".1yA-31 POLY-SILICON (P-TYPEI

S1C2· • (C'tn.".IIIoaJ

p+

REAR CONTACT (1.1)

$lcpJ
Sheet resistances CO/a) Performances
Cell
n" n p' Ioe Voe FF "IJ
(mA/cm 2 ) (IIV) (t) (t)

BSF-COMP 21 70 170
~~J~~ElE~L.E~
~ JKI7 - '
.5101 ... Uhou'.ett'
29.0 587 76.5 1J .1

, .. "...·:U

Fig. 5. Scheme and performances of a solar

cell with a composite emitter, made by the
solid source-etchback process .
.step S

iiiin:'Ili"mmr ""mIi-iiij'iii'' ' i:jiml:mmm

bmmmJii'iOC1 J 3101 wlthou' • .", Selective emitters may be efficiently
oxide-passivated ( § 4) and this is also
.sup 6
true for conventional emitters when using
an. etchback step (§ 3). The optimization of
thlS step, for polycrystalline silicon, is
an important task which is not yet fully
I::E=====.=.=='='=~=
=~::: '_m"N'- filled. It is expected that some value of
the surface doping Ns allows for the best
compromise between the value of V~, which
should decrease with Ns , and that of J sc
after oxide passivation.
Finally, if a selective emitter is
used, it is important from a practical
point of view to minimize alignment pro-
blems. In this respect the screenprinting-
etc~back process (§ 4.3) seems to be the
easlest one to be used in an industrial
environment.

ACKNOWLEDGEMENTS.
Fig. 4. Procedure used to make composite
emitters by the SiO z - screenprinting pro-
The authors wish to thank P. de
cess. Schepper, W. Laureys and P. Laermans for
their continuous effort in solar cell
processing, as well as C.Grattepain for the
5.DISCUSSION AND CONCLUSIONS. dopant profile measurement by ion analysis.
This study was funded by the contract
The demonstrated dependence of the JOUR-0036 (Multichess), granted by the
gettering action of P on the method of CEC-DG XII (Joule program).
introduction of this element into the sili-
con has no precise fundamental interpre- REFERENCES.
tation up to now. It is important from a
technological point of view: it means that
[1] S.Martinuzzi, D.Sarti, P.Torchio,
the processes using sources such as SiO z ! P
T.Buonomo, J.Gervais, G.Mathian,
or screenprinting pastes (§ 4.1, 4.3) may
1. Perichaud, 9th EC Photovol t. Solar
lead to high efficiencies only with high
quality starting material where L" Z l20llm Energy Conf. (Kluwer, 1989) 458.
(e.g. the best Italsolar wafers, or Polix [2] M.A. Green, High efficiency silicon
wafers pregettered by a POC1 3 treatment). solar cells (Trans Tech. Publ., 1987),
As-grown Polix wafers, with L" Z 60 to chapter 7.
801lm, may however be directly used when the [3] S.Martinuzzi, 1.Perichaud, J.Gervais
process to make the emitter is based on a D.Sarti, 10th EC Photovoltaic sola~
POC1 3 source or PzOs solid source. Energy Conf. (to be published) .
[4] D.Sarti, personal communication.
[5] H.Valdman, C.R.Acad. sci., 256 (1961)
246, M.Rodot, Ch.Guillaud, U.N. Conf.
on New Energy Sources (Rome, 1961).
[6] Le Quang Nam, M.Rodot, D.Sarti,
P.Loubly, J.Coppye, E.Demesmaeker
J.Nijs, R.Mertens, lOth EC Photovo1t:
Solar Energy Conf. (to be published).
[7] J.S~~ufcik, H.A. Elgamel, M.Ghannam,
J.Nl]S, to be published.

660
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

IMPROVEMENTS OF POLYCRYSTALLINE SILICON WAFERS AND CONVENTIONAL SOLAR CELLS BY A

BORON DIFFUSION STEP.

Le Quang Nam+, M.Rodot+,

D.Sarti**, P.Loubly**,
J. Coppye* , E.Demesmaeker*, J.Nijs*.

+ CNRS/LPSB, 1 place A. Briand, F-92195-Meudon

** Photowatt, 6 rue de la Girafe, F-14043-Caen
* IMEC, Kapeldreef 75, B-3001-Leuven

ABSTRACT. Industrial polycrystalline Si wafers have now

a thickness of 170-200~m and an electron diffusion
length Ln ~ 125 to 180~m after cell processing. It is
thus profitable, in terms of conversion efficiency, to
build a back p' layer inducing a back surface field.
Such a layer has been built from a B-doped CVD-sio 2 •
Solar cells including this technique have been made and
studied.

CNRS MEUDON DEPTH PROFilE

1. INTRODUCTION 10 21
IMS]F

If the base of a solar cell is thin 10 20

enough and has a high enough quality, a
pip· back contact providing a back surface
...m-E
tn t9
field (BSF) may be useful to reduce the
back surface recombination rate sn and
hence to increase the conversion efficiency --..u
(/)
10 18

[1]. In a case studied by GAROZ ZO [2], :::;

for a base thickness d = 170~m and an 0
l-
1011
electron diffusion length in the base In = e( SiOl+0{CVD): 900·C. 1 I,

100~m, the short circuit current J se is

increased by 1.9% and the open circuit
10 16 ..,..... .
oOr) : 900 C, t h
vol tage Vae by 0.6% when comparing a BSF
cell with seff ~ 0 to a cell with a back
ohmic contact (sn ~ 106 cm/s).
Presently, the industrial cells
produced by Photowatt use polycrystalline MICRONS

Si wafers of thickness 170 to 200~m. The

diffusion length is currently 60-80~m in Fig. 1. Diffusion profiles of boron from various
the as-cut wafers and is raised to sources
100-125~m after the emitter formation, (The ~lantation of BFs ~s made by J.e.Huller,
because the diffusion of P from a POCl 3 ~RS-Strasbourg)
source induces an impurity gettering
effect; the same is probably true with a
P20 S solid-source [3].
It is thus advisable to add a p. layer
at the back of the base and this becomes
more and more true increasing Lnjd. This is
m- CNRS.MEUDDN
;;'10 21 IMS3F DEPTH PROFilE (8)
the purpose of the present paper. E ?--------------------------------~
......
u
1) SIOz+B (1050·C. 12 min)
2. RESULTS OF BORON DIFFUSION. ~ 1020 : Z) SIOZ+B (940·C. 12 min)
oI- '
.
We have considered 3 possible sources 3) SIOZ.O (910·C. 12 min)

of boron incorporation: the use of the e( 1019 ; 4) SIOZ.O (850·C. 12 min)

gaseous source BBr3' a B-doped Si02 film
and implanted B atoms. The diffusion 1018
profiles obtained after an annealing step
(900'C, 1 hr) are shown in Fig. 1. Since we
do not wish to exceed this temperature too
much, we have to disregard the gaseous
source, because under these conditions it
gives a too low surface concentration Ns •
In the two other cases, the surface
concentration is satisfactory; the
diffusion is deeper, which is due to the
presence of defects (radiation-induced MICRONS
defects for BF3 implantation, interstitials
generated by oxygen diffusion for Si0 2 : B). Fig. 2. Boron profiles in polycrystalline Si from a
We have chosen to continue with the latter B~oped silica source
source, in order to avoid problems related
with radiation damage.

661
 Wafer Diffusion Diffusion Ln ii) Industrial Technique (by Photowatt)
source conditions (t.lm) - diffusion from POCl 3 source (SSO'C, 12
min),
- oxydation at 800'C in dry O2, to give a
Polix no 68 lsoA thick oxide,
- screenprinting of front and back con-
Polix Si02:B 8S0'C, l2min 69 tacts,
- antireflective coating made of Ti0 2.
" " 9l0'C, l2min 74
CNRS-MEUOON
DEPTH PROFILE
, ,,zz IMS1F
" " 940'C, l2min 79

" lOSO'C, l2min SHUT AESISTANCU Of P"" LAYlR

" 70
I) fl, ... 170ll/e

II R, ... IOniC

TABLE. 1. variations of electron diffusion

length Ln after B diffusion from a Si0 2: B
source.

It can be expected that the above-

mentioned defects ease the diffusion of '" ,8
metal impurities to the surface. In order
to see whether some gettering effect can be
caused by the B diffusion, we have made a
la's ~------r-----~r-~---.------~------~
series of LBIC measurements of Ln on as- ,2 .4 .8 .6
grown wafers and after boron diffusion. MICRONS

Typical results are shown in Table 1. They Fig.S. Boron profiles of two p+ back layers
prove that the wafer quality is improved by annealed in two different conditions:
this boron diffusion, the best annealing 1. BOO'C, 15 /Din
temperature being 940·C. The corresponding 2. 940'C, 20/Din
B profiles are shown in Fig. 2.
IMS3F DEPTH PROFILE
3. SOLAR CELL FABRICATION. I~ZI S- ________________________ ~ __ ~ _ _ _ _ _ _ _ _'

We worked on Polix wafers which had

been texturized in a basic etch and
submitted to a diffusion of P from a POCl 3
source (8S0'C, 12 min.) followed by an acid
etch (HF + HN03 ) removing l~m on each side. 1~19
These wafers were then thinned to a
thickness of 170~m. In this state their
electron diffusion length is 110-l2S~m, as
measured by LBIC, and the mean surface
rugosity on the front side is 31-'m (see
insert in Fig. 4).
CVD silicon dioxide is then deposited
on the back side, by CVD at 400·C. The
precursor gas is a mixture of SiH 4 , 02 and
B2H6 • A boron doped Si02 layer of 2sooA is
deposited, followed by an undoped layer of
2sooA. Then an annealing takes place. The
annealing conditions allow to control the .2 .4 .6 .8 I 1.2 1. " !.1Ii L8
sheet resistance of the p. layer (fig. 3). MICRONS
Solar cells are completed by one of
two alternative techniques: Fig. 4. ElDitter characteristics in the IHEC process:
1. P profile after diffusion (900 C, 15 /Din)
i) Laboratory technique (by IMEC): fr061 P2D5 through a perfectly polished surface.
- diffusion of P from P20 S (900'C, IS min.) 2. P profile after etching the highly doped
giving an emitter with a sheet resistance superficial layer.
of 21 nla, Insert shows the mean front rugosity of a
- acid etch to increase the emitter sheet texturized surface.
resistance up to 68 nlD : about O.IS~m of
the superficial, highly doped, layer is 4. RESULTS.
removed (Fig. 4),
- dry oxidation (800'C, 3S min.), producing The best conversion efficiency
a 80A thick Si02 layer, obtained with the laboratory technique is
photoresist deposition to define the 13.8% (2x2cm 2 cells), i.e. one of the best
front contact grid pattern, followed by HF values ever obtained on a Polix wafer.
etching of the oxide, still more striking is the value of the
- evaporation of front contact (Ti-Pd-Ag), open circuit voltage: Voc = 603mV. This
then of the back contact (AI), high value is a strong indication of the
- plasma deposition of a silicon nitride presence of an efficient BSF effect. A
antireflection coating (thickness 800A, confirmation of the BSF effect comes from
refractive index 2. ). the comparison of solar cells made with
different dopings of the p. layer: Ise, Voe
and are all improved if the p. layer is
more heavily doped, and the improvement of
Ise comes principally from the infrared
response (Fig. S).

662
 Wafers back p+ Ise Voe FF
Cell n° layer (rnA/cm 2) (mV) (%) (%)
3/7 28.3 583 73.1 12 . 1
3/11 28.0 587 72.5 11.9
4/7 no 27.8 583 73.4 11.9
4/11 27.5 588 71.2 11. 6

1/7 28.5 587 74.5 12.4

1/11 28.8 587 73.4 12.4
2/6 yes 27.4 578 72.6 1l.5
2/11 27.2 580 73.6 11.6

TABLE 2. Performances of industrial cells made on Polix wafers with

or without a p+ back layer.

For cells obtained with the industrial IOOIr---r-

, ---"..----,------rI- - - - .'- - - - .I- - - ,
technique, Table 2 shows results obtained
for 2.5x2.5cm2 cells cut in the center of 4
~

!~~
consecutive Polix Wafers. A p+ layer had
been formed on all wafers, but then removed
goo
" ~
on wafers 3 and 4 while it was kept on

I.. j..
wafers 1 and 2. Cell performances and
spectral response do not allow us to
conclude on the presence of a significant V
~ ~~
BSF effect.

i 40 ft!
J

\~
DISCUSSION AND CONCLUSIONS. I- II Rsh PERFORMANCES 1\ \
~ CURVE CELL n+ p + he Voc FF 'l '\ \
Our estimation of the role of the BSF 20 >lAIc .. IOV (ll (\l
effect is only qualitative at this stage.

","
In order to determine quantitatively In and 1 BSF-NS7C 68 170 28.8 587 76 .• 12.9
sn from the infrared spectral response, we 2 BSF-N$lC 68 80 lO.5 60) 74.7 !J.S
need accurate values of the front side
reflectivity, which are not yet available. 0 L I I

The theoretical value of sn [ 4], for a p. .4 .S .6 .1 .8 .9 I. 1.1

doping of 4 .10'9cm·3, is of the order of WAVELENGTH (MICRONSI
10cm/s. Even if the real sn is 10 times
higher due to interface defects, the
presence of a large BSF effect is expected Fig.S. Performances and spectral response of BSF
i f Ln ~ d. solar cells (thickness 170 um) with etched back
There is a good chance that building emitters.
the p. layer from a Si0 2 source can be Cells 1 and 2 differ only by the B profile of the
acceptable for industrial production, in p+ layer, represeJlted respectively by curves 1 and
the frame of the cost goal of 1 ECU/Wp in 2 of Fig. 3
1993. The condition In ~ d is not far being
satisfied in industrial conditions. The
availability of good production yields with
10x10x.017cm wafers is very likely soon; on REFERENCES.
the other hand, wafers with In :::: 160 to
180J,£m have already been produced both by [1] M.P.Godlewski, C.R.Baraona,
Ita1so1ar and by Photowatt (the latter by H.W.Brandhorst Jr., Proc. 10th.
submitting their Polix wafers to a special Photovolt. Spec. Cont. (IEEE, 1973)
phosphorous diffusion step [5]. 40.
Consequently, the development of [2] M.Garozzo, L.Fornarini, R.Peruzzi,
equipment able to deposit the p. layer at a J. Electrochem. Soc.134, (1987) 1807.
cadence of 1 cell per second, with a [3] J.Nijs, E.Demesmaeker, J.Coppye,
relative cost increase smaller than 3% of R.Mertens, Le Quang Nam, M.Rodot,
the total cell cost would allow to D.Sarti, P.Loubly, Proe. loth EC
introduce this new step in the industrial Photovolt Solar Energy Conf.
process in profitable conditions. (Lisbonne, 1991) to be published.
[4] J.Del Alamo, J.Van Meerbergen,
ACKNOWLEDGEMENTS. F.D'Hoore, J.Nijs, S~lid State
Electron. 24 (1981) 533.
This study was funded by EEC (DG XII) [5] S.Martinuzzi, I.perichaud, J.Gervais,
under contract JOUR-0036 (Multichess), O.Sarti, loth EC Photovoltaic Solar
inside the Joule program. Energy Conf. (to be published) .
The authors wish to thank W. Laureys
for his help in solar cell processing and
C.Grattepain for SIMS measurements of the
boron prof iles.

663
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

+ +
AN APPROACH TO THE DEVELOPMENT OF EFFICIENT 31 cm 2 AREA ALL-IMPLANTED n pp
SILICON SOLAR CELLS

Michal RUZINSKY and Stanislav BARATKA,

Slovak Technical University, Faculty of Electrical Engineering,
Ilkovicova 3, 812 19 Bratislava, Czecho-Slovakia (CSFR),
Jozef PALAJ, Tesla, Vrbovska cesta, 921 72 Pies~any, CSFR,
Pavel SEIDL, Czech Technical University, Faculty of Electrical
Engineering, Technicka 2: 166 27 Prague, CSFR
Frantisek KOSEK and Zdenek CIMPL, University of Chemical Technology,
Namesti Legii 565, 532 10 Pardubice, CSFR

ABSTRACT. Single crystalline silicon solar cells 31 cm 2

area, with all-implanted n pp+ layer structure on second
quality electronic grade silicon with AM 1 efficiency 13.2 %
(average value) have been developed. High-cost Ti/Pd/Ag
contacts , which are mostly used in high efficiency silicon
solar cells, were substituted with Ti/Ni/Ag (750 nm), enhan-
ced by dip soldering in low-cost Pb/Sn or Pb/Sn/Ag solder.
Evaporated thin film of chalcogenide glass As 2S8 , or plasma
deposited passivation oxide SiD were used as an anti-
reflection coating. Best solar x cell was characterized by
conversion efficiency 14.6 %.

1. INTRODUCTION solar cells in Czecho-Slovakia till 1988

we presented in 1121, where we reported
13.1 % efficiency for 15 cm2 area all-
Silicon solar cells are basically -implanted s·ilicon solar cells. In this
simple semiconductor devices, which have work we have tried to prepar~ e!ficient
been improved considerably during the past 31 cm2 area all-implanted n pp silicon
10 years. In fact, ten years ago, many pho- solar cells.
tovoltaic experts were skeptical that 17 %
efficient single-crystal silicon solar cell
(some cm2) could be produced. With the sub- 2. CELL STRUCTURE AND FABRICATION
sequent availability of very high-purity
material, and the increased activities in
cell modelling and design, the efficiences Cell design was based on all-implanted
have envolved not only beyond the 17-18 % n+pp+ structure. Starting wafers comprised
barier, but up to 24 % 111,121,131,141, boron-doped industrial monocrystalline sili-
151, 161, 171. These facts indicate that con obtained from TESLA Roznov (electronic
silicon cell efficiences of 25 % are possi- grade). The wafers have a diameter of 63 mm,
ble indeed. However, these excellent and 350 pm in thicknees. By technical specifi-
further improvable results were obtained cation - concerning of surface quality des-
mainly under quite restrictive laboratory cription - wafers have been second quality:
conditions (namely small-area 4 cm2 cells polished front surface free from contamina-
and with high-cost Ti/Pd/Ag contacts). It tion, dimples, orange peel; scratches allo-
is known that a cell surface increase leads wable. Back surface: etched in acid mixture,
to homogenity problems which may decrease free from contamination, saw marks. Wafer
the efficiency. edge rounded, chipping and chips up to
Very successful was therefore the 0.25 mm are allowable.
Spire Corporation approach to high efficien- Surface orientation was (100). The
cy large-area silicon cells 181. The cells nominal resistivity of monocrystalline sili-
developed by M.B. Spitzer et al. were 53 con was 2.2 - 4.7 ohmcm and 6 - 10 ohmcm,
cm2 in area and h~ve exhibited AM 1.5 effi- respectively. The selected resistivity is
ciencies of 18.1 - 0.3 % (average perfor- not of best choice for the fabrication of
mance of 25 all-implanted silicon solar the very high efficiency cells: however it
cells). Using single-crystal CZ SILSo is within the range of resistivity present-
silicon encapsulated cell efficiencies of ly used in crystalline silicon material
18 % were attained for areas in the range approaches I BI, 1111.
90 to 113 cm2 191. Table I. indicates the fabrication
Best conversion efficiency for larger proces for 31 cm2 area solar cells prepared
cell sizes of about 50 cm2 have been achie- on monocrystalline silicon - resistivity of
ved by Green et al. 151. Sixteen laser 6 - 10 ohmcm and Table II. next approach
grooved, buried contact silicon solar cells for 2.2 - 3.7 ohmcm.
(7 cm x 7 cm) have exhibited average effi- In adition to parameters listed in
ciency 19.2 % (record value 19.6 %). Note- Table I. chalcogenide glass As 2SB was evapo-
worthy is the 20.4 % conversion efficiency rated onto the front surface, as an
demonstrated for 12 cm2 microgrooved solar antireflection coating. AR-layer thickness
cells fabricated by Green's UNSW Kensington 53 nm was optimized for wavelength of 500
group, too 151. nm. Evaporation had been done from molybden
Authors of this paper ~re+dealing with crucible at vacuum 10 (exp -3) Pa, growing
development of large-area n pp solar cells rate 0.5 nm/s. Resulting refractive index
on monocrystalline silicon from the begin- was n = 2.35.
ning of 1985 only. These status of the deve-
lopment of large-area crystalline silicon

664
 Table I.
Material: 5i-wafers (1001, nominal resistivity 6 - 10 ohmcm
Implant back: 11 8+ 50 keY, 5xlo l5 ions/cm 2
Implant front: 31 P+
30 keY, 2xlo l5 ions/cm 2
Anneal in three steps (dry N2 ): 550 0 C 120 min
850 °c - 15 min
550 °c - 120 min
Evaporate back metallization: Ti/Ni/Ag (750 nm)
Pattern front using of microlithographic mask
Evaporate front metallization Ti/Ni/Ag (750 nm)
Enhancement of contacts by dip soldering (Pb 95 %, Sn 5 %)

Table II.
Material: 5i-wafers (100), nominal resistivity 2.2 - 3.7 ohmcm
Implant back: 118+ 50 keY, 5xl0 15 ions/cm 2
3l
Implant front: p+ 12.5 keY, 2xlo 15 ions/cm 2
Anneal in three steps: 550 °c 120 min (in dry N2 )
850 °c - 15 min (in dry N2 + O2 )
550 °c - 120 min (in dry N2 )
Surface passivation - evo oxide 5iO x (at temperature 380 °C)
Evaporate back metallization: Ti/Ni/Ag (750 nm)
Pattern front using of microlithographic mask
Evaporate front metallization: Ti/Ni/Ag (750 nm)
Enhancement of contacts by dip soldering (Pb 46.5 % 5n 47 %, Ag 6.5 %)

Another approach (Table II.) was Nine samples 31 cm 2 area all-implanted

carried at ion energy 12.5 keY, on mono-5i n+pp+ solar cells were fabricated on mono-
~esistivity of 2.2 - 3.7 ohmcm) with similar crystalline silicon with resistivity of
technology steps as mentioned before but 6 - 10 ohmcm, using of technology summarized
withgifferent annealing (in the second step: in Table I. e.g. with antireflection coating
850 C 15 min in dry N2 + 02), dip soldering As 2S8 • The best solar cell was characterized
in Pb/Sn/Ag (Pb 46.5 %, 5n 47 %, Ag 6.5 %) by conversion efficiency of 13.8 %.
and with plasma deposition of passivation Table III. lists the average performance
oxide SiD (instead of As 2S8 ). parameters for these cells.
The xpp junction is not such
effective as a back surface field 181, 110 I
but the p+ region does aid the formation
of a low resistance back contact. Thicknees
of anneal oxide was about 6 nm.
High-cost Ti/Pd/Ag contacts, which are
used mostly in high efficiency silicon
solar cells, were substituted with Ti/Ni/Ag
(750 nm), enhanced by dip soldering in low-
-cost Pb/Sn (Pb 95 %, Sn 5 %) or Pb/Sn/Ag
(Pb 46.5 %, Sn 47 %, Ag 6.5 %) solder, in
this work.
The grid covers about 10 % (shadow
loss) of whole surface. The maximal hight
of enhanced grid was approximately 20 ~m.
The grid-design of developed 31 cm2 area
crystalline silicon cells is ilustrated
in Fig. 1.
3. TESTING AND RESULTS
Fig. 1. The grid-design of 31 cm 2 area
Cell performance measurements were n+pp+ silicon solar cell
made under a simulated AM 1 spectrum. The
testing was carried with ORIEL model 6722
solar simulator, at room temperature.
Standard xenon lamp Canrad Hannovia 300 W
has been used as a light source. Incident
light was measured by ORIEL model 81020
solar simulator radiometer.

665
Table III. Average performance of nine efficient 31 cm 2 area n+pp+ silicon solar cells
with chalcogenide glass As 2S8 AR-coating; AM 1 100 mW/cm 2

Voc (mV) Isc (rnA) J sc (mA/cm2) FF Eff. (%)

average 572 1239 39.9 0.584 13.21

value
standard 3.5 44 1.4 0.034 0.34
deviation

The cell design and fabrication process 121 A.W Blakers, M.A. Green, "20 % Efficien-
described in Table II. was res~lting to cy Silicon Solar Cells", Appl. Phys.
14.6 % efficient 31 cm2 area n pp silicon Lett., 48, 0),1986, p. 215.
solar cell (AM 1 100 mW/cm2 and room tem- 131 A. Rohatgi, "Review of High Efficiency
perature). Next performance data of the Silicon Solar Cells", SPIE 706. Photo-
best solar cell were: voltaics for Commercial Solar Power
V = 581 mV, I = 1204 mA, J = 38.8 Applications, 1986, p. 15.
oc mA/cm2, and sc FF = 0.64. sc 141 A. Rohatgi, "A Review of High Efficien-
cy Silicon Solar Cells", Proc. 18 th
4. DISCUSSION AND CONCLUSION IEEE Photovoltaic Specialists Conferen-
ce, Las Vegas, 1985, p. 7.
151 M.A. Green, A.W. Blakers, S.R. Wenham,
The structure and fabrication of "Improvements in Silicon Cell and Modu-
31 cm2 area monocrystalline silicon solar le Performance", Proc. 9 th E.C. Photo-
cells with 13.8 % and 14.6 % (AM 1 100 mW/ voltaic Solar Energy Conference, Frei-
cm2 ) conversion efficiency have been burg 25 - 29 Sept. 1989, p. 301.
described. 161 A.W. Blakers, J. Zhao, A. Wang, A.M.
In comparison with the record conver- Milne, X. Oai, M.A. Green, "23 % Effi-
sion efficiency values mentioned in intro- cient Silicon Solar Cell", Proc. 9 th
duction of the paper, efficiency values E.C. Photovoltaic Solar Energy Confe-
of our solar cells are quite lower. The rence, Freiburg, 1989, p. 32B.
main reasons of this lower efficiency we 171 A. Wang, J. Zhao, M.A. Green, "24 %
have assumed: Efficient Silicon Solar Cells", Appl.
- non-optimized implantation parameters for Phys. Lett. 57 (6), 1990, p. 602.
substrates used in our experiments. 181 M.B. Spitzer, L.M. Geofroy, M.J. Nowlan,
Precise computer modelling of the implan- C.J. Keavney, "Performance Improvements
tation and annealin~ steps resulted in in Silicon Flat-Plate Cells and Modules"
0.415 ~m depth of n p junction, what Solar Cells, 20, 1987, p. 333.
would be more suitable for screen-printed 191 K.D. Rasch, W. Schmidt, "A 18 % Crystal-
contacts than for evaporated Ti/Ni/Ag line Silicon Cell Module", Proc. 9 th
contacts used at our cells. E.C. Photovoltaic Solar Energy Confe-
- the use of Si substrates having resisti- rence, Freiburg, 1989, p.71S.
vity in the range 6 - 10 ohmcm, what 110IM.B. Spitzer, C.J. Keavney, S.P. Tobin,
causes lowering of the FF values to the "Ion-Implanted Silicon Solar Cells with
average value 0.5S . 18 % Conversion Efficiency", Proc. 17 th
- the use of non-optimal contacts. IEEE Photovoltaic Specialists Conferen-
Evaporated Ti/Ni/Ag contacts enhanced ce, Kissimmee, May 1 - 4, 1984.
by dip soldering (low-cost Pb/Sn or IllIM.B. Spitzer, C.J. Keavney, R.G. Wolf-
Pb/Sn/Ag) used in our experiments are not son, R.J. Matson, "A Study of Ion
the best choice for high efficiency cells, Implantation and Annealing of Low-Cost
however in comparison with evaporated Silicon Sheet", Proc. 17 th IEEE Photo-
Ti/Pd/Ag and electroplated with Ag lsi, voltaic Specialists Conference,
1101,1111 these are cheaper. Kissimmee, May 1-4, 1984.
From these reasons we have seen the 1121M. Ruzinsky, S. Baratka, J. Palaj,
possible ways of increasing of the efficien- J. May, "13 % Efficient 15 cm2 Area
cy of our cells in the changes of the All-Implanted Silicon Solar Cells",
emitter design (or the use of screen-prin- Proc. 8 th E.C. Photovoltaic Solar
ted contacts without changes in emitter Energy Conference, Florence, 9-13 May
technology), surface texturing, optimiza- 1988, p. 1422.
tion of the base material resistivity and
increasing of the technology precision
discipline al all technological steps.
REFERENCES

III M.A. Green, A.W. Blakers, S.R. Wenham,

S. Narayanan, M.R. Willison, M. Taouk,
T. Szpitalak, "Improvements in Silicon
Solar Cell Efficiency", Proc. 18 th
IEEE Photovoltaic Specialists Conferen-
ce, Las Vegas, 1985, p. 39

666
 8-12 APRIL 1991 LISBON, PORTUGAL
tOTH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE

MAXIMISING MINORITY CARRIER LIFETIME IN

HIGH EFFICIENCY SCREEN PRINTED SILICON BSF CELLS

T M Bruton BP Solar, Solar House

A Mitchell 36 Bridge Street
L Teale Leatherhead, Surrey KT22 8BZ, UK

J Knobloch Fraunhofer Institute for Solar Energy Systems

Oltmannsstrasse 22
Freiburg D7800, Germany

ABSTRACT. A theoretical model was used to predict the performance

of n+pp+ solar cells as the wafer thickness was reduced. Sol~r
cells were made using aluminium and boron to produce the BSF ~n
wafer thickness from 350 to 200 microns and resistivities from
1 to 10 ohm cm. The boron based cells gave good agreement wi~h the
model but the aluminium p+ cells did not. The model was used to
demonstrate the effect of increasing minority carrier diffusion
length in the cell. The minority carrier diffusion length of the
aluminium p+ cells was measured by EBIC and spectral response. A
large increase in diffusion length was noted and the gettering
action of the aluminium was confirmed by SIMS analysis. High
efficiency cells were made in the optimum design.

1. Introduction

Economic consideration requires the use of experience has shown this to be incorrect.
thinner and thinner silicon wafers to (1) (2) Therefore the predictions of the
fabricate solar cells. Furthermore, as the model were tested experimentally.
wafers become thinner there are beneficial
effects in decreasing bulk recombination and Thickness 1 ohm cm 5 ohm cm 20 ohm cm
series resistance. However, there is also a (urn) NoBSF BSF NoBSF BSF NoBSF BSF
disadvantage in the loss of short circuit
current due to incomplete absorption of the 200 12.9 13.2 11. 9 12.7 11.0 12.3
solar spectrum and an increase in the effect 250 13.1 13.2 12.1 12.7 11.3 12.2
of back surface recombination. It is well 300 13.2 13.3 12.3 12.7 11.5 12.2
known that these negative effects can be 350 13.2 13.3 12.4 12.6 11.6 12.1
reduced by the use of a highly doped (p+ for 400 13.3 13.3 12.4 12.6 11. 7 12.1
pn+ structures) region on the cell rear
giving rise to an internal field (BSF) which Table 1: Solar Cell Efficiencies for I, 5 &
acts as a minority carrier reflector in the 20 ohm cm wafers for varying thicknesses with
base (1) (2). For an optimum BSF structure, and without BSF.
wafer thickness and resistivity are
independent variables which need to be 3. Experimental Work
determined. A theoretical evaluation was
carried out by FISE. Cells were made by BP I, 5 and 10 ohm cm wafers in thicknesses 350,
Solar using both boron and aluminium to 300 and 250 microns were obtained from
produce the BSF. A further theoretical commercial suppliers. These were processed
evaluation was carried out on the into solar cells using conventional
significance of the minority carrier processing except for the formation of the p+
diffusion length in BSF cells and following region. The wafers were thinned by 50
measurements of minority carrier diffusion microns in the process. Boron diffusions
length in BSF cells, high efficiency cells were carried out in a belt furnace using a
were produced. Merck spin-on source. Sheet resistances of
100 ohms per square were obtained. Lower
2. Theoretical Analysis values were not possible because of the
degradation in lifetime of the CZ wafers with
The model developed by FISE (3) (4) was used prolonged high temperatures. The BSF using
to model the effect of wafer resistivity and aluminium was formed by screen printing
wafer thickness on solar cells with and commercially available aluminium pastes on to
without a BSF. The model assumed a the cell rear and forming the BSF by heating
phosphorous doped emitter at 30 ohm per to 850°C for a short time.
square and an anti-reflection coating giving
4% reflection loss. The values of minority The results for the Voc versus thickness of
carrier diffusion length against resistivity wafer and resistivity are given in Table 2.
from commercially supplied wafers as measured There is reasonable agreement with the model
by EBIC were used for the calculation. The for the boron samples but the aluminium BSF
results are given in Table 1. As the wafer samples gave values which were much higher
thickness decreased Isc decreased and the 1 than the boron BSF cells and that predicted
ohm cm wafers always gave the highest by the model.
efficiency without BSF except at 200 microns
thickness. Although this would indicate that
there is no benefit in BSF cell technology,

667
Thickness 1 ohm cm 5 ohm cm 10 ohm cm Spectral Response EBIC
(um) C1c Brn Al. Clc Brn Al. Clc Brn Al.
No BSF 160 140
200 595 588 564 563 562 557 585 BSF 400 316
250 595 564 566 562 558 595
300 595 596 606 564 568 562 559 610 Table 5: Minority carrier diffusion length
(microns) for 10 ohm cm solar cells with and
Table 2: Voc (mV) for 1, 5 and 10 ohm cm without Aluminium BSF.
wafers with BSF calculated, Boron BSF and
Aluminium BSF. 6. Causes of Improved Minority Carrier
Diffusion Length
4. Further Theoretical Work
In forming the BSF, aluminium forms a
The unexpected high values of Voc (Table 2) eutectic with silicon which is molten above
obtained for aluminium BSF cells led to a 675'C. This molten alloy acts as an infinite
re-examination of the model. The parameters sink for impurities diffusing out of the
describing the back surface field were varied silicon and fast diffusers such as copper and
over a wide range but the model could not be iron are readily gettered. SIMS analysis
made to fit the experimental data for the shows a concentration gradient of these and
aluminium BSF cells. The only way that the other metallic species at the back surface of
high Voc's could be achieved was by an the silicon after the aluminium has been
increase in the minority carrier diffusion etched away (Table 6). This clearly
length in the base of the cell. Table 3 demonstrates that in addition to minimising
shows the theoretical variation of Voc with recombination of the cell back surface
minority carrier diffusion length for a 1 and aluminium actually getters impurities and
10 ohm cm 200 micron wafer. An increase in this is a major factor in producing high
Isc is also predicted (3) and demonstrated in efficiency cells. This has not been fully
Table 4. recognised previously (5).

Diffusion 1 ohm cm 10 ohm cm

length (um) Rear Rear Bulk
Edge Centre
50 573 510
100 591 531 Cobalt 11 2 < 0.1
150 601 544 Titanium 5 0.4
5 <
200 606 554 Copper 200 10 < 0.8
250 609 562 Iron 400 300 < 1.0
300 611 568 Nickel 300 100 < 3.0
350 613 573
400 614 576 Table 6: SIMS analysis of Back Surface of
Aluminium BSF cell after etching off
Table 3: Variation of Voc (mV) with minority aluminium. 3Surface concentration in 10 exp
carrier diffusion length for 1 and 10 ohm cm 16 atoms/cm .
200 micron wafers with BSF.
7. Production of High Efficiency Cells

By considering the enhancement of minority

1 ohm cm 10 ohm cm carrier lifetime due to aluminium gettering,
Calc. Meas. Calc. Meas. it is shown that higher efficiencies are
obtainable by using higher than 1 ohm cm
Without BSF 30.2 27.3 30.6 27.2 resistivity wafers with an aluminium BSF.
With BSF 31. 2 31.4 33.0 31. 5 Efficiencies of 15.6% in small area cells
(Fig 1) with an ARC have been achieved while
Table 4: Calculated and actual IZc for cells efficiencies of 14 22% (Fig 2) have been
with and without Al/BSF in rnA/cm with achieved on 100 cm cells with ARC under
increase in diffusion for BSF cell. production conditions using normal CZ wafers.
5. Verification of Increase in Minority 8. Acknowledgements
Carrier Diffusion Length
Dr A E Hughes and M Lesniak of BP Research
In order for the observed Voc's to be Sunbury, performed the EBIC measurements.
achieved, the minority carrier diffusion Many colleagues at FISE, especially B Voss,
length must approximately double as a result contributed helpful discussions.
of the formation of the BSF by aluminium.
Table 5 gives the results of minority carrier 9. Conclusion
diffusion length on samples with and without
aluminium treatment. It is clear a large Aluminium in n+pp++ BSF cells enhances
increase has occurred. There is some efficiency not only by creating a BSF but by
uncertainty in the higher measurement as the gettering impurities and significantly
diffusion length exceeds the wafer thickness increasing the minority carrier diffusion
which is outside the approximation used to length in the base. A theoretical model has
calculate diffusion length. been developed which fully predicts the cell
parameters when this increase in diffusion
length is taken into account.

668
 REFERENCES
(1) R Van Overstraeten, R P Mertens.
Physics, Technology and Use of Photovo1taics.
Adam Hilger Ltd (1986).
(2) M A Green.
High Efficiency Silicon Solar Cells.
Trans Tech. Publications Ltd (1987).
(3) J Knobloch, A Aberle, W Warta, B Voss.
Technical Digest of PVSEC5 pp 955-958 (1990).
(4) J Knobloch, A Aberle, B Voss.
9th EC PVSEC, Freiburg. p 777 (1989).
(5) Reference (2), p 142.

Amps 2.0 ,
~
,
J
,
~
i
!
,
~
i
!
'
!
i
I
'
!
~ I
Ise 1.7 Amps
Voe 601 mVolts
FF .731
Eff 15.6%
Area= 48 em 2

.24 .48 .72

Volts

Fig. 1

Voe= .61 V
Ise= 3.091 A
.8 Jse= 31.82 mA/em~2

.6 Pm.x= 1.38 I'

Vmax= .5 V
.4 I( .46 ) 2.914 A

"'"
lSI
.2 FT = .731

*VJ
o
Eff= 14.18 %

0..-.2 II Area= 97. 15 em~2

E II Illum= 100 mN/cm~2

a: -.4 ,
Ileak= .0109 A/V
-.6 Rsh= 91 Ohms
-.8 I Rs= 7.6 mOhms

-1
CD
~ ~ IS) ~
I Vo1lts (*1)

Fig. 2

669
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

ABOUT THE EFFECT OF IMPURITY CONTAMINATIONS, PROCESSING AND THERMAL ANNEALING ON THE
PV PROPERTIES OF POLYCRYSTALLINE SILICON

S. Pizzini, M. Acciarri, S. Binetti, S. Acerboni

Department of PhYSical Chemistry and Electrochemistry,
via Golgi 19, 20133 Milano.
M. Falconieri, L. Pirozzi, M. Garozzo
ENEA C.R.E. Casaccia, S. M. in Galeria, Roma.

ABSTRACT. In this paper we report the results of a cross analysis of different

polycrystalline silicon materials, carried out using LBIC and FTIR spectroscopy
measurements.
The materials investigated were those produced by Wacker, Photowatt and Italsolar.
The characterization aimed to account for the different PV properties exhibited by
the various materials in terms of composition, bulk and grain boundary properties of
the as grown and processed materials.
Al though not complete, the results offer interesting arguments for a thorough
discussion about the role played by oxygen and carbon impurities on the PV
properties of these materials.

INTRODUCTION.

The primary aim of the work reported in this defects is secondary wi th respect to that of the
paper was to make a first (although preliminary) "chemical reconstruction", originated by the
comparative analysis of the phYSical, structural segregation of oxygen and carbon.
and photovoltaic properties of three different A large scattering is however observed when
types of polycrystalline silicon, already the "average" diffusion length is displayed as a
available on the market, namely the Wacker, function of the macroscopic variables of the
Photowatt and Italsolar ones. As it is well known, system, which calls for a different recombination
they are produced using remarkably different strength of individual dislocations and GBs /7/.
growth processes, two of them being however grown Further complications arise when the as grown
with unidirectional solidification procedures. As material is heated during the device fabrication
it could be expected, they exhibit quite different process. Here, depending on the temperature and on
structural, physical and PV properties, here the time of treatment, impurities may segregate or
including the capability of being H2 cured /1/ and precipitate to a second phase compound, eventually
that of being Al gettered /2/. introducing local intepnal stresses /11/.
These differences apparently result from the When conditions for stress relaxation do
different chemical and mechanical constraints occur, other effects are possibly generated, as
/3,6/ associate to the particular growth process, dislocations, twins, and point defects /12/, all
which may affect the grain boundary (GB) and being eventually, although unevely, electrically
intragrain properties, these last being generally active.
believed to be dominated by dislocations, and, The results of experiments on polycrystalline
eventually, by precipitates. materials are therefore not easily fitted in the
It could be therefore easily understood that frame of existing models, unless all the
the ultimate properties of extended defects in a experimental variable are known and kept suitably
PV cell not only would depend on the material constant, except the one on which the system is
manufacturing process, but that they are studied to depend.
profoundly influenced by the device fabrication Therefore, we intend first to understand how
process, too. the chemical composition of polycrystalline
Why the properties of a particular material silicon is influenced by the growth process, and
are so different it is instead not a trivial how the chemical composition influences its
question, as the effective diffusion length and physical properties. Furthermore, we would like to
the conversion efficiency depend in a very understand how the thermal treatments do influence
complicated manner on all the factors considered the physical properties of a specific material,
above /7/. In fact, while to a first insigth the and how the different materials "react" to the
diffusion length Ln is shown to depend closely on same device processing, in terms of GBs activity.
the dislocation density Nd /8-9/, with a threshold The results of this preliminary investigation
effect at 10*4<'Nd -< 10*5 /3,9/, the role of GBs already appear of practical and scientific
being almost negligible, a deeper insigth shows interest.
that its functional dependence looks more
complicated /1/. It was therefore argued that the
role of structural reconstruction of extended

670
EXPERIMENTAL. In( * l=In,oo 1((0{ - j )/1. .. ~.1) (3)
where:
10XlO cm square slices of different ongln In(*) is given by the ratio In/I(baseline)
were first cutted to a smaller size, 90~ CP4 In(baseline) is the short circuit current of a
etched and then lapped to both sides using 15 urn reference sample of known diffusion length,
diamond paste, rinsed with acetone and then with C(is the absorption coefficient at a specific
doubly distilled water, before being examined for wavelength and In,CO is given by the ratio:
the oxygen and carbon content with a Fourier
Transform (Nicolet mod.MX1) IR spectrometer. The I(Ln·~OO)/In(baseline)

oxygen and carbon contents were calculated using

the conversion coefficients reported in the ASTM In this particular case, Ln was calculated from
Practices, F1l88 and F123. eq. 3, using the known value of Ln (100 urn),
The LBIC measurements were performed using a measured with the surface photovoltage method, of
system already described /11/, which permits a the as grown sample, which was used as the
lateral resolution better than 2 urn. The light reference sample. As a further approximation, In,
sources were three lasers (HeNe, lambda=632.8 nm, was taken equal to one.
GaAIAs, lambda=780 nm, GaAlAs, lambda=830 nm). All
the measurements were carried out at constant
photon density and in conditions of low injection. EXPERIMENTAL RESULTS AND DISCUSSION.
In order to measure the electrical properties of
samples in the as grown conditions, Schottky 1. Effect of the growth process on the chemical
diodes were deposited on the front side of the composition of the material.
different samples, after CP4 etch, using a The results of the IR measurements on the
procedure also described in ref. 11. different samples examined are reported in TableI.
For the characterization of the processed The Italsolar material results to be carbon and
material, 10xlO cm square as cut wafers from oxygen rich, in good agreement with our results
different manufacturers have been first CP4 reported in ref.3 relative to a material grown
etched, rinsed, and then submitted to a POC13 with the same procedure.
thermal diffusion run at 850·C for 15 min, in an We know that on this material /3/, the
open tube furnace: the sheet resistance of the average value of the diffusion length takes the
emi tters obtained in this way was of about 60 maximum value for Nc-No values close to 4-6 ppma,
Ohm/square, and the depth of about .5um. For the an excess of carbon being therefore beneficial for
ohmic contacts, the grids have been made by its electrical properties.
evaporating a t~in gold layer, and an In-Ga alloy The Wacker material presents instead a very
has been used for the contact on the back. low oxygen and carbon content; we may speculate
For the study of the effects of thermal that this material should be only slightly
treatments, a single p-type sample (W22/29), grown sensi ti ve to segregation phenomena involving
/3/ with the same technique now adopted by oxygen and carbon. If these last are capable to
Italsolar, of known oxygen (5.33xlO*17 cm-3) and passivate GBs /5/, we expect a worsening of the PV
carbon (7.75x10*17 cm-3) content was subjected to properties,with respect to the Italsolar one.
a series of successive isochronal (24 hrs) heating The Polyx material, eventually, presents a
steps from 450 to 1150 ·C. After each thermal large excess of oxygen over carbon: we have
treatment both the electrical activity of a single already shown that this kind of materials present
GB and that of a nearby grain were measured. significative GB recombination activity /5/.
It is to be remarked that after each
treatment the sample was rapidly quenched to room 2. Electrical activity of the as grown materials.
temperature. The results of these measurements The analysis of the electrical activity of
are, therefore, not representative of the the as grown materials has been carried out on
conditions adopted for the manufacturing of solar single samples of well characterized slices; the
cells. resul ts obtained are not, therefore, relevant
The LBIC data were analyzed in terms of the statistically. As it will be seen, however, they
recombination strength /5/: are reasonably well in agreement with data
reported in the literature. Also in this case the
F=l/In~ (In-I*(x)) dx (1) results are reported (see Table II) in terms of
the recombination strength of a specific extended
and of the contrast factor: defect, in addition to the contrast values, as
well as in terms of the local diffusion length Ln,
C = (In - Igb)/In (2) as determined in the neighborood of the extended
defect from the local short circui t current In,
of the GB, where In is the short circuit current using eq. 3.
measured far from the GB, 19b is the current at However, in this case a reference CZ sample
the GB and I*(x) is the LBIC profile. The minority on which the diffusion length was measured by the
carrier intragrain local diffusion length Ln was SPY method was used, and the In, co value was
calculated from the experimental values of the calculated from the data reported in ref.13,
short circuit current In measured in the centre of which permi t to assume that the short circuit
the grain, according to the equation /12/: current of the reference sample amounts to about
the 86% of the short circuit current of a sample

671
having a Ln ~ CO strength of the GB examined, which was a strongly
recombining GB already in the as grown sample,
2.1 Properties of the Italsolar material. increases after the heat treatment at 450 ·C and
On this material the GBs are mostly non then drops to negligibly low values after the heat
recombining. On a region of 3.5 sq.cm we were able treatment at 650 ·C.
to find only a single region where a strong In both cases, therefore, 650· C are a
recombination activity was found. Incidentally, critical temperature. It is known from the
this is a region where GBs are decorated with li terature, however, that at this temperature
dislocations; here, the local short circuit oxygen precipitation begins to occur at measurable
current and then the diffusion length drops to rate /14/, eventually anticipated by the formation
very low values, as it is shown in Table II. of oxygen aggregates /15/ known as new thermal
donors (NTD) and that the formation of NTD is
2.2 Properties of the Photowatt material. favoured by the presence of carbon.
On this material GBs and also twins are We may therefore speculate whether the
strongly recombining. passivation of GB at 650 ·C comes from
The material is, however, macrocrystalline, heterogeneous segregation of oxygen at GB, which
with very large grains. We expect therefore only a saturates the residual dangling bonds. In turn,
sligth influence of these extended defects on the homogeneous nucleation of precipitates (which may
overall PV performance, which should depend also occur heterogeneously at dislocation /11/) is
entirely on the bulk properties. responsible of the sudden decrease of the bulk
These last, as determined by the value of the diffusion length, according to a mechanism of
bulk current in a very large grain, are reported stress relaxation via self interstitial emission
in Table II, and look confortably close to the already discussed in ref. 11.
literature values for this material. From the known value of the oxygen diffusion
coefficient Do at this temperature, we can
2.3 Properties of the Wacker material. estimate the amount of oxygen segregated at GBs
On this material GBs and twins are from the equation:
systematically strongly recombining. (see Fig. 1)
Additionally, the local short circuit current No(s)= Ld x A x No (3)
shows a very strong dependence on the measuring where:
site, with variations of a good factor of 2. On Ld =.fDo t
this sample, the lowest values of local bulk t = time
diffusion length were measured. A = unitary surface area
No =average oxygen concentration (5.53x10*17
3. Electrical activity of the processed material. cm*-3 )
The analysis of the electrical activity of
the processed m'aterial has been carried out by which amounts to 8.04x10*12/cm2, using a value of
performing LBIC measurements, using the ReNe Do=2.45x10*-15 cm2xsec-1 (at 650 ·C) and t =
laser, on several GBs for the different materials. 8.6x10*4 sec.
The parameters extracted (recombination We can see that this value fits relatively
length F and contrast factor C) are therefore the well with the calculated interface state density
average of several data and so they are (1.6x10*13/cm2) /8/, in the hypotesis that a
statistically relevant; due to the fact that in partial reconstruction of GBs already occurred,
the Italsolar material most of the GBs are non thus supporting the hypotesis that oxygen
recombining, in this case the average has been segregation at GBs is responsible of their
performed on only 3 GBs. passivation.
The results are reported in Table III. As it
can be seen, the Wacker material and the Photowatt
one show appreciable recombination activity at the REFERENCES
GBs, with similar recombination lengths and
contrast factors. /1/ R. El Ghi tani et al., Ann. Chim. Fr. 12
In order to have a first sight of the PV (1987), 429.
perfomances which can be obtained from such /2/ S. Martinuzzi et a1. , Rev. Phys. App1. 22
materials, external quantum yield measurements (1987) , 645.
(see Fig.2) have been performed on the same cells /3/ S. Pizzini et a1. , J. Electrochem. Soc. 135,
analysed by the LBIC. As it can be seen the same (1988) , 155.
process induces different PV behaviours on /4/ L.Ammor et al.,Ann. Chim. Fr. 12, (1987), 377.
different materials. Further investigations are in /5/ S. Pizzini et a1., in "Point and extended
progress. defects in semiconductors", ed. by G. Benedek, A.
Cavallini, and W. Schroeter, (Plenum Press, New
4. Effect of thermal treatments. York, 1989) p. 105.
The results of the LBIC measurements are /6/ S. Pizzini, M. Acciarri, in "Polysilicon thin
reported in Fig.3 and 4. film and interfaces", Mater. Res. Soc. Symp.
One can observe that while the diffusion Proc., vol. 182, Mater. Res. Soc. Pittsburgh
length of the minority carriers drops (1990) p. 185.
monotonically in the heat treated samples, /7/ S. Martinuzzi et a1., in Proc. IX PVSEC,
reaching a minimum at 650·C, the recombination (Kluwer-Academic, 1989) p. 3

672
/8/ G.Mathian et al. Sol. St. El., 26, (1983), 131
/9/ B. L. Sopori, Proc. XX IEEE PVSC Las Vegas Material F C
(USA) ,1988
/10/ S. Pizzini et aI., Rev Phys. App1. 23, (urn) (%)
(1988), 101.
/11/ S. Pizzini et al., Mat. Sci. Eng. B7(1990)69.
/12/ F. C. Larche in Ann. Rev. Mater. Sci. 20 Italsolar 1.9 9.5
(1990) p. 83. Photowatt 4.1 17.5
/13/ R. H. Hopkins et al., DOE Contract
DOE/JPL-954331-80/9. Wacker 4.4 16.7
/14/ N. Inue et aI., in "Semiconductor Silicon
1981", H. R. Huff et al. eds. Proc. Vol. 81-5, The Table III. LBIC characterization of
Electrochem. Soc. Ing. Pennington, NJ (1981) processed materials
p.282.
/15/ K. Schmalzl et al. in Mat. Sc. Forum, Defects
in Semiconductors 15, G. Ferenczi ed. vols. 38-41
(1989) p. 667. ..
cJ
.6
,., I
Material C(oxyg.)
(ppma)
C(carb.)
(ppma)
o
.. 8)
,i
~

~,
,2

Italsolar 3.5+/-0.1 12.9+/-0.3 e L_I_-,"_ ,.o-&..-._'- __ '-"-'-- -'._-"_~"'--..I~~,J__ _

4013 sea SSIit 1ea Ica SBB I eeB I HI8 12C0
Photowatt 9.6+/-0.4 1.3+/-0.8 Lambda enm)

Wacker 1.1+/-0.2 0.5+/-0.2 Fig.2. External quantum yield of cells on:

A) Italsolar; B) Polix; C) Wacker .
Table I. Oxygen and carbon I.R.
determination
E
.:,.

I
- --633nm

Material F Ib In(*) Ln
<:
..J ~
, ----780nm
C
\:,-, - '--830nm
urn % 10-8 A um \, -' -
\ '." -
" ,
Itals,GB 12.6 3.7 .32 '\
5
Itals,grain 4.3 .37
\ '\
6 \ I
, \
Polix,twin1 23.3 37.6 10.5 .91 47 \
\

1
, \
Polix,twin2 50.6 61.0 \\ \
Polix,twin3 12.1 27.6 I \
\
I,

Polix,grain 9.7 .85 35 ..........

'\ --~-.:.:
Wacker GBI 88.8 36.5 1.6 .14 2 400 800 1200
Wacker GB2 37.3 38.5 anneal ing lemperalu,e (oc)
Wacker GB3 n.d. 65.0
Fig.3. Dependence of the diffusion length
Table II. LBIC characterization of as grown on the annealing temperature
materials
20

1 --633nm
.....1 ----- 780nm
/\
15 ,,/" \ - _ . - 830nm
/' \
/ \
,,/" \
\
1~ \
\
i
\
\
\
\
~=:-::::~::7'-";

o 400 800 1200

annea1,ng lemperatIR("C)
Fig.1. LBIC map of Wacker as grown
material Fig.4. Dependence of the recombination
strength on the annealing temp.

673
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

MICROSTRUCI1JRE INVESTIGATIONS AND ELECTRICAL MEASUREMENTS IN

POLYCRYSTALLINE SOLAR SILICON

H,J. MOLLER, E. RODAK, L. HUANG, M, GRIESS*

Department of Material Science and Engineering,
Case Western Reserve University, Cleveland, 44106, USA
(*) IV. Physikalisches Institut, Universitat Gottingen, FRG

ABSTRACT. The microstructural and electrical changes of polycrystalline ingot silicon after
annealing between 400 - 1100 °C are investigated by TEM, resistivity measurements and
DLTS. In particular, the interaction of fast diffusing transition elements (Cu, Ni, Co, Mn) with
grain boundaries has been studied. The metals are diffused into the specimens from a surface
source between 800 - 1000 oC and the precipitation after cooling is studied by TEM (Cu),
radioactive tracer measurements (Co), and DLTS (Mn). The distribution and size of the
precipitates depend on the cooling rate after the diffusion, and differs considerably for each
element. The vicinity of the grain boundaries (about 50 - 100 ~m) is depleted from impurities
which can result in an increase of the resistivity of the bulk material.

1. INTRODUCTION on the structure and morphology of precipitates will be presented

in this paper.
The fabrication of a silicon solar cell requires annealing
treatments at elevated temperatures, for instance during the
diffusion of the doping elements for the p-n junction or a back 2. EXPERIMENTAL PROCEDURE
surface field. These processes can change the electrical properties
of the starting material, which may be especially important for For the investigation wafers from multictystaIline ingots
pol ycrystalline silicon of lower purity for low-cost solar cells. have been cut'which were produced by casting and directional
The presence of lattice defects such as dislocations and grain solidification of putified metallurgical silicon. The
boundaries together with a rather high concentration of impurities microstructure of the starting material was characterized by
can lead to a variety of interaction processes and new electrically scanning (SEM) and transmission (TEM, HREM) electron
active defects. Since under the constraint of low production microscopy, respe.ctively. Typically, the cast silicon had a grain
costs the presence of these lattice defects has to be tolerated to a size of about 1-10 rnm and a dislocation density of about 105 -
certain extent it becomes important to study these mechanisms 1()6 cm·2. The material was p-doped mainly with aluminium and
and to identify the most dominant defects and their ramifications boron and had a specific resistivity of about 1 ncm. Besides the
on the material properties. doping elements the main impurities in the investigated material
Many of the electronic properties of polycrystalline were carbon, oxygen and a variety of metal impurities. Before
semiconductors are related to grain boundary phenomena. Recent the annealing treatment the specimens were cut, polished and
experimental investigations /1-3/ have shown that the electronic chemically etched. The electrical properties of the silicon
properties of grain boundaries can be changed by the presence of specimens were studied by resistivity, capacitance-voltage (C-V)
impurities. The transition metals are of particular interest since and DLTS measurements. For the resistivity measurements
they are very mobile even at low temperatures and introduce deep ohmic (aluminium) contacts were applied at opposite sides of the
levels in the bulk. Copper, nickel and cobalt are the transition specimens so that an average resistivity for the entire material
elements with the highest diffusion coefficients and solubilities in could be obtained. Smaller titanium - Schottky contacts for the
silicon and it has been shown that even after quenching they capacitance measurements were placed inside of larger grains to
cannot be kept in solution /4/. The major difference in the measure the bulk properties. The temperature dependence of the
precipitation behavior of these elements is determined by the electrical properties was investigated by annealing the specimens
different properties of their metal-silicides. The usually observed at temperatures between 400 and 1100 oC for about one hour.
Tj" - CU3Si phase has a large lattice mismatch with silicon, The electrical contacts were removed before the annealing and re-
whereas the structures of NiSi 2 and CoSi 2 matches the silicon applied after cooling. The changes in the microstructure,
matrix almost perfectly (mismatch below 1%). The distribution, dislocation density and grain size, were studied by SEM/TEM
size, and eventually the electrical properties of the precipitates are for selected specimens, so that this influence on the electrical
thus substantially different. Slower diffusing elements such as properties could be determined separately.
manganese or titanium precipitate less, and a substantial fraction For the investigation of the behavior of particular
remains interstitially dissolved and electrically active as transition elements, the metal (Cu, Ni, Co, or Mn) was deposited
recombination centers even after slow cooling. on opposite surfaces of the specimens by evaporation. The
The precipitation and electrical behavior of 3d elements in samples were annealed then in a quartz tube between 800 - 1100
dislocation free, monocrystalline silicon has been studied DC under flowing Argon gas and then quenched or cooled down
extensively /5-12/. On the contrary, muIticrystalline ingot silicon slowly. The annealing times were about one to two hours which
which contains grain boundaries and dislocations is much less yields a rather homogeneous bulk concentration at the annealing
investigated, though one can expect that these lattice defects can temperature. In particular, the following concentrations were
alter the precipitation behavior of some of the elements obtained: about 10 16 cm· 3 for Cu and Ni, 1013 cm- 3 for Co, and
considerably. From the perspective of the performance of solar 10 14 cm- 3 for Mn. The cobalt and manganese diffusion
silicon two aspects are important: (a) the contamination of grain experiments were carried out with bicrystals which had a higher
boundaries and dislocations can change their recombination purity. The poly- or bicrystals were investigated then in the TEM
activity and resistivity, and (b) the distribution of the impurities (Cu, Ni), by tracer measurements (with radioactive 57Co), and
in the vicinity of these lattice defects and thus the electrical DLTS (Mn). Details of the applied experimental techniques have
bulkproperties are changed. Some of the results /13-17/ on the been published elsewhere /13, 16/.
distribution of these impurities at or near grain boundaries, and

674
 Table 1. Maximum solubility, diffusion coefficient at 1100 DC and crystal structure of the metal-
silicide phases of the transition metals investigated here.

3d Elements Silicide phase Structure Cffi3X

a [cm- 3] D1100 [cm2 / s]

Cu CU3Si T\" - orthorhombic 2*10 18 1*10-4

Ni NiSi2 CaF2 (0.4% mismatch) 7*10 17 4*10- 5
Co CoSi2 CaF2 (1.2% mismatch) 3*10 16 1*10- 5
Mn MnSi 2 FeS2 (18.2% mismatch) 2*10 15 3*10-6

3. EXPERIMENTAL RESULTS most cases even for good rectifying contacts non-linear l/C2 vs.
V curves are obtained in this material. The following results were
3.1 Microstructural investigations obtained from contacts with barrier heights between 0.5 - 0.7 eV
The material that was used for the experiments was cut and in a bias range where quasi-linear behavior was observed. In
from various parts of large (> 10 kg) silicon ingots. Optical general the DLTS spectra inside of grains were rather broad and
micrographs showed a preferential orientation of the grains along composed of several single lines. A separation of the spectra into
the growth direction and a large number of low energy grain individual lines which can be related to impurities is difficult if
boundaries. The majority of the investigated grain boundaries they lie close together. In addition, the situation is complicated
were either coherent twins (L 3) with (111) planes or near since the bulk material below the contacts contained dislocations
second order twins (L 9). Etch pit measurements supported by which can also give rise to broad unsymmetrical DLTS spectra.
TEM investigations of the as-grown specimens show a rather Therefore the spectra were analyzed for the time being in terms
inhomogeneous distribution of dislocations with densities of two groups of broad lines with trap levels at ETl '" 0.33 ±
varying between 104 and 108 cm- 2. Areas with very high 0.05 eV and ET2 '" 0.55 ± 0.05 e V above the valence band. The
densities are probably close to dislocation sources which are capture cross sections varied between 1017 - 10 19 cm 2. The
activated during the cooling of the ingots by internal thermal intensity of the two lines changed as a function of the annealing
stresses. During the annealing treatment of the specimens the temperature and the results on the trap concentrations are
dislocation sources seem to remain inactive and above about depicted in fig. 2. For comparison the majority carrier
6()()OC the dislocations become more homogeneously distributed, concentration as obtained from the resistivity measurements is
however, the average density remains almost constant at about included. The main features are the disappearance of the trap
106 cm- 2_ The dislocation density is however reduced in the level ETi above about 500D C and the maximum in the
vicinity of grain boundaries within a distance of about 10 - 20 concentration of ETI between 700 - 800 DC which corresponds
!-Lm. The influence of the annealing treatment on the grain size of with the minimum in the hole concentration in the same
the specimens is negligible. temperature range.

3.2 Resistivity and DLTS measurements

Resistivity measurements at contacts inside of grains and 4. RESULTS FOR THE TRANSITION ELEMENTS
by the four-point probe technique show that the local resistivity
can vary considerably by an order of magnitude or more. Experimental details on the interaction of transition
Therefore an averaged resistivity has to be determined by elements with grain boundaries have been published previously
measurements between larger ohmic contacts and from a larger /13-16/ and only the main results which are most relevant for
number of specimens. The results for the average resistivity as a muIticrystalline solar silicon shall be presented here. Some of the
function of the temperature are depicted in fig. I. They show a basic properties of the transition elements investigated here are
maximum between 700 - 9()()oC. summarized in table I. With the exception of copper the
The presence and formation of electrically active defects precipitates form a silicon-rich phase. The maximum solubilities
was investigated by DLTS measurements on the Schottky and diffusion coefficients for Cu and Ni are rather similar, and
contacts. It was observed that the rectifying properties of the decrease for Co and Mn. Cobalt and nickel have a silicide phase
Schottky contacts degraded if they covered several grains. This which can form almost coherent precipitates with the silicon
may be due to leakage currents along the grain boundaries which lattice. TEM experiments have been carried out for copper and
become conductive after impurity segregation. The Schottky nickel using the same ingot material that has also been analyzed
contacts were further characterized by C-V measurements which by the previous electrical techniques so that a direct comparison
allows to determine the majority carrier (holes) concentration. In is possible.

IE+17
1.8
§ 1.6 E
u
IE+16 -
a 1.4 c:
~
.g IE+15 -
:~ 1.2 g
8c: 1E+ 14 -
n I~
Vi
-;;;
<) 1.0 o
0:::
0.8
u IE+13 n

200 400 600 800 1000 20 400 500 600 700 800 900 1000 1100
Anneal Temperature ( 0c} Anneal Temperature [ °CJ

Figure I. Resistivity of poly-silicon as a function of the Figure 2. Majority carrier and trap concentrations as afunction
annealing temperature. of the annealing temperature. ( • ) holes, ( EI) trap level at Er
'" Ev + 0.33 eV, ( 0 ) trap level at Er '" Ev + 0.55 eV.

675
 For the investigations of the distribution of impurities in NiSi2 is very similar to the one of silicon the diffraction patterns
the vicinity of the grain boundary techniques have to be are almost indistinguishable. Differences only occur for
employed with a high sensitivity and a good spatial resolution. reflections which are forbidden in the diamond cubic lattice but
Cobalt has been selected because a convenient radioactive are allowed in the CaF2 structure.
isotope is available which allows tracer and MoBbauer In general, random grain boundaries do not seem to favor
measurements even at very low concelltrations of about 1012 cm- the formation of nickel precipitates in the boundary plane as it is
3. Manganese is a slower diffusing element and can be kept in observed for copper, and precipitation appears to occur ma,inly in
solution after cooling, where it occupies interstitial sites and the bulk. In some cases, however, precipitate nucleation at grain
introduces deep trap levels which can be detected by DLTS. Both boundaries can be observed but the platelets still prefer to lie
techniques require however particular grain boundary and parallel to the (111) plane and thus extend into the bulk. A
specimen geometries which were only available in silicon different situation occurs for coherent twin boundaries with a
bicrystals. In both cases bicrystals containing a low energy (111) interface plane. In this case the NiSi 2 habitus plane
(coincidence) grain boundary (E9 and E13) have been used. coincides with the grain boundary plane and precipitate formation
parallel to the interface becomes possible. In fact a careful
4.1 Copper analysis of the frequent thin « 10 nm) twin lamellae in the
The main experimental results for copper which have multicrystalline silicon has shown that nickel seems to
been obtained by TEM investigations in this and previous studies accumulate preferentially along these twin boundaries in form of
/l5,16/can be summarized in the following way. Copper-silicide a very thin (only a few atomic layers thick) silicide layer. Since
(11" -Cu3Si phase) precipitates preferentially at grain boundaries they remain invisible in the conventional TEM pictures unless
and dislocations during cooling. A typical feature is that the they reach a certain thickness they can only be detected by their
precipitation occurs at grain boundaries in colonies which contain weak extra spots in the diffraction patterns. The platelets along
up to several hundred precipitates. For slow cooling rates (5 K / twin lamellae are frequently very large and extend through the
s) the colonies are about 100 11m apart and have a mean diameter whole TEM foil and it may be that the major fraction of the nickel
of about 1 - 311m. The average density of precipitates is about precipitates at coherent twin boundaries.
109 - 1010 cm-2• In addition to this result it is also observed that In addition to platelike precipitates also three dimensional
the type of the grain boundary has an influence on the precipitates occur. They are surrounded by a high density of
distribution and size of the precipitates. For instance for slow dislocations which are caused by stresses in the vicinity of the
cooling rates no precipitates are found in coherent twin particle during cooling and are due the different thermal
boundaries. expansion coefficients of silicon and NiSi2.
In quenched specimens ('" 500 K/s) the distribution of
precipitates is different, the precipitates are smaller and 4.3 Cobalt
distributed more evenly. For a general type of grain boundary The distribution of the radioactive cobalt (57Co) in the
both colonies and single precipitates form simultaneously. The bulk of a bicrystal and near the grain boundary has been studied
colonies lie in the grain boundary plane and have an oval shape by measuriIlg the intensity of the emitted radiation after each step
with a high density of precipitates inside. The mean distance of of the successive removal of thin layers parallel to the grain
the colonies is about 30 11m and the precipitate size 10 nm which boundary. MoBbauer spectra y.'ere recorded at different positions
also yields on average a density of the precipitates of about 109 - along the concentration profile. The results show that after
1010 cm-2. A larger fraction of copper precipitates now also at cooling the cobalt in the volume is predominantly present in form
bulk dislocations and in form of colonies in the bulk. of cobalt-boron pairs, whereas the precipitated plate-like cobalt-
silicide. (CoSi2) is located at or near the grain boundary. The
4.2 Nickel results of the quantitative analysis give an increase of the grain
The precipitation behavior of nickel is quite different from boundary concentration CaB (per unit area) with respect to the
the previous case. It is usually difficult to find nickel precipitates cobalt bulk concentration Co2!3 (per unit area) by a factor of 50
at all in a TEM specimen because only very few but rather large to 100. The absolute grain boundary concentrations were about
precipitates nucleate. The majority of precipitates have the shape CaB'" 10 10 cm-2 for a bulk concentration of 10 12 cm- 3. The
of thin platelets which lie on (111) planes. In some cases a results of the tracer measurements /14/ are shown in fig. 3 which
thickness of up to about 12 nm has been observed, but previous depicts the concentration profile C(x) as a function of the
results obtained by HREM at bulk precipitates also show that distance from the grain boundary. A main result is that the width
they can be only a few atomic layers thick /12/. These planar of the depleted zone is about 50 - 100 11m on either side of the
defects are similar in appearance to thin twin lamellae which are boundary and the amount of missing cobalt is accumulated at the
also frequently observed in this material and one has to confirm grain boundary. The depletion width depends on the cooling time
the nickel silicide phase in each case. Since the lattice constant of and is thus a function of the cooling rate.

NT[crrr3]
c(x) I c 0 Mn in Sil icon 975 Dc
*10 14
1.00
0.75
1.5 • •
1.0
0.50
0.5
0.25

50 100 150 200 [J.lIll] 50 100 150 200 [11 m]

Figure 3. Tracer measurements of the cobalt concentration as a Figure 4. DLTS measurements of the manganese concentration
function o/the distance/rom the grain boundary plane (at x = 0). as a function of the distance from the grain boundary plane (at x
Cooling rate after diffusion about 300 Kls. = 0). Cooling rate after diffusion about 300 Kls.

676
4.4. Manganese whereas random grain boundaries are less attractive for nickel
Contrary to the previous metal-silicon systems manganese compared to copper.
is an element which has a considerable lower diffusion The numerical results for cobalt and manganese yield that
coefficient and solubility (table 1). In addition, manganese forms the precipitation kinetics determines the width of the depletion
in equilibrium with silicon the MnSi2 phase with a rather large region. The cooling rate thus becomes an important factor in the
lattice mismatch. These factors may be responsible that a larger development of the depletion region. A model which can explain
fraction of the interstitially dissolved atoms at high temperature some of the observed features has been discussed in a previous
can be kept in solution after cooling. Previous investigations paper /17/. It takes into account the participation of silicon-self
have shown that interstitially dissolved atoms introduce several interstitials which are injected into the lattice during precipitation
trap level which represent different charge states of the same of the silicides. A consequence is that the process is controlled
species. In the n-type silicon used here one can observe two by the production and diffusion of intrinsic point defects rather
levels Mnr /0 at about ET '" 0.11 e V and Mnio/+ at about ET '" than by the (usually faster) diffusion of the metals.
0.42 eV below the conduction band edge. These trap levels have The DLTS results show that changes in the electrical
been used for the quantitative DL TS measurements. The properties of the material also occur in the bulk region. One main
measurements were performed on specimens where the grain result is the occurrence of deep (near-midgap) levels at ET2 '"
boundary intersected the surface at a very low angle « 1.5°). 0.55 eV with a concentration maximum of 5*10 15 cm- 3 at about
This geometry allowed to measure the manganese concentration 7000C. A quantitative comparison suggests that these defects
in the vicinity of the grain boundary with a spatial resolution of may compensate the majority holes and lead to the observed
about 10 11m. minimum in the carrier concentration. The second group of levels
Diffusion experiment were performed at 975°C which around En '" 0.30 eV has also been observed in other low purity
yielded a maximum bulk concentration of 1.4*10 14 cm- 3, about poly-silicon /18/. These levels disappear around 500 °C and have
93% of the maximum solubility for this temperature. The a lower concentration « 1.4*1015 cm- 3) compared to the other
specimens were cooled down slowly (4 KJs) after the diffusion. group. Since these levels consist of several lines an analysis in
The results as a function of the distance from the grain boundary terms of impurity or dislocation induced level requires more
are depicted in fig. 4 and show again that a depleted zone with a systematic investigations.
width of about 50 11m occurs. Assuming that the atoms missing It is evident that such a high concentration of deep level
in the depleted zone are accumulated at the grain boundary one defects will reduce the lifetime of minority carriers considerably
can estimate an enrichment of manganese atoms at the grain and thus the efficiency of the solar cell. The lowest
boundary by about a factor of 140. This correspond with the concentrations « 5*10 14 cm- 3) of these defects are obtained
results for the cobalt accumulation. after annealing at 1000 DC, which indicates that it may be
possible to improve the material by an annealing treatment.

5. DISCUSSION
6. REFERENCES
A comprehensive discussion of the various experimental
results has been given elsewhere /13-17/ and only some general 1) M. Aucoutourier, A. Broniatowski, A. Chari, J.L.
aspects which may be relevant for the performance as a solar cell Maurice; Polycrystalline Semiconductors, (ed. R.I.
material shall be discussed here. Although the results are still Moller et al.), Springer Proc. in Physics 35, 64 (1989)
incomplete and restricted to a few elements they show that in 2) J.L. Maurice, C. Colliex, Appl. Phys. Lett. 55, 241
multicrystalline silicon grain boundaries can have a far reaching (1989)
influence on the bulk properties. Considering the previous 3) A. Broniatowski, Phys. Rev. Lett. 62, 3074 (1989)
experiments one can distinguish three regions with different 4) E. Weber; Appl. Phys. A30, 1 (1983)
electrical properties: 5) M. Seibt, Proc. 16th Intern. Conf. on Defects in
(1) The grain boundary which can be considered essentially a Semiconductors 1990, in press
two dimensional defect with a width of a few.atomic distances 6) H. Wendt, H. Cerva, V. Lehmann and W. Palmer,
only. The trap levels which can be introduced by this lattice J.Appl. Phys. 65, 2402 (1989)
defect enhance the recombination and/or can lead to a permanent 7) G. Das: J. Appl. Phys. 44, p. 4459, 1973
grain boundary charge and a potential barrier which hampers the 8) E. Nes and J. K. Solberg; J. Appl. Phys. 44, p. 486,
flow of charge carriers in the crystal. Heavily contaminated grain 1973
boundaries may also lead to an enhanced electrical conductivity 9) K. Ryoo, R. Drosd, and W. Wood; J. Appl. Phys. 63,
along the grain boundary plane. 4440, 1988
(2) A denuded zone around the grain boundaries which 10) M. Seibt and K. Graff: MRS Symp. Proc. 104, 215
corresponds approximately to the region where transition (1988)
elements are depleted (about 50 - 100 11m on either side). Inside 11) M. Seibt and K. Graff: 1. Appl. Phys. 63, 4444,1988
the denuded region the volume is also (partly) depleted from 12) M. Seibt and W. Schroter, Phil. Mag. A 59, 337 (1989)
dislocations and other impurities such as oxygen or carbon as 13) T. Tiitken, W. Schroter, H.I. Moller; MRS Symp. Proc.
well. The electrical properties are determined by the remaining 122, 109 (1988)
bulk defects and it can be frequently observed in local resistivity 14) T. Tiitken, W. Schroter, H.I. Moller; Polycrystalline
measurements that the resistivity near grain boundaries within a Semiconductors, (ed. H.I. Moller et al.), Springer
distance of about 100 11m is increased. Proceedings in Physics 35, 108 (1989)
(3) The bulk region inside of grains where the electrical 15) U. Iendrich and H.I. Moller, MRS Symposia Proc. Vol.
properties are determined by bulk defects such as impurities, 163,579 (1990)
dislocations or precipitates. 16) U. Iendrich and H.I. Moller, 1. de Physique 51 - Cl, 197
It is evident from the preceding discussion that the (1990)
influence of the grain boundaries (and their surrounding depleted 17) H.I. Moller, U. Iendrich, Lan Huang, A. Foitzik; MRS
regions) on the overall properties of the material can vary Symp. Proc. (1991), in press
considerably. The results for copper and nickel show that the 18) V. Schlosser, 8th E.C. Photovoltaic Solar Energy
grain boundary structure has a strong influence on the pre- Conference, Italy, 1244 (1988)
cipitation behavior. For instance, coherent twin boundaries offer
no favorable nucleation sites for copper precipitates under slow
cooling conditions but are preferred by the nickel-siIicides,

677
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

IkJm AND ACCEPTOR NEUTRALIZATI~ IN K1.TICRYSTALlINE SILI~. (*)

D. Margadonna, F. Ferrazza, R. Peruzzi

Italsolar S.p.A., Via A. D'Andrea 6, 00048 Nettuno (RM},Italy
S. Pizzini, C. Acerboni, L. Tarchini, Wei XiWen(**)
Universit di Milano, Via Golgi 19, 20133 Milano, Italy
A. De Lillo
E.N.E.A. CRE Casaccia, 00060 S. Maria di Galeria (RM), Italy

ABSTRACT

In order to verify the convenience of employing low cost feedstock for realizing high performance
polycrystalline material for solar cells, two different studies were undertaken; first, the possible
influence of B-P compensation and microstructure on minority carrier diffusion lengths and, second,
the presence of B-O deep centres and their dependence on impurity concentration were investigated.
Experimental results show that material with excellent features, approaching those of single
Grystal, is produced with the directional solidification method, using both compensated and
uncompensated feedstock, thus indicating that grain boundaries and other extended defects have
little or negligible influence on minority carrier lifetime. Results in CZ material show the
presence of a dopant dependent recombination centre related to oxygen content, of still unknown
nature which needs closer investigation.

1. INTRODUCTION

The PV market is recognized to be strongly This model remains however largely speculative,
expanding, its annual increase rate being of about because of the limited amount of experimental
25%. Some 30% of the sales is already covered by evidence to support it.
polycrystalline cells, with a rather strong Nevertheless, if proved, the importance of such
increasing trend for the next years. hypothesis is readily seen for the possibility of
Interest for polycrystalline material is at last employing a low cost feedstock for high performance
building up in cell manufacturers owing to solar cells.
improved material features which allow
polycrystalline cells to appr~ach CZ based cells
in efficiency; moreover the cost to performance 2. EXPERIMENTAL
ratio is already acceptable. Other reasons that Two different experiments were carried out at the
justify such interest call for reliability and same time.
long term performance problems which amorphous The first aimed to provide evidence of the possible
silicon and thin film devices still suffer. influence of B-P compensation and microstructure on
Bearing in mind all these considerations, we have the minority carrier diffusion length; the second,
already demonstrated that the Bridgman-Stockbarger to demonstrate the presence of B-O deep centres and
technique shows many advantages over other study their dependence on single impurity
presently employed growth techniques for large concentration. For the first aim 12 Kg.
size polycrystalline silicon ingots (1,2). In fact multicrystalline silicon ingots were grown with the
not only it allows strictly directional Italsolar solidification technique, the details of
solidification, but also definitely better which are described elsewhere {1,2}.
microstructure. Thick (2 rom) 10 X 10 cm} wafers were cut from the
A further advantage, which makes this process ingot central regions to be used for the
undoubtely superior, is the possibility of determination of the integrated diffusion length Ld
employing low grade feedstock (i.e. solar grade on the whole wafer area (4) and the etch pit
silicon) because of the excellent impurity density according to the ASTM standard, spec. F
segregation ability, due to the strongly 47-88.
directional nature of the process, as is the case The properties of polycrystalline samples and the
of CZ and FZ silicon growth.(3} results of measurements are reported in Table I.
In the mainframe of a work devoted to the
improvement of such techinque, we were able to Table I
confirm earlier anticipations (4) showing that
suitable mixtures of p- and n-type silicon may Ingot Res( j*cm} Na (lE16) Nd (lE16) CR Ld(f}
represent an alternative, low cost feedstock for 207 0.14 25.0 1 88
PV applications, because of the occurrence of B-P 210 2.2 0.65 1 155
compensation processes. The latter occur up to a 211 2.2 0.65 1 131
critical amount of total acceptor content and act 223 1.0 1. 50 1 148
positively on the minority carrier diffusion 224 0.8 2.50 1 120
length, which is, in CZ material comparable or 225 1.8 0.80 1 150
even larger than that of uncompensated material of 228 7.3 0.85 0.65 8 190
the same free carrier concentration.(5} 230 4.2 0.80 0.48 4 135
The authors of ref. (5) proposed a model to 243 0.33 25.0 8.0 2 80
explain such behaviour, based on the hypothesis 244 0.36 25.0 15.0 4 76
that B-O clustering occurring in boron doped 245 0.42 25.0 18.0 6 97
silicon is inhibited by B-P complex formation (B-O 246 0.50 25.0 20.0 10 110
clusters acting as recombination centres). 252 2.0 1.5 0.50 2 140
253 2.5 1.5 0.90 4 160

678
Some samples are donor compensated with different 3. DISCUSSION
compensation ratios CR = (Na+Nd)/(Na-Nd) and it is Experimental results show that the minority carrier
demonstrated that such compensation has little diffusion length depends on the excess acceptor
influence on the diffusion length values. Infact concentration (Na-Nd). On the other hand, in the
(see fig. 1) diffusion length depends on the excess (Na-Nd) range investigated (lE15 ~ 2E17), single
acceptor concentration (Na-Nd), as already shown crystal exhibits a similar behaviour, as already
for single crystal samples (5,6). Experimental Ld reported in literature (5,6). In fact the diffusion
data show a dispersion which is probably related to length decrease (with increasing (Na-Nd)) is caused
the dislocation density. by a decrease in the minority carrier lifetime,
which is always below the corresponding Auger
Ld luol
limit. (7) Such behaviour has been explained in the
coo
,8
past (7,8,9) on FZ material with the presence of
vacancy-like defects reducing minority carrier
lifetime. In addition, no evidence of the presence
150 of deep level showed up from DLTS measurements.
On the contrary, our measurements on CZ material
show that a deep level, whose concentration is
100 proportional to boron and oxygen content is present
in all examined samples. Such recombination centre,
responsible for lifetime reduction, is of still
unknown nature, but its concentration seems to be
proportional to both boron and oxygen concentration
via the following expression:
NI"'" Ilt/ c. ' JI

10' 16 10' 17 10' 18 N (0.53 eV) =N x N

T 0 B

as expected for the case of a B-O cluster.

The polycristalline material we investigated shows

Single crystal CZ and FZ samples with variable a similar behaviour in Ld, but data are somewhat
boron to oxygen ratio were provided for the second scattered owing to the dependance of minority
aim by Wacker Chemitronic, and used to perform DLTS carrier diffusion length on the EPD. In fact, in
measurements with a Polaron S46000 modular system . agreement with Sopori and Pizzini(10,6), in the
Results are reported in table II and we point out examined experimental EPD value range (2 ~ 3.5E4
that only a single deep level at 0.53 eV was EPD/cm2), the diffusion length varies dramatically
observed on all CZ crystal samples, in contrast for any small change in EPD density. Moreover,
with the two observed by the authors of ref. (5). samples with higher Ld values, at fixed (Na-Nd),
generally have lower EPD values. Such strong
dependance on microstructure masks any possible
Table II effect of the creation of neutral B-P complexes
competing with the recombining B-O centres.
Ingot Res(Q*cm) Na (lE16) 0(cm- 3) C(cm-3) Nt(cm-3) In any case it clearly appears from the analysis of
experimental data (fig. 1) that high compensation
CZ /23 32.0 0.04 7 . 6E17 1.lE16 3.3E12 ratio values do not affect the electronic
CZ /4 18,0 0.08 6.3E17 <0.5E16 1. 2E13 properties of the material.
CZ /23 2.0 0.70 7.6E17 <0.5E16 7.8E13
CZ /50 0.2 11. 00 n.d. n.d. 5.0E1S
CZ /12 0.01 850 . 00 n.d, n,d. n.d, 4. CONCLUSIONS
The first conclusion is that in a "good"
FZ /20 1.S 0.95 8.2E16 3.0E16 2.SE12 polycrystalline material grain boundaries and other
extended defects have little or negligible
The results of DLTS experiment$, reported in fig,2, influence on the lifetime of minority carriers. The
show a linear relationship between the 0.53 eV deep polycrystalline material studied in the present
level and boron concentrations, at fixed oxygen work shows such features in as grown conditions,
concentration. its behaviour being comparable to single crystal
Furthermore, on FZ sample such deep level seems silicon,
also to depend on oxygen concentration at fixed The second one is that material performances at
boron content. In all samples the density of deep high boron concentrations are controlled by a deep
level is sistematically about two orders of recombination centre, whose concentration is
magnitude lower than boron concentration. determined by dopants and oxygen content.
It appears finally that even at high R values donor
10' 16
compensation with phosphorous does not affect the
[B'OI lco- J)
electronic properties of the material. There is
experimental evidence indicating how the formation
of B-P complexes can compete with B-O recombinating
10'15 centres. Further study are needed to definitely
clarify this point.
In any case the "free" oxygen concentration must be
kept as low as possible.

10'13 5. ACKNOWLEDGEMENTS

[81 (at/c.'31 One of the authors (S.Pizzini) is sincerely

indebted to Dr. W. Zuhlener and Dr, J. Dietl of
10'15 10'16 Wacker Chemitronic for the delivery of the single
crystal samples used in DLTS experiments.

679
6. REFERENCES

(l)M. Beghi et al. "Polycrystalline and amorphous

semiconductors" P.Pinard, S.Kalbitzer EDTR,
Les Editions de Physique, (1984) pag. 181
(2}M. Gasparini et al. Proc. XVI IEEE PVSC, S.Diego
(USA), (1982) pag.74
(3}D. Margadonna et al.Proc. 8-th EPSEC, Florence
(IT), (1988) pag.1254
(4}D. Margadonna et al.Proc. 9-th EPSEC, Friburg
(FRG), (1989) pag.750
(5)W. Kruehler et al. Proc. 8-th EPSEC, Florence
(IT), (1988) pag.1181
(6}S. Pizzini et al. J. Electrochem. Soc. 131,
(1984), pag.2128
(7}D. Huber et al. Proc. 5-th Int. Symp. on
Silicon Mat sc.and technology, Boston (USA)
(1986), pag.45
(8}P.T. Landsberg et al J.Appl. Phys. 56 ,(1984)
pag. 1696
(9}J.G. Fossum et al. Solid State Electronics,
25 , (1982) pag.741
(10)B.L. Sopori Proc. XX IEEE PVSC, Las Vegas
(USA), (1988)

(*) Work partially supported by CEC Contract

n. JOUR/0036/C (GDF)
(**}Permanent adress: Physics Dept., Dalian
Univ. of Technology (China)

680
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

LIMITING FACTORS IN THE EMITTER AND BASE REGION OF DIFFUSED SOLAR CELLS
MADE FROM CRYSTALLINE LOWER PURITY SILICON

Viktor Schlosser
Institut fUr Festkorperphysik der Universitiit Wien and
Ludwig Boltzmann Institut fUr Festkorperphysik
Strudlhofgasse 4
A-lOOO Wien (Austria)

ABSTRACf. On different samples of lower purity crystalline p-silicon n+p-junctions were prepared by
a conventional phosphorous diffusion step. From the measurement of the short circuit current density
and the open circuit voltage under different ilIimination conditions the current losses in the solar
cells due to diffusion across the pn-junction and the recombination within the space charge region
were studied as a function of the temperature. The temperature dependence of the electron diffusion
length was determined from spectral response measurements under monochromatic ilIiumination. Point
defects which are due to iron were detected with capacitive deep level transient spectroscopy in all
our samples. A comparison between the recombination rate within the depletion region and the minori-
ty carrier diffusion length in the base which was in the range between 5 ~ and 70 lim show that both
are essentially governed by the same impurity distribution. The diffusion current was found to be
limited by Auger recombination in the emitter which causes a shrinkage of the mobility gap up to
0.18 eV thus restricting the internal losses due to diffusion to current densities of more than
lxlO·7 Am-=2.

1. INTRODUCflON ferent samples of lower purity silicon. Their properties

Crystalline silicon commonly used for solar cell are summarized in table I. For comparison reasons devices
manufacturing is Boron doped p-type silicon in the resis- were also prepared from high purity polycrystalline samp-
tivity range from 0.2 Oem to 10 Oem. During the last les as delivered from Wacker Heliotronic under the trade
years large effords were made to obtain silicon suitable mark SILSO which were p-type. The average resistivity was
for solar cell fabrication from a metallurgical meltstock about 5 Oem. In addition single crystal substrates cut
by different approaches [ 1,2 ]. The material we used in from Czochralsky pulled high purity p-silicon - the re-
our experiments was prepared by such an unconventional sistivity ranges between 6 ilcm and 9 Oem - were used. The
method. The resistivity ranges between 0.03 Oem and 0.5 phosphorous diffusion was done by first coating the front
Oem and the samples were p-type silicon with a rather side of our samples with a phosphorous glass. Then the
large amount of residual impurities. Purpose of this work diffuion was carried out in a furnace at 91SoC for 30 min
was to exemine the light induced photocurrent and the corresponding to the conditions of a single step diffu-
internal current losses due to diffusion across the sion from an unlimited source. The original active cell
pn-junction and due to the recombination which takes area was between 1 cm2 and 4 cm2 • For the space charge
place within the space charge region of the pn-junction. spectroscopy and the determination of the current losses
The potential for industrial applications was kept in as a function of the temperature small sized mesa struc-
mind which requires a low cost production step for the tures were prepared out from these solar cells. This was
formation of the large area pn-junction. Therefore a done in the case of DLTS studies in order to use a capa-
simple phosphorous diffusion step from a glassy layer citance bridge from Hewlett Packard which is limited to
which was deposited onto our samples was carried out in capacitors of less than 2 nF. In the case of the i(V)
order to obtain a shallow step junction profile. Electri- characteristics the superimposing of the current mecha-
cally active point defects were characterized by means of nisms due to recombination and diffusion by the ohmic
deep level transient spectroscopy [ 3 J. The conclusions contributions due to shunting and series resistance ef-
we draw from the temperature resolved determination of fects were minimized. In order to further reduce ohmic
the three current components were made with respect to contributions the i(V) curves were obtained from short
the concentrations of point defects which have been de- circuit current, i , versus open circuit voltage V
sc ' oc'
tected in our devices. measurements at different illumination levels. The advan-
tage of this method is that the series resistance, R can
s
be neglected since for open circuit conditions the cur-
2. EXPERIMENTAL rent and hence the product i x R iszero. The relation
s
Diffused solar cells were prepared from three dif- between isc and Vocthen is given by the internal current

681
 TABLE I: Properties of lower purity silicon we used in our experiment
Producer crystal Systems Siemens Munich
Crystal growth HEM 2xCZ + 2xFZ 2xCZ I
Crystal perfection polycrystalline single crystal
Resistivity [ncm] .13 - .25 .12 - .18 .025 I
Type/orientation P/<100>1 ) p/<l11>
e-Diffusion length [/lm] 7 - 15 30 - 50 2 - 5 I
1) Single crystal section have been examined
HEM tleat Exchanger Method, CZ •.. Czochralski pulling
FZ Float zone processing

losses as expressed by Eq. (1). 4

area: 0.16 cm 2
-i
sc
= idiode = i0exp(qVex!InkB1) - i
0
(I) T = 289.4 K
2
i diode •••diode loss current
diffusion
n...diode ideality factor
"<
T... temperature
E
........ 0
<.>
kB ... Boltzmann's constant '"

The diode idealty factor, n, has a value of about 2 c: -2

for the recombination current within the space charge slope .49- 1
region. The scaling factor, i 0' is proportional to the
-4
intrinsic carrier concentration, ni, the depletion layer ~recombinatio n current
width, W, and inverse proportional to the recombination
lifetime of holes and electrons, 'tsc' within the space - 6~~~~~~~~~~~~~~~
charge region. The second current loss mechanism due to 8 10 12 14 16 18 20 22 24
diffusion also obays an exponential law. The diode idea- (QV at Ike T) [ - )
lity factor however is approximately I and the scaling
factor io is proportional to ni 2 which is responsible for Fig. I : A plot of the values of the short circuit current
the extremely large dependence on the temperature. In and the open circuit voltage under different levels
of the incident light power.
order not to confuse the scaling factor of the two mecha-
nism furtheron the factor describing the recombination 80 0

"
term will be denoted as i 01 ' whereas the diffusion compo-
nent will be described by i02' A typical result of a
measured isc(VoJ-curve is shown in fig.l. In the case of 40 -40
the diffusion current the two sides, emitter and base .....
~
region respectively, contributes differently to the the
total diffusion current. The Auger recombination process '"E
I 0 - 80
u C>
which takes place in the emitter due to the excess of ...... ........
phosphor atoms causes a shrinkage of the mobility gap, '" l>
n
5 - 40 - 120 '3
I
109. Since the intrinsic carrier concentration depends
c
'"
exponentially on the ratio of the activation energy ex-
pressed by, 109, and the thermal energy, kBT, an Arrhenius -80 - 160
plot of i02versus kBT will be significantely different
when the bandgap is reduced by M. In fig.2 the result is
\
shown for the diffusion current derived from measurements -120 -200
on a device made from lower purity silicon . The triang-
0 100 200
(keTl- 1 [eV - 1 )
les are representing the measured values. For comparison
the second power of the intrinsic carrier concentration
is displayed as dashed line. The differences in the tem-
Fig.2: Arrhenius plot of the diffusion current density
perature dependence easily can be seen. Since the Arrhe- for a sample of lower purity silicon.

682
nius representation does not display the dependency as a The minority carrier diffusion length was derived
strait line themagnitude of the bandgap narrowing, till, from spcctral response measurements under monochromatic
was derived from calculating the temperature dependence illumination. The experimental set up is described else-
of the diffusion current with varying values of till. The ~here [ 3 J. Leading to values of the electron diffusion
result is represented by the solid line and was achieved length, Ln' given in table I for lower purity silicon.
for till=O.l8 eV. The same data evaluation was carried out The polycrystalline samples were exhibiting values for Ln
for pn-junctions made on polycrystalline and on Czoch- between 30 ~m and I00 ~m. In the case of the high purity
ralski pulled high purity p-silicon. The results are samples the photoresponse was limited by the sample
summarizied in table II. At room temperature the contri- thickness of about 350 ~m. Therefore the diffusion length
butions from the base region and the emitter were calcu- only can be estimated to be larger than half the sample
lated and shown together with the experimental data in thickness. By numerical solution of Poisson's Equation
fig.3. for the doping profiles present in our samples we could
calculate the space charge layer width and as a conse-
quence derive 'ts/rom j02 ( = i02 normalized to the area
TABLE II: Mobility gap narrowing due to Auger re-
combination in the emitter of the pn-junction plane). In all devices which we have
characterized the value for 'tsc suits agreeably well to
Material resistivity [Qcm] aE leV]
the determined electron diffusion length in the base of
lower purity S1 0.2 0.18
our solar cells. This is a strong indication that recom-
polycrystalline S1 4.0 0.15
bination within the depletion region as well as the mino-
high purity Si 6.0 0.10
rity carrier lifetime in the base region is governed by
the same impurity distribution. In order to find out
N
which point defect dominates the recombination deep level
I

• transient spectra were taken from our samples.In the case

E
~
of lower purity silicon the results are given in table
N
III. In order to find out a correlation between the mea-
.~
l TOlAL DIFFUSION CURRENT
:>- /' sured electron diffusion length they were plotted as a
t--
......
en
z
/' function of the point defect concentration. In the case
of the hole trap denoted with H6 in table III a good
~ 10-8 /
t--
z / - FRAC110N FROM THE BASE correlation was found as can be seen in figA. Together
UJ
a: with the hole trap H always two electron levels, E and
a: 6 1
~
u / E2 were seen. Due to the numerous levels however we could
z
~ 10- 9 not resolve the peak height of our capacitance spectra
en T 29SK
~
lL.
which is determined by the concentration of the point
lL.
..... defect within the space charge region .
D

.1 1.0 10
RESISTIVITY [ 0 em]

3. DISCUSSION
Fig.3: Measured (circles) and calculated dependence of
the diffusion current density on the resistivity of
p - silicon. Both, DLTS, as well as photocurrent measurements

TABLE III: Energy levels detected by means of DLTS

Trap ~E reV) CT [cm 2 ] Remarks

H .40 5X10- 15 due to iron
1
H .43 lX10- 18 interstitial iron
2
H
3
.10 1x10- 23 -
H
4
.19 lX10- 21 -
H .28 1X10- 2o preparation induced
5
H
6
.32 2X10- 19 -
H
7
.50 5x10- 15 -

E1 .29 2X10- 17 correlated with H

6
E2 .54 - .61 8X10- 17 - 2X10- 13 correlated with H
6

683
 tributes a small fraction - about 30 per cent in the case
10Q .--------------------------.
of high purity silicon - to the total diffusion current
E density. The diffusion current therefore is mainly deter-
:s..
mined by (i) a significant mobility gap shrinkage and
:I:
~ (ii) hole recombination due to Auger processes in the
to
Z
W highly doped emitter. An estimation of the lower limit of
-.J
Z the loss current density thus is about 1 x 10-' Am-2 for
:=:
Vl
10 solar cells of p-type silicon in the resistivity range
:::>
LI...
LI... between 0.1 ilcm and 10 ilcm. The influence of the recombi-
......
C nation current density within the space charge region as
a
II: well as the minority carrier diffusion length in the base
~
U
LU is dominated by residual point defects. The steep
-.J
IJ.J pn-junction profile however is leading to a very narrow
depletion layer width, W, which is typically less than
10 15 0.1 l-1m at zero bias voltage. Since 't was found to be
H6 - CONCENTRATION [ cm-3 1 RC
around 1 l-1sec the recombination current density in our
devices was in the range of about 2xlO-3 Am- 2 •
Fig.4: The dependence of the electron diffusion length on
the concentration of the hole trap H6 for different
samples of lower purity silicon.
4. CONCLUSIONS
indicate that in all the samples of lower purity silicon
we have characterized the minority carrier lifetime is
From the studies we have made on the photocurrent
limited by the same kind of point defect which causes the
and the current losses in solar cells of lower purity
hole trap H6. Although the origin of this deep lying
crystalline silicon we conclude that there is only a
level - 0.32eV - is uncertain it is probably due diffu-
little chance to obtain high output solar cells made by a
sion and redistribution of interstitial iron during the
low cost high throughput technology.
high temperature pn - junction formation [ 3 ]. DLTS
studies on MIS structures which were heat treated at
different temperatures show that the concentration of HI
5. REFERENCES
which is due to interstitial iron is present in concen-
trations which far exeeds the solid solubility of iron
1. C.P. Khattak, F. Schmid, P.H. Robinson and R.V. -
and is about 100 times greater than the concentrations of
D'Aiello, Proceedings of the 16th IEEE PV-
HI observed in the pn-junction devices. This can be ex- Specialists Conference, IEEE, New York, 1982, p.128.
plained by an unintentional gettering step. However the 2. A. Ludsteck, H.J. Fenzl, Proceedings of the 5th EC
same way the concentration of HI increases near the sur- Photovoltaic Solarenergy Conference, D. Reidl, Dord-
recht, 1983, p.946.
face the the concentration of the point defect H6 incre-
3. V. Schlosser, Proceedings of the 9th EC Photovoltaic
ases within the space charge region of MIS and Solarenergy Conference, Kluwer Academic Publishers,
pn-junction devices. The electron diffusion length in our Dortrecht, 1988, p.1244.
samples of lower purity silicon is typically in the range
of 1/10 of the wafer thickness thus causing a significant
reduction of the output efficiency. As a simple conse-
quence the ratio of the electron diffusion length and the
thickness should become larger by either increasing the
minority carrier diffusion length or by decreasing the
solar cell thickness. In the fITst case a higher purifi-
cation of silicon is necessary. The latter case will make
it necessary to improve optical light trapping in the
solar cell by multiple light path methods in order to
absorb the incident light to more than 90 per cent in the
useable spectral range. From the results we made with a
simple phosphorous diffusion step for the formation of
the pn - junction we conclude that the current losses due
to diffusion are nearly exclusively governed by the pro-
perties of the emitter whereas the base region only con-

684
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

TRANSPORT PROPERTIES IN THE EMITTER OF THE SILICON SOLAR CELLS: A

COMPARISON BETWEEN A NEW ANALYTICAL MODEL AND THE EXPERIMENTAL
RESULTS.

L. Abenante, G. Arabito, S. La Motta, L. Pirozzi

ENEA Casaccia, s.p. Anguillarese, 301 ROME
G. Di Francia, M. Garozzo
ENEA CRIF,Via Vecchio Macello, Portici, Napoli

ABSTRACT.A new semi-analitycal model for the transport of minority

carriers in heavily doped emitters has been developed and the
results are compared with those obtained by numerical models.
various samples have been processed under different phosphorus
diffusion conditions and SIMS profiles have been used as input data
for the computer analysis. Experimental internal quantum yield has
been compared with the model results.

1. INTRODUCTION In this paper we present the

results of a new analytical model,
The modelling of the emitter is whose mathematical details will be
considered one of the most difficult presented elsewhere. A comparison
topics in silicon solar cell design between these results and preliminary
(1). It requires, in fact, the experimental data is reported.
solution of the transport equation for
minority carriers, taking into account
the strong dependence of the transport 2.THE ANALYTICAL MODEL
parameters on the position inside the
emitter. This dependence arises from
the carachteristics of the solid state An analytical solution to the
diffusion of phosphorus in silicon and trensport equation can be found when a
in particular from the high doping relationship between electric field
effects. As a consequence the involved and minority carrier diffusion length
mathematical functions sharply vary can be established, as it has been
together with the electrical shown by Selvakumar (7), even if his
characteristics of the emitter itself. solution refers to a very narrow range
Since an analytical solution of of physical conditions.
the transport equation cannot in In our approach the emitter is
general be found, a number of divided into a suitable number of
different mathematical approaches have regions, where a similar relationship
been followed ( see for instance 2-S) . still holds. The so obtained solution
The numerical one starts from a trial applies to the whole emitter, with no
solution and the equation is solved limitations on the kind of
iterativelly, either by means of an recombination or emitter shape.
integral approach or by simple Of course, the expressions for the
approximate expressions. In this case emitter dark saturation current Jo and
the solution is often strongly the point quantum efficiency Qy(x} are
dependent on the starting function, so different in the different regions of
that many steps are needed to the emitter, but the analytical
adeguately define the physical expression is everywhere the same.
situation even if this one slightly For instance, the excess hole density
differs from the initial hypothesis. t.p has always the following
The so obtained solutions generally expression:
better describe transparent or t.pi=Ai sinh (fdx/LPi) +Bi cosh (fdx/LPi)
quasi-transparent emitters or, on the
contrary, "dark" emitters. where the subscript i refers to the
On the other hand, the attempts to i-th emitter region and Lpi is the
find an analytical solution are hole diffusion length in that region.
limited by the strong hypothesis on A complete description of the
the coefficients of the transport transport inside the emitter, with
equation (6). So, for instance, exact particular regard to Jo and Qy, can be
solutions have been obtained when the obtained by applying the continuity
minority carrier recombination is due conditions at the boundaries of each
merely to the Auger effect. emitter region.
The situation is still more In Fig. 1 the simulated Jo is
complicated when the aim of the model reported as a function of the emitter
is to optimize the emitter design, depth when a normal distribution for
because, in this case, a large range the doping profile is considered with
of variation for the emitter depth Xj a surface impurity concentration of
and for the doping profile has to be Ns=10"20 cm"-3 and a surface
considered. recombination velocity of S=10"6 cmls

685
is assumed. The results of our model 3.EXPERIMENTAL
are compared with the numerical
solutions given by euevas-Balbuena and
by Vitale. In order to experimentally test
The calculated emitter quantum our model, solar cells with different
efficiency at A=400 nm is shown in emitter configurations have been
Fig. 2 as a function of the emitter fabricated.
depth, assuming the same doping Doping runs were carried out in an
profile and two different values of S. open tube furnace, using phosphorus
Also in this case the model is oxychloride as dopant reagent on FZ,
compared with the numerical results p-doped (B), 1 ncm silicon under the
given by euevas-Balbuena. The very conditions reported in Tab.I.
good agreement, both for Jo and Qy, After the diffusion process, the
between our results and those obtained emitter surface of some samples of run
by using numerical approaches can be 42A were passivated by thermally
considered as a first important check growing a 120 A thick silicon dioxide
of the assumptions made in developing layer. The thickness and the
our model. stoichiometry of this layer was
For a given emitter configuration, ellipsometrically checked.
the discrepancy between our model and As above mentioned, in order to
the "true" physical situation is performe a reasonable test of our
connected to the artificial model, particular attention has to be
subdivision of the emitter in regions payed on the uniformity of the
and obviously strictly depends on the diffusion along each wafer. An
number of these regions. Neverthless accurate sheet resistance map of each
the model allows to quickly converge sample gives a good estimation of the
towards an exact description: uniformity. Iso-contour map (Fig.3)
referring to the emitter configuration has been performed in each sample and
of Fig. 1, by increasing the number of the maximum inhomogeneity value was
considered regions from two to five about 8%.
the discrepancy is reduced from more 4 cm2 solar cells have been
than 10% to less than 3%. fabricated by using the central part
When the results of a theoretical of the wafer, where the sheet
model have to be compared with the resistance inhomogeneity was less than
performances of a real emitter, 2%. These results, of course, do not
possible errors due to the ensure the same uniformity level for
inhomogeneity have to be carefully both doping level and profile.
taken into account, expecially when Neverthless they have to be considered
average measurements are employed. as a minimum requirement for a
We have used our model for reasonable comparison between
evaluating the uncertainty on Qy and theoretical and experimental data.
Jo arising from a 10% inhomogeneity in SIMS profiles have been then
junction depth and surface carried out on the emitters, in order
concentration. The calculations, to have the exact doping profiles to
performed by assuming a doping profile be used as input data for the model.
approximated with a double normal As expected, the observed doping
distribution (8), show that the same profile respected the typical
percentage variation of Xj and Ns is
observed for Qy and Jo. e~~~V!O~~r:a~eP~~~~~~;~:ti~~ ~~I~~~~O
Different carrier transport models cm"-3 and a junction depth ranging
do not show appreciable discrepancies from 0.5 to 2.0 ~m.
from our results, at least when small Finally, solar cells have been
variations are taken into account. fabricated using evaporated Ti!Pd!Ag
As a consequence, great attention and Al respectively as front and back
has to be payed on the homogeneity contacts.
during the junction diffusion
process,if calculated and experimental
data have to be compared.

Run Tl (ee) T2 (ee) t l (min) t2 (min) 02 (%) POC13 (%) nO

42A 850 850 30 5 3.9 0.12 68

42B 850 920 30 120 3.9 0.12 18

43A 920 920 30 5 1.2 0.12 15

43B 920 920 30 120 1.2 0.12 7

Tab. I. POCl3 diffusion parameters. Tl and T2 are respectively

the diffusion and the drive in temperatures. In all cases
the nitrogen flux was 3 l!min.

686
4.RESULTS AND DISCUSSION diffusion condition.
A very good agreement between
theoretical foreseens and measured
Absolute internal quantum yield of data has been found for the various
the fabricated samples has been devices.
measured in the wavelenght range
350-400 nm, where it is essentially
sensible to the emitter properties. In 6.REFERENCES
particular, QY(A) in this range can be
used for calculating the surface (1) J.A.Del Alamo,R.M.Swanson "The
recombination velocity, as reported in Phisics and Modelling of heavily doped
Ref.9, and this calculation is very emitters";
accurate when the emitter depth is IEEE Trans.Elec.Dev. ~ 1878 (1984)
already known. The comparison between (2) H.C.de Graff,J.W. Slootboom
the experimental and theoretical data and A.Schmitz "The emitter efficiency
has been then performed, by using SIMS of bipolar transistors"
doping profile and S as input data for Solid State Electron.2Q 515 (1977)
the model. Fig. 4 shows the very good (3)E.De Castro, M.Rudan "Integral
agreement obtained for sample 42A with equation solution of minority carrier
and without surface passivation. transport problems in heavily doped
The same figure demonstrates the semiconductors"
effectiveness of the surface IEEE Trans.Elec.Dev. ~ 785 (1984)
passivation process, which reduces S (4) A.Nagakawa "One dimensional
from 5*10 A 5 to 10 A 4 cm/s. This fact is device model of the npn bipolar
also an indirect proof for the absence transistor including heavy doping
of relevant dead layer at the surface. effects"
Surface recombination velocity, Solid State Electron.22 943 (1979)
theoretical and experimental internal (5)H.T.Weawer,R.D.Nasby "Analysis
quantum efficiency at A=350 nm, where of high efficiency silicon solar cell"
the only contribution is given by the IEEE Trans.Elec.Dev. 2a 465 (1981)
emitter, are reported in Tab. II. (6) A.Cuevas,M.A.Balbuena "Review
of analytical models for the study of
highly doped regions of silicon
Sample S (cm/s) Qytheor Qyexp devices"
IEEE Trans.Elec.Dev. ~ 553 (1989)
42A/a 10 A 4 0.64 0.63 (7)C.R.Selvakumar "Simple general
42A/b 5*10 A 5 0.43 0.45 analytical solution to the minority
43B 10 A 6 0.10 0.11 carrier transport in heavily doped
semiconductors"
Table II. Comparison between J.Appl.Phys. ~ 3476 (1984)
experimental and theretical quantum (8)G.Eranna,D.Kakati "Process mod-
efficiency. Sample 42A/a has elling of Phosphorus diffusion in
passivated emitter. Silicon.A new model"
Solid State Tech. 12 116 (1984)
(9) G.Di Francia,M.Garozzo and
The difference between Qytheor and L.Sardi "Non destructive determination
Qyexp is well within the experimental of Xj and Sp in homo-junction solar
uncertainty for each considered cell"
sample. The poor responce of sample 19th IEEE Phot.Spec.Conf. New Orleans
43B is mainly due to the very deep 513 (1987)
emitter (2/.lm) and therefore to the
emitter bulk recombination, rather
than to the high value of S.
In conclusion, taking into account
experimental uncertainty and possible
discrepancies due to sample
inhomogeneities, also the comparison
with the experimental data confirm the
validity of our model.
In particular the effect of sample
inhomogeneity on the comparison
between theoretical and experimental
data is object of a work still in
progress.

5.SUMMARY

The results of a new analytical

model for the transport of minority
carriers in heavily doped emitters
have been reported and compared with
those obtained by other authors.
Theoretical and experimental Quantum
Yield have been compared for various
solar cells fabricated under different

687
 -II
+

-12
.,o
'"
o
.J

- 13
.•- -·--·.'o-
;,-~-- · -1: 2- '-'1 :;;- ' - 2
xJ ( Uln J

Fig.1 Emitter dark saturation current

versus junction depth.Comparison of
our model with Vitale (+) and
Cuevas-Balbuena(*) simulations.
Fig.3 Iso-contour map (2% of relative
variation) performed on a 3' sample
from run 42B.

- - - - _ .._- - ----.-.---
Oy .9

1
~
.8

. /<b)
+
.7
1
I
/ +
>-
C-
.6 I
o

'" .5 ./ I
!:
L
~
-'
c
.4
+
~
1
.....
1.& 2 .3

Fig.2 Calculated internal quantum .2

efficiency at A.=400 nm versus junction
depth for S=1000 cm/s and S=100000 . 1
cm/s. (*) Data by Cuevas-Balbuena.
o ~~_~_~_~~_~_~
350 360 370 380 3g0 i00
Wavelength ( nm )
Fig.4 Q.y.comparison between our model
and experimental data (+) for samples
obtained by run 42A. (a) passivated (b)
no passivation.

688
10TIl EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A FLEXIBLE MEASUREMENT SYSTEM FOR R&D IN INDUSTRY

V. E. Manfnez, M. J. Saenz, J. R. Gutierrez, I.C. Jimeno and E. Perezagua*

Dpto. Automatica, Electr6nica y Telecomunicaciones,

E.T.S.I.I.I.T. de Bilbao. Universidad del Pais Vasco
Alameda de Urquijo sin. 48013 Bilbao. Spain.
* ISOFOTON S. A.. C/Miguel Angel 16. 280lO Madrid. Spain.

ABS1RACT: A solar cell sorting system developed between Isofot6n S.A. and the Universidad del Pais Vasco is
descri~. This syste~, which ~s of an industrial nature, aimed at enabling companies to carry out quality control
wo~, IS, at the same ~e a fleXible and totally programmable system which makes it apt not only for the basic work
required, but also as an m-plant measurement laboratory, capable of doing I-V measurements under a wide range of
temperature and lighting conditions.

l.INTRODUCTION
2.1. Measuring platform and temperature control.
Isofot6n S.A., the Spanish solar cell maker, was founded The cells to be measured are placed on a copper
in 1981. The albedo collector panels and bifacial cells, developed thermostated base of 15Ox150mm, and vacuum-secured. The
in the Universidad Polirecnica de Madrid, are the products which base is machined so that the vacuum is applied in two different
represent the work of this company. These range from nickel- areas; this allows fastening cells which are as big as the base and
metalized bifacial cells of the past to the textured and screen- as small as 5x5mm.
printed metalized cells of today. In 1986, Isofot6n introduced its The base moves in and out of a light-tight compartment,
monofacial technology and at present work is being done to put integrated into a standardized rack, thanks to a pneumatic drive
the final touches to a technology for manufacturing high- system.
efficiency cells. These continuous changes are not possible Inside the compartment the good electrical contact
without a great deal of research and a high analytical capacity in necessary for measurements is provided by the base itself
the technology implemented. (contacts after the cell), as well as two rows of floating contacts
The usual work required by industry in its classification on the illuminated side. Each row consists of five floating
systems is based only on the measurement of currents at a pre- contacts (see figure 2) which allow good adaptation to the
established fixed voltage under stable conditions of temperature metalization bus, thereby minimizing errors in the estimation of
and lighting. Nevertheless, this single measurement does not serial resistance of the cell. Of the ten floating contacts, nine are
enable the internal parameters of the cell to be known, which are used for introducing current into the cell and one for measuring
necessary to control the technology. On the other hand, research the voltage.
teams have systems which are virtually manual, designed for The copper base is maintained at the required temperature
carrying out exhaustive measurements of the cell under test by means of the use of 4 thermoelectrical devices based on the
which allows practically all the internal parameters to be Peltier effect (peltier cells). The set of devices used is capable of
obtained. Work of this type could require one day of work for a extracting or providing the base with up to 200 W of calorific
single cell, it being virtually impossible to process large power, at full performance.
numbers. Temperature control is made by means of supplying the
The aim of this paper is to present an industrial sorter Peltier cells via an electronic regulator and a LM335-type silicon
capable of carrying out more extensive work than that of sorting, sensor, and the system provides a PID-type control.
in order to be able to obtain some type of additional information It is envisaged that the system will be capable of
on the cell being sorted. This requires, in turn, a change in the establishing a temperature variation range of about 2O'C to 80'C,
conception of the system used for making additional with variations in atmospheric temperature of from 10'C to 40'C.
measurements, to make it adequate for dealing with large The required temperature is established by means of
numbers of cells. programming and communicated to the control system via GPm
This system has been developed by the Universidad del bus. The time estimated for a temperature change of 5'C,
Pais Vasco and Isofot6n S.A .. The system described below including stabilization, is from 15s to 20s.
allows cells to be sorted as well as performing several different
measurements on 100% of cell production, without any need for 2.2 Lighting source and irradiance control.
introducing additional quality control devices. The source of illumination for the system is based on
dichroic halogen lamps (4 x 75w + 1 x 50w) the light of which
provides a color temperature of 3100 K. With this configuration
2. THE SOLAR CELL SORTER it is not possible to obtain an AM 1.5 spectrum, but on the other
hand, it is the only simple way of obtaining a rapidly variable
A large part of the sorting system described here is light source.
integrated into a standardized rack, requiring only the supply of As in the case of the measurement platform, the light
electrical energy and pressurized air for its operation. source is built into a standardized rack. This allows the light
The system can be operated manually, the operator having source module to be easily interchanged with other light modules
the job of placing the cells to be measured on a thermostated (for example, systems based on Xenon lamps, laser beam scans,
platform, as well as indicating the measurement conditions. The etc.).
system will respond in this case with the information obtained The range of irradiance provided by the source is from
from the cell, as well as the type to which it belongs. The system 2mW/cm2 to 200 mW/cm2. The surface illuminated in this way
can also operate totally automatically by means of the use of a is of 150x15Omm with a uniformity of over 10%.
robot for manipulating the cells. The operator will have the job of The point of light is controlled by means of the supply of
choosing, by means of a tactile screen, the type of measurement the halogen lamps via an electronic regulator and a silicon
to be made during successive cycles. photodiode and the unit has PI control. The stability which is
Figure 1 shows a general diagram of the solar cell sorter obtained in this way is better than 1%. The irradiance level is
in which the main parts of the system can be seen. The estimated established by means of programming, this being communicated
operating cycle is more than 5 cells per minute. to the control circuit via GPID bus.

689
2.3. Electronic load. The criteria for the sorting process which is normally
In this sorting system, cell polarization conditions are used involves the series connection of the cells, although
established by means of an electronic load capable of operating in. sometimes it is also used with cells which are later grouped in
four quadrants. The electronic load used is a KEPCO, model parallel. With series interconnection, the current which crosses
BOP 2O-10M, totally programmable via GPIB. The operational each of the cells is the same. In this way, if a single cell showed a
range is ±20V and ±1OA, allowing a step of almost 0.5mV in a limitation in current, it would cause a limitation in the current of
subrange of ±2V. It also allows cell polarization conditions to the whole module (the panel behaves like the worst cell).
change in under 2ms. Sorting is then carried out by measuring the cell current
under standard conditions (100mW/cm2 and 25'C) at a pre-
2.4. Data adquisition system. established voltage close to that of the point of maximum power
The data adquisition system is responsible for measuring and by grouping the cells into types with typical 5% current
and transmitting to the computer the voltage and current variations. With this system, a minimum power is guaranteed in
measurements which are determined in the cell, as well as values the modules. Nevertheless, should significant variations occur in
for irradiance and working temperature. The data adquisition the voltage of the point of maximum power, Vmax, of different
system, model HP 3421A, is totally programable via GPIB. cells, the method described would lead to the the incorrect
The system used is a rapid voltmeter (up to 30 grouping of cells. These Vmax variations will be caused by
measurements per second), which integrates the voltage differences in the characteristics of materials used in
measured, offering a noise rejection of up to 80 dB. The manufacturing the cell (e.g., if polychrystaline silicon is used) or
resolution of the voltmeter can vary from 3 1/2 to 5 1/2 digits, by tolerances in the manufacturing process (significant deviations
with measurement scales which range from O.3V to 300V in DC. are often obtained in the series resistance and parallel resistance
of the cell).
2.S Computer and auxiliary systems. As the current of the cells drops sharply for voltages
A PC-compatible computer is responsible for controlling higher than that of the point of maximum power, the criteria
the sorting system. It has a 20 MByte hard disk for storing data, which is normally applied involves measuring the current in the
as well as a 3.5" disk drive. minimum permitted voltage of the point of maximum power, so
The MS-DOSTM operating system, version 4.0, is used that the majority of the cells are not rejected, even at the cost of
and Microsoft Basic™, version 7.0 is used as programming losing information on Vmax-Imax. Making 2 or more V-I points
language. The programming language which has been chosen in the cell would enable the point of maximum power to be
allows structured programs to be prepared, as well as the creation estimated with greater accuracy and could lead to different cell
of small performable modules and the possibility of executing grouping criteria from those described here. The system
routines written in other MicrosoftTM languages (FORTRAN, C, described in the previous section, allows this type of
Pascal, Macro Assembler). measurement to be made at high speed without any difficulty.
The computer is also responsible for receiving the
measurements made by the system, either for instantaneous 3.2. Control of technology.
processing and real time decision taking, and for their "in situ" In the normal operation of an industrial solar cell
storage and statistical processing or by other similar computer manufacturing process, it is advisable to apply a quality control
systems. which goes beyond a simple sorting process. It involves being
Communications between the computer and the able to obtain information on the correct operation of the
equipment used for making measurements are established via the manufacturing process itself.
GPIB bus. This bus which has asynchronous series byte/parallel The way to obtain this information is by means of the
bit type transmission characteristics, allows the transfer of determination of the internal parameters of the cell (series
programming and measurements with the required speed. resistance, parallel resistance, saturation currents, etc.). To do
A PLC is responsible for pneumatic drives and the this, the usual method consists of processing I-V lighting, I-V
supervision of availability signals and possible errors given by darkness and Isc-Voc variable lighting curves. By making this
the temperature and irradiance control systems, as well as other type of measurement during the customary sorting process and
auxiliary event detection systems (vacuum obtained, contacts maintaining statistical control of the values obtained, it would be
established, end of cycle, etc.). possible to carry out a satisfactory quality control for the
The PLC communicates with the computer via a RS-232- manufacturing process as well.
C line, for the necessary consultations throughout the operational The system presented here is capable of making the range
cycle. of measurements described in the last paragraph. Nevertheless, it
A robot can be installed for handling cells outside the is necessary to optimize the system of measurements to be made,
system, giving the system the speed required for optimizing the at the cost of losing accuracy in the values obtained, so that this
operational cycle time. can be compatible with dealing with large numbers of cells,
representing 100% of the production at the moment when these
2.6. User interface. are sorted.
A tactile screen is the normal means of communication
with the user. By means of this screen, it is not only possible for 3.3. Support for technological changes.
the system to communicate the results obtained in the different As has been said in the previous section, the exhaustive
measurements but also for the operator to establish those measurement of a number of cells per day is desirable to obtain a
measurements and tasks which are required of the system at more exact knowledge of their internal parameters. This would
specific moments. provide accurate information on the state of technology used in
The system also has an ink-jet printer which a?0ws the their manufacture.
graphics and statistics based on the measurements carned out to The sorting system which has been developed possesses
be printed. operational characteristics (number of cells sorted per cycle)
The transfer of the data obtained in the different which would enable it to do this work with a number, still to be
programmed measurements to other pieces of equipment, outside determined, of cells, without interfering with the daily work rate.
the plant is made via 3.5" disks with a capacity of 1.4 MBytes. The measurements to be made with these selected cells
A mode~ could be adapted to the system, should this be may be programmed by research teams with access to the sorting
necessary, to enable communications by telephone line. system without too much difficulty thanks to the software used.
The data obtained could be processed "in situ", or stored for
later, more detailed analysis.
3. SYSTEM UTILITIES
3.1. Cell sorting. 4. CONCLUSIONS
As the solar cells which are obtained industrially can vary
considerably with respect to their characteristics (of even more Quality control in industry requires equipment such as is
than ±15%) it is necessary to group these before connecting them presented in this paper, capable of carrying out the basic sorting
together to form a panel. This allows power losses due to the tasks which normally required in industry, as well as additional
interconnection of cells with widely varying characteristics to be measurements leading to the control of the technology.
reduced to a minimum. Strategies regarding the systems in which additional

690
 measurements leading to a control of the technology must be 5. ACKNOWLEDGEMENTS
made are open to further research.
With systems of this type and appropriate measurement Thanks to Fernando Hernando for the development of the
procedures, even the sorting techniques in use could be varied, GPIB communications and programming card used in irradiance
obtaining new sorting modes for solar cells. and temperature controls.

TOUCHSCREfN

~~==~=~~:~=::T~~ TEAUOST~TED PLATFOA W

FHOTOOIOOE

o
PELTlfAS CEllS ANO TIEMPtRATUAf SENSOR

TENPEAATUAIE CONTROLLER

P L. . C • ROIOT CONTROLLER

Figure 1. Schematic diagram of the sollar cell sorter.

, - - -- - Fl.OATINO CONTACTS HOI.DER

COPPER

PHOTOD1001E

'--,.,p.' - - - - - - i l - - - - MIC AOC V L HO

I EA

PELTIER CELL
__ ~"-J.

(fOOIEO.OI

IASf HOLDEA ,-~~_~~_ _ _ _ _ _ _~TfMPEA.TURE SENSOR

(E .... O .01

, 6 ' ' ' - - - - - - - -- -- ","CUU M PIPIES

HEAT $IHK-----.J ' -- - - - - - --8AND CYLINDER

Figure 2. Sectioned view of the Measuring platform

691
taI'H EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON,PORTUGAL

663 mV OPEN CIRCUIT VOLTAGE EPITAXIAL THIN FILM SILICON SOLAR CELLS

A.W. Siakers*, J.H. Werner, E. Sauser, and H.J. Queisser

Max Planck Institut fOr Festk5rperforschung

Heisenbergstr 1, D 7000 Stuttgart 80, FRG.

ABSTRACT: Thin silicon films have been grown epitaxially on silicon

substrates by the liquid phase epitaxial process. Solar cells made on these
layers grown on heavily doped silicon substrates have displayed open circuit
voltages as high as 663 mV (AM1.5, 25 C). This value is believed to be the
highest Voc observed on an epitaxial or other thin film silicon solar cell by a
large margin. A high Voc is a prerequisite for a highly efficient thin film silicon
solar cell. This result has applications to both space cells and to low cost
terrestrial cells.

appropriate, prior to cell encapsulation by etching or by

1. INTRODUCTION
other means to avoid loss of light into it. There are
possibilities for monolithic interconnection within modules
Since 1983 the efficiency of the best crystalline
silicon solar cells has increased from 17% to around 23% for all three substrate types.
The potential efficiency of a thin cell is higher than
(AM1.5, 100 mW/cm 2) (1). Efficiencies of 25% appear for a thick cell for the range of crystal quality likely to be
possible with current technology. There are several used in commercial production. The Voc of a solar cell will
reasons for this rapid improvement. The most important
increase as a wafer is thinned as long as base layer
has been the increased attention paid to obtaining low
recombination dominates the total recombination current.
surface recombination currents, both at the electrical
The increase is small unless the cell thickness is less than
contacts to the cell and away from them. Other important
the minority carrier diffusion length. At small thicknesses
developments have included improvements in cell
the Voc saturates because surface recombination
processing to maintain the high starting minority carrier
lifetimes of the wafers and the introduction of effective dominates the recombination current. Recombination at
light trapping schemes. A large gap has now opened grain boundaries in polycrystalline silicon will also be
between the efficiency of laboratory and commercial cells, suppressed when the wafer is thinned since the area of
with the latter usually in the range 10 to 14% efficient. The grain boundaries declines. Using a thin cell will lead to a
reason for this gap is partly due to the lower quality wafers reduction in the amount of light absorbed even when a
used and partly due to less than optimum processing. light trapping scheme is incorporated in the cell design.
In recent years there has been a large research However, thin cells will have a carrier collection efficiency
effort to produce solar cells in potentially low cost thin approaching 100 % as the diffusion length will be several
films of strongly absorbing non-crystalline-silicon times larger than the cell thickness. Thus the J sc of cells
materials. However, none of these materials have been can actually increase as the cells are made thinner,
able to match the combination of efficiency, stability, non- although it will eventually decline at small thicknesses.
toxicity and raw material abundance of crystalline silicon. From the above discussion it can be seen that there is an
In addition, it has come to be realised that crystalline optimum wafer thickness whose value depends on the
silicon can also be used in a thin film form. Commercial quality of the silicon, of the surface passivation and of the
solar cells are mostly made on single or polycrystalline light trapping scheme used.
silicon wafers of about 400 micrometer thickness. The The upper bound on the efficiency of 1 ohm-cm
cost of the substrate wafer is a large fraction of the total silicon cells as a function of wafer thickness, minority
cost of the solar module. Fabrication of thin polycrystalline carrier lifetime and diffusion length is shown in fig. 1.
silicon cells (20 to 50 micrometers thick) has the potential Reducing the cell thickness below 20 micrometers yields
to lower the substrate costs or to increase cell efficiency no significant efficiency advantage. The limiting efficiency
or both. as a function of material quality can be translated into a
practical efficiency goal by allowing about a 2 %
1.1 Thin silicon cells for terrestrial use (absolute) efficiency reduction to account for reflection
The substrate cost of a silicon solar cell may be and parasitic resistance losses. Fabrication of a 20%
lowered by fabricating cells on high-quality polycrystalline efficient polycrystalline silicon solar cell requires a
epitaxial layers grown on a low cost substrate such as minimum lifetime in 1 ohm-cm material of about 5
cheap silicon ribbon, a ceramic material (2) or glass (3). microseconds. Some solar cells processed on 1 ohm-cm
The substrate would be used as an epitaxial template and polycrystalline wafers have displayed Voc's in excess of
supporting substrate, but would play little part in the
620 mV (4). which implies carrier lifetimes of over 20
electronic operation of the cell. It could be removed, if microseconds .

• Present Address: School of Electrical Engineering, University of New

South Wales, Kensington, Australia.

692
 than conventional space cells without light trapping. In
24 ~-------------------------,
addition, the higher Voc expected in a thin cell means that
\ 30 ~sec the efficiency declines less rapidly with temperature
(300 ~m)
~---
increase.
~ 22 \10~SeC
o ~ ___ (175~m)

D 2. LIQUID PHASE EPITAXY

C
::J
tiS 20 t
3 ~sec
It is necessary to establish practical ways of
1 ~sec / ' (100 ~m) producing high quality, thin layers of crystalline silicon,
>-. (55 ~m)
u and to fabricate highly efficient solar cells on them. The
C
QJ method chosen in this work to produce thin silicon layers
u 18 is liquid phase epitaxy (LPE) (7,8). LPE is a process that
involves the saturation of a molten metal such as tin,
w indium, bismuth, gold or gallium with silicon and the
subsequent epitaxial deposition of silicon from the melt
16 onto a suitable substrate, usually as the substrate
temperature is slowly lowered. LPE allows for a deposition
10 100 of silicon at temperatures in the range 400 to 1000 C, well
Cell Thickness (~m) below the melting point of silicon. The epitaxial method is
relatively Simple, and being an equilibrium process
Figure1: The upper bound on the efficiency of p-type 1 produces low levels of lattice defects.
ohm-cm silicon solar cells as a function of base The substrates used for our epitaxial growth were
layer thickness and minority carrier lifetime polished single and polycrystalline silicon wafers. Early
(diffusion length in brackets). An upper bound work has focussed on single crystal substrates in order to
on the Voc of 710 mV is assumed due to study the quality of the LPE grown layers without the
surface recombination. complications of grain boundaries. Epitaxial layers for
solar cell fabrication were grown directly from the
1.2 Thin silicon space cells substrate, and not from via holes in oxide masks.
Thin silicon solar cells have an application as space The silicon layers were grown either in a tipping boat
cells. An important factor limiting the lifetime of space craft system or in a centrifugal system. The tipping boat system
in some orbits is damage to the solar panels by radiation. consists of a quartz tube mounted in a horizontal axis
The radiation causes damage to the silicon crystal, and furnace in such a way that the tube can be rotated. A
the cell efficiency declines as the diffusion length of graphite boat attached to the quartz tube is used to
minority carriers in the silicon substrate declines to less contain the melt and the substrate .. When the tube is
than the thickness of the substrate (normally 100 to 250 rotated the melt flows onto the substrate, where it remains
micrometers). If the substrate is made very thin (10 to 20 until the tube is rotated back to its starting position. The
micrometers) then the cell can absorb more radiation furnace is equipped with controllers which allow the
damage before an unacceptable reduction in diffusion temperature to be ramped or held steady. The substrate
length occurs. An important additional advantage of thin area is 4.3 cm 2. Hydrogen, purified in a palladium
silicon space cells is that the power-to-weight ratio will be diffusion cell, is used as the growth ambient.
several times larger than in conventional silicon space The centrifugal system consists of a graphite crucible
cells. mounted on a rotating, magnetically suspended rod
Conventional space cells are not fabricated on located in a vertical axis furnace. Metal solutions can be
thinned silicon wafers because of problems with wafer moved onto and off the substrate by varying the speed of
breakage, the need to passivate cell surfaces and the rotation The design of the crucible allows for several
need to incorporate a light trapping scheme. Wafer different solutions to be moved successively onto and off
breakage can be controlled by processing the unthinned the substrate by programming a series of different
wafer to form a solar cell, and then chemically thinning the rotational speeds into the controlling computer (9).
wafer just prior to bonding the cell to a glass cover slide. The melts used in this work usually consisted of
Advantage can be taken of automatic etch-stop layers indium metal, sometimes with the addition of gallium (for
buried beneath the surface if the active area of the cell is p-type silicon) or InP (for n-type silicon). Saturation of the
epitaxially grown. Passivation of surfaces becomes indium melt was performed by raising the furnace to the
increasingly important as the base layer of a solar cell is growth temperature, around 925 C, and pouring the melt
thinned. The reason for this is that the Voc must rise to onto a saturation wafer for several hours. During this time
compensate for a decline in light absorption, and hence silicon was dissolved until the melt was saturated and no
Jsc , if the efficiency of the solar cell is to be maintained. more silicon could be dissolved into the melt. The melt
was then removed from the saturation wafer and the
There have been substantial recent improvements in
furnace cooled. Meanwhile a growth wafer was cleaned
surface passivation techniques.
using the RCA method (10), finishing with a dip in
A light trapping scheme is necessary to reduce the
hydrofluoric acid to remove the native oxide. The growth
escape of long wavelength, poorly absorbed light. Most
wafer was then substituted for the saturation wafer and
present day space cells do not incorporate a light trapping
the furnace reheated to the growth temperature. The
scheme due to the fact that infra red light with energy less
furnace was held at this temperature for an hour and a
than the silicon bandgap is also trapped, and eventually
half to allow equilibrium to be reached in the melt and to
absorbed at metal contacts or by free carriers, leading to
ensure removal of residual native oxides from the silicon
an increase in cell temperature and a consequent drop in
wafer by the hydrogen atmosphere. Growth commenced
cell efficiency. However, recent cell designs incorporating
with a supersaturation of half a degree to prevent
light trapping are characterised by very low levels of such
dissolution of the substrate. The furnace was cooled at a
parasitic absorbency (5,6), and hence will not run hotter
rate of about 50 degrees per hour. After growth the wafer

693
was removed and residual indium removed in acid. A
typical layer thickness was 15 micrometers.
The layers of silicon grown on (100) oriented, 710
I
polished single crystalline substrates were generally
uniform, smooth and shiny over the entire wafer, with local 15 microns
variations in thickness of less than 10% of the layer 690
thickness. Silicon grown on polished polycrystalline
substrates clearly revealed the individual grains, each of
:;-
which had a different growth rate and morphology.
Substrate resistivities were nominally either 0.5 ohm- .
.5- 670

cm or 0.035 ohm-cm, with boron doping densities of 3 x '"

~
0
1016/cm3 and 1.4 x 10 18/cm3 respectively. Some of the >
.;; 650
0.5 ohm-cm single crystal substrates were subjected to a
~
heavy boron diffusion at 1120 C for 4 hours prior to <:;
epitaxial growth, leading to a sheet resistance of 0.4 ohms
per square and a high surface boron concentration. The
."
a.
0
630

purpose of this procedure was to confine most of the light

generated minority carries to the epitaxial region.
However, the large boron doping density at the substrate 610
surface led to a marked reduction in the quality of the
epitaxial layer morphology. The surface had a wavy
appearance, unlike the smooth, shiny surface normally
seen on LPE grown layers. Heavily boron doped layers
have a decreased lattice constant compared with undoped 19 20 21 22 23 24 25

silicon due to the smaller size of the boron atom (11). The Efficiency Upper Bound (%)
epitaxial layer will therefore be under tensile stress. The
performance of solar cells on these substrates were Figure 2: The upper bound on the efficiency of p-type 1
significantly worse than on undiffused substrates. ohm-cm silicon solar cells as a function of base
layer thickness and Voc·

3. LPE SOLAR CELLS

Several cells were made with Voc's in excess of 640 mV
A limited number of solar cells were made on p-type on substrates grown in both the tipping boat and
LPE layers grown on p-type substrates using PERC cell centrifugal systems. The best cell obseNed was made on
technology (5,6). The front surface of the epitaxial layer a 20 micrometer thick layer of silicon grown in the
was diffused with phosphorus to form the junction, centrifugal system on a 0.015 ohm-cm silicon substrate,
followed by the growth of a high quality oxide passivation and had a Voc of 663 mV. The growth solution was pure
layer that doubled as an antireflection coating. A low indium. The doping density of similar films has been
silicon/metal contact recombination current was achieved measured by the C-V technique to be 1.6x10 16 indium
through the use of low contact areas and/or a heavy atoms/cm 2 (12). This corresponds to 1 ohm-cm material.
diffusion under the metal. A standard Ti/Pd/Ag grid was The recombination current of this device was calculated
used. The cell area was 1 cm 2 or 2 cm 2. from the diode equation to be approximately 0.175
The cells had low currents due to the lack of light
pNcm 2 The emitter and metal contacts in this type of cell
trapping and the high reflectance of the oxide AR coating.
are known to contribute a recombination current of at least
The main goal at this early stage was to demonstrate a
high Voc. The Voc of a solar cell is a good measure of the 0.015 pNcm 2 (1), leaving 0.16 pNcm 2 from the rest of
the cell. Recent work at Stanford University has allowed
electronic quality of the base layer. It is often a better the recombination current from heavily doped regions to
guide than efficiency in assessing the potential of a solar be deduced (13). Their technique involved measuring a
cell design, as improving the electronic performance of a
cell is more difficult than improving the optical value of JO / ni 2, which was converted to a value of JO
performance. A high Voc is a prerequisite .for a high using a value for ni of 1.45 x 1010 /cm 2. A more accurate
efficiency thin film silicon solar cell in order to offset the value has recently been determined for ni, of 1.01 x 10 10
low J sc caused by a significant loss of infra red light.
/cm 3 (14). The heavily doped substrate is equivalent to a
Figure 2 shows the theoretical upper bound on solar cell
very deep boron diffusion, and is likely to contribute
efficiency as a function of Voc and base layer thickness.
recombination current to the cell as shown in figure 4 of
Cells made on epitaxial layers grown on substrates reference (13). Using the lower value of ni (14) the likely
heavily diffused with boron had poor performances
contribution of the substrate to the total cell recombination
relative to undiffused substrates, with the best Voc
current is at least 0.07 pNcm 2, leaving at most 0.09
obseNed being 603 mV. This poor performance is
consistent with the poor morphology mentioned above. pNcm 2 from the epitaxial layer. This corresponds to a
A few cells were made on epitaxial layers grown on diffusion length in excess of 250 micrometers.
0.5 ohm-cm wafers. Voltages of up to 627 mV were The previous highest published Voc obseNed in an
obseNed. Unfortunately the control wafers in these LPE grown solar cell is 600 mV (6), corresponding to a
batches (wafers without epitaxial layers) displayed low recombination current an order of magnitude higher than
Voc's, around 630 mV, so it was difficult to draw in the best cells made at MPI. Some silicon cells have
conclusions about the epitaxial layer quality. been made at the University of New South Wales on gas
The most interesting results were from cells made on phase epitaxially grown silicon on a heavily doped
epitaxial layers grown on heavily doped substrates. substrate with Voc around 625 mV (15).

694
4. DISCUSSION Dimensional Structures", edited by R.F.C. Farrow,
S.S.P. Parkin, P.J. Dobson, J.H. Neave and A.S.
The fabrication of cells with high Voc is a prerequisite for Arnott (Plenum Publishing Corporation, p.171, 1987)
highly efficient thin crystalline silicon solar cells. The 10) W. Kern and D.A. Puotinen, RCA Review .6., 187
observation of a Voc as high as 663 mV demonstrates (1970)
that the crystal quality of silicon grown at MPI by the LPE 11) F.H. Horn, Phys. Rev . .aI, 1521 (1955)
technique on heavily doped single crystal substrates is 12) W.H.Appel, PhD thesis, Max Planck Institut fOr
good. In fact, at most half of the recombination current in Festkorperforschung, Stuttgart, FRG (1985)
this cell came from the epitaxial layer, with the substrate 13) R.R. King, P.E. Gruenbaum, R.A Sinton and R.M.
contributing most of the balance. Swanson, Proc. 21 st Photovoltaic Spec. Conf. 227
The application of this work to space cells is immediately (1990) ,
apparent. The radiation tolerance of such a thin cell will be 14) AB. Sproul, M.A. Green and J. Zhao, Appl. Phys.
substantially better than a conventional cell (16). It is Lett., .QZ, 602 (1990)
possible to bury etch stop layers beneath the epitaxial 15) M.Taouk, private communication
laver so that etch-removal of the substrate just prior to 16) T. Markvart and AW. Blakers, "Radiation Resistant
encapsulation beneath a cover slide becomes relatively Space Cells", Report to European Space Agency
easy. This would avoid loss of light into the substrate and (1990)
increase the power-to-weight ratio of the cell severalfold,
while allowing for a low breakage rate during processing.
A standard light trapping scheme such as the inverted
pyramid approach (5,6) could be easily incorporated into
the cell. antireflection and
The presence of grain boundaries in the substrate creates
passlvating oxide
quite different morphologies in epitaxially grown layers.
The effect that this will have on solar cell performance
requires further investigation. The fact that the grain
boundaries are very far apart compared with the epitaxial
layer thickness will lessen the difference between cells
made on polycrystalline and single crystalline substrates. epitaxial layer
If satisfactory epitaxial cells can be made on p-type, 1 ohm-cm
polycrystalline wafers then the way would be open to the
attractive possibility of using low cost, low quality silicon
ribbon as a substrate.

ACKNOWLEDGEMENTS single crystal silicon substrate 300pm

p-type, 0.015 ohm-cm
We acknowledge the contributions of B. Bez, E. Czech , E.
Feyer, R. Stiefel and E. Silier to this work. We gratefully
acknowledge the generous financial support of the
Generalverwaltung of the Max-Planck-Gesellschaft. AWB
was supported during this project by the Alexander von
Humboldt Stiftung. Figure 3: Structure of the 663 mV open circuit voltage
solar cell

REFERENCES

1) AW. Blakers, "High Efficiency Crystalline Silicon

Solar Cells", in: Festkorperprobleme/Advances in
Solid State PhYSics Vol ~, ed. by U. Rossler
(Vieweg, Braunschweig 1990), p 403
2) A M. Barnett, D. H. Ford, R. B. Hall, C. L. Kendall
and J.A. Rand, Proc. 9 th Photovoltaic Solar Energy
Cont., Freiburg (1989)
3) S.H. Lee, S. Healy, T. Young and M.A. Green,
Materials Letters, .a, 53 (1990)
4) S. Narayanan, S. R. Wenham and M.A. Green, Proc
4th Int. Photovoltaic Science and Engineering Cont.,
Sydney (1989)
5) AW. Blakers, A Wang, A.M. Milne, J. Zhao, X. Dai
and M.A. Green, Proc 4th Int. Photovoltaic Science
and Engineering Conf., Sydney (1989)
6) A.w. Blakers, A Wang, A M. Milne, J. Zhao and
M.A. Green, Appl. Phys.Lett., .55, 1363 (1989)
7) R. Linnebach and E. Bauser, J. Crystal Growth .QZ,
43 (1982)
8) E. Bauser and H. Strunk, Mat. Res. Soc. Symp.
Proc. Xl, 109 (1985)
9) E. Bauser, in "Thin Film Growth Techniques for Low

695
 POSTER SESSION 3C
Hybrid and Utility PV
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

LIME PHOTOVOLTAIC BACK-UP SYSTElI FOR RURAL NUCLEUS

WITH LOW-QUALITY ELECTRIC SUPPLY

S. Gracia Navarro
K. Erill Perel16
Compania Sevillana de Electricidad, S.A.
Avda. de la Borbolla, 5 - 41004 - SEVILLE (Spain)

ABSTRACT. In every country there may always be many a far-fetched rural

nucleus of inhabitation where electricity distribution lines have become
both insufficient to cope with an unceasingly increase in load demand, due
to industrial and human development, and deteriorated to such an extent as
to be unable to support the demanded quality service standards, any
longer.
In most instances, end-of-line voltage at peak hours may even fall under
legal limits.
A conventional correction would probably consider line replacement but an
alternate reasonable answer under view of the renewable energies might
also consist in the erection of a Photovoltaic backing-up station, feeding
energy at any point near the end of any such old and deteriorated lines,
to help keep line-end voltage within legally rated values.
Such a PV system might also be dramatically useful to:
a.- Flatten load characteristic
b.- Optimize network voltage
c.- Optimize PV costs
d.- Analysis of performance on available photovoltaic equipment
e.- Advance in the knowledge of such systems, specially upon line failure
f.- Take over minimum services in case of line failure.
g.- Integration of PV generation into conventional networks.
With this in mind, COMPAiilIA SEVILLANA DE ELECTRICIDAD, S.A. (CSE)
together with CIEMAT-IER and ISOFOTON, all with a high degree of
experience in photovoltaics, have presented to the THERMIE PROGRAM the
proposal for a Project consisting in a "160-KWp GRID-CONNECTED LINE
PHOTOVOLTAIC BACK-UP SYSTEM FOR LOW QUALITY ELECTRIC SUPPLY" as commented.

i.-INTRODUCTION

Unfortunately, a great many sites might

usually be found that meet all of the unwanted Average billed power' 2400 Kw Cailada. de Cane pia
conditions of an old, poor and insufficient Average daily energy' 8100 Kwh/dia
distribution line feeding far away nuclei of
1100 Kval!
inhabitation and not supporting any quality , Telephone Co.
service standard. : dl50Kvai
In the case of COMPAiilIA SEVILLANA DE is,aUghterhOUS8 Gas Station
ELECTRICIDAD, S .A., generating and supplying 7 Km
electricity to the regions of Andalusia and . 1200 Kval 1200 Kval
Extremadura in the whole of Southern and Western 1250 Kval
Spain, among the different studied sites liable ~Maria Topares
to have the proposed backing-up 160-KVp Junction
photovoltaic station implanted, one that meets all
of the necessary conditions is located at the 1100 Kval 125 Kval Other
No r t h - Ea s t 0 f the pro v inc e 0 f AIm e ri a , in 1100 Kval
Substation
Andalusia. 23 Km/25 Kv
That electrical distribution line runs from
"Vera Nueva" sub-station to as far as the 18 Km 5 Km
furthermost power consumption point village of Figure 1. Old distribution line to be backed-up
"Topares", some 85 Kms. away, both in the province
of Almeria (Andalusia - Spain).
The line backing-up Photovoltaic system
would be feeding PV energy into the knot that
comprises the main feeding line and the lines To study the characteristics of the line at
branching off toward the rural nuclei of "Maria", the above mentioned junction point, both power
"Cailada de Cailepla", "Topares" and some other demand data and voltage variation data have been
near-by loads, as shown in the diagram of Fig. #1. carefully recorded all along anyone typical day.

699
 Power demand curve characteristic is shown
in figure #2. Solar Irradiance Power Syetem

A.C. Ouipui
380 V
500 Load (Kw)
SIor·o·
450
400
350
300 380 V Reoulated
Conetent A.C. Output
380 V
250
200 Low quality A.C.
150 Tap- Tranatormer
Input
100
50 25 K.
O +---~----~----~--~----~--~
o 4 8 12 16 20 24 Figure 4. System block diagram
Time
3. - SCHEDULE
Figure 2. Average load characteristic at line junction
The estimated project schedule would be:
Phase 1 Design MAY, 91 - SEPT . 91
Phase 2 Procurement OCTO. 91 - MARCH 92
The curve plot ted in accordance with the Phase 3 Installation APRIL 92 - SEPT. 92
data recorded for voltage variations at line Phase 4 Test/Running OCTO. 92 - NOVE . 92
junction point is shown in figure #3. Phase 5 Monitoring DECE. 92 - NOVE. 94

4.- FINAL

The main target of the Project is the

30 Voltage (Kv) design, installation and demonstration of a
photovoltaic system connected to the conventional

-
25 r-________________~ mains supply network in such a way as to be able
20
----_/ to back-up any old and deteriorated distribution
lines supplying far-away nuclei of dwelling, when
any such lines might prove insufficient or unsafe
15 to guarantee standard quality service to its area
10 of influence.
Further, the Project also means to attain
5 the following objectives:
a.- Flattening of load characteristics
O r---~----~----._--~----~---J b.- Network voltage optimizing
o 4 8 12 16 20 24 c.- Optimization of photovoltaic costs
Time
d. - Analysis of performance upon currently
available photovoltaic equipment (inverters,
Figure 3. Average daily voltage at Input node
controllers, cells, batteries, etc.)
e.- Study of reliability of photovoltaic
systems, specially upon line tripping and other
line failures.
2.-CHARACTERISTICS OF PROJECT f.- Integration of photovoltaic generation
into conventional networks.
The proposed grid-connected PV-system Once all of these studies are carried out
basically consists in one 160-Kwp PV generating SEVILLANA is confident that this project may aid
system containing 3,000 solar modules (arrayed to prove that photovoltaics-based systems
into 200 parallel-connected strings of 15 series- connected to conventional electricity networks may
connected modules) which will be connected through be of the utmost importance as a means of line
controlling and regulating sub-systems into a 240- backing-up to guarantee the necessary service
element (2 x 120 series-connected elements of quality as an alternative to deteriorated line
1,300 Ah, in parallel), stationary Pb-acid replacement.
battery. Furthermore, SEVILLANA is aware of being
The system also comprises 8 x 20-KVA, DC/AC accepting the challenge of investigating and
power conditioning inverters (2 load-synchronized, developing new PV technologies and components
auto-regulated ones and 6 slaving ones) and 1, specially in the field of better and mO're
250-KVA, 25,000/380-220 V., tap-transformer for reliable inverters capable of being connected to
voltage r egulation . conventional networks with maximum efficien cy,
The PV-system will be run on the continuous reliability and performance.
monitored mode by means of an automatic data We are also confident that the conclusions
acquisition system . that may derive from this Project shall be of
The system block diagram is shown in Fig.#4. paramount importance and universally acceptable.

700
laTH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

SCENARIO FOR INTEGRATION OF PV SYSTEMS IN-

TO MULTIPLE-CARRIER GRIDS WITH INTEGRATED
ENERGY STORAGE

Bent Sm-ensen

Technical Director, Professor of Physics

COWIconsult, Consulting Engineers and Planners
Parallelvej 15, DK-2800 Lyngby, Denmark

ABSTRACT. Several integrated energy supply systems are described, based on com-
mon grids and a diversity of energy sources, including photovoltaic systems. First, it is
argued that decentralized rooftop PV systems may coexist with other energy sources, if
the grid system is suitably constructed. Then, a multiple-carrier concept is developed,
based on combined heat and power production from PV systems, integrated into grid
systems for electricity, fuel and heat, that accept energy from several sources. These
modes of solar energy utilization bears promise of considerable cost reductions, as
compared with central solar cell arrays and separate solar thermal collectors, and a fair-
ly smooth transformation of current energy supply and distribution systems into the
new ones seems possible.

1. INTRODUCTION System efficiency

(watts deHvered divided by watts of insolation)
The price of a PV systems may be divided 8 10 12.5 14 (%)
into a module cost, including the pure solar cell cost,
16
and what of sometimes called the balance of system
costs, which include in part power conditioning costs
(with costs of DC and AC subsystems and switch- ; 14
/ Centml PV plant, total cost
yards), and in part site-dependent costs. The latter
consist in the case of a central PV plant of land cost,
t 12

site preparation and civil engineering costs, the cost l Site costs (may be Bvolded
.... 10 for rooftop systems)
of support structure and mounting of modules, plus ~
infrastructure such as roads, fences and power line
connections. Figure I shows a European estimate of
the future trend of these cost components (1). The US
-8!
o
8
.
'.
'.
l-
(I)
S •...., ......
Department of Energy estimates a more rapid devel- 0
u
opment, reaching the lowest price level given in Fig- 4 ...............
"

ure 1 already in year 2000 (3,4). ...................

Module cost . "',
..........
".
".
It is evident, that both historically, and ac- 2 .............
cording to the extrapolations into the future, the bal-
ance of system costs cannot be expected to decrease 0
I
as rapidly as the cost of the PV modules themselves. 1900 1990 2000 2010 2020
This implies, that central PV plants are most interest- YEAR
ing for regions with an abundance of cheap, marginal FIgure 1. Trends and projectIOns of e cost ofsolar
land, while densely populated regions, such as most cell central plants and their components (1,2)
of Europe or Japan, would have a greater chance of
making decentralized PV systems economically decentralized solution,
viable, e.g. by integrating the panels into rooftops.
In case a building is constructed with this in 2. DECENTRALIZED PHOTOVOLTAIC SYS-
mind, no site-dependent cost may be involved, be- TEMS
cause the solar panels replace other roofing materials,
and because the installation should be no more diffi- Figure 2 shows the general layout of a grid-
cult than installing conventional roof tiling or other based electricity supply system, including rooftop PV
roof structures. This is the basis for favouring a installations as well as centralized renewable energy

701
 Local storage
Local wlnd power Figure 2. Electricity supply system
Large-scale, and.oollop
centralized hydro, pholovollalc with centralized Rnd decenirRlizixl
wlnd power and Inatallalion& renewable energy inputs, Bnd with
sola, photovoltalc
power planls Storage at eneraystorage facilities (6)
consumers b.

Pumped
hyd.o (Inter-)nallonal g,ld

Hyd,o High-voltage
slo,age central grid "
,eservolrs ""
I

Ballery storage I
I
12V building
I power systems
Synthetic fuel 11 D-24DV local
Fuel - based production and networks
power production fuelato.age
or combined
heat and
power ELECTRICITY SUPPlY SYSTEM
production Fossilluel
sto,age

inputs (hydro, wind and PV), where these are practi- agement on various levels may be invoked. There will
cal and the resources adequate. be more "local traffic" on the grid lines, due to the de-
Assuming that the use of fossil fuels are di- centralization of power production, and more "Iong-
minishing for environmental reasons (both pollution distance traffic· due to the need for power exchange
and greenhouse gas emission problems), and that nu- between systems with larger shares of uncontrollable
clear fuels are omitted (due to environmental hazards, inputs. This increased traffic need not entail grid
e.g. connected with the possibility of large accidents), strengthening, because it may be counteracted by in-
then such electricity supply systems will eventually creased efficiency of end-use conversion, leading to
need components of energy storage, in order to cope smaUer overall supply requirements. Because the pur-
with the variable renewable resources. These could be suit of energy efficiency saving as much as 60% of
the hydro instaUations with seasonal reservoirs (e.g. current energy use is cheaper than the energy dis-
found in Northern Europe), or pumped hydro stores. placed at current prices (5), and because alternative,
Alternatives would be synthetic fuels (biogas, liquid more environmentally benign energy sources are at
biofuels, hydrogen) produced from surplus renewable best as costly as present energy supply, then energy
energy sources. efficiency should have fIrst priority in any energy po-
One advantage of large, interconnected grids licy.
is, that demand patterns are different in different soci- Figure 2 indicates the possibility of in-
eties, such that a certain amount of load-leveling au- stalling 12V power systems in buildings, in addition
tomaticaUy manifests itself when power inputs any- to conventionall00-240V power lines. This would al-
where into the grid can be used anywhere else. The low the PV power conditioning system to be simpli-
other advantage is, that inputs from variable rene- fied, and would be convenient in association with
wable resources such as sun and wind exhibit better several modes of energy use coming into use, notably
time distribution curves, when geographical locations in connection with battery operated appliances (e.g.
far apart are integrated into the same system. In other computers, consumer electronics), which could be at-
words the probability of failing input diminishes with tached to the 12V grid without the need for a trans-
increasing size of the network. However, in the case former. The option of charging batteries when a sur-
of solar inputs, this primarily smooths out occur- plus of power is available also contributes to load-
rences of heavy overcast situations, while an integrat- levelling or rather, in the case of a renewable energy
ed system connecting many time-zones is required in system, to loads following more closely the produc-
order to avoid loss-of-power at night. tion (6). Several other appliances could be operated
In order to utilize a network such as the one on low-voltage grids, as it is e.g. done routinely on
depicted in Figure 2 in an optimized way, load man- many house boats and yachts.

702
 (a)

Coal

........--__~ E,pc'ri c
vehlctes

/ .
Natural gas _ .--I.'. _ .. JG;;l ....~ Ot her Uges 01
~ natural gas

(b)

Wi nd energy I cenlral
photo-..oltaic converters -,*"~

+
I ./
\ ._.-L_. \ '_.~ Olher uses of
hydrogl'n

E lec!<icily gr i d
Figure 3. Outline ofmultiple-carrier grid
systems. At top a system (a) similar to sys-
Hyd rogen p l pell nt'
tems presently in place, in the middle a sys-
Oistricl heating lint'
tem (b) Rllowing centr8lized Rnd decentral-
FUl't cell unit
ized reneUlJlble energy inputs, and one tbM
DJRy develop nRturslly trom tbe current sys-
Combined heat and power (CHP) planl tem(7).

• Electrfc boll tery s!orl'

o Community heat Slore
o Indlvidua I users
• do with solar collec tors

3. COMBINED NETWORK SOLUTIONS This may also be so, if PV systems are introduced.
The electricity conversion efficiency of current sys-
In most industrialized societies, there is a d~ tems are of the order of 10%. There is a possibility of
mand for space heating, and in any case for hot water. future systems reaching values of 20-30%, but most
The efficiency of current, fuel-based systems is high- likely, the most cost-effective PV systems will be
ly improved by combined heat and power production. based on amorphous silicon and still have pretty low

703
 Wind ---------tr-:-----:--,
Uncontrolled
Figure 4. lflriom
conversion routes ror integra-
Solar radiation ----31...._ _ _ _---1 tion ofrenewsble
energy SOIU'CCS in-
>----~ Elec lrieity
to supply systems
_Li_g_n_e_ou_s_ _ _
L. biomass store
;---...;J:M-I Gasi Ii l'r
~.-.-. with energy stor-
age components
,, (7).
Non-lignl'ous Digeslt'r ". - - - - - .. Hea t
Fuel
biomass storE' cell

conversion efficiencies. The rest of the solar energy tween heat and power may be changed if required, by
captured by the cell is rejected as heat at a tempera- producing heat from surplus power through heat
ture somewhat elevated above the ambient one. pumps (7).
This makes it possible to combine a PV cell Figure 3a shows the system used in many
with a solar thermal collector. Most PV cells require a present societies, with components that form a transi-
certain temperature interval in order to achieve their tion between this and the future system of Figure 3b.
best conversion efficiency, and several current cells Some of the conversion routes from renewable energy
have decreasing efficiency at temperatures noticeably to heat and electricity are shown in Figure 4.
above ambient ones. This means that heat should be
extracted at a fairly low temperature, but not neces-
sarily excluding used for hot water or space heating REFERENCES
(e.g. using an air-based distribution system). An alter-
native would be to select solar cells optimized for (1) Coiante, D. and Previ, A. 1989. The Italian PV
operating at elevated temperatures, in order to be able Program Proc. 4th Int. PV Solar Energy Conversion
to pick up heat say in the range of 50-80·C, similar to Conference.
those of current heating systems.
In either case, the design would involve (2) Kern, E. 1984. Proc. 17th IEEE PV Specialist
mounting the solar cell on a finned back-structure, Conference, p. 1196. New York.
where a fluid or air could be passed over the heated
back surface in order to collect the heat. A combined (3) U.S. Department of Energy 1988. Report
PV and thermal collector could be substantially DOEICH 10093-19 DE 88001141.
cheaper than the sum of separate units for the two
purposes. Furthermore, the rooftop area of most (4) Royer, J. and Adkinson, D. 1990. Proc. 5th Int.
buildings (including side areas for high-rise build- PV Solar Energy Conversion Conference, pp. 149-
ines) is in fair relation to either the heat or the power 153.
use of the building, but rarely sufficient for support-
ing a reasonable amount of both types of panels side (5) S0rensen, B. 1991. Selected Experiences on En-
by side. Therefore, the combination of solar heat and ergy Conservation and Efliciency Measures in other
power production may be a necessity for reaching a Countries, thHt l1J8y be relevant to the AllStralian Ef-
high coverage of the building'S energy needs. /brt to reduce the Greenhouse Etrect. Australian De-
Combined solar panels will also require partment of the Arts, Sport, the Environment, Tourism
backup or storage, in order to be able to meet demand and Territories.
at all times. This can again be supported by suitable
grids, in combination with energy storage componen- (6) S0rensen, B. 1991. Advanced Electricity Storage
ts. Figure 3b shows such a grid system, containing Technologies. Ch. 22 in "Fuels and electricity from
long-distance integration of electricity and fuel (e.g. renewable sources of energy' (Johansson et ai, eds.),
hydrogen) networks, plus local district heating grids United Nations, New York.
capable of supplying back-Up heat for buildings. Con-
version of surplus power to storable hydrogen, and (7) S0fensen, B. 1984, Energy Storage. Ann. Review
storing surplus heat in community size heat stores of Energy, 9:1-29.
(water, rock or phase transition materials), make it
possible to meet power and heating needs, when the
renewable inputs are inadequate. The balance be-

704
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PV/DIESEL SYSTEMS ON HOUSEBOATS

T.C.J. van der Weiden, C.P.M. Dunselman, K. Hoekstra, K. Blok

ECOFYS Research and Consultancy
Biltstraat 11 0
NL - 3572 BJ Utrecht
The Netherlands

ABSTRACT. At present, the application of solar energy for domestic electricity consumption has already proven
profitable in the Netherlands for local autonomous systems with back-up diesel generators. Owing to the fact
that a PV generator and a diesel generator are available simultaneously, battery expenses can be restricted,
while control can be better attuned to the battery. In addition to economic factors, considerations regarding
comfort (noise and odour pollution) are also significant in the introduction of these systems.
With support from the CEC-DGXVII and the NOVEM, fifteen systems are to be installed and monitored. The
systems installed so far are functioning according to expectations and satisfactorily for the users.

INTRODUCTION expensive industrial batteries are not used in all cases, since
semi-traction batteries are already being used in a number
In view of anticipated price developments for modules, of existing diesel/battery systems, and practical experience
expectations are that application of PV/diesel systems with a has shown positive results. In this respect, another important
power of several kilowatts will be profitable for the whole of aspect is the fact that the initial costs required may constitute
Europe in a few years' time. The European potential for these an obstacle to the purchase of the PV system.
kinds of systems is substantial, considering the number of A third difference is a result of ship mobility. Comparatively
isolated houses (estimated at approximately one million (1)). cheap power is available while a ship is underway.
As a preparation to the large-scale introduction of these Furthermore, modules are positioned almost horizontally on
systems, fifteen existing diesel systems on houseboats are sailing ships, whereas the tilt angle employed for non-sailing
being equipped with a solar generator and a controller. Set- ships is adjusted so as to maximize output.
up and expected economics of the systems will be discussed The systems mainly consist of standard, commercially
in this paper. In addition, experiences and provisional results available products. The controller, however, has been
with regard to the performance of installed systems will be specially designed for this project, in view of the fact that no
presented. controllers were available for PV/diesel systems with- powers
up to 2 kW that provided sufficient control signals for diesel
generator operation.
SYSTEM DESCRIPTION

Figure 1 provides a diagram of the system. The system CONTROL STRATEGY

configuration is not the same for all 15 boats. Various types
of modules have been employed (30-cell mono-crystalline The main objective of the control strategy is to protect the
and 36-cell poly-crystalline). battery from over{dis)charging. When discharging, control
A second difference is the type of battery used. Two types takes place at a current-compensated voltage level. When
are used: industrial stationary batteries (type OPzS) and charging, control takes place at a temperature-corrected
semi-traction batteries (type TV-marina). The more voltage level. Upper and lower limits for battery voltage are
set at 28.2 V and 23.5 V, respectively. Moreover, after a
number of charging hours, the eqUipment switches to a lower
voltage level (float charge voltage of 27 V). The result is that
batteries are more efficiently used in comparison with their
utilization in a conventional diesel system, where it is not

-
economical to give the diesel the task of float charging.
The inverter will not always have sufficient power for
supplying maximum required power. In those cases when
2" v de power demand is considerable (washing machine), the diesel
generator will also be activated, irrespective of whether or not
sufficient energy is available.

ECONOMICS
I IICCU

--r- On the basis of simulation calculations a simple payback time

has been determined. The payback time is specified here as
the ratio between initial costs and costs saved annually. The
following assumptions have been used. Attainable production
Figure 1 Diagram PV/diesel system in the Netherlands totals 1.2 kWh per Wp available [3]. A

705
 consumption pattern has been drawn up on the basis of COMPLETED ACTIVITIES
available equipment and annual consumption. The patterns
are comparable to the pattern for domestic use in single- Apart from the dimensioning, the specification of the controller
family dwellings. Chief differences in comparison with normal and the installation of systems, much attention has been
domestic electricity use include energy requirement for the given to publicity, since the main objective is to prom?te
water supply installation (fresh water pump) and the fact that market penetration of these kinds of systems. Data are being
average consumption (900 kWh/year on average) is much monitored in all systems. Two ships have yielded reliable
lower than in most households. Within the SOMES model the data regarding the use of the system for several months.
performance of the solar generator, the battery and diesel System performance is discussed below on the basis of these
generator are simulated [4]. The diesel generator efficiency data, as well as user experience.
has been assumed to be 6%. Test Reference Year data for
De Bilt have been used as meteorological data regarding
irradiation and temperature. PERFORMANCE
Simulation calculations carried out by the SOMES programme
yielded a utilization value of 0.7 kWh per Wp per year. At The systems which have been installed, are functioning
present, the cost of diesel fuel amounts to NLG 0.86/1itre satisfactorily and according to expectations.
(35.5 MJ/litre). On the basis of these data, a minimum of Initial controller problems (malfunctions occurred twice) now
saved costs can be determined for a system with a 400 Wp seem to have been solved. The controller's own consumption
PV generator at NLG 400.-/year. This can be regarded as a depends on the operational situation. Consumption is minimal
minimum, since it has not been taken into account here that in the event of energy shortage, amounting in that case to
the diesel generator within the PV/diesel system may have a approximately 1.8 W. One of the systems had to be adapted
higher efficiency, because, during a large part of the year, its in order to realize integration with an existing wind turbine.
use can be limited to those moments when a relatively high The glass of one of the modules was broken by an antenna
power can be employed. Nor have the extended life span of hitting the alumina side of the module. This module has been
the battery or the reduced wear and maintenance costs for replaced. Initially, Ah counter accuracy proved insufficient.
the diesel generator been considered here. Adjustments were made, and it now seems possible to
The investment in expansion primarily concerns the solar balance the energy balance to within 1%.
generator and the battery controller and their installation, and
will cost approximately NLG 10,000 (excluding VAT) for a 400 Table 1 represents the output of two systems.
Wp system. This would set the payback time at 25 years.
Yet, the situation in the Netherlands is considerably more The coverage rate as presented in the table has been
favourable due to national and regional support, so that calculated by dividing PV production by consumption.
payback times of 6 to 14 years can be realized. It can be Accordingly, the fact that part of the energy taken from the
concluded that profitable applications for diesel systems batteries may have been provided by PV has not been taken
expanded with PV generators are already available at the into account, yet this has no effect on the average annual
present time. However, it is difficult to accurately determine coverage rate.
the payback time of the investment in advance. In practice, The PV generator proves to be able to provide 10% to 20%
results will strongly depend, among other things, on the of the electricity demand during the winter months. An
quality of existing components and on the way in which the average annual coverage rate of 50%' is expected on t~e
system is operated. basis of calculations. Output strongly depends on the way In
Initial costs still prove to be a problem for many users. which the system is used. Consumption intensity and use of
Financing or leasing arrangements may provide a solution. the diesel generator are the principal factors here. The cause
of the difference in results between systems A and B cannot
be determined on the basis of weekly data. Results of
analytiC monitoring will provide a better insight into this
matter. These results will be incorporated into the user
manual.

Montih PV Wind Diesel Demand Coverage

Ah Ah Ah Ah PV%

~ystem A 11/90 61 -- 518 651 9.4

(24 Volt) 12/90 84 -- 945 1038 8.1
01/91 182 -- 807 1065 17.1
TOTAL: 327 -- 2270 2754 8.4

System B 11/90 531 117 2726 3531 15.0

(12 Volt) 12/90 225 145 3428 3689 6.1
01/91 525 284 3290 4032 13.0

TOTAL: 1281 546 9444 11252 11.4

Table 1 Energy flows in two PV/diesel systems during three winter months. Both systems have 400 Wp PV
generators. System A is a 24 V system, while system B is a 12 V system.

706
USER EXPERIENCES

Even during the winter months, a clear advantage is reported.

The availability of the PV generator reduces the time needed
to operate the diesel generator and noise annoyance. A
practical problem may arise from the height of the boat.
Installing of the PV generator on top of the ship is not always
acceptable in view of bridge heights.

FURTHER RESEARCH

The systems will be monitored for two years in accordance

with EC guidelines for global monitoring. Furthermore, five
systems will be monitored more extensively pursuant to the
guidelines for analytic monitoring [2]. Research results will be
employed for further optimization of dimensioning and use of
PV/dlesel systems.

REFERENCES

[1] Schmid, J. et ai, A 220 Volt AC photovoltaic power

supply for remote houses, ISE, D-Freiburg,
Proceedings of Eighth E.C. Photovoltaic Solar Energy
Conference, Italy, 1988.
[2] Siaesser, G., Guidelines for the Assessment of
Photovoltaic Plants - Document A - Photovoltaic
System Monitoring, CommisSion of the European
Communities, DG for Science, Research and
Development, Joint Research Centre, Ispra, Italy,
1988.
[3] Alsema, E.A. et ai, Elektriciteit in Nederland met
zonnecellen, Rijksuniversiteit Utrecht, Vakgroep
NW&S, 1988.
[4] Slok, K. et ai, SOMES, a simulation and optimization
model for autonomous energy systems, Rijks-
universiteit Utrecht, Vakgroep NW&S, juli 1988.

707
wrn EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

A PHOTOVOLTAIC-HYDROGEN-FUEL CELL ENERGY SYSTEM:

PRELIMINARY OPERATIONAL RESULTS

P,A. Lehman and C.E. Chamberlin

Environmental Resources Engineering Department
Humboldt State University, Arcata, CA 95521 U.S.A.

ABSTRACT. We report preliminary operational results for a photovoltaic (PV) energy system
which uses hydrogen as the storage medium and a fuel cell as the regeneration technology. The
system installed at the Humboldt State University Telonicher Marine Laboratory consists of a 9.2 kW
PV array coupled to a high pressure, bipolar, alkaline electrolyzer. The array powers the
Laboratory's air compressor system whenever possible; excess power is shunted to the electrolyzer
for hydrogen and oxygen production. When the array cannot provide sufficient power, stored
hydrogen and oxygen are furnished to a proton exchange membrane fuel cell which, smoothly and
without interruption, supplies the load. We report measurements of PV module and subarray
performance and electrolyzer current-voltage behavior and include an analysis of the match
between these two components. We detail modifications made to the electrolyzer to accommodate O2
collection and describe the monitoring and control systems being developed and tested.

1. INTRODUCTION 1.2 Objectives of Project

The objectives of the Schatz Project are:
1.1 Description of Project • to assess the storage efficiency of hydrogen when
The Schatz Solar Hydrogen Project began in used as a medium to store solar electricity
the fall of1989. It is a stand-alone photovoltaic energy • to assess the use of a proton exchange membrane
system that uses hydrogen as the storage medium fuel cell as a means of regenerating electricity
and a fuel cell as the regeneration technology. Its from stored hydrogen and oxygen
goal is to demonstrate that hydrogen is practical • to design, test, and utilize a computer based control
storage medium for solar energy and that solar system which will allow for efficient component
hydrogen is a reliable and abundant energy source integration and provide for reliable, unattended
for our society. operation
A schematic for the system is shown in Figure • to monitor operating and environmental
1. It is installed at the Humboldt State University parameters to chronicle system performance and
Telonicher Marine Laboratory and the Lab's air to allow development of a simulation model
compressor system, used to aerate aquaria, is the
load. When PV power is available, it is used directly 1.3 Contents of Paper
to supply the load. Any excess power is supplied to The installation of this system is nearing
the electrolyzer which stores the electricity as completion. In this paper, we report testing of the PV
hydrogen and oxygen gases. When the array cannot modules and subarrays, testing of the electrolyzer,
provide electricity, the stored hydrogen and oxygen and discuss the match between the two. We also
serve as fuel and oxidant for the fuel cell which discuss interesting and unusual aspects of the design
provides uninterrupted power. and installation and plans for the monitoring and
control systems.

Air 2. PV ARRAY
Compressor
2.1 Description
4 The PV array consists of 192 Arco M75
modules configured into 12 independent subarrays.
~ Inverter Each subarray consists of 16 modules, wired in 8
series pairs for 24 VDC operation. The subarrays are
electrically isolated from each other by 60 amp
Schottky diodes. The nominal power rating for the
array is 9.2 kW.

2.2 Module Testing

In an effort to minimize mismatch loss in the
array, we determined the I-V curve for each module
[1]. The modules were tested outdoors under clear
sky conditions. Insolation was measured with an
Epply PSP pyranometer and module temperature
was measured with a fast response thermocouple
affixed to the back of the module. A capacitive
Figure 1. Schematic of the photovoltaic-hydrogen- charging measurement system was used with data
fuel cell energy system at the Schatz Solar Hydrogen logged at 100 Hz. Details of the measurement process
Project are given in [1].

708
 Results of the module tests are given in These measurements were also taken under clear
Figures 2 and 3 below. Experimentally determined sky conditions. The Epply PSP pyranometer
correction factors were used to adjust the results to measured insolation; a thermocouple on each of two
1000 W/sq-m and 47°C. Conditions for insolation and modules measured temperature. The temperature
temperature were close to these values, so the readings were averaged.
corrections were generally quite small. The I-V curves for 4 of the sub arrays ranging
from #1 (best) to #12 (worst) are shown in Figure 4.
The curves demonstrate that sub array performance
is fairly constant; the curves differ by less than 2
amps at any voltage.
30

25

30
.
>-
20

.
;
~
<T
.;;
1$
2$

';j 20
Q.
10
!
c 15
~
~
U 10
o
Pm. . (Watt.)

Figure 2. Histogram showing distribution of module 10 15 20 25 30 35 40

power outputs corrected to 1000 W/sq-m and 47°C Vo lt. (VDC)

Figure 4. Subarray I-V curves corrected to 1000 W/sq-

m and 47°C

., The performance of the subarrays is

approximately 10% lower than would be predicted
based on the individual module I-V curves.
i 41
Mismatch loss accounts for less than 1 % of this
!. •• difference [1]. Most of the loss is accounted for by an
increase in series resistance and a decrease in
E
Q. parallel resistance of the subarray resulting in a
It
lower fill factor for the subarray. The series
resistance is increased an average of approximately
It
0.13 ohms, presumably by the wire and connections.
The parallel resistance is reduced to approximately
'7 ~~ __ ~-L~~~~~~~ __~~~
15% of the value to be expected from the individual
. 0' .1 • 10 20)0 10 7010 to II .t .... 11 .'1
Percenl :5i Pm.x modules, from 180 to 26 ohms. A more precise
quantitative analysis of this effect is in preparation.
Figure 3. Cumulative distribution of module power 3. ELECTROLYZER
outputs corrected to 1000 W/sq-m and 47°C
3.1 Description
These M75 modules are rated at 48.0 W at 1000 The electrolyzer is a Teledyne Energy Altus 20,
W/sq-m and 25°C. When derated for operation at a high pressure, bipolar, alkaline electrolyzer. It
47°C (using our experimentally measured was chosen because it is the only commercial
correction), their output should be 44.0 W. As can be electrolyzer available in the correct size (7.2 KW
seen from the figures, the modules produced maximum, 6.0 KW nominal) which supplies gases
significantly lower power than their rating indicated. pressurized to 750 KPa, thus eliminating the need for
In fact, 68 of the 192 modules (35.4%) produced less supplementary compression. The electrolyzer
than 90% of their rating. The mean power output of module consists of 12 cells in series and operates at a
39.87 W is barely over 90% (90.6%) of the rating. nominal 24 VDC. The module contains an electrolyte
Clearly, this is a fact any designer should take into of 25% (by weight) potassium hydroxide.
account in sizing an array.
The modules were assembled into subarrays by 3.2 Electrolyzer Testing
matching their output currents at 14 VDC at 1000 The electrolyzer I-V curves were obtained by
W/sq-m and 47°C. The voltage was chosen because it passing PV generated current through the module.
represents a typical operating point for the modules. The current was measured with an Amprobe Hall
Effect ammeter and the electrolyte temperature was
2.3 Subarray Testing measured by a fast response thermocouple affixed to
We also determined I-V curves for the the exit pipe through which the electrolyte circulates.
completed subarrays. These were measured with the The I-V curves as a function of temperature are
subarrays installed in the array field, but the shown in Figure 5 below. The efficiency (based on the
measurement was taken at the power distribution higher heating value of the hydrogen produced)
center. This means that as much as 35 meters of varies between 70% (125 amps, 20°C) and 91% (50
wire (#2 USE wire) and various electrical connections amps, 100°C). Under nominal conditions (240 amps,
are between the subarray and the measurement. 60°C), the efficiency is 74%.

709
 Two factors must be taken into account when
300 designing such a system: 1) the voltage decrease in
the array as PV module temperature increases and
250 2) the voltage increase as the electrolyzer ages, as
shown in Figure 6.
~ 200 The match between the array and electrolyzer
""E can be shown by superimposing the possible range of
:! electrolyzer operating voltages on the subarray power
150
~

~
curves. This is shown in Figure 7.
~
U 100

50
600
I! • I
0
11 20 21
• ~b.,n~ t
••. ~ - ........ b6rr.~ ..• -
: I.· (I , 1
..... !'_...... -J.!..l'... - i"'- ".1''''--- 1---"
22 23 24 25 2_ 500

.
Vollo (VDC) .~b.,n.,," ! ~: ! •• l
- •--5..b,.., '. " _.+' . >......... i !I...........;-..\.,..........I,......._..
i
1...........

._ . .J_I . . . . . .I. . .... .~i_;. _. __. .L! .........I-......+ ..I"

400
Figure 5. Electrolyzer I-V curves at varying •
~ ~ ..-.:-..-
temperatures
;0
300
i
I ".
I· I In.. Ii I
.~~.
: I
!
... j 1

Figure 6 below shows a comparison between 200 ··-·····t~·,··-t-·--·- .I-···-·~nae

. Of ~1K" ;r.·~~:··-·--
the I-V curves furnished by the manufacturer and r l ....g • • ..-c. '2OCI ~,.,. ...

measured under similar conditions in our 100 ·---{ ---···-t·-·--····t·...···~..·i~~~·r-~~ i _.._...• -........
laboratory. As can be seen, the agreement between I' 1 I iii
our data and Teledyne's data is excellent. Also 0
0 5 10 " 20 25 30 35 40
shown in Figure 6 is the expected voltage increase of VollO (VDC)
3.3 volts as the module ages. The old module
represents an operating life of approximately 25,000
hours. As discussed below, this performance Figure 7. Subarray power curves showing the match
degradation with age must be taken into between the subarrays and the electrolyzer
consideration in system design.
At nominal operating conditions, the array
and electrolyzer match very well. The worst of
circumstances will occur with an old, cold
electrolyzer and a warm array. Because of the cool,
250 maritime climate, the array will never get much
-...
E 200
:!
hotter than the 47°C used in the figure. Even under
these circumstances, the operating voltage would not
exceed 29 VDC. This would result in a power loss of
~ 150 approximately 10% which would not be a serious loss.
~
U
100 3.4 Modifications to the Electrolyzer
The Altus 20 is designed to deliver pure
50 hydrogen for industrial purposes. In normal
operation, DC power is provided by a power supply
and the oxygen gas produced is vented as an
30 unwanted byproduct. Since our fuel cell requires
Volt. (VDC)
oxygen (and since any fuel cell will work more
efficiently using pure oxygen as the oxidant), it was
necessary for us to modify the oxygen delivery system
Figure 6. I-V curves for the electrolyzer at 60°C to enable its collection and use. These modifications
showing comparison of measured values and are shown in Figure 8.
manufacturer's specifications
Tube-in·tube
3.3 Matching the PV Array and Electrolyzer heat exchanger
Various choices are possible in coupling the
PV array and the electrolyzer. Significant work has 02 from
been reported by the Hysolar project [2] on the Coales·
eleClrolyzer cing
enhancement in performance that is achievable by
tracking the maximum power point and then using filter
Water return
DC to DC voltage conversion to match the electrolyzer to electrolyte
voltage. Though enhancement is possible, its small reservoir
magnitude plus the extra complexity and cost led us
to choose direct coupling for our system.
Since the array and electrolyzer are directly
coupled, it is crucial that there be a good match
between the electrolyzer's operating points and the
array's maximum power points. These will vary as
conditions change but, in any circumstance, the
electrolyzer must not be allowed to operate at a
backpressure Water return
voltage much higher than the voltage at maximum to eleclrolyte
power, Vmp. The steep decline in power beyond Vmp regulator
reservoir
would be a serious loss.
Figure 8. Modifications of the oxygen delivery system

710
 The main goal of the delivery system is to The computer is equipped with an
remove small amounts of hydrogen (0.5-2%) which uninterruptible power supply. In the event of a
are present in the oxygen stream due to manifold power failure, the system will also go through a
electrolysis. The removal is necessary to prevent safety shutdown before turning itself off. Further, all
overheating and other damage in the fuel cell. The relays and solenoids are configured so that they will
hydrogen is removed by catalytic oxidation in the fail in their safe position.
deoxo catalyst bed. The gas entering the bed must be
noncondensing and must be preheated to initiate 5. MONITORING SYSTEM
oxidation. Once the catalyst has begun working, the
exothermicity of the reaction makes the catalytic 5.1 Description
process self sustaining and heating is stopped. An extensive monitoring system will measure
and record performance. A separate Macintosh
It is also necessary to remove water in the gas computer will handle data logging. The following
stream twice. This is accomplished by using a tube- experimental variables will be recorded:
in-tube heat exchanger with chilled water circulated • temperature: PV modules (3), ambient air,
through the outer tube as a coolant. In this way, the electrolyzer, fuel cell, hydrogen and oxygen storage
dew point of the gas stream is reduced to tanks
approximately BOC. This is done before the catalyst to • current: electrolyzer, battery, fuel cell, inverter, AC
insure that the gas is noncondensing and after load
oxidation to remove the product water before the • voltage: electrolyzer, battery, fuel cell, AC load
oxygen is delivered to storage. The coalescing filter is • mass flow: hydrogen and oxygen out of the
installed upstream of the catalyst to trap any aerosol electrolyzer and into the fuel cell
particles that may contain traces of KOH; it has been • pressure: hydrogen and oxygen storage tanks
used in such an application previously [3]. The KOH • other: insolation, windspeed and direction
would poison the catalyst. Finally, a backpressure These variables will be recorded on a real time basis.
regulator is necessary to separate the high pressure
in the electrolyzer from the varying pressure of the 5.2 Objectives
storage tank. The objectives of the monitoring program are to
The hydrogen delivery system is simpler since determine:
the hydrogen is pure as it comes from the • overall storage efficiency
electrolyzer. However, it too has a coalescing filter to • array, electrolyzer, fuel cell, and inverter
trap KOH (in the unlikely event that some KOH could efficiencies
reach and damage the fuel cell's acid membrane) • temperature behavior of the array, fuel cell, and
and a condenser to eliminate most of the water from electrolyzer
the saturated gas stream. Heat exchangers in both • effect of environmental conditions on system
gas lines are configured to deliver the water back to performance
the electrolyte reservoir. • real time behavior of the system
These objectives will aid a major goal of this
4. CONTROL SYSTEM research: to develop a simulation model to enable
prediction of system performance from
4.1 Description environmental conditions. Such a model will be
This system is designed for unattended useful in optimizing efficiency and in future system
operation. A Macintosh computer system will design.
handle all control and safety functions. Using
various transducers the system will monitor current
outputs from each of the 12 subarrays, the AC load 6. ACKNOWLEDGEMENTS
current, battery voltage (a small battery serves as a
buffer to prevent frequent switching), hydrogen and The authors gratefully acknowledge generous
oxygen storage pressures, and a number of safety grant funding from Mr. L.W. Schatz of General
interlocks. These include: 1) the presence of Plastics Manufacturing Co., Tacoma, WA, U.S.A.
hydrogen gas in the laboratory or in the hydrogen and the assistance of engineering students Jim
hood (all hydrogen equipment is contained in a Zoellick, Ron Reid, Gian Pauletto, Tim Murphy, and
continuously vented hood), 2) unusual electrolyzer Tom Herron.
parameters, 3) ventilation in the hydrogen hood, 4)
any out-of-range operating parameters reported by a 7. REFERENCES
separate monitoring system, and 5) activation of the
emergency stop switch. 1. J.I. Zoellick, Testing and Matching Photovoltaic
Control is accomplished mainly through use of Modules to Maximize Solar Electric Array
relays which enable the system to individually switch Performance, Senior Project, Humboldt State
each subarray to either the electrolyzer or the load or University, December, 1990.
to disable it by shorting it out. If PV power is not
available, the system will switch to fuel cell 2. A. Brinner, H. Bussmann, W. Seeger, and H.
operation. If gas pressures are insufficient to operate Steeb, Operation results of a 10 kW PV-electrolysis
the fuel cell, the system returns to utility power. system in different coupling modes. Proceedings of
the 8th World Hydrogen Energy Conference,
4.2 Safety Functions Pergamon Press, Oxford, 1990.
If any safety interlock is tripped, the system
will effect an orderly shutdown. This includes 3. M. Miller, Teledyne Energy technician, private
closing gas solenoid valves which will isolate the communication.
hydrogen and oxygen storage tanks and returning an
error message indicating the reason for shutdown.

711
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

GRIDCONNECTED 30 kW PHOTOVOLTAIC POWER PLANT IN FINLAND

P

Christer 'Nyman
Tecnical Research Centre of Finland
Semiconductor Laboratory
P.O. Box 34, SF-021S1 Espoo, Finland

ABSTRACT

A 30 kW gridconnected photovoltaic power plant has been built at the latitude of 60 o N. The
projectPis a cooperation with IVO Ltd, NAPS Ltd, Helsinki University of Technology, Tecnical
Research Centre of Finland and University of Turku. The PV-power project is a part of the Ne-
mo programme for advanced energy systems and technology sponsored by the Ministry of Trade
and Industry, Energy Department. The results of the PV power plant show the possibilities,
reliabilities and economical aspects of photovoltaic power plants in Finland.
The aim of the PV power plant in Finland is to get information of large scale PV applica-
tions, energy production, system and module reliability at this latitude. The part of the Se-
miconductor Laboratory in this project is to design and set up the data acquisition system.

1. SYSTEM DESCRIPTION One hourly output signal controls the inver-

ter profile. The inverter profile have hour-
The array consists of,crystalline modules 27 ly values, calculated from earlier insola-
kW and amorphous modules of 3 kW connected tion data, which control the inverter to
to Pan .lnverter and atterles
b' .Ph capaclty
Wlt . have 70 % of the total daily energy output
of 2860 Ah. The produced energy is supplied of the modules. The values are the same eve-
to the public electric supply system. The ry day but changed every month. If the bat-
inverter consists of three inverters of 10 teries are charged and the momentanly energy
kW rated power each. The small amount of in- output of the modules is higher than the
solation from October to February chase the calculated inverter value, all the energy
system to run only one of the inverters to output of the modules is directed to the in-
achieve a better efficiency of the inverter. verter.
The data aquisition PC and system controller
works independentlY of each other. A perso-
nal computer, with two AID converters to ac- 2. MEASUREMENTS
hleve 32 differential input channels, col-
lects the data. With the PC it is possible The measurements and input parameters of the
to run different load profiles for the in- horisontal and module plane insolation are
verter where the energy produced by the mo- used to calculate the efficiency of the pa-
dules is either supplied to the batteries or nels. Variations in the efficiency due to
through the inverter to the grid. The auto- temperature can be calculated from ambient
matic data collection is made by a commer- and module temperature measurements and wind
cial data acquisition program together with speed on the modules. The temperature of the
subprograms for clock routines, file opera' batteries is also measured. Measurements of
tions and calculations of the collected da· the module voltage and current are used to
tao calculate the power of the modules. The sta-
te of charge in the batteries can be calcu-
To operate the two axis tracking system two lated by battery current, battery DC voltage
output signals are calculated from the mul- and battery charger current. The inverter DC
tipyranometerdata (University of Turku) current and AC energy is measured for the
which give the direction of maximum insola- input and output power of the inverter. To
tion. The tracking panel has 8 modules with get absolute energies, the runtime of the
a maximum power of 96 W. This panel can be power plant is measured. The calculations
compared with a fixed panel also with a ma- are shown as monthly average and monthly sum
ximum power of 96 W. of the measurements.

712
3. RESULTS efficiency is presented in table II. In fi-
gure 1 the daily output energy of the array
3.1 PV array performance in June 1990 is presented.

The total array output energy as monthly

summary is presented in table I. The capaci- Table II.
ty factors of the modules are also presented Crystalline, amorphous and total monthly
in the same table 1_ The monthly capacity average module efficiency. (5/90 - 3191).
factors are calculated by dividing the mont-
ly total energy output of the modules by the Effici ency
energy output had the modules produced rated Month c-Si a-Si total plant
power over a month period. 5 10.8 2.7 9.4
6 10.9 2.6 9.4
Capacity = Tot. module energy (kWh)* 100 7 11. 1 2.7 9.6
factor period hours* plant rating 8 10.2 2.5 8.9
9 10.8 2.6 9.4
The resource capacity is calculated by the 10 9.1 1.9 7.8
same formula, where the total energy is mo- 11 5.0 1.3 4.4
dule plane insolation and rated power 1000 12 1.3 0.5 1.2
W/m'. The relative capacity is the module 2.6 0.7 2.3
capacity factor divided by the resource ca- 2 8.2 1.7 7.0
pacity factor. If the modules had converted 3 9.1 1.4 7.7
all the available insolation at the maximum
power point to electricty the relative capa- The efficiency of the modules will decrease
city or module usage would be 100 %. Now significantly in winter due to the high vol-
the relative capacity shows the differernce tage of the load.
of the average module power from maximum po-
wer point. This is also verified by compa-
1800 - - - - -
ring the average module voltage and current
' 000
with the specifications of the manufacturer.
"00

"00

,000
Table I.
.oo
The total monthly module energy output, ca-
pacity factor and relative capacity of the t .oo

30 kW plant .(5/90-3/91)
p ~ : -----1

Month Energy Capacity Rel. Capacity

kWh" "

5 3561 15.9 77% Figure 1.

6 3588 16.6 78% Daily output energy of the array
7 3127 14.0 79% July 1991. Inkoo plant.
8 2642 11.8 73%
9 1599 7.4 78%
10 963 4.3 62% 3.2 Array output power versus irradiance
11 227 1.0 34%
12 21 0.1 10% The hourly mean array output power against
1 81 0.4 19% hourly mean irradiation in array plane in-
2 511 2.5 54% dicate how consistently the plant operates
3 1093 4.8 63% near its maximum efficieny. In figure 2 the
graph of july and in figure 3 the graph of
In winter when the insolation is low the september of output power versus irradiation
difference between actual module output po- are plotted. The modules have operate near
wer and maximum power point is significant their maximum efficiency in September much
and the relative capacity is small. The vol- better than in July.
tage of the load should be smaller in win-
tertime. The monthly average of the module

713
 Table II I.
JO Dai l y average of the module plane irradia-
tion, module output energy, inverter output
energy and system efficiency of the month.
(5/90 • 3/91).
syst.
1$
Month I rr. mod out inv out eft.
10 kWh/m'/day %
5 4.9 0.46 0.44 75.1 %
6 5.1 0.48 0.43 67.9 %
7 4.2 0.41 0.38 81.5 %
1000
8 3.9 0.34 0.27 77.5 %
9 2.3 0.21 0.17 68.3 %
Figure 2. 10 1.6 0.12 0.07 8.4 %
The output power versus irradiation 11 0.7 0.03 0.04 o%
July 1991. 12 0.2 0.00 0.02 o%
1 0.5 0.01 0.03 72.7 %
2 1.0 0.07 0.06 46.8 %
JO
3 1.8 0.14 0.13 50.0 %

When the system efficiency is zero the ener-

:
.,:f gy from grid to batteries has been greater
......
I than the inverter output energy to the grid.
. ~~,;",:, In January the inverter output energy is
,.
10
much higher than the actual module output
energy because the inverter controlling va'
lues are calculated from earlier average in'
solation data.

The hourly average module output energy, in·

Figure 3. verter output and calculated values for the
The output power versus irradiation inverter profile of the month are presented
September 1991. in figure 4.

3.3 System performance "

"
The efficiency of the system performance is I,
calculated by dividing the inverter output 10

energy to the grid by the module output

energy. The energy supplied from the grid to
charge the batteries is excluded from the
inverter output energy in the calculations
of the monthly average of the system effi-
ciency. Energy for warming up the house of
the control systems is monitored but not
used in the calculations. The system perfor-
mance in summertime would have been better Figure 4.
if the regulations of the batteries had been The hourly average of module output ener'
changed before summer. Now there has been gy, inverter output and calculated values
unnecessary charging of the batteries during for the inverter profile. July 1990.
the summer, which affects on the system ef-
ficiency. The table III shows the daily ave'
rage of module output energy, inverter out- The calculated inverter profile follows the
put energy and system efficiency of the hourly insolation. The total daily inverter
month. output energy is decreased to 70 % of that
of the corresponding module output energy.
These values have been calculated from ave·

714
rage insolation data over a ten years pe- these modules are not near the maximum power
riod. The aim of the inverter profile now is point ,but changing by the insolation. The
to maximize the use of the inverter corres- values usable of this system are the insola-
ponding to the insolation. Later the profiLe tion ratio calculated from the moduLe cur-
will be changed to simulate a load corres- rents.The small insolation ratio in December
ponding to the users. is due to mechanical problem with the trac-
king system and also snowlayers at the modu-
3.4 Tracking and fixed module. les.

In table IV the insolation ratio of the fi-

xed and the tracking module is presented.
Tracking module gives according to the inso- 4. CONCLUSIONS
lation 32 % more output than the fixed modu-
le during the measured period. The module relative capacity, which also
shows the differnce of actual power and ma-
ximum power point or usage of rated power,
Table IV has low values only three month in the win-
The insolation ratio of the fixed and the ter. The moduLes have produced 17413 k~h
tracking module. (5/90 - 3/91). from May 1990 to March 1991, the invertter
output energy to grid is 15236 kWh during
Month Insolation ratio the same period. The calculated inverter
5 1.25 output energy from May 90 to March 91 is
6 1.22 18348 kWh. This value should be 70 % of the
7 1.19 module output energy through the period. The
8 ,1.39 32 % more actual inverter output energy is 22 % lesst-
9 1.49 with han the calculated inverter output. The in-
10 1.43 tracking solation during this period is about 20 %
11 1.52 less than average. The use of photovoltaic
12 0.94 power plants in Finland is possible in small
1 1.40 scale and local applications for the period
2 1.37 from February to November. Data has been
3 1.25 collected with a rate of 99.8 % since May
1990 to March 1991.
In figure 5 the hourly average of the trac-
king and fixed module energy output of May
is presented.

20 --------------
II

11
-----'-- I,
, ",I .. - .......

I
,
.0 ...L
I
I

Figure 5.
The hourly average fixed and tracking
module energy output of May 1990.

The two axis tracking system and for compa-

rison a fixed PV system are connected to a
constant resistive load. The power point of

715
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DESIGN TOOL PHOTO FOR SIZING OF HYBRID POWER

SYSTEMS: PROGRAM VERIFICATION

L. M. Manninen and P. D. Lund

Helsinki University of Technology
Department of Technical Physics
SF-02150 Espoo, Finland
tel. (int) 358-0-4513213, telex 125161 HTKK SF,
telefax (int) 358-0-4513195

ABSTRACT. The purpose of this paper is to summarize the verification stu-

dies of the simulation program package PHOTO. Data from a small PV test
system measured during 8 days with averaging time of 2 minutes gives
information about the short timescale performance of the model. Long-term
energy balance comparisons are made on the basis of 5 month data set of
hourly measurements from an experimental 30 kWp grid-connected photovoltaic
plant. The goal of the work has been to expand the analysis from the
performance of single system components to the performance of a complete
system, which consists of PV-array, wirings, controller, battery storage
and load. The results show that the model is well suited for performance
predictions of photovoltaic systems with battery storage.

1. INTRODUCTION and battery was 0.96 V and wl.rl.ng resis-

tance was 0.17 ohms. The peak of the ap-
Program package PHOTO is a user plied load profile was concentrated bet-
friendly analysis and simulation tool for ween 15.00 p.m. and 20.00 p.m., and the
photovoltaic and hybrid power systems maximum load current was about 1.8 A. The
with optional battery storage and back-up shunt regulator voltage limit was 14.0 V
generator. Main features of the model: and initial state of charge (SOC) of the
Time step simulation battery was 100 %.
The battery model makes it possible to
implement control strategies based on 2.2 Measured data
battery state of charge and voltage Data logging and load control was
All subsystems and controls modelled accomplished through PC with A/D and di-
Solves the performance of a complete gital I/O cards. Measurement frequency
system including controls was 1 Hz and averaging time 2 minutes.
Accounts for subsystem interactions Solar insolation on array plane, PV-modu-
The photovoltaic array model parame- Ie temperature, battery temperature and
ters can be determined using parameter load power were used as input data for
fitting procedure FITSOL. If measured the simulation.
hourly weather data is not available, it The simUlation model could also use
can be provided by means of a stochastic measured ambient temperature and wind
generation routine WESIM, which uses ge- speed to calculate the PV-module tempera-
nerally available monthly weather data as ture, but wind speed signal was not avai-
input. The model has been discussed ear- lable when the measurements were perfor-
lier in /1,2/. med.
This paper summarizes the status of Measured PV-module current, battery
the verification studies of the model. So current, load current and battery voltage
far only PV-systems with battery storage were used in the comparisons between si-
have been analyzed. mulation results and real performance.

2.3 Simulation principles

2. SHORT TIMESCALE PERFORMANCE The simulation program input was
adjusted to correspond closely to the
This section deals with short time- operation of the test plant. The simula-
scale measurements gl. vl.ng information tion time step was 2 minutes.
about the dynamic performance of the sys- In the simulation model the voltage
tem. The verification data was acquired of the battery is monitored. If it ex-
from our small PV test plant. ceeds the upper limit, which is 14.0 V in
this case, battery current is limited to
2.1 System configuration such a level that corresponding voltage
For the 8 day test period in May equals the defined upper limit and exess
1990 presented here, the reference system current is dumped.
consisted of one 47 Wp single crystal
module (Arco M7 5), one sealed 12 V, 25 2.4 Main results
Ah (C lO ) battery based on lead-calcium Main results of the comparisons bet-
technology (Support, Neste Battery Limi- ween simulated and measured performance
ted), shunt regulator with adjustible are shown in Tables 1 and 2.
voltage limit and variable resistance Mean bias error MBE, mean average
controlled by the data acquisition PC for error MAE, root mean square error RMSE
the hourly load profile simulation. and DELTA presented in these tables are
Estimated value for the constant calculated as follows:
voltage drop in the diodes between module

716
 SOC of the period is about 50 % of the
nominal capacity.
( 1)
Table 2. Cumulative system balance
over the 8 day measurement period.
Notation: [Q]: cumulative charge, [GJ:
(2)
cumulative gassing current, Le. cur-
rent losses in the battery, [OJ: deg-
ree of discharge at the end of the
measurement period, [E]: cumulative
energy.
RMSE= -1 LN (X!tlm~-X_)2 (3)
Ni=l SIMULATED MEASURED DELTA(7-)
PV-MODULE,[O]. ............. (Ah) 80.8 82.1 -1.5
N
L(x.FxlIIMJ,t)
REGULATOR,[Oj ............... (Ah)
BATIERY,CHARGE,[Oj ........ (Ah)
23.6
37.9
24.0
37.5
-1.8
1.0
DELTA = 100. _i=_l_ _ __ (4) BATIERY,DISCHARGE.[O]. .. (Ah) 39.7 39.0 2.0
N BATIERY.[Gj .................... (Ah) 1.B - -
LXIIIMJ,t BATIERY,[Dj .................... (Ah) 3.6 0.9 -
i=1 LOAD,[Oj ......................... (Ah) 59.1 58.3 1.4
where X. 1m ,1 is the simulated value of ti- PV-MODULE,[Ej .............. (Wh) 1101.5 1114.2 -1.1
mestep i, xmea ,1 is the measured value of REGULATOR,[E]. .............. (Wh) 329.0 335.9 -2.1
timestep i and N is the number of simula- BATIERY,CHARGE.[Ej ........ (Wh) 518.4 506.7 2.3
tion time steps. BATIERY,OISCHARGE,[Ej ... (Wh) 494.6 493.4 0.3
LOAD,[Ej ......................... (Wh) 746.3 746.3 0.0
Table 1. Mean bias error (MBE) , mean
average error (MAE) and root mean
square error (RMSE) for the 8 day mea- The degree of discharge (DOD) of the
surement period. battery gives the number of Ah's between
current SOC value and full battery (i.e.
SOC= 100 %). Measured DOD at the end of
MBE MAE RMSE
the period is obtained by charging the
PV,CURRENT ...... (A) -.007 .035 .053 battery until float conditions have been
DUMPED,CURRENT .. (A) -.002 .040 .158 reached and subtracting the calculated
cumulative gassing current from the cumu-
BATIERY,CURRENT.(A) -.002 .068 .159 lative measured charge current.
LOAD.CURRENT.. .. (A) .004 .005 .010 This method gives only an approxima-
BATIERY,VOLTAGE.(V) -.110 .227 .261 te result, since SOC= 100 % is not a sin-
gle well-defined point, and battery per-
PV.POWER ........ (w) -.068 .463 .707 formance varies from cycle to cycle. This
DUMPED,POWER .... (W) -.037 .556 2.206 can also be seen from Table 1. Subtrac-
BATIERY,POWER... (W) .056 .900 2.175 ting the discharged Ah's from the charged
Ah's gives the minimum of the DOD at the
end of the period, since the gassing gur-
The differences between simulated rent is then assumed to be zero. In this
and measured current and power values are case the result is 1.5 Ah. In the real
small when compared with the maximum cur- charging process the result should have
rent and power output of the PV-module been higher, whereas the result is only
(about 3 A and 50 W, respectively). 0.9 Ah. The simulated value 3.6 Ah seems
The MAE of the battery voltage is 38 reasonable.
mY/cell. The result is good, considering The simulated system performance
that the battery model is as general as agrees well with the measured data. The
possible, L e. it is valid over a wide errors are comparable with the measure-
range of current, temperature and SOC ment accuracy. The method is thus a sui-
values. The effect of the voltage error table tool for estimating the short ti-
on the system energy balance is small, as mescale performance of photovoltaic sys-
can be seen from Table 2. tems.
The accuracy of the voltage model
can usually be improved for certain spe-
cial conditions (e.g. constant temperatu- 3. LONG-TERM PERFORMANCE
re, small currents, high SOC values) by
limiting the validity range. For this 8 This section discusses long-term
day test period a better result could energy balance comparisons between simu-
have been obtained by using this specific lation results and measurements from an
data set to fit the parameters of the experimental grid-connected photovoltaic
battery model. This is not sensible, sin- plant that is situated on the southern
ce this specific system configuration and coast of Finland.
the special conditions prevailing during
this particular period would have been 3.1 System configuration
over-emphasized. The 30.2 kWp photovoltaic array con-
Cumulative calculated gassing cur- sists mainly of C-Si modules (Arco M660) .
rent is only about 5 % of the total char- There is also a 1.9 kWp field of amor-
ge put into the battery during the period phous modules included (Chronar CSB.13).
(Table 2). The value is low due to the The tilt angle of the modules is 45 deg-
battery type and low voltage level of the rees. The rated 10-hour capacity of the
shunt regulator. About 30 % of the PV- battery storage is 2860 Ah, and there are
module production is dumped through the 53 cells connected in series. The energy
voltage regulator. The calculated minimum capacity is thus around 300 kWh.

717
 The system also includes a program- 3.4 Main results
mable system controller, a battery char- Main results of the comparisons bet-
ger and an inverter suitable for grid- ween simulated and measured performance
connected applications with nominal power are shown in Tables 3 and 4.
of 30 kW. The battery charger is utilized
to prevent prolonged operation in deeply Table 3. Mean bias error (MBE), mean
discharged state, and it is not used in average error (MAE) and root mean
normal operation. Estimated value for the square error (RMSE) for the 5 month
constant voltage drop between array and measurement period.
battery is 0.85 V (diodes) and calculated
equivalent wiring resistance is 0.0086 MBE MAE RMSE
ohms. The hourly load profile of the PV.CURRENT... ... (A) -.25 3.87 9.09
plant was adjusted on a monthly basis
during the measurements to correspond to BATTERY,CURRENT.(A) -.01 4.28 9.47
the typical average solar radiation pro- lOAD.CURRENT .... (A) -.63 1.68 3.47
file of that period. BATTERY.VOlTAGE.(V) .42 2.85 4.03
PV,POWER ........ (w) 48.65 402.82 1043.01
3.2 Measured data
The hourly average values for a 5 BATTERY,POWER ... (W) 48.88 485.59 1097.81
month period starting from May 1, 1990 lOAD,POWER ...... (W) -32.11 126.31 330.02
were used in the comparisons. The total
insolation on array plane of the period
was 607 kWh/m2, which is about 54 % of The differences between simulated
the total expected yearly insolation. and measured current and power values are
Solar insolation on array plane, small when compared with the maximum cur-
ambient temperature, battery temperature, rent and power output of the PV-array
wind speed, battery charger current and (about 220 A and 30000 W, respectively).
load power were used as input data for The MAE of the battery voltage is 54
the simulation. mV/cell, which is more than in the short
Measured PV-array temperature, array timescale comparison. In this case cer-
current, battery current, load current tain voltage model refinements are re-
and battery voltage were used in the com- quired. The effect of the voltage error
parisons between simulation results and on the system energy balance is small, as
real performance. can be seen from Table 4.

3.3 Simulation principles Table 4. Cumulative system balance

The simulation time step was 1 hour. over the 5 month measurement period.
The simulation model uses measured am- Notation: [Ql: cumulative charge, [Gl:
bient temperature and wind speed to cal- cumulative gassing current, i. e. cur-
culate the PV-module temperature. rent losses in the battery, [El: cumu-
Photovol taic module parameters were lative energy.
fitted to measured performance curves
using program FITSOL included in the pro- SIMULATED MEASURED DElTA(%)
gram package. Inverter efficiency curve
PV-ARRAY,[O] ................. (kAh) 128.6 129.5 -.7
was also supplied to the simulation prog-
ram. DUMPED.[O] .................... (kAh) 1.4 - -
In the simulation model voltage and BATTERY,CHARGE.[O] ........ (kAh) 45.9 48.1 -4.7
SOC of the battery are monitored. If both BATTERY.DISCHARGE.[O] ... (kAh) 35.5 37.7 -5.9
values exceed their upper limits, battery BATIERY.[G] ..................... (kAh) 10.5 - -
current is limited to such a level that CHARGER.OUTPUT.[O] ....... (kAh) 14.0 14.0 .0
corresponding voltage equals the defined
AC-lOAD.[Q]. .................. (kAh) 130.7 133.0 -1.7
upper limit and exess current is dumped.
From the operational point of wiew INSOLATION,[E] ..... (kWh!sq.m.) - 607.2 -
this method approximately corresponds to PV-ARRAY.[E] ................. (kWh) 14929.3 14750.6 1.2
the control strategy used in the real DUMPED.[E] .................... (kWh) 177.2 - -
reference system where current is cont- BATTERY.CHARGE,[E] ........ (kWh) 5494.1 5588.3 -1.7
rolled via disconnecting subsequently BATTERY.DISCHARGE.[E] ... (kWh) 3752.5 4026.2 -6.8
separate groups of photovoltaic module
DC-lOAD,[E] ................... (kWh) 14617.9 14735.9 -.8
strings if upper limits of both voltage
and SOC have been exceeded. AC-lOAD,[E] ................... (kWh) 13139.0 13139.0 .0
The measured number of PV-fields
connected to the system by the controller The simulated and measured current
is used as input for the simulation. This efficiencies of the battery are 77.3 %
is justified, since in grid-connected and 78.3 %, respectively. For energy ef-
applications all fields of the array ficiency we get 68.3 % and 72.0 %, res-
should be kept connected during normal pectively. The values are obtained by
operation. The control strategy of the dividing discharged Ah's or Wh's by char-
plant was not optimized to take this into ged Ah's or Wh's. The simulated and mea-
account at the time when the measurements sured inverter efficiencies are 89.9 %
were performed. The fields should be dis- and 89.2 %, respectively. The calculated
connected to protect the battery only minimum SOC of the period is about 40 %
during system failure or when the inver- of the nominal capacity.
ter cannot accept the peak output power In an ideal case, the dumped energy
of the array. should be zero in the simulation, since
Also the battery charger current is the disconnection of PV-fields is taken
given as input for the simulation model, into account. However, the simulation
since battery charging from the grid is shows that some exess energy is being
not performed during normal operation. dumped. The discrete time step used in

718
the simulation causes some error, since The goal of the work has been to
the fluctuations within one time step expand the analysis from the performance
have to be neglected. Part of the de- of single system components to the per-
viation may be caused by the fact that formance of a complete system, which con-
the system controller doesn't work with sists of PV-array, wlrlngs, controller,
same time step as the simulation. The battery storage and load. The results
error is only 1.1 % of the total PV-array show that the model is well suited for
current production. short- and long-term performance predic-
The battery current and the battery tions of photovoltaic systems with batte-
voltage for the first 50 days of the pe- ry storage.
riod are displayed graphically in Fig. 1. At this stage, however, the effect
The effective internal resistance of the of battery ageing is neglected, and no
battery model seems to be too high, i.e. thermal model for the battery has been
the simulated battery voltage is too high implemented. Also, the range of applica-
during charge and too low during dischar- ble battery types could be expanded and
ge, even though the trend of the measured the battery voltage model could be re-
voltage curve is reproduced well. fined, even though the current program
The calculated PV-module temperature version reproduces the actual system
follows the measured value well. Typical- energy balance quite well. Further stu-
ly the error is less than +-2'C. The si- dies are thus required.
mulated PV-module temperature is calcula- In future also hybrid systems inclu-
ted according to /3/. ding wind turbine, backup generator and
The simulated system performance hydrogen storage will be studied.
agrees well with the measured data. The
method is thus applicable for long-term
performance prediction of photovoltaic 5. REFERENCES
systems.
1 L.M. Manninen and P.o. Lund, Dynamic
Simulation and Sizing of Photovol-
4. CONCLUSIONS taic and Wind Power Systems, Proc.
9th E. C. Photovoltaic Solar Energy
A leading feature in the development Conference, Freiburg, Fed. Rep. of
of the simulation model PHOTO has been to Germany, pp. 546-549 (1989).
consider the system as a whole, i.e. to
model all the subsystems and controls, 2 L.M. Manninen, P.D. Lund and A.
and to account correctly for the subsys- Virkkula, PHOTO - A computer simu-
tem interactions. lation program for photovoltaic and
This paper summarizes the status of hybrid energy systems. Report TKK-F-
the verification studies of the model. A670. 1990.
Two data sets are used, one for the ana-
lysis of short timescale dynamic perfor- 3 M. K. Fuentes, A simplified thermal
mance, and one for the analysis of long- model of photovoltaic modules. San-
term energy balances. dia National Laboratories, SAND85-
0330 (1985).

200

ISO
~ 100
::I
rrI
~ 50
0
c
;u
;u
M
z -50
-I

~ -lOa
- 150
I I L I I i I I I I t
15 30 <5 DAY

140

lOa

90L-----------------------------------------------------------------------
I I I I i i I
15 30 45 DAY

Figure 1. The battery current and the battery voltage for the first 50
days of the 5 month period.

719
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

DEVELOPMENT POTENTIAL OF SMALL-SCALE

GRID-CONNECTED RESIDENTIAL PHOTOVOLTAIC SYSTEMS

K. Peippo and P.O. Lund

Helsinki University of Technology
Department of Technical Physics
SF-02150 Espoo, Finland

ASTRACT. 1-5 kWp grid-connected and building-integrated PV-systems are

considered most potential applications for wider future use of PV. The
objectives of this study under the Finnish Advanced Energy Systems National
Program (NEMO) are to identify and utilize the possibilities for improvements
and cost reductions of the system. These include optimal integration of the
PV-array in the building and the combination of PV-production and efficient
electricity end-use to improve the overall system economy. In the first phase
a 3 kWp pilot plant has been installed in Helsinki University of Technology
Educational Centre (60 N), Finland. Present day technology optimally
0

integrated in an electricity efficient building may reduce the investment

cost enough to attract a limited market although a major decrease in module
prices is still necessary for market penetration.

1. BACKGROUND The means to achieve these goal s

include:
Photovoltaic power systems have reached • Full-scale instrumented pilot plant
the state of economic competiveness in • Development of simulation tools
small-scale remote and special • System studies and optimization
applications. However, in order to increase • Component development
their share in future power production, the
PV-systems must be grid-connected and
integrated in the power system.
Small-scale grid-connected and building ,1/
integrated PV-systems are considered most
promising as an intermediate stage between -0-
stand-alone and large utility systems.
Following advantages may be identified:
/1'
• Building-integration reduces construction
costs.
• Simple design: no batteries of back-up
power.
• Due to higher retail prices, the value of
electricity is higher for the consumer than
for utility thus lowering the threshold for
investment.
• PV can be combined with efficient end-use
for better overall system economy.
A possible future scenario would then
be a partly dispersed energy system with
numerous co-producers having two-way
electricity flow between the utility grid
feeding their surplus electricity to the Figure 1: The concept of grid-connected PV-
grid and withdrawing energy from the grid "ystem
during low insolation. The concept of grid-
connected residential PV-system is shown
schematically in Fig. 1.
3. EXPERIMENTAL: 3 kWp PILOT PLANT
In the first phase a 3 kWp pilot system
2. OBJECTIVES AND APPROACH has been integrated in an existing building
as a reference installation. To demonstrate
The main objectives of the three year the applicability of the technology, proven
study (1990-1992) are: components in series production were
• Demonstration of PV in a northern climate selected for the plant. The preliminary
• Insight into the technology investigations of the availability of
• Recommendations and guidelines standard components, cost of the system and
• Identification of improvements and cost the typical residential demand of
reductions electricity indicated that a nominal system
Evaluation of the combined effect of size of the order of 3 kWp was appropriate
efficient end-use technology and PV-system for the pilot plant (Fig. 2).
• Design and possible realization of an
advanced system

720
 PV-ARRAY
54
Arco Solar
M55
2862W

- -- - -- -
1 environmental : - - - - - - - - 1
south·lacing, tilt 3SO
(001 mounted. retrot~
1 Insoletlon horizontal. OIfTay plane, 1
1 temperall.lre amblent,module 1

I 1
I ......~v.ol.t~~ •.~~e.n~ ••• • ~ _ •••
1 ~-,--~

I I INVERTER
I 1 Photoelectric
1 1 SI-3000
I 1 3000W
I I
1 • . . . !n~~~e~ AC·power 1
1 .. . •.• . .'1 ' .• •. . . '"----,----'

I 1
1 1
lesldentialloads:
1 I LOADS Ilghlino. appliances
1 I
1 • •e~~~y. t~~ ~rid
1 ..... 'j1 ' ...... .

I 1
I 1
UTILITY
1 PC I GRID
1 • . . . . . . . .' I
I 1
1 I
1 I
I INSTRUMENTATION I
I ~~__~ I
I I
I I
- - - __ _ J

Figure 2: 3 kWp pilot plant

4. SIMULATION APPROACH accurate pV-module and inverter models were

'sed to estimate the power production of
4.1 The Model the PV-array.
A photovo1taic and hybrid plant
simulation program PHOTO /1/ was used to
predict the operation of the grid-connected 4.2 Energy Balance of the System
system. An electricity-efficient The simulation resulted in the energy
residential house was considered the balances in Fig. 3. The annual production
reference installation with the annual of the 3 kWp system equals the annual
demand of electricity reduced from the 1988 demand of the electricity-efficient
Finnish average of 4500 kWh to 2300 kWh reference application with about 25% of the
(excluding heating). This reduction is production directly fed to loads and the
considered feasible by the introduction of rest "stored" in the utility grid. However,
the best commercial technology for lighting due to the annual and diurnal mismatch of
and appliances /2/. The average hourly the supply of solar energy and the demand
electricity consumption profile of a of electricity in Finland, this
typical Finnish house-hold were used as an dimensioning principle may not be
ec~nomicly. optimal as it results in large
input with the appropriate scaling to
account for the efficiency improvements. grld-fractlons (share of production fed
The hourly measured insolation and into the grid) . Based on further
temperature data of the Helsinki Test simulations, Fig. 4 shows the share of the
Meteorological Year (TMY) together with produced PV-electricity that can be
immediately used in loads resulting in
maximum economic benefit for the PV-user.

721
 kWh/month
LOAD E ERGY FRO " PV
D PV SURPLUS ENERGY TO GRID

2 3 4 5 6 7 8 9 10 11 12

month
Figure 3: Monthly elect ric i ty balance o f th e slmulated
. s yste m.

1
5.2 System optimization
.8 With no batteries or back-up power the
remaining question is the sizing of the
inverter with respect to the PV-array rated
.6 capacity. With the typical inverter
Direct efficiency curves rapidly decreasing on low
power, the theoretically optimal nominal
loads .4 capacity of the inverter would be slightly
lower than the rated capacity of the PV-
.2 array . Of marginal intere st may be the
choise of optimal inclination angle to
improve economics by reducing the summer
0 t i me surplus production.
0.0 0.5 1.0 1.5 2.0
PVl10ad 5.3 End-use efficiency
There is little point in considering
Figure 4: Share of PV-electricity directly the applicati on of PV unless the most
fed to loads as a function of system efficient electricity end-use technology is
dimensioning. applied. Following the guidelines of /2/ it
is assumed that the present average
residential demand could be reduced by
approximitely 50 % with minor additional
5. IMPROVEMENT POSSIBILITIES cost. If the total cost of electricity
services (purchaced electricity, efficiency
In the following, various aspects for measures and possible PV-system) is to be
improving the system performance and maintained, cheap negawatts would allow a
economy are considered. The pilot plant fraction of the services to be produced
configuration with typical Finnish also with higher PV production cost.
residential load and Helsinki weather is As an example, the "affordable" PV
used as a reference . share of the total demand is calculated as
a function of the PV-electricity price
5 . 1 Micro-climate assuming 50 % efficiency improvement with
The insolation conditions are, of the average cost of $0.025 per saved kWh.
course, uncontrollable, but the variation The cost of grid-electricity is $O.l/kWh
of local weather conditions still makes it and the price of surplus electricity
possible to situate the first systems on supplied to the utility grid is $0.03/kWh
better locations compared to the standard reflecting the avoided costs of the
reference weather. For instance, in Finland utility .
the annual insolation on coastal areas and
in the archipelago may be up to 10 % higher
than in inland on the same latitude.

722
 1.5.----------------------------. 5.7 Combined effect of measures
In Table 1, indicative calculations are
performed for a 3 kWp system by estimating
1.2 the effect of the suggested improvement
aspects. Both the cost of produced PV-
.9 electricity and the virtual cost of
electricity services are calculated. Also,
PV/load the total PV investment cost is compared to
.6 the typical construction cost of alSO m2
residential house ($ 200 000 in Finland) .
With optimal integration of the system
.3 during the construction and possible
investment aid for promoting the first
systems, the total PV investment could be
reduced to approximitely 10 % of the total
.2 .4 .6 .8 1 construction cost. This may be enough to
create a small market for the systems.
PV cost $/kWh
Fig. 5: Maximum PV-fraction maintaining tCle 6. CONCLUSION
cost of electricity services with 50
efficiency improment at $0.025/kWh. Various measures to improve the cost-
effectiveness of grid-connected residential
PV-systems with present day technology
5.4 Optimal integration combined with efficient end-use can bring
Compared to the retrofit of standard the PV-system investment cost to a level
PV-modules in a building, the well-designed that can make these systems attractive to a
integration of unframed PV-laminates in the limited market also in northern latitudes.
building during the cost ruction potentially However, a major decrease in module cost is
result in major savings both in materials still a prerequisite for large-scale market
and labour, especially if conventional penetration in a cost-effective way.
building components can be replaced with
PV-elements. In the future, the integration REFERENCES
wich no PV-specific costs should be aimed
/1/ L.M. Manninen et al. A Computer
at. simulation program for hybrid energy
systems. Helsinki University of Technology,
5.5 Advanced technology Dept. of Technical Physics, Report TKK-F-
Finally, the future option for A670, 1990.
inexpensive mass-produced PV-modules is
still the most potential source of improved /2/ T.B. Johansson, B. Bodlund, R.H.
cost-effectiveness. A goal of $l/W is Williams (eds.). Electricity, Efficient
cons ired feasible for large-scale a-Si thin End-Use and New Generation Technologies and
film production. Their Planning Implications. Lund
University Press, Lund, Sweden 1989.

Table 1: Indicative cost break-downs for 3 kWp systems.

Reference 1990 Optimal Integration Optimal Integration, Optimal integration, Advanced

·efficlent end-use efficient end-use, technology
40~investmenl aid.

PV-array $ 19500,00 15000,00 15000,00 15000,00 3750,00

Inverter $ 4500,00 3750,C~ 3750,00 3750,00· 1875,00
integration $ 5000,00 2500,00 2500,00 2500,00 0,00
engineering etc $ 5500,00 1000,00 1000,00 1000,00 500,00
VAT, investment aid $ 7066,27 0,00 0,00 -8900,00 0,00
total Investment $ 41566,27 22250,00 22250,00 13350,00 6125,00
fraction of total construction costs 0,21 0,11 0,11 0,07 0,03
PV production kWh/a 2550,00 2550,00 2550,00 2550,00 2760,00
load-fraction 0,35 0,35 0,25 0,25 0,25
demand kWh/a 4500,00 4500,00 2300,00 2300,00 2000,00
cost of Improved efficiency $IkWh 0,00 0,00 0,02 0,Q3 0,00
price of surplus electricity $IkWh 0,03 0,03 0,03 0,03 0,08
cost of grid electricity $IkWh 0,10 0,10 0,10 0,10, 0,13
cost of PV-electrlcity $IkWh 1,39 0,77 0,77 0,49 0,15
~Sl of electricity services $IkWh 0,86 0,51: 0,47 0,31 0,11

5.6 Social aspects

As an environmentally benign
alternative with considerable future
potential in energy system. To advance the
technology and create a market a set of
subsidies can be considered justified in
the introductory phase. In Finland the
investment aids for the introduction of new
technology on other areas have been up to
40 % of the investement and could be
applied for PV, too.

723
tafH EUROPEAN PHOTO VOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

EIGHT YEARS EXPERIENCE WITH THE 300 kWp SOlAR P<MR PLANT ON THE
ISLAND PELLIaM (FRG) AND THE EXTENSION TO A 1.25 MIl HYBRID SYSTEM
(ADDITIONAL 300 kWp PHOTOVOLTAIC AND 600 kW WIND ENERGY)

C. Kruse, H.J. Lowalt, K. Maass

TELEFUNKEN SVSTEMTECHNIK GMBH
Industriestrasse 23-33
0-2000 Wedel / F.R.G.
Telephone: +49 4103 60 3823
Telefax: +49 4103 60 4701

ABSTRACT. In the year 1983 the 300 kWp PV Pilot Plant was built on the North Sea island of Pellworm with
financial support of the CEC, BMFT (Ministry of Research and Technology) and TELEFUNKEN SVSTEMTECHNIK GMBH
(former AEG Aktiengesellschaft). The community of Pellworm disposed a site of 28,000 m2, still used as
graze land for sheep. Since October 1989 the utility board of Schleswag is the new owner and operator of
the whole plant. The solar power plant has been operated for 7 years under the harsh environmental condi-
tions of the North Sea climate to supply the recreation centre of Pellworm in a stand-alone operation mode
with batteries. In 1988 additionally three Wind Energy Converters, 33 kW each, were installed for direct
grid connection in order to supply more energy to the recreation centre in spring and autumn. In 1991 the
PV-array will be extended to 600 kWp and in addition a wind energy converter of 550 kW will be installed.
Then a 1.25 MW Hybrid System will be in operation on the island of Pellworm.

1. INITIAL DESIGN Two self-commutated inverters of 75 kVA each

are used. According to the momentary load demand
1.1 Summary the inverters can be switched on or off individu-
The converted electrical energy of the 300 ally controlled by the superimposed microproces-
kW PV-plant (Fig. 1) supplies the recreation cen- sor system. The inverters' internal control cir-
tre of Pellworm in a stand-alone mode. Surplus cuitry is provided for parallel operation of
energy is fed into the grid. To guarantee the these two inverters. It allows each inverter to
stand-alone operation mode, a battery with a 10- be started and stopped at any time. Actually
hour capacity of 6,000 Ah is installed. To real- these start and stop signals are generated by the
ise an operation as buffer battery and a simulta- monitoring system according to the energy demand
neous charging, the battery is divided into two of the recreation centre and the efficiency opti-
operational parts. The grid connection is possi- misation.
ble for the consumer as a back-up energy supply. Additionally each inverter has its own moni-
As a superimposed control and monitoring sys- toring unit. A number of sensors are distributed
tem a microcomputer is integrated. To control the over the ci rcui try. Thus every operation status
300 kWp peak power the solar generator is divided can be displayed. If any failure is detected, the
in 21 switchable groups. inverter is switched off automatically before any
For the energy storage lead acid batteries damage occurs. The monitoring unit stores all in-
are used. Each individual cell has a rated volt- coming failure signals according to their chrono-
age of 2 V, therefore 173 cells had to be connect- logical order. These signals can be read out by
ed in series to provide the necessary system volt- service personnel. The failure signals and their
age of 346 V. The battery is divided into two sequence order give useful hints for correction.
parts so that it is possible to work with one As inverter for the grid feed-in a line com-
hal f of the battery as a buffer battery, while mutated type with two six-pulse thyristor bridges
the other half is charged in a defined way. in parallel is used. This inverter can operate
wi th max. 1,000 Amps DC at an output voltage of
380 V (3-phase). In the reverse direction an emer-
gency charging for an eventual battery recondi-
tioning is possible.
As already mentioned, the automatic opera-
tion mode is accomplished by a microprocessor sys-
tem. To use the converted energy in an optimum
way the microprocessor has to take over all
switching and control decisions. This is only pos-
sible by a very complex monitoring and control
system, so all essential values and signals of
the whole plant are collected by the microproces-
sor.

Figure 1: Bird's Eye View

of the 300 kW Power Plant

724
1.2 Data system of the type DAM 800 is installed.
Improving Energy Output
Rated Peak 'Power: 300 kW The energy management is modified . Special-
Nominal DC-Voltage: 346 V ly, the operation parameters of the battery are
Solar Generator Field changed and the two-busbar system with its two
- Overall Area: 28,000 m22 separated battery blocks is converted into a one
- Structure Area : 16,500 m2 busbar system (Fig. 2). Special care has to be
- Module Area: 4,500 m taken of the treatment of the aged battery cells.
Building : 15 m x 10 m Monitoring Panel/Load Management
Battery Capacity (10 h) : 6,000 Ah In the recreation centre a monitoring panel
Self Commutated Inverters: 2 x 75 kVA for the indication of significant plant operation-
Output Voltage: 3 x 380 V ± 2% al data with remote control switches for the oper-
Frequency: 50 Hz ator's attention is installed . An automatic
Total Harmonic Distortion : 3% max . switching system is set-up to supply the recrea-
Line Commutated Inverter: 450 kVA tion centre at night time with cheap electricity
Output Voltage: 3 x 380 V from the public grid. The load demand of the rec-
DC - Input Current: 1,000 A reation centre increases constantly. The energy
Transmission Line delivery to the grid and recreation centre is
- Length: 1,300 m shown in Fig. 3. Indications to influence the
- Voltage : 20 kV load demand are given by the Monitoring Panel.
Modem
1.3 Basic Operation Experience Information about the plant status and mea -
The effort of the former owner and operator sured data are available on the video terminal at
of the PV-plant, the community of Pellworm, re- the plant . The adaptation for the modern is done .
sulted in a total different operation mode as A Telemodem type TM 1200, IAWD is installed.
originally implemented. Due to a better benefit,
if all converted energy can be directly used by
the consumers, more and more consumers have been
installed into the recreation centre in order to
prevent any energy supply into the publ ic grid.
The reimbursement by the utility as about 0.09 OM
per kWh. On the other hand the commun i ty had to
pay about 0. 25 OM per kWh for purchased energy.
With electrical water heating systems in the
beginning followed by electrically driven heat
pumps of each 25 kW and 15 kW power demand, the Power
computer aided operation philosophy was perverted. Codmonlng
and
All hardware equipment like inverters, Dlstrlbullon
switch-gears and PV-array were very reliable . The
technical availability reached as much as nearly
100%.
Often the control system could only act as
safety system for the batteries in order to
switch off the whole system to prevent the batter-
ies from deep-discharge. This happened normally
when one had forgotten to switch-off the big con-
sumers manually in the evening after the regular
opening hours of the recreation centre with its
swimming pools and restaurant. In the next morn-
ing both battery halves, connected to the two Figure 2: Block Oiagramme of the
busbar system, were discharged. Small Hybrid System
Energy was supplied to the public grid only,
if the big consumers had not been switched on
over a longer period.
kWh (,n thousand)
One of the operation philosophies was to
take special care of the batteries ' treatment. 30 28.7
After a system shut-down due to deep-discharge of 26,1
both battery halves the only possible operation 25
mode was to recharge the batteries. For at least
8 to 12 hours (in summertime) the PV-array could
not be used to supply any consumer. 20
Due to the above described operation chang- 16.8
es, the computer aided monitoring and control sys-
tem had to be modified in order to optimise the t5

kWh-output of the plant by increasing the over-

all reliability and availability by simplifica- 10
tion of the operating modes. 8.5
Wi th the financial support of the CEC Fol-
low-on Programme and based on the long-term expe- 5
3.4
2,3
rience of TELEFUNKEN SYSTEMTECHNIK with this 0.8 0,3
plant, the following improvements are performed.
Jan. Feb Mar Apr May Jun Jul Aug Sap Oct Nov. Dec
1.4 Improvements
In the frame of the CEC Follow-on Contract
the following improvements on the 300 kW Pilot [=::::J Grid Supply !::::::~:::::n To Recreation Centre
Plant Pellworm are performed:
Software Splitting
This work is done to organise a separation
of the system control eqUipment and the data ac- Figure 3: Energy Delivery to the Grid
quisition system . A separated data acquisition and Recreation Centre

725
2. THE HYBRID SYSTEM
2.1 Summary
An optimum of experience with large PV
plants and/or wind converter plants will be
earned by a close cooperation with the local util-
ity board Schleswag. In the first step three wind
generators of 33 kW each (see Fig. 4) were in-
stalled in 1988. Schleswag will be the future own-
er and operator of the whole solar/wind plant.
The community of Pellworm authorised us to use
the site south of the existing plant for the real-
isation of our new project (Fig. 1 and 5). This
site will be used in future for agricultural as-
pects, too.
All necessary infrastructure can be mini-
mised due to an optimum use of the existing equip-
ment. Now the capacity of the 7-year old lead ac- ", ==0
~
=
id accumulator is at the end of its life time and ',,= =
will be recycled. This will reduce the new plant- ',=
\c::=;,
control and acquisition system. The battery room
inside the existing building will be sufficient ......... New Field
300 kWp
to build up the new switch-boards and monitoring
room for visitors.
Wi th respect to the costs a new generation
of industrial inverters will be modified. Due to
minimised grid influences our intention is to
adapt a 12-pulse line commutated inverter. This Figure 5: Map of the new
inverter type seems to be an optimum solution to Hybrid Plant Pellworm
overcome the utilities' requirements with respect
to harmonic distortions. For weak grids (e.g. at
the end of long transmission lines) and a power
supply by large PV plants this is a most impor- 20 kV Grid of Pellworm Island
tant argument.
The newest module types PQ 10/40 HRD 50 of
TELEFUNKEN SYSTEMTECHNIK will be foreseen. Con-
nected to a DC voltage of up to 900 V the opera-
tion voltage will be higher than at all other ex-
isting pilot plants before. In combination with
wind energy converters, which will have their max-
imum energy output at this site in autumn, winter
and spring, the ti It angle of the PV structures
wi 11 be optimised for the summer season. So the
tilt angle will be lower than at the previous
plant. The result will be a reduction in struc-
ture height and consequently the row distances
can be shortened. The block diagramme of the
1.25 MW Hybrid Power Plant of Pellworm is shown Generalors L-_---! Generator Generator
in Fig. 6. 1 - 3 4 5
(asynchron) (synchron) (asynchron)

300 kW 3.33kW 300 kW 300 kW 250 kW

in 1983 In 1988 in 1991 in 1991 in 1991

Figure 6: 1.25 MW Hybrid

Power Plant Pellworm

2.2 Design Data of Extension to the Hybrid System

Rated Peak Power PV 300 kW
DC System Voltage up to 900 V
Line commutated inverter
(12 pulse) 300 kVA
Wind Energy Converter
synchronous generator
Figure 4: 33 kW Wind Energy AC/AC converter with
Converter DC intermediate circuit 300 kW
Wind Energy Converter
asynchronous generator 250 kW (Option)

726
3. SUMMARY AND PROJECT GOALS
Existing 300 kW PV Power Plant
• Consumer: recreation centre
Summer season operation
Stand-alone mode of operation with
batteries
• Surplus solar energy supplied to the grid
Extension of the Opening Time of the
Recreation centre
• from spring to autumn
• therefore installation of 3 x 33 kW
Wind Energy Converters
Extension to 1.25 MW Hybrid System
• NEW OWNER: SCHLESWAG / utility company
• Battery: End of life time - recycled
New Project Goals and Improvements
• Optimisation of energy output over the
whole year with grid supply only
Space minimisation and decrease of field
losses
New module type PQ 10/40 HRD 50
efficiency> 12,5%
system voltage up to 900 V DC
qualified by
Joint Research Centre Spec. 503
Low cost constructions including earthing
and lightning protection, foundations,
cables and labour.

727
10TH EUROPEAN PHOTOVOLTAlC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON,PORTUGAL

AUTONOMOUS PHOTOVOLTAIC SYSTEM WITH AUXILIARY SOURCE FOR

BATTERYCHARGING

V.A.P. van Dijk~, E.A. Alsema~, R.A.W. Albers~, T. Degner P, H. Gabler P, E. Wiemken P
U Department of Science, Technology and Society, University of Utrecht
Oudegracht 320, NL-3511 PL Utrecht, The Netherlands
II Department of Physics, University of Oldenburg, P.O.Box 2503, D-2900 Oldenburg, Germany

ABSTRACT. A Photovoltaic/Battery /Gasoline-Generator system for the autonomous supply of an annual

load of 1000 kWh was set up and operated for several months. The system has worked reliably, first
experimental data are shown in the paper. By numerical modelling of the realized system a detailed
annual energy flow chart was calculated, the results were compared to the empirical data. The system
model was used to test the effect of different operational strategies on the overall system behaviour.

1. INTRODUCTION trol strategy that ensures a reliable electricity supply

at the lowest cost.
A photovoltaic system that consists of a PV generator
and a battery storage can supply any load with a high
degree of reliability, provided that the system has been
sized appropriately. However, when only small yearly
energy shortages are required or in climates charac-
terized by periods with prolonged low irradiation, a
PV /battery system will be a relatively expensive sys-
tem, because part of the installed capacity is only used
for a small period of time. In such a case the addition
of a fossil fuelled generator to the PV /battery system PY'SU8RRRRYS L£RO RCIO MOTOR/
BAlfERV
can be economically attractive. The PV generator and 1: 9GOWp
2~ V
GENER~TO~

1kW.1
the battery can have a smaller size and their capacity ~OOAh

will be effectively utilized. The fossil fuelled generator

can fill in periods of prolonged high energy demand
or low irradiation. Moreover, the generator will be Figure 1 The layout of the experimental PV /battery
operating only for short periods, resulting in a long system with the gasoline generator.
lifetime and small environmental nuisance. In simula-
tion studies with the computer model SOMES, the per- The DC system voltage is 24 V. The 960Wp PV gener-
formance of PV /battery / generator systems was com- ator consists of 24 RSM 40 crystalline silicon modules
pared to PV /battery and generator/battery systems which are divided into 4 subarrays, directly coupled to
and their economic attractiveness for a household ap- the battery. Charge regulation is done by switching
plication was demonstrated /1/. one or more of the subarrays and by disconnecting the
The inclusion of a motor generator in a PV /battery inverter, if the battery voltage reaches a high respec-
system makes the construction of the system and its tively low level. The gasoline generator acts as a bat-
operation more complicated. In order to study the be- tery charger; once it has started at a low battery volt-
haviour of such a system in practice, an experimental age level, it runs continuously until the battery reaches
PV /battery system with a DC gasoline generator has a preset high voltage level again.
been built at the University of Utrecht, the Nether- System research was started in July 1990; up to Febru-
lands. A schematic layout of the system is shown in ary 1991, the system has mainly been operated as a
figure 1. Electricity is supplied via an inverter to a AC PV /battery system. During this period, research has
load which enhances the applicability of the system concentrated 011 the functioning of the system without
concept. The load is a simulated household load of gasolille generator (section 2) and on the automatic
1000 kWh/year and a maximum power of 500 W. The starting of the gasoline generator (section 3).
entire system is controlled by a Programmable Logic Parallel to the experimental work and the first mea-
Controller, enabling a flexible implementation of con- surements, extensive numerical modelling of the system
trol strategies. Extensive data monitoring and acqui- was carried out at the University of Oldenburg (sec-
sition of one minute mean values of the energy flows in tion 4,5 and 6). Simulation calculations of the system
the system and of the irradiation are carried out. The behaviour are a valuable complement to experimental
purpose of the research is to develop an optimized con- work because:

728
 Loss of data through failures in the data acqui- irradiation has been corrected for the temperature of
sition system or through technical failures in the the reference cell. The used PV energy is the energy
energy system itself may be compensated. that has actually been produced by the PV array. The
difference of the used PV energy and the final DC en-
- System modelling will help to distinguish energy
ergy equals the battery losses.
losses in the system due to non-ideal component
behaviour from losses caused by a mismatch be- The following comments can be made on the perfor-
tween system components. For principle reasons, mance of the PV array and the battery:
most mismatch losses can not be measured exper-
imentally in the realized system. PVarray
- A system model may be used to test different op- The contribution of various mechanisms to the devia-
erational strategies under identical operating con- tion of the PV array output from the output under
ditions. Standard Test Conditions (STC), averaged over the
- A system model may be used to investigate the measuring period, has been quantified:
operational behaviour of an energy system nnder - 4% deviation may be due to spectral mismatch;
different external conditions (different loads and the 4% difference between the array plane irra-
different meteorological conditions) or under dif- diation measured with a pyranometer and with
ferent system layouts. a reference cell, however, may also be caused by
reflection losses and orientation errors. The ref-
The autonomous PV /battery /generator system which
erence cell measurements have been corrected for
is described above was modelled with the block-
the temperature of the reference cell;
oriented simulation system INSEL /2/ which was de-
veloped at the University of Oldenburg. - 8% due to deviations from the standard irradi-
ation of lOOOW/m 2 and module temperature of
25°C; the energy output of the PV array in the
130
&1 ..gy Wh) maximum power point was calculated from the ir-
radiation as measured with the reference cell and
'~]
100 -
the module temperature on basis of the PV mod-
110 _ .
ule I-V curves;
80
70
60
14% due to mismatch of PV generator and bat-
so tery; this estimate was derived in the following
40
30 way. The PV generator output energy using a
20 fixed temperature coefficient of -0.3%;oC above
10
0 25°C. Subsequently, the hourly averaged value of
.lui Oct Nov o..c Jan Fob
the PV generator output at 25°C was calculated
month
and plot ted for each mouth as a function of the
1_ EM?P."'.'" epy.ust<l 0 e....og •.• ", hourly averaged irradiation. The plot showed a
straight line and a number of points falling below
this line. The straight was assumed to represent
Figure 2 Performance of the experimental PV /battery the energy efficiency curve of the PV generator.
system from July 1990 until February 1991. The ac- The scattered points were identified as cases of
cumulated monthly values of the PV generator energy disconnection of one or more subarrays. The PV
output in case it had been operated with an ideal maxi- array output without the effect of subarray discon-
mum power point tracker (Empp,ideal), the actual array nection can now be estimated from the measured
output (EPVused ) and the energy that has been sup- array plane irradiation and module temperature;
plied to the inverter (Eslorage,exit) are shown.
- 8% due to disconnection of subarrays; this is the
difference between the PV output array taking
into account all foregoing effects and the actual
2. EXPERIENCES WITH THE PV /BATTERY PVoutput.
SYSTEM
Battery
The system without the gasoline generator has been
functioning reliably: in the test period the system The system storage consists of a 24 V ABSOLYTE
and the data acquisition operated almost continuously. sealed lead acid battery. The 400Ah battery executed
Figure 2 shows the monthly accumulated values of the the equivalent of 23 full cycles with an overall energy
ideal mpp energy, the used PV energy and the energy efficiency - the quotient of the battery output energy
supplied to the inverter. The mpp energy (the energy and the battery input energy - of 94%. The high en-
produced by the modules under actual conditions if ergy efficiency can be attributed to a specific aspect
they were operated at their maximum power point) was of the battery performance over the measuring period,
calculated from the array plane irradiation as measured namely the strong variation of the bat tery voltage with
with a reference cell and the measured module temper- the battery current. Due to insufficient initial charg-
ature using the PV module I-V curves. The measured ing, only part of the battery plate material is used.

729
We found apparent resistances for charging and for dis- is operated at about 2/3 of its rated power. Figure
charging of about O.lD, which is much higher than ex- 4 shows the performance of the system on a day, in
pected for this battery. As the switching voltages for which the generator ran twice. At 12:05h the gasoline
disconnection of the arrays and the load were not cur- generator started running: it stopped when reaching
rent compensated, the battery was never fully charged the upper switch point shortly after the load became
nor discharged. In March an equalizing charge was zero. The gasoline generator ran for a second time in
given to the battery for four days, after which the ap- the evening. During the daytime, the output current
parent resistance of the battery returned to a more of the gasoline generator is different from the output
normal level. current in the evening due to the combined electricity
production of PV array and gasoline generator.
3. EXPERIENCES WITH THE GASOLINE
GENERATOR

The gasoline generator is a 1kW Honda battery-

charger with automatic voltage regulation. The max-
imum gasoline generator current can be manually set. 40

The energy efficiency decreases with decreasing max- Iinverter input I

30
imum generator current: while the energy efficiency $
with the generator set at 2/3 of the rated power is ] 20
9.4%, it is only 6.6% at a setting of 40%. The gasoline
):S
I r---. I h

} I I 1
.10
generator voltage as a function of the supplied current
is shown in figure 3 for various settings of the maximum 010
12 14 16 18 20 22 24
current level.
40
l PV generatorJ
40

r----- I

1 1
30

Imaxi
rl'l. ~ .A
I

L--..A.." I
~. .--.;
- 12 14 16 18 20 22 24

~ 1420wl ~ I gasoline generator I

30
-
~ ~\
$
~ 20

10 10
\~
mill ):S
Imedium I 12 14 16 18 20 22 24

10 20 30 40 I""- .j battery voltage I -t

25
DC Current (A) -
~
j, \..... ~ ~
~ 24
~

Figure 3 The I-V curves of the gasoline generator. The

"):S> 23
~l ~
solid lines were provided by the manufacturer and in-
dicate the behaviour with the current limit set at the ~ I
"'\
12 14 16 18 20 22 24
minimum, at a medium, and the maximum current.
hour
The dots are measured minute data at an average gen-
erator output of 420 W respectively 660 W.
Figure 4 The performance of the PV /battery system
Originally, this type of generator is started manually. and the gasoline generator (March 1991). Shown are
For this system, it has been adapted for automatic minute data from 10 until 24 hours. The start and
starting by the addition of a starting motor, a starter stop voltages for the gasoline generator are 22.2 V and
battery and an electrical switcl~ for the choke. As it 25.7 V. The PV arrays, that had been disconnected
appeared that in this setup the generator did not start previously, were switched on manually at 10:35h and
reliably in cold weather, a new procedure was devised: the load at 10:40h. At 12:05h the gasoline generator
during the start ether is injected into the air inlet of started running: the generator current gradually de-
the generator from a spraying can that is activated crea.ses, due to the increase of the DC voltage. When
electrically. Choking is no longer required now. This the inverter input current becomes zero, the entire
setup operates reliably. In March, the system was op- gasoline generator current is fed into the battery calls-
erated including the gasoline generator. The battery ing a sharp rise in the battery voltage. As soon as the
voltage at which the generator is started, was set at stop voltage of the gasoline generator is reached, the
22.2V, the stopping voltage at 25.7V. The generator gasoline generator is switched off.

730
4. NUMERICAL MODELLING OF ENERGY Calculations were also performed for a photovoltaic
SUPPLY SYSTEMS generator made up of 24 modules R&S RSM 40
(Prated = 960 W, module area 1l.54m 2 , efficiency at
The model which simulates the PV /battery / generator STC 8.32%, 36 cells per module).
system calculates an annual energy balance of the sys-
tem for time intervals of one minute. Input to the
5. THE ENERGY FLOWS IN THE SYSTEM
calculations are meteorological data measured as one-
minute averages in Oldenburg, North Germany. A calculated annual energy flow chart of the system
Fig.5 shows the model assumptions for the fuel con- is shown in figure 6. Less than 5% of the energy flow
sumption of the gasoline motor-generator set and the intercepted by the solar generator array is finally uti-
efficiencies of the DC/ AC inverter. The data have been lized to cover the load, energy losses occur at all steps
extracted from the manufacturers data sheets. The of energy conversion. The different loss mechanisms
daily load pattern given in fig.5 peaks at noon and shall be discussed in detail:
in the evening and may thus be typical for a house- Under ideal conditions, a solar array of a certain size
hold application. The low permanent load which we and an optimized fixed tilt angle may make use of the
assume is unfavourable to the inverter operation be- annual insolation Eirradiation,ideal. Due to non optimal
cause it forces the inverter for many hours of the day geometrical orientation or to possible shading effects
into low efficiency part load operation. The load itself the absorbed energy Eirradiation,real will in general be
is weakly dependent on the battery voltage, two simu- smaller than the ideal value.
lation runs for different system layouts may therefore
result in slightly different values for the load.
® G
ro--- .----
E\~:;~i(lt ion

'/
~~~~
a - >-
U
a
a
3' 40370 "11;70 El~::;diQtion
g 3)60
~
c- a
Q)
..,a "0

-
~
U 0
a 0 ~:mo
a'"
0
0 ---'
w N
N "- ~ 2705
0 0
P rated=1,2 kW ~
0 ~ 2b~5 ::! 22?O EPVused
0 0 0

500 1000 0 0.5 6 12 18 24

P Gen [w] P aut/ P rated Time [h]
~ 5150 ~ 5130 EPVused ~ Emot
Figure 5 The model assumptions for the simulation
calculations. The fuel consumption of the motor-
generator (left) and the efficiency of the DC / AC in-
verter are data sheet values. On the right the figure
3510
:'520 E5tora9" ~
gives the load pattern which is reproduced day by day. 20 20 tg~~~- ~ Esrorage
The assumption of a permanent load of about 65 W
I~O ctlorgif/g-__
170 11 KI
leads to long periods of part-load operation of the in- lasses
verter. VOltage- ~
175 175 losses ----~

The battery is described by a 'Shepherd model', the ~mo it 4775 Estorage exit
model parameters were calculated for a battery of the (j 2'tbO ~ 2135 EPVstorage e.it
type VARTA-bloc. The battery voltage is limited by
Inverter-
two controllers: the PV- subarrays are disconnected 1115 1115 losses
stepwise if the voltage exceeds 27.4 V. If the battery
is charged by the generator, charging is first done at §, 3615 i:i 3b~O Elo,,"
0-'1> Eload
the rated power of the motor-generator; if the battery ~ISS5 ~ 1';:'5 EPVload ",Q

~e LOP.D
5--
voltage reaches the limit of 28.8 V the generator power
L,
F~O.s1 F·O.1t5 0'\'

is reduced to realize a constant voltage charging.

The photovoltaic generator is described by a two diode Figure 6 The annual energy flow chart of the system.
model, model parameters for the temperature depen- The calculation was performed for two PV generators.
dent 1-V-curves were derived for German AEG PQ System I : AEG PQIO/40 modules, System II R&S
10/40 modules and Dutch R&S RSM 40 modules. The RSM40 modules. Due to energy losses in all steps of
standard system configuration which was investigated energy conversion only 51% (56% resp.) of the rated
has 24 PQ 10/40 modules (Prated = 922 W, module energy production of the PV generator is used in the
area 11.85m 2 , module efficiency at STC 7.78%, 40 cells load. F is the share of the PV power 011 the total energy
per module) and a 400Ah battery. The system control production.
sets the motor into operation if the battery state of
charge (s.o.c.) falls to less than 20% of its rated value, The absorbed energy is multiplied by the rated effi-
the battery is then recharged to 90% S.O.Coo ciency of the photovoltaic module at STC (data sheet

731
value) to calculate Erated,ideal. This value is used as of a system which has no solar generator and no bat-
a reference for the evaluation of the subsequent energy tery storage and where the load is covered by a gasoline
conversion steps. motor-generator for the 8760 hours of the year (fig.7,
The deviation of the temperature of the modules and upper right). The fuel consumption is reduced by a
the absolute value of the radiation in the real operating factor of 2.3, the number of operational hours brought
conditions from the Standard Test Conditions is taken down to less than 1500 hours through the integration
into account in the calculation of the energy Empp,ideal. of a 400 Ah battery storage. A photovoltaic generator
Spectral effects are not yet included into the calculation (Prated = 922W) added to the system reduces the fuel
of Empp,ideal as all calculations are based on pyranome- consumption by another factor of 1.8. The fuel con-
ter measurements. Empp,ideal would be produced if the sumption and the motor operation time do not react
modules were operated at mpp-conditions with an ideal sensitively on a variation of the storage size as long as
mpp-tracker (no converter losses). The system under the nominal storage capacity is equal to or bigger than
consideration has no mpp-tracking converter, the di- 200 Ah (a 200 Ah storage battery is able to cover the
rect coupling of the PV -modules with the battery re- load for 1.5 days if no other power sources are avail-
sults in losses through a system operation in non mpp- able).
conditions.
The photovoltaic power thus produced (EPVutilizable)
is in principle utilizable to the system, however it will
not be used completeley. Energy production from so-
lar radiation and energy consumption have a different
time pattern, the mismatch between production and o
o noPV _
consumption can not be compensated completely by a 2 c '0' nosloroge!
battery storage. .Q ' "
g 0. ~ I
The total energy production in the system is the sum E '-"
Il) 87(;0
:J h
of EPVused and the energy provided by the motor- o en
o c o
generator Ernot . The share of the photovoltaic power o o o

~
U OPV.t'
on the total energy production is the solar fraction o 0

F (F = Epv..sed/(EPVused + Emod). ,
~ 8 '" -
The total energy production is further subject to losses
in the battery storage and in the DC/ AC-inverter. 04-____- .____- .____--.
Fig. 6 shows the quantitative results of an energy flow a 1000 2000 3000
Motor Operation Time (h)
calculation for two system layouts which are identical
up to but excluding the PV-generator. Simulation run
I shows results for a system with the AEG PQ 10/40 Figure 7 Fuel consumption and operation time of the
generator, run II is calculated with the generator of backup motor is shown as a function of the storage size
R&S RSM 40 modules. Although the rated values of (numbers in the figure give the storage size in Ah). In
the generator powers are very close, the different mod- this calculations, the motor recharges the battery from
ule layouts lead to different values of EPVmpp,ideal and 20% state of charge of the nominal storage capacity to
EPVutilizable resulting in solar fractions of F = 0.45 90% state of charge. The figure includes two systems
(system I) and 0.51 (system II). System II would thus of reference: A system without PV and battery - the
have an annual fuel consumption which is only about motor runs continuously all the year - and a system
90% of the fuel consumption of system I. without PV generator but including a battery.
In the evaluation of CEC-photovoltaic demonstration
projects it was found to be of value to characterize
the systems behaviour by two numbers, 'Performance The solar fraction F of the system was calculated for
Ratio' and 'Yield' /3/. The performance ratio is the different storage capacities and two different control
system efficiency in the broadest sense (EPVioad di- strategies for the motor-generator operation. The re-
vided by EPVrated,ideal), the system yield is consid- sults are plotted in fig.8. The control strategy which
ered as the number of hours of operation per day at is applied to calculate the results labelled 'a' starts the
Prated which would give the same energy output as the motor if the battery state of charge falls to 20% of its
recorded integral value for that day. Assuming a tech- rated value. The battery is then recharged continu-
nical availability of 100%, performance ratio and yield ously to its full capacity (100% s.o.c.). If the charg-
are calculated to be 51 % and 1.35 for system layout I ing of the battery by the motor is terminated at 90%
and 56% and 1.49 for system layout II. s.o.c. the system behaviour improves; F is increased
by about 10% (results labelled 'b'). The effect is due
6. VARYING THE STORAGE SIZE AND THE to the voltage control which limits the system voltage
CONTROL STRATEGY during the charging and thus limits the motor output
at the end of the charging process. Part load opera-
To demonstrate the power of system modelling we show tion of the motor increases the fuel consumption and
in fig.7 and 8 the result of an investigation of the ef- increases the motor operation time; the utilization fac-
fects of the storage size on the system behaviour. !\.S a tor [or the photovoltaic power is reduced and thus also
point of reference, we calculated the fuel consumption the solar fraction F.

732
 (EPVv.sed/ EPV;deal,rated) of 79%, the measured value
for the 8 months of operation is 64%. This discrepancy
is in part due to the behaviour of the battery voltage
control which results in different losses through the dis-
<D
a connection of PV-modules in the simulated and in the

..
real system. A detailed analysis of the different mecha-
nismn which are responsible for the deviations between
a measurement and simulation will be performed in fu-
ture.
N The simulation calculations show clearly that the lay-
a
out of the RSM 40 module is better adjusted to a direct
a coupling with a 24V lead-acid battery than the AEG
O~__. -__. -__. -__ r--. PQ 10/40 module. The voltage of the RMS 40 modules
o 100 200 300 400 500 at mpp conditions is close to the voltage of the battery,
Storage Capacity (Ah) unnecessary mismatch losses are avoided.
An optimized control strategy for the system which
minimizes the gasoline consumption and the operating
Figure 8 The figure shows the solar fraction F plotted hours of the motor-generator and which guarantees a
versus the battery size for two different control strate- long operational lifetime of the battery has not yet been
gies of the motor: In curve 'a' the motor stops at 100% worked out in detail. Numerical modelling and simu-
state of charge of the battery while in 'b' the battery lation validated with experimental results will be the
is charged only to 90% state of charge. Strategy 'a' tool to solve the problem.
should result in higher battery lifetimes, but it reduces
F due to long motor operation periods at high voltage ACKNOWLEGDEMENT
levels where the PV subarrays have been blocked. The
effect is evident at a storage size of about 100 Ah when The Dutch research activities were made possible by
a periodically daily motor operation is observed. the financial support of NOVEM, the Netherlands
Agency for Energy and Environment. The experimen-
tal system in Utrecht was designed and constructed
For both control strategies, the fraction F is a smooth
by Ecofys Research and Consultancy Utrecht, The
function of the stora.ge size. At a storage capacity of
Netherlands.
100 Ah however, F shows a sharp decrease. A closer
examination of the effect reveals that in this system
layout the motor is operating periodically at a 24 hours REFERENCES
period, starting in the morning when the batteries are
exhausted. Around noon, when the motor operation /1/ van Dijk, V.A.P., E. A. Alsema, K. Blok and
is finished, the storage is filled and the power produc- W. C. Turkcnburg. 1989. Performance and Costs of
tion from the PV-arrays can not be stored, the PV PV /Battery Systems with a Gasoline Back- Up Gener-
power which is actually used reaches a minimum. Once ator. Proceedings of the Ninth E.C. Photovoltaic Solar
the problem is understood, these unfavourable operat- Energy Conference held at Freiburg 1989.
ing conditions may be avoided by changing the control /2/ Luther, J. and J.Schumacher-Grohn. 1991.
INSEL - A Simulation System for Renewable Energy
strategy of the motor-generator.
Supply Systems. Proceedings of the Tenth E.C. Pho-
tovoltaic Solar Energy Conference held at Lisbon 1991.
7. CONCLUSIONS /3/ Kaut, W., G.Blaesser and W.B.Gillet. 1989.
Recent Results fom the CEC PV Demonstration Pro-
A Photovoltaic/Battery system with a gasoline genera- gramme. Proceedings of the Ninth E.C. Photovoltaic
tor as backup energy source is an interesting technical Solar Energy Conference held at Freiburg 1989.
solution for the supply of loads in the range of 1000
to several thousand kWh per year. A demonstration
system has gone into operation at the University of
Utrecht; the first technical problems arising from an in-
correct initialization procedure of the battery and the
problems with the cold-start of the motor generator
have been overcome.
The measured data from the first months of operation
show the high stora.ge efficiency of 94% for the lead
acid battery. The simulation calculations which were
performed independently predict a mean annual energy
storage efficiency of 90%, a value which seems more
realistic for a battery which is operated in the full range
between 20% and 100% state of charge.
The model calculates an annual PV -array efficiency

733
 10TH EUROPEAN PHOTOVOLTAlC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Integration of Photovoltaics into Electric Utilities: Issues and Approaches

Robert H. Annan James E. Rannels R, An il Cabraa1
Office of Photovoltaic Technology Division Meridian Corporation
Solar Energy Conversion Office of 4300 King Street, Suite 400
U.S, Department of Energy Solar Energy Conversion Alexandria, VA 22302
1000 Independence Ave. S.W. U.S. Department of Energy U.S.A.
Washington D.C. 20585 1000 Independence Ave. S.W.
U.S.A. Washington D.C. 20585
U.S.A.

Th~s p~per presents an over~iew of the key issues addressed in the comprehensive strategy
belng ~mplemented by the Offlce of Solar Energy Conversion atthe U.S. Department of Energy
to achleve the deployment of 1,500 MWp of U.S. PV systems worldwide by the year 2000.

Background 3.2 Distributed On-grid Applications

On February 20, 1991, the U.S. Secretary of Energy Use of grid-connected distributed PV at substations
issued the National Energy Strategy which employs a
balanced approach to meet the President's directive and buildings to reduce local ized peaks are now
to "bequeath a 1egacy to the next century of a' being considered in the U.S. A study by Pacific Gas
cleaner, more prosperous and yes, more secure and Electric, a leading util ity in the U.S. has
America." Consistent with the overall strategy, the shown that distributed PV at selected substations
Office of Solar Energy Conversion (OSEC) of the U.S. will have a favorable benefit-cost ratio before
Department of Energy (DOE) has prepared an aggres- 1995.(7) Use of PV for peak load shaving in commer-
sive plan for rapid and extensive deployment of PV cial buildings is being tested in the New York area
systems by addressing technical and market issues, by Niagara Mohawk Power Corporation in their Demand-
and industry concerns. The OSEC is responsible for Side Management PV Demonstration Project.(8)
solar conversion technology -- photovoltaics (PV),
solar thermal electric, and biomass power (excluding 3.3 Central Generation
landfill gas and municipal solid waste) -- utiliza-
tion in the electric power sector. The DOE, utilities, and U.S. PV companies have been
supporting bulk power generation demonstration
2 Power Supply and Demand projects since the early 1980s. The PVUSA project
will have over 1 MWp on-line in 1991.(9) A total of
Between 1990 and 2000, about 100 GW of additional 20 MWp of grid-connected, central ized PV systems
capacity will be needed by U.S. util ities, of which have been installed in the U,S. Public Service of
almost 20 GW are as yet unplanned.(l) In addition, Colorado is planning to build a 14 MWp PV power
there is an undetermined potential for demand-side plant ~s part of its repowering of the Fort St.
management. While few commitments have been made to Vrain nuclear power plant.(lo)
solar conversion technologies per se, there are
significant opportunities for supply-side, as well PV at today's prices is not cost-effective for bulk
as demand-side applications. power generation. However, preliminary analyses
have shown that the 14 MWp system in Colorado would
Probable capacity additions overseas, expects to add only about $0.0007 /kWh to the rate base. (11)
dwarf that in the U.S. with over 400 GW to be added For comparison, the average revenue for the utility
in the 1990s. If no other alternat i ves are pro- is approximately $0.06/kWh.(12)
vided, coal and large hydro will meet most of these
needs.«2),(3)) 4 Goal
3 Opportunities for PV The goal of OSEC is to deploy 1,500 MWp of U. S. -
made PV systems worldwide by 2000 (Fig. 1). Cur-
The opportunities and the experience base for photo- rently, there is approximately 100 MWp of U.S. PV
voltaics (PV) applications in the utility sector in systems installed worldwide. Corresponding annual
the U.S. are growing. Sixty U,S. electric utilities production of PV systems by U.S. companies would
have implemented PV projects, many of the systems then exceed 400 MWp by 2000.
are interconnected with the gri d. (4) PV app 1i ca-
tions in the residential, commercial, and industrial Off-grid applications will continue to dominate the
markets are expanding rapidly. market. In the mid-1990s, utilities will start
becoming major suppliers of PV for off-grid utility
3.1 Off-grid Applications customers and operations. As a result of legisla-
tion, improved economics, and greater awareness, the
PV is clearly a mature technology for off-grid U, S. Government purchases wi 11 increase, with the
applications. Nearly 12 MWp of 14.8 MWp of U,S. PV Department of Defense as a major customer. (13)
modules sold in 1990 were for off-grid applications Utility use of PV for peak load reduction will also
in the U.S. and overseas.(S) The remaining modules increase in the 1at ter part of the 1990s. Some
were used for low power consumer electronics, and expansion of the bulk power market is expected due
for on-grid applications. to regulatory pressures caused by environmental
concerns, and because of the interest among some
Utilities are also using PV for powering utility ut i lit i es to increase thei r operating experi ence
operations such as sensors, microwave repeaters, with PV.
cathodic protection, and lighting.(6) This market,
although small at present, is far from saturated.

734
 u.s. PV Systems Market Development
U.S. Department of Energy Goal
Cumulative MW Installed
-------
1.500

1.250
Conlumer Electronics
1.000 Utility Sulk Pewe.
e:: Utility Operat'n/ DSM
Ld U.S.lSt.le Oov't
750
C' ROIldantla""g.lc.
_ Indultry / Comme.c lal
500
_ Ov •• a.as

250

o
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Figure 1 - U.S. DOE Projected Market for U.S. PV Systems

S Strategy 6 Technical Issues and Approaches

The OSEC strategy i nvo 1ves coord i nated act i on on Closer communication with utilities and other users
three fronts -- technology development and valida- is needed to understand and respond to thei r re-
tion, market conditioning, and developing projects- qu i rements. The Solar Energy Research Inst itute
- which combine to address existing barriers in the (SERI) is estah~ishing communication links with
utility marketplace: Public Utility Commissions (PUC) and utilities by
setting up coordinating councils. The OSEC is also
• Technology Development and Validation. The goal p1anni ng a seri es of workshops di rected at ut i 1i-
of this element is to prove the technical feasi- ties.
bility, reliability, and manufacturability of PV
systems for utility applications through re- Hi gh quality, and complete solar resource data is
search, develapment, and demonstration (RD&D). needed. The OSEC is sponsoring a worldwide resource
assessment to help develop information necessary for
• Market Conditioning. The goal of this element matchi ng technol ogi es to resource characteri st i cs
is to cultivate the environment needed for and availabillty.(I4)
increased penetration of PV into the power sector PV performance, re 1i abil Ity and cost must be i m-
through the accomplishment of two subgoals: (1) proved if PV is to access a 1arger share of the
address market barriers in order to allow needed
capital to flow; and (2) leverage the regulatory power market, especially, the huge bulk power mar-
and policy arenas to encourage a more equitable ket. An RD&D program addressing these issues is
consideration of PV. being implemented.(IS) The OSEC assists in improv-
ing performance, reliability and cost-competitive-
• Project Development. The goal of this element is ness of PV through basic and applied research and
development. A multi-technology development path
to assist in building a strong, self-sustained involving thin films, crystalline silicon, and
U.S. PV industry through support of commercial concenty;;tors is supported to minimize risks and
project activities that will ensure the timely maximize benefits.
availability of technologies, lead to increased
product ion, reduced product costs, and steady
market growth.
Fig. 2 represents the key players involved In each
aspect of the strategy.
Th is coord i nated approach goes beyond t rad It i ona 1
Department of Energy utility sector activities that
had focused primarily on technology RD&D, to encom-
pass the broader range of issues that will ultimate-
ly impact widespread deployment and commercializa-
tion. Developing a secure, stabilized, and growing
market at prices that ensure profitabil ity to in-
dustry is fundamental to the OSEC strategy. The
following sections discuss key issues addressed by
the strategy.

Figure 2 - Strategy Framework

735
PV.,systems development activities tests and eval- Codes and standards acceptable to electric util i-
uates PV systems; addresses issues such as dis- ties and the engineering community are essential if
patchabil ity, power qual ity, and safety; and helps PV use is to become widespread. The OSEC is work-
users and industry develop field experience that ing with the PV industry, IEEE, Power Engineering
must precede large scale deployment. A key program Society, and the Electric Power Research Institute
is the PVUSA project. It is assisting in increasing (EPRI) to ensure that suitable codes and standards
utility experience with PV as well as helping the PV are established. Recommended design practices that
industry understand the needs of the utility market- comply with the U.S. National Electric Code are now
place. avail able. (23)
Most PV manufacturing facilities employ early 1980s 8 Project Development Issues and Approaches
technologies and much of the performance improve-
ments seen at the laboratory have yet to move into Informat i on on the extent and regui rements of the
the production 1 ine. The OSEC initiated the 5- market for PV is needed. The OSEC has 1aunched a
year, $55 million PVMaT (PV Manufacturing Technol- market assessment project that will identify and
ogy) Project. Through PVMaT, OSEC will assist characterize the market for PV systems worldwide.
industry advance PV manufacturing technologies.(16)
Utilities are reluctant to embark on projects in-
7 Market Issues and Approaches volving new technologies due to perceived risks. A
key strategic thrust of OSEC is a number of project
A lack of information, or misinformation is often a development efforts that will generate the momentum
major barrier to consideration of PV. The OSEC is necessary for achieving the year 2000 goal of 1500
launching a major communications effort directed at MWp of PV. The OSEe will provide technical assis-
util ities, investors, regulators, legislators throu- tance to reduce the technical risk of projects, and
gh the PUC and Utility Coordination Councils. The will offer cost-shared funding. The solar industry
OSEC hopes to work closely with the National As- and ut il it i es or other users wi 11 be partners in
sociation of Regulatory Util ity Commissioners in these project development efforts. The solar in-
th is effort. dustry provides the technology and services. The
util ities, investors, or other users provide the
The OSEC is also documenting successful investments project sites, cost-shared funding, and, most im-
in renewable energy by utilities or independent portantly, a commitment to invest in larger numbers
power producers. The 1essons 1earned from thi s of projects in the future. The following are some
analysis will be useful to the PV industry in devel- of the projects being considered or implemented
oping future projects. jointly by the DOE, industry, and utilities:
The utility financial structure, and accounting • Federal agency projects with Department of
practices have biases against technologies such as Defense, and Department of the Interior, Power
PV which are capital intensive, but have low opera- Marketing Administrations, etc. A plan for
tions and maintenance (O&MI costs. For example, installing 100 MWp of PV by 1996 is being prepar-
capital intensive technologies are penalized because ed by the Department of Defense.
they are subject to sales and property taxes, and
tax benefits are delayed due to depreciation schedu- • Investor-owned utility, public utility, and
les.(17) In contrast, fuel costs can be expensed in independent power producer projects. Example:
the year they were incurred. The OSEC in coopera- Pub 1i c Servi ce of Colorado Ft. St. Vra in
tion with utilities is investigating utility financ- 14 MWp PV project
ing strategies, and is developing approaches to Alaska Energy Authority for PV-diesel hybrid
remove these and other biases against PV and other village electrification
capital intensive generators. Idaho Power and KC Electric Association for
PV water pumping.
The OSEC is also investigating power sector invest- Niagara Mohawk Power Corporation for PV
ment strategies of government agencies such as Power demand-side management
Marketing Administrations (PMA) , and the Rural PG&E for substation peak load reduction, and
Electrification Administration in the U.S. These power conditioning system evaluation
organizations have invested heavily in capital Maui Electric Company for the IEUP
intensive technologies in the past, and their ex- Early utility applications project with EPRI
periences could be valuable to PV. and PG&E.
Existing utility planning methods either do not • Overseas project development efforts through the
consider PV and other renewables, or have implicit Committee on Renewable Energy Commerce and Trade.
biases against them. The OSEC is participating in
integrated resources planning (IRP) activities to 9 Benefits
develop improved utility capacity planning method-
01 og i es. The DOE is exami ni ng capac i ty expans i on This strategy assists in developing sustainable and
planning and production costing tools used by utili- growing markets for PV products in the utility
ties to determine how PV and other renewable energy sector, thus enab1 ing firms to be profitable, im-
techno1 ogi es can be properly represented in these prove products, expand production, reduce costs, and
models. The DOE is also participating in regional achieve steady growth. By the year 2000, this
utility planning exercises. For example, the DOE strategy helps develop a strong U.S. PV industry
took part in two workshops on integrated resources wi th sales expected to exceed $1 bill ion (1991$)
planning in Hawaii, as part of the DOE Integrated annually. If this trend continues, by 2010 U.S.
Electric Utilities Program (IEUP).(18) industry could supply 15 GWp of PV systems. In the
year 2000, U.S. PV systems could displace 5 million
Seventeen states currently require environmental barrels of oil equivalent annually, thus eliminating
external ities to be included in util ity resource the production of over 2 million tons of CO 2 •
selection.((19),(20» For example, rulemaking in
Nevada is expected to add about 1 cent/kWh for 10 Next Steps
natural gas-fired generation, about 5 cents/kWh for
coal-fired generation, and over 14 cents/kWh for The detailed OSEC action plan is under preparation.
diesel generation.((21),(22» Ten other states have OSEC will continue to meet with its partners in the
plans under consideration. power sector, industry, research community, investor

736
groups, and state and federal government agencies to 16. Department of Energy, • DOE Launches Energy
ensure that the common objectives and concerns of Efficiency and Renewable Energy Initiatives," DOE
the key players are addressed, and to enlist their News, January 26, 1990.
support.
17. Rueger, G.M., "Testimony Before the U.S. Sen-
The strategy development exerci se is des i gned to ate Committee on Energy and Natural Resources Re-
expedite technology development and deployment. It garding Titles III and IV of the National Energy
will not stifle or delay on-going activities. Wo,rk Security Act of 1991 Concern i ng Energy Effi ci ency
that commenced in pri or months wi 11 cont i nue un- and Renewable Energy," February 26, 1991.
abated and in addition, several new initiatives have
been launched. The detailed action plan is expected 18. Public Utilities Commission, "Instituting a
to be ready by September 1991. Proceeding to Require Energy Utilities in Hawaii to
Implement Integrated Resources Planning," Docket No.
References 6617, Order No. 10458, January 10, 1990.
1. North American Electricity Reliability Council, 19. Cohen, S.D., J.H. Eto, C.A. Goldman,
"Electricity Supply and Demand for 1990-1999," J. Beldock, and G. Crandall, "A Survey of State PUC
Princeton NJ, October 1990. Activities to Incorporate Environmental Externa-
lities into Electric Utility Planning and Regula-
2. Moore, E.A., and G. Smith, "Capital Expendi- tion," Lawrence Berkeley Laboratory, LBL-28616, May
tures for Electrical Power in the Developing Coun- 1990.
'tries in the 1990s," Industry and Energy Depart-
ment Working Paper, Energy Series Paper No. 21, The 20. Finnell, J., "Status Report on Environmental
World Bank, February 1990. Externalities,· Meridian Corporation report to the
U.S. Department of Energy, March 4, 1991.
3. Energy Information Administration, "Interna-
tional Energy Outlook: Projections to 2000," De- 21. Nevada Public Service Commission, "Re Rulemak-
partment of Energy, Washington DC. ing Regarding Resource Planning Changes Pursuant to
SB 497," Docket No. 89-752, November 19, 1990.
4. Solar Energy Industries Association, "Solar
Energy Uses in the Utility Sector," Washington DC, 22. DeGroat, K., "Rate Impact of Nevada Resource
September 1990. Planning Changes to SB 497,· Meridian Corporation
draft report to U.S. DOE, March 6, 1991.
5. Maycock, P., "World Module Shipments Increase
15.4%: European Shipments up 25%," PV News, 23. Wiles, John, "The Photovoltaic Answer Book,"
Vol.10:2, February 1991. Southwest Technology Development Institute, Febru-
ary 1991.
6. Jennings, C., PG&E's Cost-Effective Photovol-
taic Installations, Twenty First IEEE PV Special-
ists Conference, Volume II, Page 914, 1990.
7. Shugar, D. S., "Photovoltaics in the Distribu-
t i on System: The Eva 1uat i on of System and Di s-
tri buted Benefits," Proceedi ngs of the Twenty-
First IEEE PV Specialists Conference, Kissimmee, FL,
May 22, 1990.
8. Maycock, P., "Niagara Mohawk Power Corp Plans PV
Demonstration Project," PV News, Vol. 9:2, June,
1990.
9. Hester, S.L., T.U. Townsend, W.T. Clement,
W.J. Stolte, "PVUSA: Lessons Learned from Startup
and Early Operation," Twenty First IEEE PV Special-
ists Conference, Volume II, Page 937, 1990.
10. McCormack, R., "World Events Set the Stage for
Solar Energy Revival," New Technology Week, Vol 5:4,
January 22, 1991.
11. Kozak, L., "PV and Sol ar Thermal Revenue Re-
quirement Evaluation," Meridian Corporation, un-
published typescript, February 1991.
12. Public Utilities Reports, "The PUR Analysis of
Investor-owned Electric and Gas Utilities," Ed:
Susan M. Johnson, 1990.
13. U. S. House of Representatives, "Mi 1itary Con-
struction Appropriations Bill, 1991," 101st Con-
gress, HR 5318.
14. Riordan, C., D.R. Myers, R.L. Hulstrom, "Spec-
tral Solar Radiation Data Base Documentation,·
SERI/TR-215-3513A, January 1990.
15. U.S. Department of Energy, "Photovoltaics
Program Plan, FY 1991 - FY 1995," Review Draft,
December 12, 1990.

737
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

lOOkW GRID CONNECTED PV-INSTALLATION ALONG RAIL

INFRASTRUCTURE IN SOUTHERN SWITZERLAND - A FEASIBILITY STUDY

R. HachIer. Th, Nordmarm

TNC Consulting AG CH-7000 Chur Switzerland
Phone ..41 / (0)81220753. Fax ..41/ (0)81226130

ABSTRACT, Thepotential oflarge grid connected PV -installations using existing infrastructures

along motorways and railway lines without the need of extra land resources in Switzerland was
evaluated in 1987/1988, A fITst l00kW pilot installation on the soundbarriers of amotorway in
the Swiss Alps was constructed in 1989.

Using existing infrastructures for PV -installations will help to reduce the unit price of electrical
energy. The plan to build 3 more similar l00kW installations will demonstrate reductions in
engineering and product costs.

As a follow-up to this fITst installation. the authors describe the feasibility of a second
installation on the district heating pipeline along a railway line near Locarno in southern
Switzerland. A preliminary study of the electromagnetic interference between the DC PV-
installation and the nearby high voltage electric railway line is also being carried out at a test
site. Insights gained from this preliminary study and from the contruction and operation of the
fITstinstallation on the soundbarriers along the Motorway Chur-Bellinzona (N13) will influence
the specifications of components for the second l00kW grid-connected PV -installation.

1. FEASIDIUTY STUDY OF A 1OOKWP GRID-CONNECTED 1.2 Meteorological and Climatic Data. Forecast Yield.
PV lNST ALLATION ON THE DISTRICT HEATING PIPELINE Given the outline of the horizon. no loss in yield is expected.
ALONG THE BELLINZONA-LOCARNO RAILWAY LINE. The solar modules are aligned to the south. with a 14° deviation to
west. If the panels are set at an inclination of 45°. the following
The project to set up grid-connected PV -installations using the annual values will be obtained:
existing infrastructure along motorways and railway lines involves
the construction of four lOOkWp grid-connected installations. The
first of these was constructed in 1989 on the N13 motorway in
Grisons. The second installation is pi armed along the Bellinzona-
Locarno rail way line. using the foundation and supports of a district _00
heating pipeline which is due for reconstruction. The feasibility of 110 ...
this installation is currently being examined. 1.60,00

IMI,OO
UO,OO O IG "1>.10%1
1.1 LocatioIL 100.• !lI1.P.WhImII
The location of the pi armed installation is on the north side of "'00 . &lop.,,\,\ W]
the east-west orientated valley. along the single-track Bellinzona- 4<100
Locarno railway in Ticino. A district heating pipeline runs parallel .0.00
to the railway. This pipeline delivers the heat generated by a refuse 20.00
incineration plant to a nearby paper factory. ..00

Period IG IA fed Into

uUllty
(kWh) (kWh) [kWb]
Summer 970 956 76272 60%
Winter 390 S97 SlOSS 40%
Year 1360 1553 127327

Fig. 2 and Table 1 Solar irradiation and forecast yield from the
planned location for a lOOkWp grid-connected installation

Fig .1 siluaJiono/lhe plannedPV·installaJion using rail infrastructure

738
The wind is a key factor affecting the output of the proposed
installation. On the one hand. it influences the static of the
foundation and suppon framework. On the other hand. it contributes
to the movement of the emissions from the refuse incineration plant,
to the west The prevailing wind directions are parallel to the axis of
the valley. This is reflected in the frequency distribution.

Acimut Partition Sum

from 1"1 to [°1 Dav [%1 Nil1ht [%]
"East" 80 130 15.1 327 47.8
"West" 260 300 19.1 5 24.1
Prevalllnil Wind Directions 34.l 37.7 71.9

Table 2 Wind frequency distribution o/the prevailing wind

directions

Fig.4 View of the Support Sructure of the PV-Installation Using

1.3 Existing Infrastructure and Wrastructure to be Constructed. Rail Infrastructure
The existing district heating pipeline is not a suitable base for
the panels. It is due to be replaced in 1995. in the context of a
complete renovation of the refuse incineration plant A section of the
new district heating pipeline could be constructed in 1992. so that the Connecting to the grid will not present a problem. since several
planned PV installation can go into operation in the proposed transformers are located nearby. DC power will be converted in one
timescale. The construction of the pipeline could take into account or several inverters and fed into the grid without energy storing in
the specific needs of the installation. batteries.

Figures 3 and 4 show a possible support structure. Please note the

1.4 Project Prerequisites.
elements of the construction which are provided for in the planned
pipeline renovation.
It is hoped to implement this pilot project as quickly as
possible. Before the installation can be constructed, though. several
preconditions. principally of an organisational type. need to be met.
The main factor is the timing of the replacement of the district
heating pipeline which is currently planned for 1995. If the project
is to be implemented in 1992. then the planning of the district heating
pipeline needs to be brought forward. It should be possible to
advance the construction of the suppon elements along the length of
the proposed PV installation. so that the project deadlines can be
met. A further factor to be taken into account is that operational
problems may arise in 1995. when the new pipeline is defmitively
mounted and the old pipeline demounted.

1.5 Progress to Date

steel profile The feasibility study will be completed in May 1991. Climatical
and meteorological data had been acquisited 1988-89. Researches
had been carried out concerning the technical feasibility. various
possible prerequisites. and possible timescales. Result of these
researches is. that it should be possible to implement the project in
1992. if the construction of the supPon elements along the length of
height the proposed PV installation will be advanced.
180mm

Fig. 3: Cross-Section ofthe PV-Installation on the Support Structure

oftlie District Heating Pipeline
739
2. PREUMINARY STUDY OF THE APPUCATION OF
PHOTOYOLTAIC TECHNOLOGY ALONG THE SWISS
RAILWAYS.
7J.oon RhB
~::!"::~RI , ...--
S~1D=

A preliminary study of the application of Photovoltaic Technology

along the Swiss railways is currently being carried out The aim of
this study is to establish the extent to which grid-connected PY
plants and the installations of a nearby railway line are mutually
compatible. The study fmdings will be directly applied in the
specifications for the tender of the inverter and solar cells for the
proposed 100 kWp grid-connected installation along a stretch of the
railway in Ticino.
Fig. 5 Situation of the test installation

2.1. Method of Approach.

The main focus of the investigation is on the interference
between the grid-connected PY -installation and the nearby high-
voltage electric railway. One of the questions addressed is ,,Does the 2.4. Other Researches
railway interfere with the PY installation?". In order to answer this Besides researches of electromagnetic interference between
question, the electro-magnetic field will be determined. Another rail and PY systems the following researches are carried out:
question to be addressed is : ,,Does the PY installation interfere with
the railway?" In order to answer this question, the electrical and • New assembly techniques of frameless lOOW solar modules on
magnetic disturbances which arise from a grid-connected PY- light substructures.
installation need to be examined. • Pollution of module surface: brake sediments from the trains.
• Blinding and sight restrictions of engine-drivers.
Transmitters and receivers of electromagnetic waves include, on the
one hand, the long electric cables and large solar module arrays
required by the project, and on the other hand, railway cables, feeder
wires, signal conduits and rail vehicles. The work will be carried out
in cooperation with the ABB EM! Control Centre.

A further point to be covered is the development of an earthing

concept and the design of a surge protection system to protect the
installation, for example, from a nearby lightning strike.

2.2. Computer Simulation and Laboratory Measurements.

Following a comprehensive computer simulation and various
measurements in the laboratory, the findings will be verified in a
field experiment As part of this experiment, a small grid-connected
PY -installation will be setup along arailway line. Along the selected
stretch of the railway, similar constellations to those in Ticino are to
be found: parallel to the railway line which has two different rail
systems (normal and narrow gauge) is a district heating pipeline and
a high-voltage electricity cable. The two sites differ,however, in that Fig. 6: Cross-section of the test installation
the existing district heating pipeline along the test site cannot be
utilised as a support structure for the PV installation. In spite of this,
new assembly techniques will be tested insofar as possible.

2.5. Progress to Date

2.3. PV 1.5kWp Grid-Connected Test Installation. Work planning is going on now; the laboratory measurements
The measurements at the PV test installation will take the will be carried out in May 1991, the measurements at the PY test
following parameters into account: installation will be carried out in sununer 1991. Results will be
available to the end of 1991.
• Grid-connected PY Installation:
- Harmonic wave spectrum of the inverter.
- Length and constellation of all cabling and electrical shielding.
- Earthing concept and surge protection.

• Electric Railway:
- Harmonic wave spectrum of power lines.
- Constellation of power lines, signal lines and electrical shielding.

740
3. FINDINGS IN RELATION TO THE FIRST Table 3 compares material and installation costs of the motorway
INSTALLATION ON THE SOUND BARRIERS ALONG THE installation, constructed in 1989, to new installations using jumbo
CHUR-BELLINZONA MOTORWAY (N13). modules and decentrally placed inverters.

3.1 Operational findings.

The concept underlying the [lISt installation proved to be a COItI or lOOkW Grid Conneded PV -Inltallatlolll (SFR.)
success when the installation became operational. This concept can IlIItaJlation NI3 Jumbo modulea deeentraJ Inverter.
generally be applied to a further installation. Once various errors had motorway 200W "Jumbo modules
been eliminated by August 1990, the reliability of the inverter has Components real lzed 1989 planned planned
been good. The remaining components of the installation have to solsrceUs 832000 860000 860000
date contributed only minimally to system breakdowns. inverter iDe\. conl&iner V4000 V4000 140000
connectors DCIAC 13000 13000 42000
group wiring 257000 130000 130000
rl8ld wiring 370000 369600 332000
Energy Flowchan 100% = DC Energy ground fundlmenl&tion 65000 0 0
support ItruclW"el 151000 151000 151000
8l25°panelswrlWce framel 121000 90000 90000
temperature ,lI"IIlSPOrts 28000 21000 21000

IlIIIaIlalion Sum 1'111'000 1'908'600 1'766'000

Table 3: Cost Comparison ofthe Motorway N131nstallation to New

LoN .. SaiD" ShUI DoWll(2.7")
Installations Using Cost Saving Elern.enls
Loa .. p.,d W"""in&(2.6'1o)
Diode LoNoo(O.l")
Cabl. Lou .. DC(1.o..)
1.0-<1.3")
Mill%lAU:b. Sb.din& by PIanu)
Leooa(l.5")
SlIIIldby Leo .... iDYMa' (0.31.)
D. . AoquiJilioD eoo.UII)pUoD(I.3")
Supply'" Grid(SS .''')
Cabl. Loa .. AC(J.o..)

Fig. 7: En.ergyflowchart 1990oflOOkWp MotorwayN13 InstallaJion

This installation shows that the inverter, in particular, is a key

component which must meet high standards, if the overall reliability
is to be increased.

3.2 Cost Appraisal.

The components of the installation were examined to establish
whether savings would be possible, apart from those which would The 100kW Grid ConnectedPV -Installation Using The Soundbarriers
arise from a drop in the unit price of PV modules. Along A Motorway In The Swiss Alps.

Important rmdings for cost savings: Literature:

• Simplifying of the cabling and support structures.
• "Jumbo" modules (over l00Wp each) cut cabling costs. 1] Ninth E.C. Photovoltaic Solar Energy Conference
• Several industrially produced inverters decen1rally placed ins tead Freiburg, September 1989
of one single inverter.
• Mounting on new infrastructure elements with includedca1culation 2] Photovoltaik-Nutzung 1990
of statics. ZUrich, January 1990

741
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PV-ELECTRIC UTILITY INTERFACE ISSUES, UNDERSTANDING THE PROBLEM OF

HARMONIC CURRENTS IN THE GRID SYSTEM

Marina Tortoreli
Center for Renewable Energy Systems
6 Frati St., Fousia, GR-194 00, Koropi, Greece
01l-301-662-6462-Fax

Paul E. Russell
Arizona State University
Tempe, AZ 85287-6506 USA
602-965-2012-Fax

ABSTRACT. Interfacing photovoltaic systems with the ac grid system requires the
inversion of that dc from the photovoltaic system, usually through the use of solid
state devices in cyclic switching arrangements, in perfect inverters. The currents
inserted into the grid system from the inverters necessarily contain harmonics of the
grid frequency which will be combined with other harmonics that have been
developed elsewhere in the grid and will be distributed throughout the power system.
These harmonics may have serious detrimental effects on many components of the
delivery system and the loads. Their distribution and the changes in the network
configurations that will yield a tolerable distribution can be defined through use of the
HPF3PH (Harmonic Power Flow, 3-PHase representation) simulation program.

1. INTRODUCTION 1.2 Effects of harmonics

Harmonic effects include:
1.1 Harmonic Sources a) increased heating losses in distribution
Interfacing photovoltaic (pv) systems with the ac transformers, lines and motors
grid requires the inversion of that pv dc power through b) communication interference with adjacent
the use of solid-state inverters. The currents inserted telephone lines or power line carrier
into the grid system from the inverters necessarily systems which may result to relay mal-
contain harmonics of the grid frequency which will be functions
combined with other harmonics that have been devel- c) more than one zero crossing every half
oped elsewhere in the grid by other devices with cycle in the current or voltage waveform,
nonlinear characteristics. Such devices may relate to in- which may result to the malfunction of
dustrial applications, such as electric arc furnaces and equipment sensitive to it, such as unstable
rectifiers used as battery chargers, motor drives, etc. operation of zero voltage crossing firing
However, consumer equipment used extensively devices, false tripping of relays and
throughout the distribution system such as motor con- breakers, errors in measuring equipment,
trol, light dimmers, television sets, computers, etc are etc
also responsible for a part of the "ambient" harmonic d) excessive peak voltages that may cause in-
voltage levels which can be in the range of 1-3% of total sulation stresses in underground cables,
harmonic distortion (VDF). The Voltage Distortion motors and power factor correction
Factor (VDF) and the Current Distortion Factor (CDF) capacitors and possible failure of some
are defined as: equipment
e) resonance conditions that enhance the

Jfv&
above mentioned phenomena [1], [2].

h=2
VDF=
VI (I a) 2. THE HP3PH SIMULATION PROGRAM

and 2.1 Models

Noting the harmonic sources in a power network

JI:i
and the various effects on power components and
loads, it is clear that estimating the harmonic
h=2 penetration is very important. Measurements of har-
CDF= monic voltages and currents at different locations of the
11 (Ib)
network help in understanding the harmonic
Photovoltaic inverters and wind generator converters propagation problem. However, power engineers need
will add up to the existing problem of harmonic dis- to know the best points for connecting a large non-
tortion as their contribution to service the system load linear load, if there exist such an option, or where to
place their power factor correction capacitors, or what to
will increase in the future.
expect in case of a fault, so that they can set their

742
protective relays and rate their circuit breakers. superimposed to the fundamental. va(t) is expressed by
Computer simulation programs, like the HPF3PH, are a a Fourier series as follows :
more economic and flexible way to perform this type of
analysis.
Other existing simulation programs assume a
Va(t) = Y2 L Vh cos [hoot + !ph] , (6)
h
balanced three-phase network and base their analysis on
a single-phase equivalent. Unbalanced conditions can where h is always an odd integer number, due to the
be studied by the use of symmetrical components [3]. voltage waveform halfwave symmetry. The
However, the HPF3PH is a true-phase harmonic power corresponding current iaW is represented by a square
flow program that models three, two and single-phase waveform lagging the voltage waveform by an angle u.
distribution lines and underground cables by their The amplitude of the square wave is Id. The current
nominaHt circuit at fundamental and odd harmonic ia(t) is written as a Fourier series in the following form:
frequencies up to the nineteenth. Capacitors are
modeled by their per-phase fundamental admittance
value multiplied by the corresponding harmonic order
ia(t) = ill..
7t
IciLh Ohlh cos[hoot-h(u+ooto+1[)].
3
(7)

number h, while transformers are modeled by their

The phase control (delay) angle a is measured from the
leakage impedance, modified for harmonic frequencies.
zero-crossing angle rot o, calculated by a Newton-
Star-star connections are considered. Harmonic per-
Raphson iterative method by the following expression:
phase load impedances (Y-connections are assumed) are
approximated by the parallel combination of a resis-
tance RL and a reactance hXL. RL and XL are calculated
by the real (P) and the reactive (Q) load power values (8)
(specified in the program for all system load buses as in
any regular load flow program) and the corresponding
calculated fundamental bus voltages as follows:
where v is the iteration number. The initial value of
roto is taken equal to : -1t/2-C[l1, the zero-crossing angle of
RL= ~2, (2) the fundamental voltage component. The dc voltage
Vd can be considered as the voltage across the terminals
XL = ~. (3)
of a battery, or across the terminals of a pv array with a
smoothing capacitor in parallel. The average value of
Nonlinear loads are modeled by three different Vd (t) is expressed as :
versions in the program (0,1 and 2). For version 0, at
fundamental frequency, inverters are modeled as loads
generating real power and consuming reactive power.
For harmonic frequencies, they are treated as harmonic
current sources, where the harmonic current can be a = 3iJ..
7t
Lh Oh-h1 Vhcos[h(u+wto+1t)3 + <Ph] . (9)
percentage of the corresponding fundamental current
specified by the user. The per-phase fundamental Knowing va(t) and a, Vd is computed. Similarly,
current injection I is computed by the following knowing Vd and va(t), the delay angle is calculated
equation: iteratively. Based on the converter control circuitry, Vd
(version 1), or u (version 2) are specified.
I=-P -j?, (4)
The dc current Id is computed from the following
relation:
V

where V* is the conjugate of the phase voltage phasor.

The harmonic phase current injection Ih, for a square
current waveform, is: The de power Pdc <negative for inverter operation) is
specified. The ac power received or delivered by the
non-linear load is equal to the specified dc power,
which is constant.
Actually in the program for a bus where a
where lIh is equal to 1 for h = 1,5,9, .... and lIh is -1 for h nonlinear load is connected, two parameters need to be
=3,7,11 ..... specified. The dc power (Pdc) and the dc voltage Vd
For the other two versions, models for single or (version 1) or the firing angle a (version 2). At the v-th
three-phase, center-tap controlled converters (rectifiers iteration all the harmonic voltage components at the
or inverters) are considered. In Figure 1, a three-phase inverter bus and all other system buses are known.
line commutated converter is shown. The transformer Therefore, the zero crossing angle (roto) and the firing
leakage inductance is assumed very small and therefore angle or Vd can be computed. Then, Id is calculated and
the commutation interval is neglected. Also, Ld is large the corresponding harmonic current components are
enough, so that the current id(t) is almost ripple free. computed. For these current injections new harmonic
For inverter operation, the firing angle a has to be voltages are found and the process continues until
greater than 900 and Vd is negative. Consequently, the convergence occurs [4]. Similar models exist for single-
per phase ac current ia(t) lags the voltage vaW by an phase controlled converters.
angle u. Figures 2 and 3 show the voltage and current Although this method does not describe the exact
waveforms for an inverter with a delay angle a equals rectifier or inverter current waveforms, it corresponds
to 1500 and the corresponding Vd and Id values for the to the worst case of harmonic current injection. The
hypothetical case where a 20% 5th harmonic is specific parameters of the non-linearity are not needed.

743
Also, the number of calculations is much less than it generation system is an infinite voltage source at
would be if the exact current waveforms were found fundamental frequency. The distribution system is
and the Fourier series of those were taken. considered unloaded at fundamental frequency, which
is a common practice for fault calculations. However,
2.2 The Solution at Steady-State Operation the load admittances are taken into account for
The current injections are related to the bus calculations at harmonic frequencies [5).
voltages by the following matrix equation:
3. POSSIBILITIES FOR RESONANCES IN THREE·
PHASE NETWORKS

Utilities use capacitors and regulators to correct

the power factor and to keep the bus voltages within
specified limits for various load conditions. So-
(11) phisticated algorithms to optimaly select the location
and the size of the capacitors are used. However, to
where, optimaly choose the capacitor location and size, the
harmonic distribution under this capacitor config-
uration is appropriate to be taken into account. It is well
[Ibus.1Il = [[11.111 [Iz.1Il ... [Ij.h] ... [IN.1Il f ' (12) known that the capacitive reactance decreases with
frequency, while inductive reactance increases with
[Vbus.1Il = [ [VI.1Il [V2.111 ... [Vj.h] ... [VN.1Il ]T , (13) frequency. When a resonance occurs, the inductive
reactance equals the capacitive reactance and the
for h=l, 3, 5, ., .. M. The [Ybus-h) matrices are formed equivalent impedance is equal to the resistive part for a
by the series impedance matrices [Z') and the shunt series circuit or to infinite for a parallel combination
admittance matrices [Y') relating each phase node (bus) with zero resistance. Since there is always a resistance in
current with the corresponding voltage [4]. The node
all practical networks, this equivalent impedance is not
current vectors [Ij-h] and the phase voltage vectors [Vj-
infinite but relatively large. In the actual distribution
hl can be (3x1), (2x1) or Ox1) sub-vectors, depending on network, series, parallel or both types of resonances may
the number of phases of bus j. The admittance of a occur but, it is difficult to realize where exactly the
linear load is included in the [Ybus-hl for h not equal to resonating branch is located. Some simulation
one. Therefore [Ij-h) is equal to [0] for linear loads and examples for a distribution system in which only one
h=3, 5, ..... M. Note that the magnitude and angle of capacitor bank has been connected and with just one
the swing bus fundamental voltage are known, while harmonic current source present in the system have
the harmonic voltage components are not known. been studied. In these cases, the harmonic voltage
Clearly: could be calculated as the product of the corresponding
harmonic current times the corresponding driving
point impedance. For different capacitor sizes, at a
particular frequency this impedance could appear as an
The solution is obtained by the Gauss iterative equivalent R-L-C series circuit or an equivalent R-L-C
method. At iteration (0), [Vbus-l)O is initialized as for parallel circuit. In both cases, the imaginary part was
the conventional power flow program, while [Vbus-h)O practically zero, while the resistive part was much
for h=3, 5, . . . . .M, are initialized to [0). The larger in the second case. As a result, the harmonic
components of [Ibus-h)O, are calculated from the voltages were much higher in the second case. The 7th
specified P and Q or from the equations describing the harmonic voltage at the harmonic source bus was
non-linear loads. Following the same process as for the 0.0167 p.u. in the second case, while it was only 0.0020
conventional power flow problem [Vbus-lJ1 is calculated p.u. in the first case. It is interesting to note that in the
by using Equation (14), until convergence for the parallel resonance equivalent situation, the 7th
fundamental power flow occurs. Then the vectors [Vb us- harmonic currents flowing in the system lines were
hP for h=3,S, .... M, are computed from the same much higher than those injected from the source (0.036
equation. The new voltage vectors [Vbus-hP for h=3,5,.. and 0.008 p.u. respectively). The increased current flows,
. .M, are compared with the corresponding vectors of characteristic in parallel resonance, bring up the system
the previous iteration [Vbus-h)O . If their difference is less losses. Insulation damage could be also expected at the
that the tolerance, the process stops. Otherwise, the capacitor bank considering that the corresponding
process is repeated, until convergence occurs. harmonic voltage was higher there than at the source
in both cases.
2.3 The Solution at Faults Figure 4 illustrates the 7th harmonic voltage at
Single, two or three-phase to ground faults are the capacitor bus, for all phases, as a function of the
simulated by short-circuiting the corresponding bus- capacitor size. It is interesting to observe that the dis-
node, where the fault occurs. The short-circuit is equiv- tortions of phases band c were greater than that of
alent to the pre-fault voltage VI in series with an equal phase-a, especially at near resonance conditions, even
but having an opposite sign voltage source (-Vf), where though only phase-a harmonic current injections were
Vf is considered a distorted waveform expressed as a considered. The last was due to the mutual coupling of
Fourier series for all harmonic frequencies under con- the distribution lines. The maximum and minimum
sideration. The fundamental and the harmonic bus harm. voltages occurred for capacitor values close to the
voltages during faults are equal to the pre-fault ones that made the driving point impedance at the
voltages, plus those due to the fundamental or source bus look as the equivalent of a series or a parallel
harmonic voltage components of -VI. It is assumed that resonance.
the system equivalent at the substation transformer
connection point towards the transmission and

744
4. CONCLUSIONS REFERENCES

The propagation of harmonic currents injected by [1] J. S. Subjak and J. S. McQuilkin, "Harmonics,
nonlinear devices may cause harmonic voltages higher Causes, Effects, Measurements and Analysis : An
at points far from the source. Harmonic levels at a par- Update." IEEE Transactions on Industry Applica-
ticular frequency may become unacceptable for certain tions. vol. 26, No.6, Nov./Dec. 1990, p.p. 1034-
capacitor configurations due to resonances. The reso- 1042.
nance quality factor greatly depends on the system load [2] IEEE Working Group on Power System
conditions. Simulation studies help in understanding Harmonics, "The Effects of Power System
the harmonic propagation problem for different Harmonics on Power System Equipment and
capacitor configurations, load conditions and Loads." IEEE Transactions on Power Apparatus
penetration levels of nonlinear generations (pv and Systems. vol. PAS-I04, No.9, September
inverters) or loads. The impedance at the substation bus 1985, pp. 2555-2563.
looking towards the transmission system, at any [3] D. Xia and G.T. Heydt, "Harmonic Power Flow
harmonic frequency is an important factor in the Studies Part I - Formulation and Solution." IEEE
propagation of harmonics. Even though this impedance Transactions on Power Apparatus and Systems.
can be mainly estimated from the fundamental vol. PAS-lOD, No.3, March 1981, pp. 1307-1315.
impedance of the stepdown transformer as Rs+jhXs, it [4] M. Tortoreli, P.E. Russell and J. A. Morris, "A
may be far from the estimated value, when a resonance Three-Phase Harmonic Power Flow Program for
occurs in the transmission system at a particular Distribution Feeders" 18th Annual Conference
harmonic frequency. on Modeling and Simulation, Pittsburgh,
Setting limits on the harmonics injected by the Pennsylvania, April 1987.
distorting sources is necessary and engineers have [5] M. Tortoreli, Harmonic Distortion in Power
worked towards this direction [6]. However, it must be Distribution Systems. Ph. D. Dissertation,
noted that harmonics from different sources may add Arizona State University, December 1987.
up (depending on the phase angle of the harmonic cur- [6] EPRI AP /EM-3124, Interconnecting DC Energy
rents and the resulting voltages) or even enhanced due Systems: Response to Technical Issues.
to resonances. Therefore, distribution system planners
should look at the system impedances as well.

1d 5-6 1-6
"'t(indeqreesJ

1-2 3-2 J-<4 5-4

Fig. 1. A three-phase line commutated inverter. Fig. 3. Voltage waveforms for inverting operation
with a delay angle a equals to 1500.

LJ ~. ~ t-- t--
I.
iJl)
l/
f\ f j
I

- - t--

-22 0
0.>1,,+90 <.>1 0 .. 150
wl (in .a9'"s)

s.q",nee ot Oiod"CoMuetinn

Fig. 2. Voltage and current waveforms for the Fig. 4. 7th harmonic voltage at the capacitor bus-#194
converter of Fig. 1. operating as an inverter. of the Alhambra distribution system due to a
single -phase harmonic source at bus-#56.

745
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERFNCE 8-12 APRIL 1991 LISBON, PORTUGAL

DESIGN OF A HYBRID PV!DIESEL SYSTEM

FROM THE ANALYSIS OF THE PRESENT PV SYSTEM PERFORMANCE
IN A DAIRY FARM IN POZOBLANCO, SPAIN

A.GONZALEZ D.MAYER
C. I. E. M. A. T. Ecole des Mines de Paris
Insituto de Energia Renovables Centre d'Energetique
Avenida Complutense, Rue Claude Daunesse
Madrid 28040 Sophia Antipolis
Spain 06560 Valbonne - France

ABSTRACT: In a first phase a PV-System was installed in a farm near Pozoblanco Cordoba, Spain, to supply electricity for cows
milking, water pumping, lighting, etc, financed with funds from the DG XII Program of the CEC and from the CIEMAT-IER of the
Ministry of Industry and Energy of Spain.
A simulation model adapted to this installation has been built to analyse the different components behaviour and analyse the system
operation. The results of the system performance are presented in terms of efficiencies.
In a second phase a diesel engine will be added to allow on one hand more reliability in the system performance as well as more
flexibility to the user and on the other hand to increase the expected life of batteries.
A general description of this hybrid PV/Diesel system will be presented in this paper.

1 • INTRODUCTION
Peak ISC(A) Voc FF(%) IMAX VMA
Power (V) (A) X
A PV -System has been installed in a dairy farm in Pozoblanco, (W) (V)
Cordoba, Spain, to supply electricity for cows milking, water
pumping, lighting, etc.
MAX 46.9 3.3 20.0 0.79 3.1 15.8
This project began the 01/01/88 and officialy finished the
30/06/89 and has been financed with funds from the DG XII MIN 42.2 2.9 18.9 0.69 2.8 14.1
Programme of the CEC and from the CIEMAT - IER of the
Ministry of Industry and Energy of Spain. AVE 44.3 3.1 19.5 0.73 3.0 15.0
The "Centre d'Energetique d'Armines" of the "Ecole des Mines S. D. 0.82 0.08 0.22 0.02 0.30 0.30
de Paris" has collaborated in the assessment of the installation
during the contract duration.
Due to the unexpected fast increasing of the user's exploitation Table 1 : Characteristics of PV-modules installed in
and consequent changes in loads, it has been necessary to Pozoblanco, Cordoba, Spain
install a diesel engine. By taking into account, that for this kind
of application the installation of a back-up generator would
improve the reliability of the whole system as well as would The whole PV system is divided, as it has been said above, in
increase the replication factor, a second phase was proposed three subsystems which components characteristics are given
and approved within the frame of the JOULE Programme with below:
the first aim of optimizing the design of hybrid PV/Diesel
systems to use in similar installations. * Subsystem A (24 volts)
This work will be carried out with the collaboration of Annines PV-Modules Re~lator
which in this case figures as co-proposer.
-Manufacturer: ISOFOTON manufacturer: ATERSA
The complete description of the PV-System and the loads (ARCO Technology) - Model: RS-130T
connected, the analysis of the system operation, as well as the - Nr. of modules: 30 - Nr. of steps: 2
future hybrid PV/Diesel system will be presented in this - Peak power = 1.4 Kwp (1.35)
document. - Wiring: (2 in serie) x 7 in parallel
+ (2 in serie) x 8 in parallel

!lattm Inyerter CAuxiliruy)

2 - GENERAL DESCRIPTION OF THE WHOLE
SYSTEM - Manufacturer: TUDOR - Manufacturer: AMBAR
- Type of element: 7 EAN 100 - Rated power: 300 W
2.1 PV ·System - Nr. of elements: 12 (serie) - Type: Ferroresonant
(sin-- Capacity: 1120 Ah wave)
The PV-Array consists of 269 M75L PV-modules (ARCO
Technology) from ISOFOTON, divided in three independent * Subsystem B (108 Volts)
sub-arrays connected to three different PV-subsystems : A, B
and C the more important characteristics of which will be given PVModules
below.
- Manufacturer: ISOFOTON - Manufacturer: JEMA S.A.
All these PV-modules were tested in the IER laboratory and (ARCO Technology) - Total power: 2 + 4 Kw
before their installation, were selected to obtain the minimum - Nr. of modules: 225 - Input
dispersion of parameters in every serie associations (strings). - Peak power = 10.6 Kwp (10.0) . Min. Voltage :97,5 V DC
- Wiring: 2 x [(9 in series) . Max. Voltage : 130 V DC
Table 1 shows the first characteristics of these PV modules x 8 in parallel] + (9 in series) . Rated volt. : 110 V DC
after testing. x 9 in parallel - Output
. Three-phase 220 V AC

746
 - Technology * Subsystem B
- Manufacturer: 11JDOR .DC/DC converter (chopper)
It supplies the electricity for the following loads:
- Type of element: 10 EAN 100
- Vacuum pump(s) for milking of cattle
Intennediate voltage: 310 V DC
- Nr. of the elements: 54 (serie) · Motor power - 2 CV/ cos j = 0.8
FUn DC/AC converter PWM - Milking cooling tank
- Capacity: 1600 Ah · Compressor motor power = 4 CV ;
cosj = 0.8
Regulator · Fan Motor power =0,4 CV ; cos j =0.8
- Manufacturer: ATERSA Pumping milk motor power =0,5 CV ;cos j =0.8
- Model: CRM-03
- Nr. of steps: 3 * Subsystem C
* Subsystem c (pumping system) It supplies the electricity to pump the water needed for the farm
exploitation, by means:
PV-Modules
- Two SP 4-8 Grundfos pumps working alternatively
- Manufacturer: ISOFOTON depending on the water level of the two wells where they are
(ARCO Technology) - Manufacturer: GRUNDFOS
- Power: 1000 W installed.
- Nr. of modules : 14 Average total height: 22 m
- Peak power = 0.66 Kwp - Input
- Wiring : (7 in serie) · Rated Voltage: 100 V DC Envisaged theoretical flow: 10 m3/day
x 2 in parallel - Output
· Voltage: three-phase
AC/60V 2.3 Monitoring system
· Current: 10 A
· Frequency: 6-60 Hz
The monitoring system consists of several sensors (i.e. :
2.2 - Loads Pyranometer, calibrated cell, anemometer, shunts, etc) located
through the different equipment of the installation and 23 (25)
transducers (KAINOS) of signals conditioning located in two
The PV -system described above supplies electricity to the panels where every necessary parameters to evaluate the whole
different loads which are described below. PV-System are conditioned in the same current range (0-20rnA)
proportionally to their actual values.
Be~ore the input to the acquisition equipment (DATATAKER),
* Subsystem A which has been programmed to take the reading of 23 channels
eac~ period of 30 s, saving the average, minimum and
It supplies the electricity for the following loads: maxImum value of every channel in each period of 10 minutes.

- Farm lighting F~m the DATATAKER, all the accumulated data during a day
· Total power installed: 380 W Wlll be transferred to the PC where are saved in diskettes which
· Ave. simultaneity factor: 0.5 are removed for the user every weeks.

- Milk shaker (via inverter 300 W) The process of communication between the DATATAKER and
· Power: 250 W the PC has a duration of approximatively 10 minutes and the
PC is started automatically for the DATATAKERjust d
- Monitoring system uring 20 minutes per day,n avoiding therefore a too high daily
· Sensors + Datataker + PC-computer consumption of the whole data acquisition system.

§±m= -----'-,....I II
r----'
9 , ~I

til
0o 24VDC
i----r----I
°
R~GUlA~OR
B

c
II I I I I 0-'----l
PUMP SYSTEM

° CURRENT SENSOR WATER

TANK

o
o VOLTAGE SE.NSOR

POWER SENSOR
WELL 1
FLOW
METER

figure 1 : Diagram of the whole PV systems and loads

747
 2.4 • Diagram of the whole PV.System and when: :
loads AO : Array output (kWh)
S : Total array area (m2) : 12 m2
The diagram of the whole PV-System, loads, sensors and da~ GI : Planar insolation (kWh/m2)
acquisition system to monitor the whole system performance IS II : Invener input (kWh) .,
shown in the figure 1. RCP : Evolution of the reference module effiCiency With
the cell temperature

3 - 108 VOLTS SYSTEM OPERATION ANALYSIS July iAugust Sept. Oct. Nov. Global

Only the 24 volts subsystem was instrumented as decided. The

experimental values were recorded at a 10 minutes time step for AC consumption 358 1216.5 1069 937 690 4315.5
the following period : (kWh)
Array output 256 938 843.5 960 497 3494.5
7 days in July 1989 (kWh)
24 days in August 1989 Deficit (kWh) 171 537 951.9 226 307 1692.5
26 days in September 1989 Coefficient of USE 0.52 0.57 0.58 0.76 0.56 0.60
27 days in October 1989 RCE(%) 6.5 6.4 5.75 6.35 7.4 6.2
23 days in November 1989 REGE(%) 5.4 5.6 4.75 5.35 6.5 5.3
CUEP (%) 67 64.8 57.25 63 71 62
The software that has been developped to analyse the system
performances is based on the simulation code HERMINES PV,
composed on the one hand of models libraries related to data Table 2 : Simulated performances of the 108 Volts subsystem
bases gathering the real characteristics of classical components
and on the other hand of routines dedicated to the system The coefficient of use expresses the ratio of the pan of the
control and management consumption ensured by the solar energy to the whole energy
required. It has a mean value of 62 % during the test period.
It has been validated on the 24 volts subsystem Ill.
A first obvious explanation of this rather low value is the
In the case of the 108 Volts subsystem, no data were recorded. underdesign of the system related to the load consumption.
The only measurements available are those reponed by the However, a detailed analysis of the system operation shows
farmer related to the mean daily consumption of the milking and that the disconnection of half of the PV array occuring almost
cooling machines. every day and during three hours represents another factor
which has an influence in this coefficient value by minimizing
The increase of the cattle from 48 to 70 cows was accompanied the array production.
by an increase of the consumption coming from an increase of
both the power and the time required (fig. 2). The figure 3 indicates the mean duration of the different
disconnection periods (load, generator) over the day.
Power (W)

6000 Po" .. (w)

t-- p--

5000
11.+~
tbCOrtllCll1
mtlsurtd

4000 >-- ~

3000

2000

1000

o
o 8 10 12 14 16 18 10 22 24
D~ y limf (hours) o 2 4 6 8 10 12 14 16 18 20 22 24
daytime (hours)
Figure 2 : theoretical and real load profile
Figure 3 : Time and duration of the disconnection periods
In that case, it is not difficult to conclude that the 108 Volts over the day
subsystem is under designed.
The presence of PV array disconnections allows to explain the
The simulation of its behaviour will give us some information low value of both the coefficient of use and the effective
about the best way to find a solution to such a prob!em. . conversion efficiency (about 6 %).
In order to simulate the whole system operatIOn ~Ithout
experimental data giving any information on the behaViOur of In order to reach the autonomy, a PV system should be
each component, we kept some of the characteristics already composed of a 14 KWp generator and a 3920 Ah battery
determined for the 24 Volts subsystem. storage.
The results concerning the monthly and global simulated However it is difficult to know exactly how long the autonomy
performances of this system during the monitoring period an: would be insured.
presented in table 4 in terms of efficiencies as dermed below: If the farmer decides to go on increasing his cattle, the
photovoltaic cannot be the right solution if it remains the single
RCE : effective conversion efficiency source of energy. Another solution would be to use the diesel
RCP : potential conversion efficiency but not only as a backup generator.
REGE : effective global energetic efficiency
: use of producible energy coefficient For such an application where the consumption is liable to
CUEP evolve, an hybrid system including a photovoltaic and a diesel
RCE II generator would present a wide range of operation that is to say
AO would fit to imponant variations of load consumption or
RCE= S x GI CUEP=RCP REGE=S x GI
insolation.

748
4 - GENERAL DESCRIPTION OF THE PROPOSED because of the new already mentionned devices (milking
HYBRID PVIDIESEL SYSTEM machines + cooling).
4.1 - Antecedents and future The existing inverter would be used for small AC loads and the
old cooling device still in operation.
Due to the failures of the transformer-less inverter, especially
designed for this project and dedicated to the milking There will not be any changes in the subsystem C.
machine (2 CV) and to the milk cooling, it was a necessity to
install a diesel engine to avoid serious damages for the user's Finally, the capacity of the monitoring system will be increased
exploitation. to be totally adapted to the new design of the installation.
At the present time, the inverter is in order but can only be used The PV/Diesel system layout is shown in figure 8.
to supply energy to the milking machine since a new cooling
device (6 CV) has been bought by the user and is connected to
the diesel generator. Moreover, the farmer intends to substitute
the milking installation for a more powerful new one (6 CV). 5 - SECOND PHASE OBJECTIVES
After having considered all these evolutions in a two years The second phase program can be divided into three parts :
period, a study on the optimization of a PV/Diesel system
design to be used in similar applications has been proposed and - Continuation of the data monitoring during the whole period.
approved within the frame of the Joule Program and pending The results obtained will provide without any doubt many
the contract signature. answers to some of the problems pertaining to such an
installation.
4.2 - Description of the proposed PV/Diesel
system equipment implemented - Optimization of the plant operation which includes:
The performance analysis of the PV system has shown that the • the optimal sizing of PV/Diesel systems applied to such
subsystem A (24 Volts) is slightly overdesigned. So, the application, depending on the exploitation scale and the load
increase of energy requested due to new lightings will be management.
supplied by this subsystem without any modifications. • Study of different possibilities of equipment connection to
allow the eventual exploitation size increase.
Regarding the subsystem B, a diesel generator (16 KVA) will
be connected to the batteries via an electronically controlled - Analysis of economical viability of PV/Diesel systems for
rectifier of 20 KVA. It will charge automatically the storage if farms located in remote areas.
necessary maintaining in any case a minimum battery state of
charge in order to increase the storage lifetime. REFERENCE

It is quite obvious that the diesel generator will be able to /1/ D. Mayer, F. Elalaoui
supply electricity to every AC loads in case of inverter failure. "Pozoblanco photovoltaic dairy farm Phase I -
Installation sizing - System Operation Analysis"
Besides, a new and more powerful inverter (6CV + 6CV) of Final report, February 1990
the same technology will be provided for as soon as possible

r----------,
I ~I SMALL
I • - _ ACLOADS
I
~
6kW
=
• I 2 kW I
I = I OLD
B I 1=71 i n COOLING
I .1L::::fI'lI" TA N K
:L.. _ _ _ _ 4 kW I
______ _ ( 4 CV)
PRESENT INVERTER

r----------.., FUTUR
I ~MILKING
:12-15kW·~ MACHINE
~
(6CV)
• I = I
6 kW I
I = I NEW
I _J=7t_LIL
• COO LI NG
IL:::..fiI..J
TA NK
L ______ ~k~_J (6 CV)
FUTUR INVERTER
• CURRENT SENSOR
+
TO AC LOADS

oo VOLTAGE SENSOR

POWER SENSOR

Figure 4 : Diagram of the future PV/Diesel system and loads

749
10TH EUROPEAN PHOTOVOLTAlC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON,PORTUGAL

PV SYSTEM GENERATION ANALYSIS, 0 & M AND COST EVOLUTION

HIDROELECTRICA ESPAROLA S.A. EXPERIENCE
by A. Pantoja, J. Garcia and A. de Julian
HIDROELECTRICA ESPAROLA S.A.
Research Deparment
Aptdo. 458-C 28080 Madrid. Spa1n.
Telephone (1nternat1onal-34) 1/577-65-00
Telefax (1nternational-34) 1/576-71-26
Telex 23786 HEKD E

- Abstract -

The present paper presents the operating experience of the

experimental power PV systems of HIDROELECTRICA ESPAAOLA
S.A: its 0 & K costs for PV stand alone systems as well as
grid connected systems. It analyzes also the cost
evolution since initial installation due to the
optimization of the operation and maintenance program,
implemented in order to reduce costs and assure continous
service to the user, keeping a high reliability and quality
of the service.

r-
Operotion hours (h)
- C<1d connectad PV oyolatn -

HIDROELECTRICA ESPAAOLA S.A. presents

its PV system experience with more than l'~
',1

100.000 hours of operation. It is u

important to emphasize the low level of ','
the 0 & K cost, less in % than with '.'>
conventional generation systems.
1
I.,
1.1
1
The paper is divided in three areas:
- Generation Analysis.
- 0 & K cost.
- 0 & M cost
maintenance program.
evolution and
.::~g"
..
0,'
0,_

..
.,>
The analysis have been performed for
grid connected PV systems and stand • I, lo

alone systems.

GENERATION ANALYSIS: The first figure Genf:rotion losses (t.I ....h)

show the first
aproximately 1000
year operation with
hours uncorrect
- Grid con_bod PV oy.tam -
--,
operation and its distribution for the
diferent parts (solar array, power
conditioning,otherl. In this way, the
figure show too how after optimization
of the operation and maintenance program
these hours can be reduced 60 %.
The second figure show which is the
generation losses and its evolution, 19
KWh energy lossed for the first year to
5 KWh energy lossed in the forth year of
operation.
I.
Both figures are about grid connected
PV system.
Stand alone systems present a high
I.
y.....
.
reliability and quality of service.

750
o & H COST: This analysis allows to know o & K cost evolution and aalntenance
the different components of 0 & M cost proqram: The next figures show the
in the PV systems. Total costs are cost evolution with PV system life. We
divided on the following areas: think that the tendency cost will be
1-1,5 t initial budget for gr1d
- Personell. connected system and 2-2,5 t for stand
- Equipment. alone system.
- Subcontractors.
- Other costs.

OtkM cost O.tM total annual cost evolution

- GrId connected PII system -

u
' ..
,..
1.7
~
e
.
.s.. '"I.'I.>
. 1.2
.E I.'
:>

;;
:>
:5
I
.....'
.....'
0.7
~
~
Ie
..'
0.>
'....2
,. . Year
:II ..

- Stand alone PII system -

ur-
- Stand olOtM PV 1)'1t.wn -

S\lbconITector (0.0:1:)
~
c
.s•
...
•
I
•":>
.5
2.11

,.,
;;
:£
'..
~
...
- ~~
I(

P ...oneIl (49.6:1:) ,. . Veor

.. ..

751
 10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Half a dollar per kWh am still dleap electricity

c. Daey 0U\Nens, Provincial Energy Bureau, North Hollam

P.O. Box 3007; 2001 IlA Haarlem, Hollan:l

ABSlRACl'. An analysis is presented for the costs of electricity

produced by an autanc:m::A.Js solar cell system in canbination with
efficient electric a~liances, as they are used in houses am small
offices. If efficient a~liances are used, an average household will
use about 700 kWh per year, 0CIIpCIZ'9d to abaIt 3000 kWh it is usiIg on
an "average today. Efficient awliances are Dm'e expensive. However,
the extra cost per kWh saved is abaIt the saJre as the cost of the fuel
needed to produce a kWh in a power plant. '!he 700 kWh per year can be
produced by an autanc:m::A.Js solar cell system with an area of 3 m2
(efficiency 19%). '!he analysis holds rwghly for the area between 400
northern am soothern latit1xle. A cc.uparison has been made between the
total costs of a grid connection am the costs of an autanc:m::A.Js solar
cell system. If all costs are capitalized am taken together, a grid
connection will amoont to abaIt $ 6000 (00 fuel costs are included).
'!he costs of a solar cell system to produce 700 kWh per year - with
today's prices - are abaIt the same or even lOlOer. '!he lOCldu.lar nature
of solar cell systems makes it possible to start with a certain number
of a~liances, dE!pen:lhg on the amoont of lIDIIeY one can spend.

Introduction
A carparison has been made between the costs of an last ool\.Dllll shows the awlianoes wdl are
already
electricity SUWly system ani an autanc:m::A.Js solar or can be pIt on the market in a relatively short
cell system. period. Efficient awliances are Dm'e expensive.
'!he considerations hold for small users of However, if these extra costs are calculated per
electricity (3000 until maxim.nn 8000 kWh per year) kWh saved (difference between oolUllll 1 ani 3),
am. the area between 400 rnrthern am. scut.hern they are abaIt the same as the cost of the fuel
latit1xle. '!he followiIg aspects have been needed to produce a kWh in a large power plant.
analysed:
1. Demarxi for electricity. Especially the use of Table 1. Annual electricity oonsunption of
the efficient electric awliances is eq.hasized. danestic awliances in 1988 (kWh/year);
2. costs of a grid connection. All costs of the Danish data
oonstruction am maintenarx:e of an electricity best advanced

SUWly system are included. average available ~

3. costs of an autanc:m::A.Js solar cell system.
4. o:mparison of these costs in connection with refrigerator 350 90 50
the demarrl for electricity. freezer 500 180 100
washirr;J machine 400 240 115
1. Efficient electric arpliarpes ventilation 500 275 75
In Table 1 an overview is given of the electricity lightiIg 800 150 150
used for electric awliances (Danish data) (1). misoellaneaJs 350 200 200
'!he first ool\.Dllll shows the average electricity (radio, 'IV,
used for these awlianoes in 1988. '!he second etc.)

ool\.Dllll shows the best available awliances ani the total 2900 1135 690

752
For the efficient awliames mentiooed in Table 1 produced I:1j the cells can be used, the efficieooy
ally aIxA.tt 700 ld>b is needed 0ClIparE!d with aIxA.tt of the cells lILISt be aIxA.tt 19% to obtain the 690
3000 ld>b these awliames are usin;J 00 an average kWh needed (Table 1). nus efficieooy is elqleCt:ed
today. before the year 2000 far mass produced
polycrystalline solar cells. en the laboratory 24%
2. Grid oannectioo is already obtained far crystalline solar cells.
To make a c:x:mparisoo bebIeen an au1:cn:lDoos solar
cell system am a grid CCX1l'leCti00 far both the Ihe use of 60% of the electricity produced I:1j the
capitalized cxsts have been calculated over a cells seems realistic, because:
period of 20 years (no fuel costs are incllDld). - a relatively large part of the electricity
Ihe costs far a grid are SllIJIIIarized in Table 2. As produced I:1j the solar cells can be used directly
they hold far a densely pq:W.ated area in Hollan:i, (refrigerator, freezer, ventilatioo, etc.), also
they may hold far a:trj area. sane kin:i of load management is possible (freezer
ooly \VOrkln;J durin:] daily hcm's etc.)
Table 2 • Average capitalized costs far a grid of a - the battery will have a high efficieooy because
small oonsumer of electricity in a densely dlargin:] and disdlargin:] are sla./ly
pc.pllated area in Hollan:i. It holds far a new grid - bebIeen 400 northern and southern latitme the
(dollars 1989). No fuel costs are incllDld. average daily insolatioo is rather equable. Ihe
choice of the inclination of the solar panels can
Q:lnnectioo to the grid of a small oonsumer increase the equability.
(3000 kWh/y) $ 380
Distribltioo grid (10 kV, 380/22OV) $ 1350 Table 4. Yield of 3 m2 solar cells in a sunny
Transport (50 kV, 150 kV) $ 470 climate (kWh/year). Insolatioo 2000 JdtJ./m2. Y•
Offices, cars, wrltshqls, etc ~ Percentage of electricity used: 60%.
Total $ 2650
efficieooy Yield
In table 3 nat ooly the costs far the grid, J:ut cells JdtJ.
also the costs far the power plant, persamel am 19% 690
maintenanoe are' given far a small oonsumer.
It is seen that a grid connectioo cost 00 an Ihe capitalized cost of a system is today alxut
average al:x:ut $ 6000. $ 9.25 per Wpi nearly half of this aJDmt is far
the controller, battery, CCX1l'leCtions, etc. 1M:"in:]
Table 3. All costs (capitalized) far a grid the next 10 years this aJDmt may go down to $
oannectioo of a small oonsumer. No fuel costs are 4.35 (2). In Table 5. the capitalized cxsts far a
inclOOed. solar system is given to produce 690 JdtJ..

Grid etc. (Table 2) $ 2650 Table 5. capitalized costs far an autcn::IIDJs solar
Power and 380 kV transport $ 1500 cell system to produce 690 JdtJ.. System efficieooy:
Personnel am maintenanoe $ 1950 60%.
Total $ 6100 1991 2000

If ooly 700 kWh is used instead of 3000 kWh the capitalized costs $ "5310 $ 2515
costs will ooly be slightly less and will be al:x:ut
S 5000. Yearly costs
(20 Yi 4% int.) $ 390 $ 185
3. Electricity prtrlI'9ffI by an autaTptg,Js solar
cell system $ 0.56 $ 0.27

In Table 4 the electricity generated I:1j a solar

cell system with an area of 3 m2 has been Ihe yearly cost of a system will be al:x:ut $ 390.
presentedi 2000 kWh per m2 per year is often nus aJDmt is rooghly the same a small oonsumer
received in the area bebIeen 400 northern and is payin:] today in an industrialized c:oont:ry far
southern latitme. When 60% of the electricity the 2900 JdtJ. (Table 1 first ooll.lllU'l). HoIriever, as

753
ally 690 kWh is used in case of the use of
efficient ClRlli.arlc2;, it means that a kWh will
cost al:loot half a dollar.

4. Ccl!parison

In Table 6 a carparison has been made of the

investment costs of an aut.onc:.mnJs solar cell
system with the costs of an electricity stgIly
system (grid oonnectiat) •

Tabel 6. catparison of the investment costs of an

aut.onc:.mnJs solar cell system with the cost of an
electricity stgIly system if 690 kWh is used.

Electricity stgIly system

(small user of electricity; $ 6000
no fuel costs irx::ludsd)

1991 2000

aut:olxm:xlS solar cell system $ 5310 $ 2515

As can been seen in Table 6. aut.onc:.mnJs solar cell

systems are cheaper than a grid oonnectiat.

Conclusion
system - even with todays
An aut:olxm:xlS solar cell
prices - is cheaper than a grid oonnectiat if
efficient electric apparatus is used. '!his holds
far a small consumer of electricity (sdlools,
shops, small offices, farms etc.) in a SI.Il1l1Y area.
'!he costs are capitalized CNer a period of 20
years am hold far a new grid oonnection.

Literature

1. Electricity, Efficient En:l-use am NeW

Generation Technologies am '!heir Pl.annin:]
Inplications. T.B. JetJansson et.al.,
Lund University Press, Lund 1989.
2. Cheap electricity with aut:onc:m::us solar cell
systems
c. Daey D..Mms, Provin:::ial Energy Bureau, North
Hollard (to be pJblished) •

754
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

TISO: 4 kW EXPERIMENTAL AMORPHOUS SILICON PV POWER PLANT

D. Chianese, M. Camani, P. Ceppi and D. Iacobucci

Dipartimento dell'ambiente
Sezione energia e protezione dell' aria
CH - 6500 BELLINZONA (TI)

ABSTRACT
A 4kWp amorphous silicon (a-Si) PV system has been working
now since three years at the experimental solar plant TISO.
Research is carried out under the sponsorship of the Swiss
Federal Office of Energy and is aimed to investigate the long
and short term behaviour of thin film PV modules operating
under real environmental conditions.
Additionally to the 4kWp plant, a test facility has been set
up which allows particular experimental work and trainig to
be done on small scale systems and on single modules.

1. INTRODUCTION previous papers (1-4). Here we just recall

the main features of the installation.
A 4kWp/195Vm a-Si photovoltaic power
system and a test stand for thin film PV
modules have been assembled in spring 1988
at the Scuola Tecnica Superiore in Lugano, 2.1 4KW SYSTEM
Switzerland, Modules of three manufacturers are
The 4kWp array is connected to the integrated in the generator. Modules of
public power grid (220V/50Hz) through an the same brand are wired in serie and the
automatic inverter unit and is intended to strings are connected in parallel to a
provide in field operating experience of single inverter unit.
a-Si PV components integrated in a A summary of the mani features of the 4 kw
relatively large PV generator. system follows:
Modules of three manufacturers were
integrated in the PV-generator. - Total nominal power: Pn = 3900 Wp
The plant is equipped with a thorough - Modules: 96 x 30Wp' ARCO SOLAR G4000
data acquisition system which continously 48 x l2Wp' CHRONAR, CSB.13.E
monitors the behaviour of the modules replaced with
working under real conditions.
26 x l2Wp' CT13
The data collected are processed off
22 x 20Wp' SOLAREX, SA20
line giving information about medium and - Strings: 14
long term behaviour of the components. - Operating voltage: 195 Vm
The generator has been working now for - Tilt angle: 55°, adj. from 30° to 55°
more than 1000 days.
The test stand is intended for the Table 1: Main features of the 4 kW system
execution of special experimentations on
single modules. Particular instrumentation
was devolopped along with the test
procedures. 2.2 TEST STAND

The test stand is set up on the same

premises as the 4 kWp generator and is
2. DBSClUPTION 01' '1'BB PLaNT
engineered to be adaptable for different
Technical details of the plant and tasks.
first results have been described in Fourtytwo modules may be mounted and

755
electrically loaded with specially temperature vs. time.
developped equipment. (' CJ (X]
60 5.O'-t-'-..L-.Ll..-L....J--L---LLLLLJ....L.I..L.LLLLLJ....L.I....L.LLLJ-,..-
The data acquisition system may be
55
instructed to collect data from the test
50
stand as well as from the 4kWp power 45
system. 40
A control cabinet near to the test site 35
contains all needed terminals. 30
25
20
15
3. RBSULTS AND OPDATDlG BXPBRIENCB 10
5
o
3.1 MEASUREMENTS UNDER STANDARD CONDITIONS .......,
Fig. 1: Efficiency of the modules under
Three modules out of the 4 kWp
real working conditions and ambient
generator were taken as samples and
temperature vs. time
underwent specific tests at standard
conditions before and after exposure.
Following points are worth noting:
The efficiency of the modules as
delivered, was about 5% higher than the - after a strong initial (first 70 kWh/m2 )
degradation the ROC-efficiency stabilizes
nominal value.
at about 7S±S% of the initial value;
After being exposed at a global
radiation energy of 1000 kWh/m (by january
2 - degradation and regeneration is cyclic
1989), an efficiency loss between 28% and and follows the evolution of the ambient
34% (referred to the efficiency measured temperature (daily mean values). During
before exposure) was observed on the winter the efficiency decreases whereas
modules of all three manufacturers. it increases again during spring (see
After 2700 kWh/m2 (by march 1990), near fig. 1).
to the bottom of a seasonal
degradation-regeneration cycle (see 3.2), The output current of each string is
the efficiency loss measured was between continously monitored. At the beginning of
31% and 35% (again referred to the operation the relative difference in
efficiency measured before exposure) . current among strings with the same modules
The voltage at the maximum power point was less than 4%. After 1000 days the
(MPP) has decreased about by 20%. Half of difference amounts to 7%, indicating a non
the loss has occurred during the first 15 uniform degradation of the modules.
days of exposure (abt. 70 kWh/m2 ). The resulting dispersion of the
As a result after about 200 days of characteristics may be the reason for
power production, none of the modules additional power loss in the strings.
tested met the nominal power declared by
the manufacturer. However the net Infrared images of the three types of
degradation seems to have stopped, leaving a-Si modules deployed in the grid-connected
the modules degrade and regenerate array were recorded.
seasonally in accordance with their working No hot spots due ~o reverse-bias power
dissipation were detected.
temperature.
The terminal boxes on the rear surfaces of
the modules caused local temperature
Measurements at standard conditions were
increases of approximately SOC.
carried out by the staff of ESTI (European
Solar Test Installation, c/o JRC-Ispra).

3.3 FAILURES

3.2 4 KWp PLANT

The modules integrated in our system
The efficiency of the PV generator is were not reliable enough: electrical
monitored under real operating conditions isolation, mechanical structures and
(ROC-efficiency) . Figure 1 shows the internal wiring, being the most evident
behaviour of ROC-efficiency and air weaknesses.

756
 [»
Arco Solar G4000 had the lowest failure 5.0 +-__--L_ _ -L_ _-L_ _.-.-1_ _--I-
rate with 3% of the installed 96 modules.
50% of Solarex's SA20 went out mainly
4.0
because of internal wiring problems: the
connection between one terminal box and the
active area was broken (see fig . 2) , 3 .0 CHRONAR CT13
"
probably due to excessive mechanical stress
caused by the deformation of the module 2.0
under windload and thermal cycles . A
couple of modules had their glass cover 1.0
broken. 7 November 1989 13 March 199 1

0.0 -jL-----,----r----,---~--~

•
o 100 200 300 400 500
(DAYS)

Fig. 3: Efficiency of the modules CT13

under real working conditions vs .
.·Siliron
time

-
at standard conditions) .
The efficiency loss recorded in the
field since the first day of operation is
- ('I) .... less than 22% of the initial value.
As a result the efficiency specification
Fig. 2: Schematic representation of the seems to be met by the new modules.
broken wire in a SA 20 module

Owing to the poly carbonate frame, the SA20 3.4 TEST STAND
modules didn't have any isolation problem.
The modules from Chronar France gave A first set of tests was performed with
the most problems, mainly caused by the available devices.
deficient electrical isolation at the Fourteen modules of every manufacturer
required operating voltage (Vm = 195V) and have been thoroughly tested at standard
by poor manufacturing. All modules had to conditions before and after exposure.
be replaced. The test procedure is listed below :
A first batch of new modules (CT 13) 2 modules are stored in a room, as
are being tested now. reference;
These new modules have a very thin 1 module is exposed in short circuit
polycarbonate/resin frame intended only to conditions (I.c);
give protection to the edge of the glass - 1 module is exposed with no load (U oo );
structure. - 7 modules are operated at the maximum
The active area has been kept back about power point (MPP) using a specially
5 mm from the frame . The module does not developped electronic load circuit. This
have any terminal box, but a small modules are exposed to different amounts
connector and can be mounted on a typical of solar energy .
simple structure with a special washer.
The new modules are performing The data recorded on the test stand
satisfactorily since 500 days (see fig. show that half of the efficiency loss
3) • (about 15%) occurs during the first 30
Some minor problems were encountered with kWh/m2 of global radiation on the module.
the stability of the edge connectors. The remaining 15% is lost during the next
The plot of efficiency (ROC) vs . time 1000 kWh/m2 •
matches the one presented in figure 1 . It is interesting to note that modules
The efficiency of the new modules exposed without load degrade more .
before exposure was on the average 17% The MPP-voltage Vm decreases with
higher than the value specified by the efficiency loss (see fig. 4) .
manufacturer (laboratory measurement taken After 1000 kWh/m2 , a variation of 20%

757
 [%]
4. CONCLUSIONS
3
2 The plant has been operating since more
1
than 1000 days.
0
-1 The a-Si modules tested, showed a very
-2 strong efficiency loss during the first 5
-3 to 10 days of exposure (ca. 30 kWh/m2
-4
global radiation energy on the module
-5
-6 plane) .
-7 After about 7 months (ca. 1000 kWh/m2 )
-8 the efficiency seems to be stabilized at
-9
about 75±5% of the initial value. A
-10
0 10 20 30 40 50 60 70 80 seasonal degradation-regeneration cycle
[kWh/m2]
which follows temperature variations has
Fig. 4: Percentual degradation of the
been detected.
MPP-voltage Vm vs. global
After ZQO days of operation the nominal
insolation. The single points of
power specification is not met any more.
the plot are referred to different
The reliability of the modules
modules.
integrated in real large systems at higher
voltage (195Voc ) showed to be a major
from the nominal value was noted on the
problem. In the best case (most reliable
modules of all three manufacturers.
module type) a failure rate of 3% was
This effect must be taken into account by
observed.
the power conditioning unit loading the
The modules with polycarbonate frame
system.
did not have any isolation problem.
Figure 5 shows the evolution of the
efficiency of the same modules during the
same time.
One notices that the efficiency decreases 5. ACKNOIILEDGEMBNTS
smoothly; some modules show an opposite
variation of MPP-current and voltage.
This project is financially supported by
the Swiss Federal Office of Energy.
['J
O+---L-~---L--~~--~~~~

-2
-4 6. REFERENCES
-6
(1) M.Camani et al. ; TISO:4 kW Amorphous
-8
Silicon, Utility Interconnected Power
-10
Plant; 8th EC Photovoltaic Solar
-12
Energy Conference, I-Florence, May
-14
1988.
-16 ARCOG400

-18
(2) M.Camani et al. : Utility
-20+---~~---,---,--,---,--,---+
o 10 20 30 40 50 60 70 80
Interconnected 4 kW Amorphous Silicon
[kWh!m2] PV Power Plant; 9th EC Photovoltaic
Solar Energy Conference, D-Freiburg,
Fig. 5: Percentual degradation of the effi-
September, 1989.
ciency vs. total global insolation

(3) M.Camani et al.: TISO: Installation

photovoltaique de 4 kWp avec modules au
silicium amorphe, raccordee au reseau;
Photovoltaik-Nutzun~ 1990, Zurich,
Januar, 1990.

(4) M.Camani et al. : Annual Project

Reports: Lugano, 1988,1989,1990

758
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Influence of Different Revenue structures on the Economy

of PV Plants in Selected Countries

G. Hille, F. StaiB, F. steinborn; Centre for Solar Energy- and

Hydrogen-Research (ZSW); dep. systems analysb;
Pfaffenwaldring 38-40; 7000 stuttgart 80;
Federal Republic of Germany;
Tel. 0049-711-6862-739; Fax. 0049-711-6862-349

Abstract
The rentability of PV plants depends not only on co~t, bu~ also,on revenues.
Revenues are gained if electricity is fed into a gr~d or ~f tar~ff costs for
non-delivered electricity can be taken into account. Under a study on the ap-
plication of solar plants for electricity generation ~n the Mediterr~nean re-
gion, the compensation structures of different countr~es are ascert~~ned. ,The
influence of different compensation structures on the revenues are ~nvest~ga-
ted. , d'ff
A system optimization with respect to maximum revenues can re~ult ~n a 7 ,e-
rent system design than the optimization with respect to max~m~m,electr~c7ty
production. An overview on possibilities to increase the rentab~l~ty of gr~d­
connected PV plants will be presented.

1. Introduction gained for continuous electricity pro-

duction during on-peak period. Further-
The rentability of PV plants depends more linear tariffs exist.
not only on cost, but also on revenues. If avoided tariff costs can be taken
The revenues of grid connected PV sy- into account, an increase in the value
stems are depending on the local isola- of energy production of more than 60 %
tion and the compensation structure of is possible. For a local public uti-
the local utility. European compensa- lity, cost-savings normally exceed the
tion structures differ over a wide revenues of private investors.
range.
As specific revenue structures for pri- In order to examine the influence of
vate produced electricity in Mediterra- compensation structures on rentability,
nean countries do not exist - or are a possible future photovoltaic-adjusted
not published for most examinated structure is compared with the existing
countries, the tariffs are ascertained. spli tted compensation structure system
The resulting revenues are compared of the FRG. For distinct non-linear
with those received by several compen- compensation structures, influences on
sation structures in Germany and Swit- photovoltaic system design can be ob-
zerland. served. If the ratio of winter to sum-
mer tariffs exceeds 2, an increase of
2. Tools the declination angle of the PV modules
is advisable.
The performance of various renewable Figure 1 shows daily profiles of diffe-
energy systems at different sites is· rent compensation structures in Germany
simulated with the computer program and switzerland. The on-peak and off-
SIRESTE developed by the Centre for So- peak time during one day is shown for a
lar Energy- and Hydrogen-Research. On timevariable and a linear compensations
the basis of hourly meteorological structure. If the annual average value
data, the systems output of single and of each compensation structure is taken
spatially dispersed plants is calcula- into account, a difference of about 300
ted and valued with the corresponding % can be found. The highest revenues
compensation structure. The time span are paid in the winter period at day-
normally is one year. time. If an owner can calculate avoided
costs, the tariffs for electricity from
3. compensation structures the grid are important. Usually these
tariffs are from 30 % up to more than
Countries like Germany do have several 100 % higher than revenues from the
public and private utilities, 8 big same utility. If electricity is produ-
ones, several regional and many commu- ced during on-peak tariff time, a capa-
nal ones. Therefore various structures city payment is paid which increases
can be found depending on the genera- the relatively low revenues up to 50 %.
tion system of the utility. In most The here presented "VDEW"-revenue
cases in the FRG structures exist at a structure has been most important com-
splitted system with a daily on-peak pensation structure in the FRG and was
period and an off-peak period at night jointly introduced by the three most
with lower prices in summer and higher important associations in the electri-
in winter. Capacity payments can be city sector, the VDEW, VIK and the BDl.

759
 Compen alion tructures in Germany without
Capacity·Payment: Daily Promes Tariff: in tbe MediteIT30ean Region

---
Daily Promes
r-------------------------------------, __ __
,,-
--- --
~

"
...e-. _.-
.. .e-.
f.
i __ . ._._____._______._.____. . .__ __ ~.
J.r-----------------------------~~
- -------- - -------
:10
--
... -
_ _ ._••

Figure 1 Figure 2

Many utilities did orientate their com- The here presented ideal tariffs with
pensation structures to the VOEW, espe- the average for the annual electricity
ciallY the important 8 big ones. From costs of 25 Opf/kWh area first itera-
the 1st . january 1991 German government tion and will be investigated more de-
decided by law, to increase the minimum tailed in the study on the application
revenues for elctricity from small hy- of solar plants for electricity genera-
dropower (up to 5 MW), biogas, garbage tion in the Mediterranean region.
gas and sewer gas to 13,91 Opf/kWh and
for PV and windenergy to 16,7 Opf/kWh. 4. Results and Disoussion
The highest compensation structure is
paid by the communal utility of Saar- Figure 3 shows the revenues of an in-
brticken, a linear tariff of 25 Opf/kWh. stalled 1 kWp PV located at one site in
Since this utility has a high percen- Germany. For a privately operated
tage of cogeneration plants in summer plant, annual revenues can differ by
time they are out of operation due to the factor of 2.7. A linear tariff de-
the lack of heat consumers. During this livers even more revenues than a time
period, electricity has to be bought splitted tariff with the same annual
from other utilities for higher prices. average. In the following analysis all
values will be related to the VOEW
The figure 2 shows daily profiles of compensation structure. Regional or
the strongly differing tariffs - com- communal utilities - or big industrial
pensation structures are not published consumers - usually have to buy a share
in the Mediterranean region. Coun- of their consumed electricity from
tries like Israel and Portugal do have other bigher utilities and have to pay
time splitted tariffs with on-, mid- "special tariffs ". Therefore they can
and off-peak periods, differing for earn 20 % more with electricity pro-
winter, summer and spring/autumn season duction from their own PV power plants.
as well as for working days and The revenues from the existing compen-
weekends. In other countries like Jor- sation structures in FRG of the 8 big
dan and Morocco the consumers can public utilities do not differ very
choose between timevariable and linear much, the BEWAG in West-Berlin however
tariffs. Many countries like Syria and pays 64 % more because up to now it has
Egypt do not measure the time de- been an insolated grid with higher
pendence of the electric demand but try electricity generating costs. The
to save electricity by strongly pro- highest revenues are paid in Saarbruc-
gressive tariffs for greater yearly ken with 2.7 times of the "VOEW"-reve-
consumption. In Syria the highest ta- nues.
riff exceeds the tariff for the average In Zurich, Switzerland, owners of PV
consumer three times. In many countries plants up to 7 kWcapacity are paid
like Egypt, tariffs for households are according to the P t,ariff structure,
very subsidized. which offers them revenues of 2.41 ti-
Two loadcurves for a winter working day mes of the "VOEW"-revenues. Even if
are shown. One is representative for they are paid by the compensation
the grids of the northern Mediterranean structure instead of the tariff for
countries, the other one for the plants with more than 7 kWp power, this
southern countries. In these countries would be 20 % more not taKing into ac-
the evening peak is 25 % higher than count the higher isolation in the Al-
the midday peak, but the average of the pes.
northern loadcurve exceeds the southern If a private operating PV plant is al-
one about 20 %. lowed feeding electricity into the grid
If the annual average costs of a natio- compensated with the revenues of house-
nal electricity grid including produc- hold-tariff, this can increase the
tion, transport, distribution and los- revenues in this example up to 70 %.
ses can be worked out, the connection Based on equivalent average prices for
of these average costs with the average the generated kWh, revenues can be in-
of the yearly loadcurves are the ideal creased by about 60 % using the
tariffs. photovoltaic-adjusted structure.

760
 Revenu of P for Different Tariff· tructures
Revenue of PV for Dirrerent Compensation
Structures in Gennany and witzerland
in the Mediterranean Region
r.. , .,w,",,- r... . J "W"-I

-=-
I
7U

...
..... ... III

-·eI-' COfTN~"""'(.N(ftGY. AHOH'I'tIftC)G.{H-«

~ ~~~~.,.......
K,UlCH ,, _ _ _ l ___

"-,,,,--.........-- -.-
......---- --
....... ........... ....-
~-
. ~-

Figure 3 Figure 4

Capacity payments exist only for com- the revenues could increase due to a
pensation structures. As shown by these higher on-peak during midday, caused by
calculations, revenues from tariffs are higher percentage of air conditioning.
always higher than revenues from com- The figure 4 shows as well that re-
pensation structures. cently in Spain and Portugal the same
revenues can be gained than in the
The figure 4 shows the revenues gained soutnern countries, due to higher ta-
in the Mediterranean region. In Egypt riffs in spite off 15 % lower isola-
and Jordan electricity for private hou- tion.
seholds is subsidized and therefore Capacity payments do not exist for ta-
poor revenues can be gained. But even riffs.
in Israel, receiving 50 % more isola-
tion than 30 0 south inclinated areas in The figures 5 and 6 are presenting the
Middle Europe, the specific revenues do specific revenues for every produced
not exceed the new German minimum reve- kWh. The higher specific revenues in
nues. The situation in southern coun- the Middle European countries in com-
tries is improving if big private elec- parison to the not costeffective ta-
tricity consumers are running a PV riffs in the Mediterranean countries
plant for grid-feeding. In Jordan this are demonstrated again. Syrias tariff
improves the revenues approximately for great consumers is the only excep-
220 %, in Egypt even more then 700 %. tion.
The highest revenues can be gained in In all these calculations revenues for
Syria, where the highest tariffs have the delivered power are neglected. If
to be paid. The development of the re- the reduction of the midday peak-load by
venues will be improved 50 - 1600 % if PV could be possible, higher revenues
the governments are willing to create could be gained. If data are available
costeffective tariffs. If costeffective such investigations could be done with
tariffs are compared, in the year 2005 the program SIRESTE.

pecific Reveoue of PV for Different Compen. peelfic Revenues of PV for Different Tariff·
ation tructures in Germany and Switzerland tructures In the Mediterranean Region
~----------------------~~~~~-----------, II ~

s--
. .... ....
i" ZI l

J. '. ' .
' ., ,....
'... I· ,.,. ,,
,..'
'....
o!! ~

l. L

'-ISW
.
t£JmIIII.tofIIlOLAIIIlHEAOY·AHOH"tDftOOe:""'UVJIICH
~ ~-= ~ttm.~
-
1 1 "'----""-
11'"-- ... _.. _ _ .......

Figure 5 Figure 6

761
In the figure 7 comparison of the spe- 5. Conclusions
cific revenues for different renewable
energy systems, single and spatial di- The recent possibilities to run grid
spersed hybridsystems is undertaken for connected PV plants costeffective are
the VDEW compensation structure in Ger- very limitated. This situation do not
many. One should focus ones attention differ very much between Middle Euro-
on the different energy-/capacity pay- pean and Mediterranean countries excep-
ment ratios. The PV plant is feeding ted to syria.
into the grid mainly during the on-peak In Mediterranean countries the great
period. This causes a higher capacity private consumers are rece~v~ng the
payment as compared to the other rene- highest revenues, due to there higher
wable energy systems. Hybridsystems are tariffs.
producing electricity more continuous The ra are several possibilites to in-
than monovalent systems. This causes crease the revenues:
the highest capacity payments for all If the ratio of winter to summer
compared systems. The capacity of the tariffs exceeds 2, the revenues can be
hybridsystems has been dimensioned such increased by a greater angle of incli-
they produce the same annual energy as nation.
a single PV power plant. I f hybridsy- Are the on-peak periods during the
stems are spatially dispersed, revenues evening hours, the use of a off-
can be increased by the higher capacity peak/on-peak storage especially with
payments. Therefore the owners of the pump storage stations should be calcu-
spatial dispersed PV plant have to be lated.
accepted by the utilities as one pri- If capacity payments are paid, ow-
vate producer of electricity feeding ners of spatially dispersed PV- or hy-
into the grid. brid-systems should try to sell there
electricity as one producer to the lo-
cal utility.

Compari on of Revenue for Different Renewable In the future costeffective tariffs

Energy y terns on Variou it including social costs of electricity
production would considerably improve
the profitability of PV plants feeding
into the grid.

6. References:
1) Informations of Jordan Electri-
city Authoritiesi Ammani 1990
2) RED Electrica Espana: Informe so-
bre la Explotation de Sistema
Electdco Nacionali Madridi 1989
3) EDP Portugal: sistema Tarifario
de Venda de Energia Electricai
Lis-boai 1990
4) Deutsche Gesellschaft fur Wind-
energie (DGW): DGW-Informationeni
Hannoveri 1988
Figure 7 5) VDEW: Vergutungsregelung fur
strom-Einspeisungen aus kleinen
Kraftwerken auf Basis rationeller
Energienutzungi VDEWi Frankfurt
a.M. i 1987
6) Deutscher Wetterdienst (DWD):
Testreferenzjahre (TRY) sowie Be-
richte des DWDi Nr. 147, 165i Of-
fenbachi 1983

762
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE S-12 APRIL 1991 LISBON, PORTUGAL

THE 4kW PV PILOT PLANT "MUTTLER"

J. Schadel
ascom eneray systems
Belpstrasse 23
CH-3000 BERN 14
Tel. 31 63 11 77 I Fax 31 63 42 36

ABSTRACT. This PV application describes how an old oassive microwave

repeater is replaced by a new active link.
To reach hardly accesible valleys in the alps i t is often necessary to
install the repeater stations on the top of a mountain. In our case it ' s
the "Muttler" with a heiqht of 3300 m above sea level. It is only
helicopter reachable if the weather conditions allow it. Due to the cold
climate a shelter with heat storage and a heatinq driven from a dual
output power tracker is used. A modular system architecture is necessary
to achieve the high reliability demanded for telecom applications.

1. SYSTEM DESIGN 1.5 Data acouisition

To monitor this pilot PV plant a special
The system consists of the PV array divided data channel on the RF link is installed.
in two rows and the shelter at the bottom of Every 25 sec. a new set of measurement
the repeater mast wi th the transmi t ters. values and system status is sent. The
batteries and the maximum power chargers transmitted data are pre-stored in a local
(see Figure 1). data aquisition system from where they are
collected every night via modem and a normal
1.1 Lightning protection telephone line by a PC. With this PC
The lightning protection is provided by graphics and lists of all values and for
the mast which is higher than the PV array, different time interval Is (1 day to 1 year)
by shie I ded capl es wi th an optima 1 1 ayou t can be printed. There is an 'on-line '
and by protecting devices at the input of function implemented, which allows supervi-
the shelter. sion of the actual status and values of the
system from the office desk.
1.2 Panel power The ' Muttler' PV plant will be installed
The power demand of the telecom in summer 1991 and first results will be
installation is rei. constant throughout the available by the end of the year.
year. So the panels are installed at a fix
tilt angle of 65 degrees which is an optimum
in December, the month with least available
sunlight. With the data given in I I I and the
site, height and temperature correction
factors the system is designed (in Dec.) not
only to feed the load under normal condi-
tions, but also to give a full charge to the
batteries, to prevent them from deep
discharge or long partially charged states.
The additional installed power is normally
used to heat the she 1 ter, where the
batteries are installed, to increase their
lifetime.

1.3 Maximum power trackers

The batteries are charged by 12 maximum
power trackers. They optimize the energy
flow in cold winter days and feature a
switchable output, one for charging, one for
heating. In this way the energy for the
heating is supplied only by the panels and
not from the batteries. This in fact is only
possible by the use of a power tracker
because of its capability to match a
resistive load to the panels!

1.4 Controller
The system is controled and supervised
by a mi cro-computer. I ts functions are
charge control with boost charge capability,
heating control, supervision and transmis-
sion of 37 measurements and diagnostics of
faulty elements. Figure 1: global situation

763
 1.6 Batteries 3. RELIABILITY ANALYSYS
Maintenance free lead-acid batteries
with modified vent for this height are In a further analysys the reliability of the
placed in 4 groups of six 2V cells. Each system was calculated. Due to the special
cell weights 75 kg and has a capacity of site of the system a high MTTR of 0.25 years
1000 Ah at + 5 C. This gives an autonomy of was assumed. The results showed that
more than 13 days. parallel ing redundant system components as
batteries, maximum power trackers and panels
1.7 Shelter are necessary to achieve the required high
The shelter with the dimensions 1.7 m x reliability.
2.8 m x 2.3 m comprises the batteries, the simplified block diagram:
transmitters, the power tracker and
controller rack (see Fig.2). Through a door
the shelter can be entered and room is left
for one person to work while the door is
closed. A natural ventilation is provided by
two diagonal set tubes. The shelter is
highly isolated by 25 cm thick PU panels on
all sides and the floor is covered with a 20
cm thick concrete block, which contains the
heating cables and represents together with
the batteries the heat capacitor.

HlHHH!HmHH!!1HHHlHHii!imllW!HHHlHHmm!m!lmmHH!1iWmHiiH!!!l!!mmmmnmm
Wi!!
H~m
wm
~-TP12~
jHH! - ----,
iiiWTrans- B B c B B tracker and panels batteries
mm mitter
; jii!i! 1-4 8 8 MP T 8 B failure rate
5'000 fit / 1'500 fit
Ili !il 8 8 8 8
0.25 years
MTTR
/ 0.5 years

The weather conditions were empirically

evaluated from dayly values over a 2 years
Figure 2: Shelter block diagram period:
number of
trackers
2. THERMAL SIMULATION batteries weight
weather conditions needed [Yo]
A computer simulation was used to analyse
the thermal behaviour of the shelter. The A: good 6 1 85
results showed that not only a good
isolation but also a heat storage and a B: 4 to 6 days cloudy 8 2 9
heating are required to get a flat thermal
response at a temperature of +5 C (mean C: 7 to 9 days cloudy 10 3 3
December ambient temp. is -11 C).
The input which was given to the shelter D: 10 to 12 days cloudy 10 4 3
was the real global irradiance and ambient
temperature from the neighboured mountain The availability of a redundant structure
Corvatsch in February 1986. This date shows was calculated according to the following
a temperature fall down to - 33 C ' formula:
The behaviour of the shelter with and
without a concrete heat storage and the
additional needed heat power to reach + 5 C v =
is shown in figure 3.

. p ~ n J / nf+1 n
5-------·----·--- - ---.....-----
( ,I) n J 1=0.1
l A_C(_n-_1_lA_',,_C(-:-n_-(_I-_1l_l_A_
= _(n_-_O_

MTTR i
~ -IS
;;;
,, -20
v : availability
....t· ..
A : failure rate
• )0 n : tot number of elements
-35 k : actual needed elements
-••»I----~--~,__--------~---_---l
»
)Q )Q ~ ~ ~ ~

da y s The batteries have a lifetime of approx. 10

years and a non-constant failure rate. For
(a)ambient temp. (b)shelter without heat the calculation a speculative value of 1'500
storage (clshelter with concrete (dlshelter fit was assumed which equals a MTTF of 75
with concrete and 235 W heat power years. The MTTR is assumed to be as high as
0.5 years because it s relatively difficult
Figure 3: thermal behaviour of the shelter to supervise the capacity of the batteries.

764
The results show a MTBF for the 5 CONTROLLER FUNCTIONS
panel/tracker part of 2'603 years,
batteries part of 638 years Via the serial bU5 the controller
and a total system MTBF of 500 years. communicates to the power trackers in the
power rack and to the analog measuring sets
and the digital ports in the peripheral rack
4 MAXIMUM POWER TRACKER (see Fig. 6). Via a terminal input 1 output
of commands, alarms, diagnostics, 5etting
The MPT is a 360 W switchmode dc-dc system parameters and display of all the
converter wi thou t gal van i c separation. Its measurements is provided.
features are a high efficiency down to a The controller keeps the system voltage on
solar radiation of 50 W / m2 (see Figure 4) a constant level depending on the
and a switchable output, one for charging temperature of the batteries, provides
the batteries and the second for heating the selectable boost charge times and
shelter. intervalls, switches the power trackers (one
by one) to the heating position if the
SPT VP 12-30 UIN;<l9 V uoun =12.3 V battery temperature and the state of charge

::[-i:::i::T:::f:::j::::j:_-j
allows it, monitors alarms and disturbances.
measures 37 system values like voltage,
current. temp. etc. and transmitts all the
measurements. the alarms and the system
states periodically to a non-volatile store
0.7 ·...... ··•· .. t ....···......·t·...... ···....i ..·· ..·..·..··T·......·....·"j'....·..........!................ ; in order to support the system analysis.
! ; : : :

:.: .::::::::.::::.1.:':::::::::::::::......·..:::-:]::.::::::::::··r:· ...........::[.:·:::.::::::.::; ..............·1 controller rack <--) TERMINAL

. i i ; i ; ; ;
0,4 ............... ~ ................ J......•......... .l. .... ···········+················~················i····.···········i

...............1,................ 1., ............... 1. ...............1. . . . . . . . L.............. t................ J

---) DATA
0,3
............... J................
I I I I

1. ...............1............... l................L............... ~ ................ 1

I

I serial bus
TRANSFER

::: ···············1················1··············1···············t················l················l··············1
1 : : : : : :
O.OL-----~----~------L-----~----~----~~--~
o 50 100 150 200 250 300 350
peripheral rack
OUTPUT POWER [W]

Figure 4: MPT DCIDC efficiency

~
TI
liD
The maximum power point is achieved by
controling the panel voltage to a certain
percentage of the open circuit voltage. The
tracking efficiency defined as input power
of the MPT related to the max. available
power from the PV generator. is shown in
(sensors) I
Fig.5. This measurement was done by a
specially developped maximum power point
monitor as described in 12/. All 12 MPT's
are paralleled at the output and two work power rack
together with the same panel group. In the
'Muttler' system the MPT s are equipped with
an interface card, which links them via a
serial bus to the system controller. The
temp. compensated charge voltage (in steps
of 30 mV) as well as the setting of the
output relay and the supervision of the 'ok'
signal is managed by the controller.

SPT VP 12-30

1.00 ~ ......~';O;;';;';;.;,j[ .................. .

i
0.911 ·t··············~
0,96 ···········t·····--········i················j················~···············+···············i················i

0.9<1 . . . ·. . . ·t. . . . ·. . . j........·......·;......·....·. ·f·..·..·....·..·t· ............1.............. ·!

L !
01921 --·········--t-----·····------t
i !
--_ .. ". --_ .. _-,
; .------.-~
!.. -- .. -.. -.-.--+-
1 ;
-----·i

~r~:~t~:I:~I:::I~~I::!:~1
power trackers with interface

Figure 6: supervision block diagram

6 REFERENCES
0,80 1L.____--'I______J..I____--'I______J..I____~I______I~__--'I
o 50 100 150 200 250 300 550
III MeteoNorm.Daten fuer den Solarplaner
MAXIMUM PANEL POWER [W]
1985.Bundesamt fuer Energiewirtschaft
121 A.S.Kislovski."Maximum Power Point
Monitor".IEEE Int. Telecommunications
Figure 5: MPT tracking efficiency Energy Conference 1990Record. pp283-288

765
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

Alfred Nentwich * Michael Schneeberger + Andreas Szeless * Heinrich Wtlk +

* Verbundgesellschaft, Austrian Federal Electricity Board
Am Hof 6A
A-1010 Wien, Austria
+ Ober6sterreichische Kraft~erke AG, Electric Utility of Upper Austria
BOhmerwaldstraBe 3
A·402l Linz, Austria

~!SIR~~I' Austrian electric utilities are engaged in noumerous photoYoltaic projects. The
maTn purpose of all these projects is to demonstrate the feasibility of photovolUic
electricity generation under realistic conditions with respect to insolation,
meteorological conditions, high elevation and given local infrastructure. Great impor-
tance is given to the economic assessment of photovoltaic electricity generation with re·
gard to its future competitiveness. This paper describes one of the major photovolUlc
projects of Austrian' electric utilities, namely the 30 kW grid-coupled photovoltaic plant
in the Alps of Austria, and reports on experiences gained.

2.· lQ_!~_~Y_~Q~~R_~h~~I_hQ~~R

In the past ten years Austrian elec- The Austrian Federal Electricity
tric uti I ities have been engaged in nu- Board and the Electric Uti 1ity of Upper
merous photovoltaic (PV) projects /1,2,3/, Austria have erected the first grid-
involving construction, testing and ex- coupled PV solar power plant with rated
ploitation of PV solar energy power sta- electrical power of 30 kW in the Alps of
tions. Austria. The purpose of this pilot project
Each of these PV projects is diffe- is to demonstrate the deSign, construc-
rent in design, size and equipment and tion, testing and long-term exploitation
each serves for a specific purpose (test- with the aim of gaining experience in the
ing, stand-alone power supply and grid- long-term operation of grid-coupled PV
coupled systems). As these PV projects are plants. Also the assessment of the economy
located allover the country different of the PV power plant is /I major aim. FI-
conditions exist with respect to insola- nally, it is important to acquire experi·
tion, meteorology, elevation above sea ence and know-how regarding licencing
level and local infrastructure. Particular procedures of this new type of power
emphasis has been given to alpine sites. plants, especially in the context with
Thorough account of capital and operatio· environmental impacts. The site of the
nal costs allows realistic evaluation of alpine PV plant is the mountain Loser
the present economy of PV electricity ge· located in the Province of Styria at an
nerating facilities. elevation of 1550 m above sea level. Fig.
Already a wide range of experience 1 shows the plant and Fig. 2 its principle
and know-how has been acquired along the technical design.
pat h 0 f the t e c hn i c a I and co mme rei ali n- Some of the technical data are: rated
troduction of PV solar power plants. electrical power! 30 kW (at 1000 w/m l ,

Fig. 1: 30 kW grid-coupled PV plant loser

766
 POWER SWITCH
eo AlDC 500Y 1"'.'17
=:f'
f/i
~1fo
~
SOLAR GENERATOR 1 ~ t-;X GRID
3 380/220 Y

..
ARCO M 55 I

POWER SWITCH ..J"\.,.-" L1

eo AlDC 500Y 1"''''1/
-~
~, .... V L2 ...
SOLAR GENERATOR 2
SIEMENS SM 50
if'
~ .-;X
'"
.... v L3
..., ..
POWER S WITCH
80 Aloe !lOOY
.-,
111''''11
~14
SOLAR GENEBATOR 3 1~1 I-;'X
KYOCI!RA LA aS1 J 48 INVERTER
3x10 kW
Fig. 2: Principle technical design

AM=1.5, 25°C); rated volta3e 265 V (at 3. Q~~~~I!Q~~h_~g~~hl~_~~Q_g~E~~!~~~~

MPP); operational mode: gri -coupled; so-
lar generators: 3 groups, 10 kW each, no Fig. 3 shows the monthly overall
sun tracking; DC-AC inverter: three-phase, efficiencies of the three PV systems (10
nonlinear distortion factor < 3 %, coupled kW each) in the time period from Aug. 1989
to 30 kV grid; control and monitoring to Jan. 1991. The efficiencies are in the
system: completely automatic; data range of 8 to 10 % and agree with predlc·
transmission: automatic via telephone link tions by computer simulations for this
to central computer. site.
In accordance with the project sche- Monthly electrical energy yields to
dule plant construction started in August the grid are shown in Fig. 4. Also
and the assembl·y of the solar generator in indicated are predicted electrical energy
September 1988. Plant start-up, grid- yields and down-time losses.
coupling and trial operation took place In Fig. 5 cumulative electrical
early in 1989, whereas routine plant energy production losses are depicted from
operation started in June 1989. Aug. 1989 to Jan. 1991, representing ' "
Operational results of the past two years of the nominal energy yield. An analysis
are discussed next. of the causes of system outages revea I s

(-AI)
11
10
9
I-
- ~~
b::--..
~ 1:1.
~~
',,-
//

V
1-- V r--. f'-...
r-....., I;l..
K V
~
I-- r"'-. J:t
~ b~ 1//-
8
~~
7
6
5
4
3
2
1 f - - ~ -ARCO 8,S" -e- SIEMENS 8.8" ~ KYOCERA 8.e,. f f - - -

o
Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan
I 89 I 90 91
Fig. 3: System efficiencies (monthly averages)

767
 IMWh J

3 +---+----"u---+-+-----!::;

AugSepOctNovOec Jan Feb Mar AprMayJun Jul AugSepOctNovDec Jan

I 89 I 90 91 1
! L;~ AC Output ~ Downtime lo.... - Ou tput / Stand.rd Year

~ig. 4: Electrical energy delivered to grid (monthly averages)

Other 6%

Inver ter 76% Human Error 2%

Maintenance 2%
DC Fuse 3%

AC Fuse 9%

Fig. 5: Causes for plant outages and associated cumulative

generating losses

that inverter failures have been the major * Alpine sites for PV power stations at
causes for plant down-times; high elevations above sea level have
Further experience and know-how gained been demonstrated to be suitable for the
so far are summarized in a qualitative operation of PV plants.
manner.
* Significantly higher energetic outputs
* No standardized PV solar power plant than originally anticipated have been
designs exist. Consequently, there are obtained especially during the winter
neither standardized PV solar plants seasons (little fog, snow reflection).
ava i lab Ie on the market, nor standar-
dized individual systems and components. * Construction times for alpine PV plants
Thus, each PV p I ant is to be treated have to be short (3 to 4 months).
individually with its own specific Therefore, thorough and detailed
design and erection problems. This is planning is neccessary and vital to keep
especially valid for alpine sites. set time schedules. In advance off-site

768
 preassembly turns out to reduce costs tual operational status of the individual
and construction time. PV projects and to perform assessments and
evaluations with the measured data. Fig. 6
* As lightning represents an important shows the availability of the plant's mo-
failure cause of alpine PV plants nitoring system as it increased from about
effective precautions are to be taken. 70 % at plant stattup to about 98 % pre-
sently.
* Soon after plant start-up excellent This country wide central solar data
operational reliability and plant acquisition and evaluation system is
availability can be achieved. standardized with respect hard- and
soft-ware, thus representing a major step
* Entire automatic plant operation (pro- towards reducing maintenance and repair
vided automatic flant restart features work and decreaSing overall costs.
after operationa interruptions exist)
minimizes maintenance and costs.
5. ~~~~~~~~~Q~~~!_~~hf~Qh~~~
* Maximum availability of the plant's mo-
nitoring system is desireable in order The EUREKA-prOject "ALPSOLAR" of the
to assure complete and efficient oper- European Research and Technology Coopera-
ational evaluation of the system and tion is a common project of ARGE Alpen-
know-how feedback. Photovoltaik/Austria, Lech-Elektrizitlt5-
werke AG/Germany and lmatran Volma
* Operators of PV solar power stations OY/Finland. The project aims include
should regularly exchange their experi-
ences and report about their operational * long term and thorough tests of already
results in order to support the ad- constructed PV plants and their compo-
vancement of this new technology. nents at different locations (Austria,
Germany, Finland)
4. * experimantal testing of all aspects of
f~~I~~fla~~~~~~i8~-~fq~1~1!!Q~-~~Q grid connection
A central solar data acquisition and * mutual exchange of data and experience
evaluation system for the utilities' PV gained in the course of test programmes
solar power projects has been installed.
The centra 1 computer located at Gmunden * joint thorough analysis lind assessment
in Upper Austria, is continuously fed by of acquired data and know-how
measured data from the respective PV solar
power projects. Via terminals at the uti- * development of adapted components and
lities' headquarters in Vienna, Linz systems for application at alpine and
Klagenfurt, Salzburg etc., it is fossible other sites, also taking into account
for the user to communicate on-l ne with transportation requirements (e.g. by
the centra 1 computer, to look at the ac- helicopter), handling and assembly

90 +--r~--+-~-+~~+-~

80 +-~~-+-+~~~~~~

10 +--t--+---;~.2~~~~~:

60
50
40
30

10
20 . . . . . . . .
o
M J J A SON 0 J F M A M J J A 8 0 N 0 J
I 89 I 90 91

Fig. 6: Availability of plant monitoring system

769
* establishing specifications for PV (probably about 200 roofs). Federal
plants, systems and components qualified funding per plant wi 11 be of the same
for alpine and other sites order of magnitude as in Germany (50% of
the investment costs), Whether additional
* initiating in this context appropriate funding by the provincial governments will
regulations and standards. be granted is not decided yet.
Regulations concerning the intercon-
The governing aim of "ALPSOLAR" is to nection of PV generators to the grid are
achieve cost reductions of PV power plants presently being worked out under the
along the path of commercialization of auspices of the Austrian Institute of
this renewable energy source. Standardization (Osterreichisches Nor-
mungsinstitut) and in cooperation with the
domestic electric utilities. The Austrian
6. Q!!I'=-QQ~ utilities are aware of their responibi1ity
in the engagement to develop, test and
Two years of successfu 1 operat iona 1 promote PV. Market penetration of this
experience with the first grid-coupled PV environmentally benign new technology
plant in the Alps of Austria have demon- will, however, follow economic realities.
strated that PV is a future option for
electricity generation. At presaent cost
is the limiting factor for larger ~,cale B.sfJ.!!s!!£s~
use. Both PV module and system costs 11/ A. Nentwich, M. Schneeberger, A.
have to significantly come down yet. Szeless, H. Wilk, Photovoltaic Acti-
Recently, a number of small privately vities of the Austrian Electric Uti-
operated PV plants (about 1 to 2 kW each) lities - Projects and Experiences, po-
have been connected to the publ ic grid. ster presentation, 9-th European
With this development Austria follows si- Photovoltaic Solar Energy Conference,
milar developments in some of our neigh- Freiburg, Sept. 25 - 29, 1989
bour countries I ike Switzerland and Ger-
many. /2/ A. Nentwich, A. Szeless, Photovoltaic
Presently, the Federal Ministry for Solar Power Plant in the Austrian
Research is preparing a programme of Alps, World Renewable Energy Congress,
funding small customers operated PV plants Reading, Sept. 23 - 28, 1990
(in connection with the operation of
electric vehicles) up to 5 kW rated /31 A. Szeless, A. Stumpf, H. Wilk, 30 kif
electrical power following the examples of netzgekoppeltes Photovoltaikkraftwerk
the "1000 roofs programme" for instance in - Erfahrungen bei Errichtung und Be-
Switzerland and Germany. However, trieb, 7. Internationales Sonnenforum,
Austria's programme will be much smaller Frankfurt, 9. - 12. Okt. 1990

770
 8-12 APRIL 1991 LISBON, PORTUGAL
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE

BATTERY RESEARCH IN THE TERSCHELLING PV SYSTEM

T.C.J. van der Weiden, J.H. Boumans, K. Blok

Ecofys Research and Consultancy
Biltstraat 110
NL - 3572 BJ Utrecht
The Netherlands

ABSTRACT. With support from the CEC-DGXII and NOVEM, Ecofys has been conducting research into the
Terschelling PV/Wind/battery system (43 kWp solar generator, 75 kW wind turbine, 180 kWh battery storage)
since 1987.
The battery is one of the most vital parts for autonomous operation of the system. The possibility to use pilot
cells in the control strategy deserves to be studied in more detail. If several banks are available, it seems to
be possible to detect malfunctioning cells by comparing currents. If the battery consists of a single bank,
voltage comparisons of parts of the bank may provide indications regarding the po~sibl.e mal!unctioning ?f cells.
The battery has aged faster than anticipated. An important secondary effect of ageIng IS the Increase of Internal
resistance and the unfavorable effect this has on the maximum power from the combination of inverter and
battery.

INTRODUCTION

Research at system level is important for the introduction of

PV. The battery is one of the most vital parts for the
autonomous operation of a PV system. Optimal use of the
battery is vital, both from an economic point of view as with
regard to the operational reliability of the electricity supply of
a PV system. This paper will discuss a number of aspects
that playa role in the operation and maintenance of a battery
in a PV system, on the basis of experience acquired with the

J{k1
Terschelling system.

SYSTEM DESCRIPTION ICc

The system consists of the following components (see fig. 1):

_ PV generator, 43 kWp (2748 AEG PQ/10/20/01 poly-
[3--ct-J t t
batt , batt 2 consumers
crystalline Si modules, distributed over 30 subarrays,
total cell surface 550 m2) Figure 1 PV/Wind/battery system at Terscheliling
Wind turbine, 75 kW (Lagerwey LW 15/75, height
30 m, rotor diameter 16.2 m) DECREASE OF CAPACITY
Battery, 180 kWh (2 banks of 180 Varta Bloc lead acid
cells with 250 Ahlcell) Since 1983, the batteries of the Terschelling system have
Converters completed approximately 1100 cycles, with a 50% depth of
(a) DC/AC: 60 kVA inverter system with discharge. On the basis of the number of completed cycles,
synchronous machine a residual capacity of approximately 90% can be expected.
(b) AC/DC: 35 kVA rectifier Actual capacity has been determined by recording a
(c) DC/DC: 29 local MPPTs, 1600 W apiece discharge curve (see figure 2). Prior to discharging, the
Sensors, control and data monitoring equipment battery was subjected to protra~ted charging until a state of
charge of over 95% had been achieved. Subsequently, the
System control is intended to have the system provide as battery was discharged with a current of approximately 12 A.
much of the school's energy demand as possible through On the basis of the discharge curve obtained, precise
immediate use of produced energy whenever possible. determination of the capacity is impossible. Extrapolation of
Surplus energy is stored in the battery as much as possible. the discharge curve indicates that the capacity amounts to no
When the battery is full, the surplus is supplied to the public more than 50 to 75% of nominal capacity.
grid (in case of grid-connected operation). The comparatively large loss of capacity can be explained by
At times when energy production does not suffice to meet the the fact that, on the one hand, the battery has been
demand, the deficiency is obtained as much as possible from overcharged for long periods of time at 2.35 V/cell, while it
the battery. When the battery is empty, the deficiency is was kept at 2.6 VIce II for a total of 9 days. On the other
obtained from the public grid. hand, the battery has been partly discharged for long periods
Consequently, control is also important with respect to the of time. Visual inspection confirmed the conjecture that both
use of the battery and keeping its ageing process as limited corrosion due to overcharging and sulphation as a result of
as possible. protracted partial discharge have played an important role in
the ageing process [1]. Also, the higher degree of sulphation
on the base of the negative plate was a conspicuous feature,
which was probably caused by stratification.

771
 SERVICEABILIlY OF PilOT CEllS
MeaSU'ed voltage o f the battery at
a constant diSCharge evrent (ca. '2 A)
Prevention of battery overcharging or undercharging is one of
370 Volt. M
the aspects where control strategy is significant. For batteries
consisting of more than one cell, the limits employed are
~5r-~~-'-'~"~'~"~"'~~~--~--------~ often based on marginal values per cell, without taking
differences between the cells into consideration.

~r-----------------------~~--~ Distributkln voltage batt. cells bank B

corr. 360 V-nom 13-2-90

355r-----------------------------~ 6~~----------------------------~

,0 '00 .000 4

Figure 2 Discharge curve at 12 A.

INCREASE OF INTERNAL RESISTANCE o

'.995 2.005 2.0'5 2.025 2.035 2.045
Another effect of battery ageing is the increase of internal VoU:aoe b11l1ery cell M

resistance. Figure 3 contains a representation of internal

resistance, established at various discharge current values,
by calculating the proportion between current and voltage Figure 4 Variation of cell voltages for 20 cells of bank B.
variations for successive data points (tN/~I) (sample time one
minute). The larger deviations in established values with The differences between the highest voltage cells and the
smaller currents may be explained by assuming hysteresis lowest voltage cells imply that Individual cells may be either
(polarization effects) as the cause of this phenomenon. overcharged or undercharged. Repeated measuring (see
figure 4) has indicated that, for both banks, highest voltage
Apparent internal resistance
cells and lowest voltage cells maintained their respective
at diSCharging (377 V. 24 March 1990)
positions and that this accurately corresponded with the
3~~n~.WM~lwr~~t~.~/~~--------------~ respective degrees of acidity. This indicates that the use of
pilot cells within a control strategy may be possible. However,
this must still be confirmed by measurements at the end of
2r-----------------------------~ (dis)charging, as those are the pOints where the largest

.
differences can be expected, and also because differences
are most relevant there. Furthermore, repeated measuring
. t++:.. ~
.. (e.g. every six months) will have to reveal whether selected
..... ~ cells remain serviceable in the course of time or whether
•
..
0 other cells will eventually have to be used. It must also be
examined to what extent the use of pilot cells to prevent
overcharging is advantageous for the battery as a whole,
-,
-50 -.0 -30 -20 -'0 0
since this may result in a large part of the battery remaining
partially discharged.

Figure 3 Apparent Internal resistance during discharge DETECTION OF MALFUNCTIONING CELLS

(voltage at rest 377 V).
Even in case of an ideal control strategy, it will be impossible
On the basis of the data obtained, internal resistance of the to completely prevent cells from malfunctioning. Acute
battery now appears to total approximately 0.45 0 during damage to a single cell may have considerable adverse
discharging. A comparison with the Internal resistance as effects on battery performance. Accordingly, timely detection
derived from factory discharge curves for new battery cells of malfunctioning cells is important. Figure 5 represents the
shows this to be a four-fold increase of internal resistance. difference In power production for both banks in a situation
In order to be able to make sufficient use of battery capacity where one of the banks contained a malfunctioning cell. It
in spite of high internal resistance, it is necessary to employ Iilecomes apparent that malfunctioning cells can be detected
current compensation in battery control. In case of PlC by comparing the power production of both battery banks. If
(Programmable logic Controller) application, as in the only one bank is available, detection of malfunctioning cells
Terschelling system, It Is in fact possible to monitor internal can, in principle, also be achieved by comparing the power
resistance continuously and incorporate it into the control. production of a number of groups within the bank.
As a result of the increased internal resistance of the battery, It has been studied to what extent voltage measurements of
the power that the battery is capable of supplying to the 10-cell subgroups within a bank comprising 180 cells could
inverter will often be restricted to 30 kW, owing to the be used to detect a malfunctioning cell. The accuracy of
decrease in terminal voltage at higher discharge currents monitoring proved inadequate for detection of malfunctioning
(nominal inverter capacity totals 60 kVA). As a consequence, cells in this way. This may be improved by expanding the
the increase of internal resistance in the course of time number of subgroups to be measured to 4 groups of 45 cells
greatly affects long term system performance and must be and/or enhancing sensor stability.
involved in dimensioning.

772
 Charge CU"ve Terschelhnc;, battery
selected pened ,n noverrber 1989
450 Vol' .... M 01, ...,

'!I

'0

37oL-----~--~~--~----~----~0
'000 ,oe;o "20 , ,80 '2"0 , 300

o co.........
~B

Figure 5 Chargecurves for both banks, one of which

contains a defective cell.

FUTURE RESEARCH

It is our intention to subject the existing battery to a closer

study so as to acquire better insight into battery condition and
the limits of its possibilities on the one hand, and into the
ageing mechanisms on the other hand. In addition, once the
study of the existing battery has been completed, we intend
to start the dimensioning and Installation of a new battery, as
well as the implementation of an optimized control strategy
aiming at the demonstration of a battery system capable of
meeting requirements in the long term.
In addition, the system will be expanded with a diesel
generator (50 kW). It is important to gain experience with,
and do research on an autonomous PV/diesel system, since
application of such PWdiesel systems can be expected to
become profitable in Southern Europe and Third World
Countries in a few years' time. For the study of the
autonomous use of the PV/diesel system, system
configuration will be simplified as much as possible, also in
view of enabling a reliable control strategy. Control
engineering, too, requires battery dimensioning to meet
certain preconditions.

REFERENCE

[1] Weiden, T.C.J. van der, et ai, Terschelling PV/Wind

Repair and Optimization Project, Final Report, Ecofys,
Utrecht, december 1990.

773
 SESSION 04

Compound Semiconductor Solar Cells

 THIN FILM COMPOUND SEMICONDUCTOR SOLAR CELLS:
AN OPTION FOR LARGE SCALE APPLICATIONS?

H.W. Schock
Universitat Stuttgart
Institut fiir Physikalische Elektronik
Pfaffenwaldring 47, D 7000 Stuttgart 80, Germany,

ABSTRACT. Recent achievements in the development of polycrystalline thin film solar cells are reviewed.
Properties of compound semiconductors are analyzed with respect to solar cell applications and compared with the
other commonly used photovoltaic materials. Cu(In,Ga)S~ and CdTe, presently the main candidates, are discussed
in detail. Basic problems such as the influence of defects and impurities on electronic properties of the
semiconductors and limitations of performance due to the polycrystalline structure of the film are investigated.
Advantageous properties of Chalcogenide semiconductors, e.g. low grain boundary and surface recombination, and
the favorable interaction of oxygen with the surface of these compounds are pointed out. Wide bandgap
semiconductors for the use in heterojunctions are selected. Key problems of heterojunction solar cells such as
energy band alignments, lattice match, chemical stability of the interface, and recombination of minority carriers
in the bulk and at the interface are considered. The different technological approaches for thin film deposition of
compound semiconductors for solar cells are reviewed in respect to scalability and thin film quality and hence
expected solar cell performance.

Extrapolating the present development, polycrystalline thin

film cells would give option for the future. However, there are
1. INTRODUCTION still many problems to be solved which do not appear in the
simplified picture displayed by table I. These problems are
A steady reduction of production costs for photovoltaic related to the specific elements and compounds which are not
modules has been achieved in the last twenty years. However, commonly used in industrial large scale processes.
the costs do not yet meet the expectations of the mid 70's. In Another severe problem arises from missing inputs from
order to come to a r~ breakthrough of photovoltaics in the other applications of these materials. It is well known that
energy market new low cost options have to be developed. crystalline silicon profits a lot from the developments in
One of these options may be compound semiconductor thin microelectronics. Amorphous silicon has found a lot of
film solar cells. However, new materials always have to be attention for large area electronics, e.g. thin film transistors
compared with silicon (either crystalline or amorphous). Since for display applications. III-V-compounds are the leading
these material are far ahead in the materials and device materials in optoelectronics. Taking the experience from the
technology and understanding, only substantial advantages with existing technologies (high purity crystal growth, optimized
respect to manufacturing cost and performance can bring novel doping, surface passivation, epitaxial growth techniques) it was
materials to a position where large scale applications can be more or less readily possible to make efficient solar cells.
envisaged. A list of different features of photovoltaic
technologies is given in table I:
2. FUNDAMENTAL ASPECTS OF THIN FILM SOLAR
CELLS BASED ON POLYCRYSTALLINE COMPOUND
SEMICONDUCTORS
Table I: Features of photovoltaic technologies.
Present device optimization is mainly done using
polycrystalline thin films. Only few attempts for the realization
CeU type Efficiency Stability ManuCactur. of single crystal based devices have been reported /11.
cost With Cu(In,Ga)S~ the best devices have been obtained
Cryst. Si very high excellent very high using polycrystalline thin films. There might be two reasons
(monocryst. ) 24.2 % for this:
- High quality bulk crystals are not readily available,
Cryst. Si high excellent medium - Grain boundaries and surfaces play an important role for
(multicryst.) 20% the control of conductivity type and defect chemistry of
a-Si (single fair low low the material.
junction) 12%
2.1 Polycrystalline Structure
a-Si (multi- fair fair low? As mentioned above, grain boundaries may not only have
junction) 13.7% a negative influence on the carrier collection in polycrystalline
absorber layers but can also help to control the material
III-V- very high excellent extremely
properties. Especially in a ternary material such as CulnS~
compounds 25.7% high the interaction of oxygen with specific species on the surface
Polycryst. medium good low? may influence also the bulk carrier concentration if other
compound 14.1% species (i.e. Cu in CuInS~) are enough mobile. Passivation of
thin film grain surfaces allows to fabricate efficient devices even with
grain diameters as small as 0.2 ~m. The cross-section of a

777
CuInS~ film which yields devices of more than 10%
efficiency is shown in Fig. 1. The fibrous type grains have a
~_----EC
typical length of about 1 I'm and a width of 0.2 I'm. n

conduction -.!!""'!!!!!!!!!!!....-::~t;~ - - - Er
band ------
EV
Ell
Ell space ~
charge recomb.
interface

window absorber

valence
band

Fig. 2: Band diagram of a heterojunction with a multilayer

window

A most suitable window material has turned out to be

ZnO 13/. This semiconductor has an energy gap of 3.2 eV
and can be doped in order to achieve conductivities close to
104 S/cm. Careful optimization is needed in order to avoid
losses due to infrared absorption and series resistance
(Especially in the case of red sensitive CuInS~ 14/,151.
This approach lead to the major improvements in device
performance with respect to current density and open circuit
voltage.

Fig. 1: TEM picture of the cross-section of a CuInS~ thin

film. 'O.OF~~;:::;;:;;::::=~---l
From this picture it becomes evident that efficient minority
carrier transport is only possible if the grain surfaces exhibit 30.0
very low recombination velocity. Besides a low surface state ~
density this can be obtained by an accumulation layer, i.e. <-
t
.§.
increased doping concentration at the grain surface in order to
form a minority carrier reflector. The resulting potential 200

troughs within the grains confine the transport of minority

charge carriers towards the junction. However, reduction of
the effective barrier height and hence losses in open circuit
) 10.0
ELH.illumination lOOmW1cm1
j.V V«(mV) i.. (mA/em) FF (%) ~ (%)
(I) 400 32.4 68 8.8
voltage have to be taken into account 12/. (2) 424 38.5 70 11.5
(3) 450 38 68 11.7

2.2 Heterojunctions 0 . 0 1--->---'+10"--0--+---=2O'::0---+-----,30;!:0;----+---:!::~-4

Since the materials considered for thin film solar cells are vollage (mV)
preferably direct gap semiconductors, homojunctions are not
the suitable structures for efficient devices, in particular when 100
also the problem of the formation of very shallow junction~ is
considered. In most cases the materials do not allow to realize
(2)
steep doping gradients for p-n homojunctions, therefore ,;
80

heterojunctions are the most suitable structures. The material

presently prefered for wide gap windows in combination with ~ (I)

~
60
CdTe or CulnS~ absorber layers is CdS. It provides a suitable

"."
interface and yields efficient devices. However, it has E

substantial disadvantages. An intrinsic loss occurs due to the

"
'0
a"
bandgap of 2.4 eV which cuts off the blue res~nse of the
cell. This results in a loss of about 7 mAlcm of current
density related to the AM 1.5 (100 mW/cm 2) spectrum. To 20
(I) (Cd,Zn)S/C"lnS,'!
some extent these losses can be reduced by alloying with ZnS (2) ZnO/CdS/l ',<inSc,
thereby increasing the energy gap by mainly increasing the
O~.·~O--~--~--~-~--~~--,~
energy of the conduction band edge. Care has to be taken in
order not to affect the device performance by misalignment of WilVE!Il'lIgLII (nlll)

the conduction band. The limited conductivity especially at

Zn-contents exceeding 20 % causes further restrictions for the Fig. 3: Comparison of
application as a highly conductive window. it) j-V characteristics and
A very efficient solution of this problem is the use of a b) spectral quantum efficiencies of CuInS~ based
multilayer window as schematically shown in the band diagram devices with (Zn,Cd)S standard window and ZnO/CdS
in Fig. 2. multilayer window.

778
 Table II:Present status of CdTe laboratory scale device performances.

Cell type Process for Vue jsc FF Eff. Area Ref.lLab.

absorber [mV] [mA/cm2] [%] [cm2]
deposition
lTO/CdS/CdTe ED 720 27.9 0.65 13.1 0.02 Univ.
Queensland
191
Sn02/CdS/CdTe SPL 783 25 0.67 12.3 0.31 Photon
Energy
1101
Sn0 2/CdS/CdTe ED 800 24.3 0.67 13.0 0.02 BP Solar
/111
CdS/CdTe SP 797 21.1 0.67 11.3 1.02 Matsushita
1121
(Zn,Cd)S/CdTe SP 870 23.6 0.61 12.5 0.3 Korea Adv.
Inst. /131
Sn0 2/CdTe CSVT 663 28.1 0.56 10.5 4 ARCO
Solar 1141
Sn0 2/CdS/CdTe/ZnTe ED 767 22.4 0.7 11.2 1.07 AMETEK
/151
In203fCdTe X-tal 892 21.6 0.75 14.4 0.02 Shinshu
Univ. 1161

Table III: Present status of Cu(In,Ga)Sez based heterojunctions

Cell type Process for Vue jsc FF Err. Area Ref./Lab.

absorber [mV] [mA [%] [cm2]
deposition Icm2]
ZnOfCdSfCuln(Ga)Sez 508 41.0 0.68 14.1 3.5 ARCO
Solar fI8f
ZnO/CdS/CuInSez SVR 48-3 35.6 0.67 12.4 1 ISET /191
ZnO/CdS/CuInSez PVD 452 40.9 0.69 12.4 0.315 IPE Univ.Stgt
1M Stockholm 15/
CdS/CuInSez PVD 446 38.9 0.65 11.3 0.93 SERI 1301
ZnOfZnCdSf PVD 555 35.3 0.66 12.9 0.96 Boeing 1201
CulnO.73Gao.27Sez
(Zn,Cd)S/Culno.6Gao.4Sez PVD 645 28.0 0.68 12.3 0.38 IPE Univ.
Stuttgart f211
ZnOIZnCdS/CuGaSez PVD 756 13.7 0.60 6.2 0.4 IPE Univ.
Stuttgart f22f

SVR solid vapor reaction (selenization), PVD physical vapour deposition (coevaporation)

779
 This is illustrated by figure 3a,b where the j-V The glass substrate for thin film deposition serves at the same
characteristics of devices with different window materials are time as a transparent cover for the module (this is also a
displayed. The relative improvement in device performance is general feature of amorphous Si based cells). However,
about 30% if the standard (Zn,Cd)S window is replaced by a hermetic sealing is of great importance for protecting the thin
ZnO/CdS multilayer window. A significantly higher blue films against environmental influences. Therefore, a second
response and an increased open circuit voltage is obtained. glass sheet is recommended as backside cover.
I-V curve analyses indicate a change from interface to space Table II lists recently achieved performances of laboratory
charge recombination. However, the interface between the scale devices. In most cases CdS is used as window material.
absorber and the CdS thin film depends very much on the way The relatively small bandgap limits the blue response of the
of CdS deposition 151. devices. Therefore, the thickness of this layer has to be
The main challenge for the improvement of devices is to reduced as far as possible in order to improve photocurrent
increase the open circuit voltage. In comparison to the best density. For comparison the data of a single crystal based
crystalline Si and GaAs based devices, where open circuit device is included in the last row of this table. Whereas short
voltages approximately 380 mV below the energy gap E/e circuit current density is quite high in the polycrystalline
have been obtained, in polycrystalline devices the open circuit devices, in particular when thin CdS windows are used, fill
voltage is lower than E/e by 500-700 mV (see tables below). factor and open circuit voltage are small with respect to the
wide bandgap (1.45 eV) of the absorber material.
3. PRESENT STATUS OF MATERIAL AND DEVICE Attempts to increase the bandgap by using the ternary
DEVELOPMENTS alloy (Zn,Cd)Te in order to realize tandem structures where
not yet successful because photovoltaic properties of the
A lot of compound semiconductor materials could be material are affected by alloying.
considered for thin film solar cells and it is important to
investigate new materials and to explore their potential for 3.2 Cu(In,Ga)Sez based solar cells
photovoltaic applications 16/. However, the most mature The work on Cu(In,Ga)Sez has been mainly pioneered by
developments are based on CdTe and CulnSez. Besides basic Mickelsen et al. /17/. The first efficient devices were
investigations at various university laboratories, several realized by coevaporation of Cu, In, and Se on Mo-coated
companies are involved in the development of large area substrates and applying a CdS or (Zn,Cd)S window in order to
modules. Rapid progress has been made in recent years. form a CdS/CuInSez-heterojunctions. ARCO Solar (now
Efficiencies of small devices obtained in several laboratories Siemens Solar Inciustries) were very successful in developing
are well above 10% and a short term goal of over 15% an alternative (proprietary) process based on the selenization of
appears to be realistic. metal films. A major step forward in efficiency was the
introduction of a Zno-buffer multilayer window (see
3.1 CdTe based devices Sect. 2.2). The typical device structure is schematically shown
Considering physical and electronical properties CdTe is a in Fig 5.
nearly perfect photovoltaic material. Research on CdTe based
solar cells started more than twenty years ago, efficient
devices are reported from the beginning of the 80's III 171. I hll
The first experiments were carried out using close space
vapour transport (CSVn or close space sublimation window (ZnO)
(CSS) 18/. Now an increasing number of groups is working buffer layer (CdS)
on CdTe based devices using various "scalable" deposition absorber film
technologies such as electrodeposition, spray pyrolysis and (Cu(Ga.ln)Se2
screen printing. These techniques do not require high vacuum metaJ contact
technology so that there might be advantages in low
investments for equipment. However, all films or devices need glass substra t .,
a subsequent heat treatment at temperatures of around 500 °c
in order to achieve optimum performance. Due to the Fig. 5: Typical structure of a Chalcopyrite thin film based
congruent evaporation of CdTe the composition always solar cell
stabilises at the composition according to the minimum partial
pressures of the elements over the compound. The same The quaternary alloy Cu(In,Ga)Sez gives a lot of flexibility in
accounts for CdS which is used as n-type window in most choosing the band gap of the absorber and to optimize
cases. photocurrent transport and heterojunction properties. A
A further advantage of this material is that superstrate selection of presently achieved performances of
structures can be easily realized and therefore are the generally Cu(In,Ga)Sez based laboratory scale devices is given in
used. This structure is displayed in Fig. 4. table III.
A number of research groups achieved efficiencies above
ohmic conlacl 11 %, mainly limited by the open circuit voltage. Further
improvements are only possible by in depth investigation of
CdTe
the chemistry and electronic properties of the heterojunction
CdS interface and the influence of defects in the bulk and on the
surface of the grains of the polycrystalline films. Recent
TCO investigations showed that segregation of secondary
phases considerably influences the properties of the
glass subslral films 1231 1241 1251. Therefore, deposition methods
t hv ••
have to be optimized in view of avoiding secondary phases.
The bilayer approach used for evaporated films 1261 is
evidently needed for avoiding secondary phases.
Fig. 4: Schematic view of a CdTe based solar cell.

780
4. MANUFACTURING ASPECTS Table V: Manufacturing processes for compound semi-
conductor thin fIlms.
There are significant differences between CdTe and
CuInS~ with respect to manufacturing. These differences are
due to the different properties of the materials. No other process remarks
binary phases are known in the Cd-Te system. Furthermore,
CdTe evaporates congruently so that in annealing processes
evaporation + efficient Cu(In,Ga)S~ devices
always single phase material and hence homogeneous material
physical vapour + flexibility for changing
deposition composition
is obtained. This situation changes in ternary alloys like ? large area
(Cd,Zn)Te. The phase diagram of the Chalcopyrite compounds
is much more complex and there is a strong tendency for the selenization + efficient CuInSez devices
formation of additional binary phases in particular CuxSe in "two stage - fIlm properties depend on
the case of Cu excess or even at stoichiometric composition. process" reaction between elements
Self stabilizing effects occur to some extent only on the In-rich + low cost deposit. methods for
side of the stoichiometric composition where indium could go metals applicable
into the gaseous phase via the formation of In2Se /27/. + easy upscaling
4.1 Materials
spray pyrolysis + simple equipment
Elements for compound semiconductor solar cells are not + efficient CdTe devices
? suitable for CulnSez
as abundant as silicon. However, due to only small amounts - toxic reaction products
necessary for thin fIlm cells this should be no problem even - annealing necessary
for large scale application. According to the information
obtained from geologists the earth crust contains enough. The s(:reen printing + simple equipment
only problem is extraction and purification. But these factors + efficient CdTe devices
are reflected by the selling prices. The prices of elements ? suitable for CuInSez
necessary for a I I'm thick CuInS~ fIlm are listed in table IV. - high T annealing necessary
electrodeposition + simple equipment
Table IV: Cost of materials for CuInS~ thin fIlms + efficient CdTe devices
? suitable for CuInSez
Element Purity Price for Price for
- high T annealing necessary
[%] quant. IlLm of sputtering + commercially available large
> 50 kg CuInSet, area, high throughput
(DMlkg) (DMlm) equipment
Copper 99:9999
+ suitable for window TCO
- 500 0.48 ? efficiency
Indium 99.9999 - 500 0.85
Gallium 99.9999 deposition in defined atmosphere is necessary. The processes
- 1000
also exhibit specific features with respect to structuring. Laser
Selenium 99.999 - 75 0.17 scribing is generally used. Screen printing has "intrinsic"
structuring capabilities. In table VI the status of CuInS~ and
With these actual selling prices the cost of the elemental CdTe module fabrication is summarized.
precursors would not contribute significantly to the total cost Siemens Solar Industries holds the leading position in
of modules, even if the material usage is only 10%. In producing large area (0.4 m2) CuInSez modules with an
addition to the absorber material ZnO for the window, a small efficiency of over 10%. Significant advances have been made
amount of CdS (or another compound) and Mo for the back with CdTe based modules. Stability tests performed at
contact is needed. An important share of the material costs SERI /30/ show good stability after several month of
would be the glass sheet for the substrate and the front cover. exposure.

4.2 Large Scale Processes 5. CONCLUSIONS

For real low cost production processes the throughput of
the process line has to be very high. Substrate surface areas Considering the fast and successful developments in the
have to be in the order of 1 m2. Flexible substrates may not field of thin fIlm compound semiconductor solar cells it seems
bring too many advantages compared to rigid glass substrates. to be very likely that this technology could really impact the
Glass is handled already in huge areas and quantities for also photovoltaic scenarium. Module efficiencies in the order of
quite sophisticated optical coatings of construction glass. 15% seem to be realistic for a longer time frame. However, it
An excellent review of manufacturing issues for CdTe and should be noted that these technologies need its own
CulnS~ is given by Zweibel and Mitchell /28/. Table V engineering approaches which may be rather derived from
lists presently used processes for thin fIlm solar cell large area glass coating than from processes in semiconductor
production. A crude assessment of their suitability is tried. industry.
The experience to date shows that very high efficient
CuInS~ based cells can be realized with selenization and
coevaporation. Other processes like spraying or
electrodeposition yield only moderate efficiencies. On the other
hand, CdTe based cells can be realized with high efficiencies
using various processes. However, careful annealing after

781
 Table VI: Status of CulnS~ and CdTe module development.

Module type Process for Voc Output Area Effie. Manufaet.1

absorber [V] power [em2] [%] Laboratory
deposition [W]
ZnO/CdS/Culn(Ga)S~ 25.5 10.5 938 11.2 ARCO Solar
(Siemens) 131
ZnO/CdS/Culn(Ga)S~ 4000 10.3 Siemens Solar
Ind. 1291
ZnCdS/CulnS~ PVD 1.78 0.877 91.4 9.6 Boeing

Sn02/CdS/CdTe ED 27 6.7 706 9.5 BP Solar 1111

Sn02/CdS/CdTe SPL 20.5 6.1 838 7.3 Photon Energy 1301

CdSICdTe SC 65.3 9.7 1200 8.1 Matsushita 1121

ED electrodeposition, SPL spray pyrolysis, SC scribing

6. ACKNOWLEDGEMENT 14. K.W. Mitchell, C. Eberspacher, F. Cohen, J. Avery, G.

Duran, and W. Bottenberg, Proc. 18th IEEE Photov.
Work of the author has been supported by the Spec. Conf., Las Vegas, 1985, (IEEE, New York, 1985),
Bundesministerium fUr Forschung und Technologie, contract 1395.
no. 0328059D, and the Commission of the European 15. P. V. Meyers, Solar Cells, 27, (1988),91.
Communities, contract no. JOUR 0045C. Informations 16. T. Nazakawa, K.Takamizawa and K. Ito, Appl. Phys.
provided by B. Basol (ISET), C. Eberspacher (Siemens Solar Lett. 50 (5), (1987), 279.
Ind.), R. Noufi (SERI) and J. Woodcock (BP Solar) and the 17. R.A. Mickelsen and W.S. Chen, U.S. Patent
assistance of R. Menner are gratefully acknowledged. No.4 392 451, July 12th, 1983.
18. K. Mitchell, C. Eberspacher, J. Ermer, andD. Pier, Proc.
7. REFERENCES 20th IEEE Photov. Spec. Conf., Las Vegas, 1988, (IEEE
New York, 1988), 1384.
1. W.H. Bloss, F. Pfisterer, and H.W. Schock, in "Advances 19. B.M. Basol, private communication:
in Solar Energy", ed. by K.W. BOer, Vol.4, (plenum 20. W.E. Devaney, W.S. Chen, I.M. Stewart, and
Publ. Corp., New York, 1988),201. R.A. Mickelsen, IEEE Transact. El. Dev., Vol. 37,
2. H.W. Schock, Int. Conf. Polycryst. Semicond., POLYSE, No.2, (1990), 428.
Malente, 1988, in "Polycrysta1line Semiconductors", 21. R. Menner, T. Walter, H.W. Schock, this conference.
J.H. Werner, H.I. Moller, and H.P. Strunk, Springer 22. R. Klenk, R.H. Mauch, R. Schiiffler, and H.W. Schock,
Proc. Phys., Vol. 35, (Berlin, 1989),246. submitted to late news session of this conference.
3. K.W. Mitchell, C. Eberspacher, I.H. Ermer, K.L. Pauls, 23. J.R. Tuttle, D.S. Albin, J.P. Goral, and R. Noufi, Proc.
and D.N. Pier, IEEE Transact. El. Dev., Vol. 37, No.2, 21st IEEE Photov. Spec. Conf., Orlando, 1990, (IEEE,
(1990), 410. New York, 1990), 784.
4. R.H. Mauch and H.W. Schock, this conference. 24. E. Niemi and L. Stolt, Proc. 2nd lnt. Conf. Polycryst.
5. R.H. Mauch et aI., submitted to late news session of this Semicond., POLYSE90, Schwiibisch Hall, 1990, to be
conference. published.
6. W.H. Bloss and H.W. Schock, Techn. Digest Int. Conf. 25. B. Dimmler, F. Grunwald, D. Schmid, and H.W. Schock,
PVSEC-4, Sydney, (1989), 39. submitted to the 22nd IEEE Photov. Spec. Conf., Las
7. B.M. Basol, Proc. 21st IEEE Photov. Spec. Conf., Vegas, 1991.
Orlando, 1990, (IEEE, New York, 1990), 588. 26. B. Dimmler, A. Content,· and H.W. Schock, this
8. Y. Tyan, E.A. Perez-Albuerne, Proc. 16th IEEE Photov. conference.
Spec. Conf., San Diego, 1982, (IEEE, New York, 1982), 27. J. Kessler, H. Dittrich, F. Grunwald, and H.W. Schock,
794. this conference.
9. G.c. Morris, P.G. Tanner, and A. Tottszer, Proc. 21st 28. K. Zweibel and R. Mitchell, "CuInSez and CdTe Scale-Up
IEEE Photov. Spec. Conf., Orlando, 1990, (IEEE, New for Manufacturing", SERIlTR-211-3571.
York, 1990), 575. 29. C. Eberspacher, private communication.
10. S.P. Albright, B. Ackerman, and J.F. Jordan, IEEE 30. J. Stone, Techn. Digest Int. Conf. PVSEC-5, Kyoto,
Trans. El. Dev., Vol. 37, No.2, (1990), 434. 1990,227.
11. A.K. Turner et ai, Techn. Digest Int. Conf. PVSEC-5,
Kyoto, 1990, p. 761.
12. N. Suyama, T. Arita, Y. Nishiyama, N. Ueno,
S. Kitamura, and M. Murozono, Ref. 11,757.
13. Y.S. Seol and H.B. 1m, Proc. 9th EC Photov. Sol. En.
Conf., Freiburg, 1989, (Kluwer, Dordrecht, 1989), 294.

782
lorH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

RECENT ADVANCES IN SINGLE-JUNCTION AND TANDEM

THIN FILM MODULES BASED ON CuInSe 2

C. Eberspacher, J. Ermer, C. Fredric, C. Jensen , F. Karg

R. Gay, D. Pier and D. Willett Siemens AG Central Research
Siemens Solar Industries Laboratories
P.O. Box 6032 Otto-Hahn-Ring 6
Camarillo, California 93011 USA 8000 Munich 83, F.R. Germany

ABSTRACT. Advances in the CuInSe 2 (GIS) photovoltaic technology continue

to increase large-area module efficiencies while demonstrating the
application of low cost production-compatible processing techniques.
Single-junction large area (0.4 m2) unencapsulated modules have been
fabricated with aperture area efficiencies over 10%. Key issues addressed
to achieve this performance were shunts related to localized defects and
patterning, thin film uniformity across large substrates, and the
improvement of macroscopic mechanical adhesion at the CIS/Mo interface .

1. INTRODUCTION have found that the durability of packaged modules

is correlated with the mechanical adhesion at the
The historical dominance of crystalline silicon Mo/CuInSe 2 interface. This paper reports on our
technology is increasingly being challenged by high- recent progress in controlling localized defects,
efficiency, low-cost thin film photovoltaic technol- thin film adhesion, and thin film uniformity.
ogies. Siemens Solar is actively developing thin
film technologies as a complement to its existing 3. CONTROL OF LOCALIZED DEFECTS
crystalline silicon technology base. Single- junc-
tion and multi-junction thin film products will be Localized defects have been an impediment to the
important in keeping pace with increasing market realization of large-area modules of the high effi-
demand and competitive pricing. Siemens Solar ciencies previously demonstrated on small-area test
Industries is developing thin film CuInSe 2 with a devices. Localized defects generally arise from
focus on low-cost manufacturing techniques suitable particulates or from improper patterning.
for large-area, high-power modules. Localized defects can be caused by particulate
This paper reports on recent advances in single- contamination . Particulate contamination is con-
junction and tandem thin film modules based on trolled by filtering particles from the processing
CulnSe 2 . In recent months large-area (0.4 m2) environment using well-known filtered air techniques
single-junction thin film CuInSe 2 modules with common to the mic roelectronics industry. A high-
aperture area efficiencies above 10% have been class clean room atmosphere is not necessary to
demonstrated using production-compatible fabrication fabricate high power CulnSe 2 modules, but attention
techn iques. must be paid to minimizing particle densities prior
to subsequent thin film deposition steps lest par-
2. MODULE FABRICATION ticulate contamination be embedded in the photo-
voltaic thin films.
The fabrication of CuInSe 2 modules is a simple
sequence of thin film deposition and patterning
steps. The modules are fabricated on soda lime
32.2 em
glass substrates approximately 32 cm x 128 cm x 1.9 CIS 0.4 m2 MODULE
mm (Fig. 1). The first step is to deposit a back .. 30.6 em · .
metal electrode of molybdenum. The second step is ~ Series
the patterning of the Mo electrode; this pattern is Connected
Cells
typically done using a Nd:YAG lase r . The third step
is to deposit the CuInSe 2 layer. The fourth step is
to deposit a thin (thicknes s less than 50 nm) CdS
layer . The fifth step is to pattern the CuInSe 2/CdS
layers in order to allow the interconnection between
the back and front electrodes of adj acent cells.
This second pattern is typically done using a me-
chanical scribing system using or the laser scribing
system used for the first pattern. The sixth step
is to deposit a ZnO transparent top electrode. The
final step of the basic submodule fabrication se-
quence is to pattern the ZnO top electrode. The
submodule is then encapsulated into a finished
module using a front cover glass, a transparent
pottant, and a frame.
The realization of the higher efficiencies and
lower costs possible with thin film technologies
requires that the high efficiencies demonstrated on
small-area test devices be achieved on large-area
substrates and that high power submodules be pack-
aged into durable encapsulated and framed modules .
We have found that the efficiencies of large-area Fig . 1. Basi~ design of a 0 . 4 m2 CuInSe 2 module.
submodules are often dependent on the control of The ~nsert shows the details of the inte-
localized defects and of thin film uniformity. We grated interconnect between adjacent cells .

783
 Local defects due to local severe nonstoichio- that ordinary soaps and standard high-efficiency air
metry are minimized by minimizing particle genera- filtering are sufficient to minimize the effects of
tion in the module fabrication processes. The residues and particulates provided that the initial
primary loss mechanism due to local severe non- contamination is not too severe.
sto i chiometry is believed to be the formation of A second common method of controlling interfaces
inclusions, most commonly conductive CuSe x ' that to promote adhesion is to facilitate interdiffusion
shunt the pn heterojunction . between adjacent layers . At first glance however
Localized defects due to improper patterning one would not anticipate that selenides of molybde-
become an issue when one shifts from the research num or intermetallic compounds of Cu, In, Se and Mo
lab where small-area test devices are reasonably would improve adhesion. We have found that high-
fabricated using standard photolithography tech- energy processing that promotes interdiffusion may
niques to the manufacturing floor where large-area result in improved pn junction performance, but can
power modules are fabricated using high-linear- lower the layer adhesion.
speed, multi-point patterning equipment . It is A third common method of controlling interfaces
critical to assure the complete removal of debris to promote adhesion is to insert interfacial layers.
and residues from the patterns themselves and from Interfacial layers can increase adhesion by forming
the area adjacent to the patterns. a bond to adjacent layers that might themselves not
The impact and control of localized defects are adhere well if in direct contact. Unfortunately,
discussed in more detail in a companion paper by D. the choice of interfacial layers suitable for in-
Wil l ett et al . at this conference (1). creasing adhesion at the Mo/CuInSe 2 interface is
limited by the tendency of many materials to be
4. THIN FILM ADHESION converted into selenides during subsequent process-
ing.
Good mechanical adhesion between the constituent A fourth common method of controlling interfaces
layers of a thin film structure is essential to the to promote adhesion is to mechanically texture
efficiency and durability of the completed module. surfaces to facilitate physical bonding. In photo-
The importance of adhesion is evident in the dura- voltaic devices this technique can compromise device
bili ty of encapsulated CuInSe z modules during humidi- efficiency if the interfacial texture results in
ty-freeze testing (Fig . 2a) and during outdoor higher junction active area for forward leakage
testing (Fig . 2b) (2). Simply stated, the better current or higher effective absorber layer stresses
the mechanical adhesion, the better the module that cause electronically active defects. Weighed
durability . against its deleterious effects on device perfor-
The degree of thin film adhesion required is mance , we have not found mechanical texturing advan-
qualitatively that necessary to ensure high manufac- tageous.
turing yields and module durability in the field . The second broad category of techniques for
The microscopic mechanisms linking mechanical adhe- promoting adhesion is the control of the layers'
sion and module durability are not well understood, bulk properties. The bulk properties of the con-
but the impacts of poor adhesion are quite dramatic . stituent layers is critical in promoting adhesion.
A simple quantitative measure of adhesion is possi- For example, the residual stress of the layers can
ble through tape pull tests using adhesive tapes of vary widely depending on layer deposition condi-
various pull strengths . A convenient quantitative tions . It is possible to control the phase behavior
measure of "good" adhesion sufficient to ensure high of the layers by controlling the fabrication pro-
manufacturing yields and module durability is that cesses carefully.
adhesion necessary to withstand Scotch brand #850 25 The third broad category of techniques for
oz/in . tape . promoting adhe~ion is the control of the fabrication
Techniques for increasing adhesion vary widely, process. The ~brication processes can affect the
depending on the particular materials and structures interfaces and t1;le bulk properties in the manners
being utilized. Broadly speaking the techniques can mentioned above, hut it can also affect the struc-
be divided into control of the interfaces , control ture in more subtle ways. For example, the effects
of the layers' bulk properties , and control of the of differences in thermal expansion can be minimized
fabrication process. These three broad categories by controlling thermal profiles or by differentially
are also strongly dependent on one another . heating separate layers in the structure. Another
The control of interfaces is the most common example is the control of reaction species such as
te chnique for controlling adhesion . Surfaces must the reactive deposition of Cu and In with molecular
be free of organic residue (e.g. , skin oils from Se vapor (3) or H2Se gas (4). -
human handling) and of particulate debris (e.g . , Mo The conclusion from this work on improving thin
pa r ticles or airborne dust particles around which film adhesion is that no single technique alone is
extensive "halos" of poor local adhesion may form) sufficient for CuInSe 2 photovoltaic structures, but
in order for adhesion to be high (1). We have found that a combination of techniques does provide excel-

1.06

...,
:; Group A Good Adh.slon
1.04 • Best From
Group A
-g ~ Group B Poor Adhesion
• Average 01
:Ii' Group A
'0 ,J • Group B
~ i 0.94
e::J .!!
0,82
'i
z

....
~
•
~
0
z
0.90
0.88
II: 0.86
0.84
0.3
0.82
0 20 40 60 80 100 120 140
Pmu belore 10 HF Days 01 Outdoor exposure

Fig. 2a. CuInSe 2 module stability after 10 humidity- Fig. 2b . Dependence of CuInSe 2 module outdoor dura-
freeze cycles . bility on Mo/CuInSe z adhesion .

784
lent adhesion. By use of a combination of these ture is critically important. The ZnO/CdS/CuInSe 2
techniques, good adhesion is possible on 0.4 m2 heterojunction is dominated by recombination (7,8),
substrates. The impact of improved adhesion is and the microstructure of the CuInSe 2 layer is
strikingly evident in maps of tape pull results critical in determining the density of recombination
(Fig. 3). centers and charged traps near the CdS/CuInSe 2
interface and in the CuInSe 2 depletion region, which
5. THIN FILM UNIFORMITY in turn contro 1 the open circui t voltage. The
microstructure of the CuInSe 2 layer is also critical
The translation of the high-efficiency processes in determining the minority carrier diffusion
developed on small-area test devices to large-area length, which in turn affects the magnitude and the
modules requires that the thin films be uniform voltage dependence of photocurrent collection (8).
across the large-area substrates. The primary film Layer microstructure also affects the scribing
characteristics for which uniformity is important processes used to pattern the layers into the cells
are thickness, composition, and film structure. of a monolithic integrated submodule.
The uniformity of the thicknesses of the layers The conclusion from this work on controlling thin
in a typical glass/Mo/CulnSe 2/CdS/ZnO structure is film uniformity is that one should focus on composi-
moderately important. The target thickness of the tion and film microstructure, rather than on layer
different layers is typically determined by the thickness.
functional requirements of the layer. For example The techniques that aid in controlling thin film
the molybdenum back electrode layer and the ZnO uniformity include a wide variety of "scientific"
front electrode layer must be thick enough to pro- and "engineering" insights. One technique for
vide negligible sheet resistances. In general the controlling layer uniformity is to design a self-
thickness of these electrode layers can vary ±lO'll correcting deposition process by providing a natural
without seriously affecting the junction perfor- feedback mechanism. For example, the volatilization
mance. of excess selenium from hot CuInSe 2 surfaces during
The thickness of the CuInSe z optical absorber vacuum co-evaporation effectively regulates selenium
layer must be sufficient to absorb the incoming content, and the rej ection of indium impacts the
sunlight, but thicker layers are typically used in composition of reactively sputtered CuInSe 2 layers
order to make certain that absorption is largely (4).
complete within material with good electronic char- A second technique for controlling layer unifor-
acteristics (e. g., well away from the back metal mity is to pre-mix the constituents in exact
electrode where grain sizes are typically smaller). amounts. For example, the Cu, In, and Se contents
In general the thickness of the CulnSe z absorber can be controlled in screen printing processes by
layer can vary ±lO'll without deleterious effects the pre-mixing of the metallic and compound constit-
provided that the CulnSe 2 materials quality is good uents in the initial pastes (9), and can be con-
wi thin the optical absorption depth. Thickness trolled in flash evaporation processes by the mixing
control to within ±lO'll is possible with all typical of source materials (10).
PV deposition processes, including evaporation, A third technique for controlling layer uni-
electrodeposition, screen printing, sputtering, or formity is to closely control reaction rates and the
chemical vapor deposition. formation of intermediary phases and materials. For
The uniformity of the compositions of the layers example it is generally best to avoid the formation
in a typical glass/Mo/CulnSez/CdS/ZnO structure is of binary inclusions or off-stoichiometry ternary
very important, especially the composition of the phases in the CuInSe 2 layer.
CuInSe z layer. The electronic properties of the A fourth technique for controlling layer unifor-
CuInSe 2 layer can vary widely with composition from mity is to design manufacturing equipment with
degenerate p-type to n-type (5), but under certain inherent mixing and/or feedback mechanisms. For
processing conditions strong electronic compensation example, vigorous stirring or sparging of electro-
can offset a surprisingly wide variation in composi- plating baths can minimize spatial variations in the
tion and yield comparable device performance (6). characteristics of plated layers.
The most catastrophic losses from variations in The conclusion from this work on improving thin
composition are from shunts across the pn junction film uniformity is that a combination of engineering
due to binary inclusions and metallic precipitates. and scientific insights give layer uniformity suffi-
However, if the CulnSe 2 layer composition is homoge- cient to yield submodule performance uniformity. By
neous (i.e., few inclusions or precipitates), then use of a combination of these techniques, excellent
the composition can vary ±S'll with little deleterious junction performance uniformity is achievable on 0.4
effect on junction performance (6). m2 substrates. The uniformity of performance is
The uniformity of the microstructure of the evident in a map of the open circuit voltage over a
layers in a typical glass/Mo/CulnSe 2/CdS/ZnO struc- 0.4 m2 submodule (Fig. 4) where open circuit voltage
Voe varies less than ±5% over the full area.

Poor Adhesion Good Adhesion 6. IMPACT ON MODULE POWER

.~ --- =~-
The overall impact of advances in m1n1mlz1ng
localized defects, maximizing thin film adhesion,
and controlling thin film uniformity over large

- --

--------------------
====~===

2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 910
8 0 .42 0 .44 0.46 0 .48

Fig. 3. Mo/CuInSe 2 adhesion measured using 25 oz/in. Fig. 4. Map of open circuit voltage Voe on a 0.4 m2
adhesive tape. CuInSe 2 submodule.

785
 3.----,----,----,----,----,
areas is the achievement of unencapsulated sub-
modules with aperture area efficiencies over 10%.
Figure 5 shows the current-voltage data for a 0.4 mZ,
40+ watt submodule with an aperture area efficiency
2
of approximately 10.4%. Area: 3890 em 2
Further improvements are possible with further lac: 2.52 A
advances in processing uniformity and precision. Voc: 24.6 V
Table 1 compares the characteristics of the best FF: 0.652
small-area test device, the best 0.1 mZ submodule, Power: 40.4 W
Aperture Elf: 10.4%
and the best 0.4 mZ submodule. When the area utili-
zation already achieved on 0.1 mZ submodules is
realized on the 0.4 m2 substrate size, then 44 watts
on 0.4 m2 will be possible. When the area utiliza-
tion already achieved on 0.1 m2 submodules is com- o 5 10 15 20 25
bined with the junction performance already achieved Voltage (V)
on 3.5 cm2 test devices, then module aperture area 100 mW/cm2 ASTM air mass 1.5 global spectrum, 25"C

efficiencies over 13% and powers over 50 watts on

0.4 mZ sizes will be possible (11). When this is Fig. 5. Currert-voltage characteristics of a 40.4 W,
achieved, the power density of thin film CuInSe z 0.4 m CuInSe 2 submodule.
modules will rival the power densities of crystal-
line silicon photovoltaic modules, and the impact of
the lower manufacturing costs possible with thin Table 1. Measured and predicted performance of
film technologies will be a key advantage. CuInSe Z test devices and modules."
Unencapsulated Modules
7. TANDEM THIN FILM TECHNOLOGIES Test Measurements Predictions
Device 0.1 mZ 0.4 m2 0.4 mZ
A CuInSe 2 thin film submodule can be used as a (1 cell) (55 cells) (53 cells) (53 cells)
single-junction device with a transparent protective
cover or as a multi-junction device by combining it Area b (cm2 ) 3.5 938 3890 3900
with submodules based on wider-bandgap absorber Vee/cell (mV) 508 464 464 508
materials. The most convenient technique for con- J se (rnA/cm2 ) 4l.0 39.3 37.0 4l. 0
structing multi-junction modules is to separately Vee (V) 25.5 24.6 26.9
fabricate the two submodules and to mechanically Ise (A) 0.641 2.52 2.86
stack the component submodules during the lamination Fill Factor 0.677 0.639 0.652 0.672
process to make a tandem module. Siemens Solar Eff b (%) 14.1 11.2 10.4 13 .3
Industries has been investigating tandem modules Power (W) 10.5 40.4 51. 8
based on CuInSe 2 and Si:H alloy thin film submodules
for several years. ~100 rrII/em 2 ASTM air mass 1.5 global spectrllll, 25°C.
The power densities possible with different Active area for test device. Aperture area for modules.
tandem stacks have been estimated by Mitchell (12).
Mitchell's analysis suggests that efficiencies
upwards of 20% are possible by combining CuInSe z and ACKNOWLEDGMENT
Si:H alloy junctions and that efficiencies upwards The advances presented here would not have
of 24% are possible by combining CuInSe Z with other possible without the contributions of our Siemens
polycrystalline semiconductor junctions. To date Solar Industries Engineering and Development col-
Siemens Solar Industries has demonstrated a multi- leagues and our Joint Venture associate Mr. Katsumi
junction cell active area efficiency of 15.6% on 4 Kushiya of Showa Solar Energy KK. The authors wish
cm2 (13), a tandem submodule aperture area efficiency to thank Dr. Kim Mitchell for his technical guidance
of 12.3% on 0.1 m2 (14), and a tandem submodule in identifying directions to pursue in this work.
aperture area efficiency of 10.5% on 0.4 m2 (15)
using CuInSe 2 and Si: H alloy thin films to make REFERENCES
mechanically stacked tandem modules. 1. D.R. Willett et al., this conference.
Work is continuing on semitransparent thin film 2. D.E. Tarrant et al. SERI PV Module Reliability
Si:H alloy technologies by our associates at Siemens Workshop, Lakewood, CA (1990).
Solar GmbH. A collaborative effort with Siemens AG 3. A. Rockett et al. Proc. 21st IEEE PV Special-
Central Research Laboratories is now under way to ists Cant., pp. 764-67 (1990).
investigate CuGaSe z and related polycrystalline 4. T.C. Lommasson et al. Proc. 19th IEEE PV Spe-
materials suitable for multi-junction devices with cialists Cant., pp. 1285-90 (1987).
CuInSe z· 5. R. Noufi et al. App. Phys. Lett., 45, pp. 668-
70 (1984).
8. CONCLUSIONS 6. J.Ermer et al. Proc. 21st IEEE PV Specialists
Cant., pp. 595-599 (1990).
Thin film photovoltaic technologies are important 7. R.R. Potter et a1. Proc. 18th IEEE PV Spec.
to the international PV industry keeping pace with Cant., pp 1659-64 (1985).
the rapidly growing market demand and to steeply 8. K.W. Mitchell et al. Proc. 20th IEEE PV Spe-
lowering PV module manufacturing costs. Thin films cialists Cant., pp. 1461-68 (1988).
of CuInSe 2 are well suited to both single-junction 9. T. Arita et al. Proc. 20th IEEE PV Specialists
and multi-junction PV products, and are rapidly Cant., pp. 1650-55 (1988).
progressing in both power density and module dura- 10. J. Piekoszewski et al. Solar Energy Hats., 2,
bility. Local defects, film adhesion, and film pp. 363-72 (1980).
uniformity are factors limiting CuInSe 2 submodule 11. J. Ermer et al. Proc. 4th Internat. PV Science
performance that have recently been overcome in and Eng. Cant., pp. 475-80 (1989).
order to achieve 0 . 4 m2 submodule aperture area 12. K.W. Mitchell, Technical Digest 1st Int. PV
efficiencies over 10% using production-compatible Science and Eng. Cant., pp. 691-94 (1984).
manufacturing process. Further development of wide- 13. K.W. Mitchell et al. Proc. 8th European Commis-
bandgap materials such as Si: H alloys and CuGaSe z sion PV Solar Energy Cant., pp. 1578-82 (1988).
will yield semitransparent submodules that can be 14. K.W. Mitchell et al. Proe. 20th IEEE PV Spe-
combined with CuInSe 2 submodules to yield high- cialists Cant., pp. 1384-89 (1988).
efficiency multi-junction power modules. 15. K.W. Mitchell et al. Proc. 21st IEEE PV Specia-
lists Cant., pp. 1481-86 (1990).

786
10TH EUROPEAN PHOTOVOLTAIC SOLAR ENERGY CONFERENCE 8-12 APRIL 1991 LISBON, PORTUGAL

PHOTOCURRENT TRANSPORT IN HETEROJUNCTIONS

WITH GRADED Cu(In,Ga)Se2 ABSORBERS

R. Menner, T. Walter and H.W. Schock

Institut fUr Physikalische Elektronik

Prof. Dr.-Ing. W.H. Bloss
Universitat Stuttgart
Pfaffenwaldring 47, D-7000 Stuttgart 80, FRG

ABSTRACT. The Voc of highly efficient CulnS~ based solar cells is thought to be limited by interface and space
charge region recombination. Recombination currents can be reduced by increasing the bandgap of the absorber
material, however, with a corresponding loss of photocurrent. In order to reduce the recombination currents without
loss of photocurrent inhomogeneous structures with abrupt or gradual increase of bandgap towards the heterojunction
by Ga-incorporatiOil have been investigated. The influence of effective forces on photocarrier collection due to the
position dependent electron affinity was calculated by a simple model for various types of graded gap absorbers.
"Electron mirror" structures for most efficient photocurrent collection were considered as well. The theoretical results
are in good agreement with experimental measurements on evaporated Cu(In,Ga)S~/(Zn,Cd)S heterojunctions with
abruptly and gradually increased absorber bandgaps and prove the potential of graded gap absorbers to increase
efficiency of CuInS~ based solar cells. The photocurrent collection is much less affected by the continuously increased
bandgap structures as proven by bias dependent spectral response and j-V analysis. Increased Voc was obtained without
jsc and fill factor losses compared to pure CulnS~ reference cells. The highest efficiency of a graded gap
Cu(ln,Ga)S~/(Zn,Cd)S solar cell so far is 10.2 % without AR-coating under 85 mW/cm2 ELH-illumination with
Vue = 428 mY, jsc = 30.8 mA/cm2 and 66 % f1ll factor.

1. INTRODUCTION recombination either at the interface /5/ or in the absorber

space charge region /6/ as indicated in Fig. I. This has been
Thin film solar cells on basis of CulnS~ have reached deduced from detailed temperature dependent j-V analysis which
efficiencies up to 12-14 % III f21 during the last years. indicated tunneling or injection/diffusion not to be the dominant
The performance is mainly limited due to a moderate open circuit loss mechanisms. Both the recombination via recombination centers
voltage far below the theoretical maximum. Inhomogeneous absorber at the interface and in the space charge region depend strongly
structures were proposed to overcome this voltage on the free carrier densities available for recombination.
problem 131 14/, however, only few experimental data
have been available so far. Concerning the space charge recombination it was already
shown by Sah et al. /7/ that the recombination contributes
significantly only in a narrow region where n'Tpo equals
2. THEORETICAL approximately P'Tno (T , Tno: minority carrier lifetimes in n- and
p-type material, resp.). In case of a p-n + junction this region
2.1 Limitation of the Open Circuit Voltage moves only within the absorber depletion region with applied
The limitation of the open circuit voltage Voc and the fill bias and thus the recombination increases proportional to
factor ff in a solar cell is predominantly caused by the junction (n·p)ll2 ex ,\(absorber). In contrary, interface recombination always
bucking currents opposing the photo current or by a reduced is located at the junction. In a p-n + junction the larger part of
photo current collection under forward bias. In CuInS~-based the diffusion voltage occurs across the absorber depletion region
devices spectral response measurements show only a small bias thereby causing a high hole barrier towards the interface. In this
dependence of the photocurrent. Hence, Voc is assumed to be case the hole capture is the rate limiting process and the interface
limited by high bucking currents most probably dominated by recombination is directly proportional to the hole density at the
junction.
As a consequence of the larger bandgap in homogeneous
Ec Cu(In,Ga)S~ devices the recombination currents are reduced
due to a higher hole barrier and a lower intrinsic density ni' The
open circuit voltage can be increased accordingly /8/ but the
short circuit current is reduced due to the larger bandgap of the
Ev absorber material
The intentioo of the inhomogeneous or graded absorber structures
is to reduce bucking currents by increasing the bandgap only
in regions of highest recombination and to generate the photo
current preferably in the low bandgap regions as close to the
heterojunction as possible. It is desirable that the increased bandgap
(Zn,Cd)S CulnSe2 affects only the valence band edge and causes no conduction band
offset or reduced field in the conduction band which may affect
C16 the photocurrent collection. Several related wider bandgap materials
(e.g. CuGaS~, CuInS 2 etc.) have to be investigated in this respect.
Fig. 1: Band diagram of a (Zn,Cd)S/CuInS~ heterojunction with However, the introduction of gradients in materials with different
paths for interface (1) and space charge region recombination (2). lattice constants may be a source of dislocations and the importance
nIP' PIP = electron and hole densities at the interface of structural properties is evident in case of graded structures /9/.

787
2.2 Modeling of Photocurrent Collection in Graded Gap Absorbers of the conduction band /1 0/. High recombination at the back
contact and a total extraction of excess minority carriers at the
In order to get an insight into the photocurrent collection heterointerfuce due to high electric fields were assumed as boundary
of graded gap absorbers, modeling has been done in a first order conditions. The linear recombination term is only valid for non
approximation. The combination of the transport and the continuity inverted regions, i.e; regions where the electrons are still minorities.
equations using a simple linear recombination model yields In case of a n +p junction inversion occurs at the heterointerface.
Hence, recombination is limited by the density of holes and equation
d 2n dE, dn n- no (1) is not valid in this region. An absorber depletion width of
D -+1' (n-+E -)+Gph - - - = 0 (1)
• dx 2 • dx 'dx t. 600 nm was assumed as well as diffusion lengths of 800 nm for
CuInS~ and 200 nm for the Cu(In,Ga)S~ to account for higher
which had to be solved numerically. Ee in that case denotes the recombination due to dislocations in the graded regions. Absorption
effective electric field for electrons corresponding to the slope data for CuInS~ /11/ were shifted according to the bandgap
of the quaternary graded material.
...... 0 5 0 . . - - -- - - - - - - - - - - - - - ,
Three different types of absorbers were considered - the pure
>., CuInS~ (type I), a gradual (type II) and a nearly abrupt (type
III) gallium incorporation towards the heterojunction. The tentative
>-
band diagrams are depicted in Fig. 2. In case of the abrupt change
'"a:w of the bandgap (type III) a negative slope occurs in the conduction
~ - 0 50
band resulting in a negative effective electric field for the
photogenerated electrons even without forward bias. The band
offsets in the conduction and in the valence band were taken from
Turowski et al. 112/. The resulting electron density for 790
-1 50
nm illumination was calculated and is shown in Fig. 3. In case
--TYPE I
of the graded gap absorbers the slope of the electron density in
---TYPE II
---TYPE I II
the diffusion region (x > 600 nm) is larger due to the reduced
light absorption in the depletion region. The spike of the electron
- 2 50 -I----I---+----+---t-----+---I
o 00 o 50 1 00 1 50 density in the type III device results from the assumed band
x [um] discontinuity to the gallium containing material. The electron
transport from the CuInS~ is hindered by a negative effective
Fig. 2: Band diagram for different types of graded absorbers. electric field in the conduction band. In order to overcome this
The location x is measured from the heterojunction. region the electrons accumulate and diffuse due to the high gradient
towards the window. The high density of electrons increases
...... 0 20
recombination and therefore reduces the photocurrent. The calculated
...
If)
photocurrent densities are displayed in Fig. 4. The principles
z explained above can be clearly seen. An only slightly reduced
:;,
photocurrent for the continuously graded absorber (type II) and
II>
a:
0 15 1\
«
II
II - - T YPE I a significantly lower current for the rather abrupt increase of
I I ---TYPE II the bandgap (type III) is observed. The situation for the absorber
I I
III - - - TYPE III type ill might be even worse, as the accumulation of the electrons
z
w
0 10
I
I I
I occurs exactl y at the position where the density of recombination
0
I I
Z I I centres might be increased due to the lattice mismatch at the
0
IX
I
I I _---- CuInS~-CuGaS~ interface. It is obvious that this is only valid
t-
u
o 05
I
I
\
/'
/ /, . . . - - - - - - for electrons generated in the low gap material. Electrons which
w I \ /' - - - ... - -
-' are generat