Scribd
Upload
320 views

MicroelectronicsPackagingHandbookSemiconductorPackaging 1

This document provides the foreword and preface for a three-part handbook on microelectronics packaging. The foreword discusses how electronics has become the largest industry and how electronic packaging is a key enabling technology for reducing system size and cost. The preface outlines how the first edition of the handbook addressed the need for information on packaging technologies and discusses new factors like cost and size reduction that have become important. It also notes several paradigm shifts in packaging technologies and systems since the first edition.

Uploaded by

ياسر الرفيعي
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
320 views

MicroelectronicsPackagingHandbookSemiconductorPackaging 1

This document provides the foreword and preface for a three-part handbook on microelectronics packaging. The foreword discusses how electronics has become the largest industry and how electronic packaging is a key enabling technology for reducing system size and cost. The preface outlines how the first edition of the handbook addressed the need for information on packaging technologies and discusses new factors like cost and size reduction that have become important. It also notes several paradigm shifts in packaging technologies and systems since the first edition.

Uploaded by

ياسر الرفيعي
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1061

MICROELECTRONICS

PACKAGING
HANDBOOK
MICROELECTRONICS
PACKAGING
HANDBOOK
Semiconductor Packaging
PART II

Second Edition

Edited by
Rao R. Tummala
Georgia Institute of Technology
Eugene J. Rymaszewski
Rensselaer Polytechnic Institute
Alan G. Klopfenstein
AGK Enterprises

"
~.

KLUWER ACADEMIC PUBLISHERS


BOSTON/DORDRECHT/LONDON
Distributors for North, Central and South America:
Kluwer Academic Publishers
101 Philip Drive
Assinippi Park
Norwell, Massachusetts 02061 USA
Telephone (781) 871-6600
Fax (781) 871-6528
E-Mail <[email protected]>

Distributors for all other countries:


Kluwer Academic Publishers Group
Distribution Centre
Post Office Box 322
3300 AH Dordrecht, THE NETHERLANDS
Telephone 31 78 6392 392
Fax 31 786546474
E-Mail <[email protected]>

. , Electronic Services <http://www.wkap.nl>

Library of Congress Cataloging-in-Publication

Microelectronics packaging handbook / edited by Rao R. Tummala,


Eugene 1. Rymaszewski, Alan G. Klopfenstein.
p. cm.
Includes bibliographical references and index. Contents: pt. 1. Technology drivers --
pt. 2. Semiconductor drivers -- pt. 3. Subsystem packaging.
ISBN 0-412-08431-7 (pt. 1) -- 0-412-08441-4 (pt. 2) -- 0-412-08451-1 (pt. 3)
l. Microelectronic packaging--Handbooks, manuals, etc. l. Tummala, Rao R., 1942 --.
TK7874.M485 1996 96-37907
621.381 '046--dc 21 CIP

Copyright © 1997 by Chapman & Hall


Second printing by Kluwer Academic Publishers 1999

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system or transmitted in any form or by any means, mechanical, photo-copying, recording,
or otherwise, without the prior written permission of the publisher, Kluwer Academic
Publishers, 101 Philip Drive, Assinippi Park, Norwell, Massachusetts 02061

Printed on acid-free paper.


FOREWORD

Electronics has become the largest industry, surpassing agriculture,


auto, and heavy metal industries. It has become the industry of choice
for a country to prosper, already having given rise to the phenomenal
prosperity of Japan, Korea, Singapore, Hong Kong, and Ireland among
others. At the current growth rate, total worldwide semiconductor sales
will reach $300B by the year 2000.
The key electronic technologies responsible for the growth of the
industry include semiconductors, the packaging of semiconductors for
systems use in auto, telecom, computer, consumer, aerospace, and medical
industries, displays, magnetic, and optical storage as well as software and
system technologies. There has been a paradigm shift, however, in these
technologies, from mainframe and supercomputer applications at any cost,
to consumer applications at approximately one-tenth the cost and size.
Personal computers are a good example, going from $500IMIP when
products were first introduced in 1981, to a projected $IIMIP within 10
years. Thin, light portable, user friendly and very low-cost are, therefore,
the attributes of tomorrow's computing and communications systems.
Electronic packaging is defined as interconnection, powering, cool-
ing, and protecting semiconductor chips for reliable systems. It is a key
enabling technology achieving the requirements for reducing the size and
cost at the system and product level.
The recent paradigm shifts in packaging processes such as direct
flip-chip attach to organic board, the ability to increase wiring in the
organic board away from expensive drilled technology by the direct depo-
sition of thin organic and metal films, the achievement of reliability without
herrneticity previously achieved only in ceramic packaging, and the poten-
tial integration of all passive components within the interconnect board are
expected to lead to revolutionary products in all segments of electronics.
To describe the status and future developments in these technologies,
editors Profs. Tummala and Rymaszewski and Mr. Klopfenstein have
assembled an outstanding team of 74 packaging practitioners from across
the globe. Together, they produced in three parts the Microelectronics
II-vi FOREWORD

Packaging Handbook, an unmatched book needed by industry and univer-


sities. It is equally appropriate both as an introduction to those just entering
the field, and as an up-to-date reference for those already engaged in
packaging design, first and second level packages and their interconnec-
tions, test, assembly, thermal management, optoelectronics, reliability,
and manufacturing.
I applaud the authors and editors for their great contributions and
hope that this book will help focus the attention of outstanding faculty,
students, and industry researchers on the complex packaging challenges
that they face in the 21st century.

Bertrand Cambou
Motorola Senior Vice-President
Sector Technology
Semiconductor Productor Sector
PREFACE

"This book reflects a need based on an ever-increasing realization


by the microelectronics community that while the semiconductors continue
to be improved upon relentlessly for performance, cost, and reliability,
it is the packaging of these microelectronic devices that may limit the
systems performance. In response to this need, the academic community
began to ask its industrial counterparts what packaging is all about, and
what scientific and technological areas should be pursued in collaboration
with industry. As a result of these discussions, a number of universities
in the United States and throughout the world already have research and!
or teaching programs in various aspects of packaging, ranging from
materials, assembly, electrical, and thermal modeling, thin films, and
many others. The multidisciplinary nature ofpackaging technology clearly
poses challenges not only to this community, but also to industrial col-
leagues who will have to use scientific fundamentals from a cross section
of disciplines to bring about advanced products." That's the way we
began the preface of the 1st edition of the Microelectronics Packaging
Handbook in 1989. The first sentence is true today; however, two addi-
tional factors have become more important: cost and size. This is readily
apparent when one considers consumer electronics. The use of packaged
microelectronics has increased tremendously in a great variety of applica-
tions-television, computers, communications, navigation, automotive,
avionics, medical, and so forth. Perhaps the biggest change is the degree
to which universities have embraced the field of microelectronics and
established major research and teaching programs. Also, the growth in
coverage of microelectronics packaging by technical societies-both in
technical symposia and in tutorial sessions-has been dramatic.
While the book is not intended to be a classic textbook for any
course in packaging, it does attempt to define what microelectronics
packaging is all about, including the current state of the technology across
all engineering and scientific disciplines, and the fundamental areas that
could impact the industrial needs. It also provides a comprehensive source
of information on all aspects of microelectronics packaging.
II-viii PREFACE

Packaging of electronic circuits is the science and art of establishing


interconnections and a suitable operating environment for primarily elec-
tronic circuits to process and store information. Accordingly, microelec-
tronics packaging in this book is assumed to mean those designs and
interconnection technologies necessary to support electrically, optically,
thermally, mechanically, and chemically those semiconductor devices with
micron and submicron dimensions that are often referred to as integrated
circuits. Electronic components and their packaging are the building blocks
for a vast variety of equipment. Customer demand and competition among
equipment suppliers result in continuing enhancements and evolution of
these building blocks, particularly in terms of cost, performance, quality,
and reliability.
Since publishing the first edition of this book, a number of paradigm
shifts have taken place in the electronics industry. For one, the emphasis
on systems has changed from mainframe computers to personal computers
and portable, wireless systems requiring ultra low-cost, thin, light, and
portable packages. Second, the semiconductor technologies have shifted
emphasis from bipolar to CMOS. A number of paradigm shifts have taken
place in packaging as well. Direct chip attach was possible in 1989 only
to such inorganic substrates as alumina and silicon. Today, chips can be
bonded and used directly on printed wiring boards. The board itself is
undergoing major change, allowing very high density wiring by photolitho-
grapic processes in contrast to mechanical drilling, which became expen-
sive and obsolete. The very high reliability that was achieved previously
only with ceramic packages is now beginning to be achieved with organic
packages and boards without hermeticity. This book is written with these
changes in mind and consistent with the needs of the industry.
This collection of books is organized into three parts. Part 1 includes
chapters 1 through 6 and covers the driving forces of microelectronics
packaging-electrical, thermal, and reliability. In addition, Part 1 intro-
duces the technology developer to aspects of manufacturing that must be
considered during product development. Part 2, chapters 7 through 14,
covers the interconnection of the IC chip to the first level of packaging
and all first level packages. Part 2 also includes electrical test as well as
sealing and encapsulation technologies. Part 3, chapters 15 through 20,
covers board level packaging as well as connectors, cables, and optical
packaging. The general overview of packaging is repeated in each part
as chapters 1, 7, and 15.
The main problem of creating a handbook in any field is obvious:
by the time the book is completed, much of the information in it may
require updating. This is particularly true with microelectronics packaging,
which is one of the fastest, if not the fastest, growing of all technologies.
The first edition was a beginning and this 2nd edition is long overdue.
Further development and refinement, based on the comments and sugges-
PREFACE II-ix

tions of the users, can appear in future editions. The book in its present
state reflects what we believe the community wants currently. We are
exploring new ways of providing the information on a more current basis
including electronic publishing and interchange.
The first edition was written entirely by IBMers. We tried to include
as many non-IBM technologists as possible and to minimize the use of
IBM jargon. This 2nd edition includes 74 authors, each an expert in his
or her own field, from many companies, universities, and countries. We
believe that this second edition provides a very representative and compre-
hensive look at the field of microelectronics packaging.
Any handbook requires the dedication of a number of individuals
involved in writing, typing, graphics preparation, manuscript reviewing,
copyediting, and publishing, and managing all these operations in such
a way that the final book is available in a timely manner. Above all, a
free and stimulating attitude on the part of all the participants is necessary.
In addition to the chapter authors, we would like to acknowledge the
work of Debra Kelley in helping to keep us on track and for her efforts
in preparing some of the manuscripts. It should be pointed out that exten-
sive use of the Internet permitted us to work together more easily and
cost effectively. Our thanks to Jim Geronimo and Barbara Tompkins
for the preparation of numerous drafts, extensive copyediting, and their
willingness to be sure that all appropriate authors had timely copies to
review. Also to Kristi Bockting and the staff at WorldComp for the
monumental job of incorporating all author comments into the final "cam-
era-ready" manuscript. Our greatest thanks go to our wives, Anne Tum-
mala and Jean Rymaszewski, and Mary Ann Klopfenstein for their patience
and full support. We thank Bertrand Cambou, Motorola Senior Vice-
President for the insightful Foreword.

Rao R. Tummala
Eugene J. Rymaszewski
Alan G. Klopfenstein
Part 2

TABLE OF CONTENTS*

FOREWORD II-v
PREFACE II-vii
CONVERSION FACTORS II-xxvii
SUMMARY CONTENTS II-xxxi

PART 2. MICROELECTRONICS PACKAGING HANDBOOK:


SEMICONDUCTOR PACKAGING II-I
CHAPTER 7. MICROELECTRONICS PACKAGING-AN OVERVIEW 11-3
7.1 INTRODUCTION 11-3
7.1.1 Packaged Electronics 11-8
7.1.2 Packaging Functions and Hierarchy 11-11
7.1.3 Evolving Trends 11-14
7.1.3.1 Product Applications 11-14
7.1.3.2 Semiconductors 11-16
7.1.3.3 Future Packaged-Electronics Considerations 11-17
7.2 TECHNOLOGY DRIVERS 11-21
7.2.1 Wireability, Number of Terminals, and Rent's Rule 11-22
7.2.2 Electrical Design Considerations 11-24
7.2.2.1 General 11-25
7.2.2.2 Electromagnetic Properties of Signal Wire 11-26
7.2.2.3 Signal Degradation 11-27
7.2.2.4 Cross-Coupling Noise 11-27
7.2.2.5 Switching (M) Noise 11-28
7.2.2.6 Signal Reflections 11-29
7.2.3 Electrical Performance-Power Distribution 11-29
7.2.3.1 Rigorous Analysis 11-30
7.2.3.2 Very Rough Estimate of Acceptable Lead Inductance
L = L,... 11-31
7.2.4 Thermal Design Considerations 11-33
7.2.5 Reliability 11-35
7.2.6 Manufacturability and Quality 11-37
*See Summary of Contents for 3 part set, p. xxxi.
II-xii TABLE OF CONTENTS

7.2.7 Testability II-38


7.2.8 Memory Packaging I1-40
7.2.9 Personal Computers and Workstations II-42
7.2.10 Portable and Mixed-Signal Equipment II-44
7.2.11 High-Performance Processors I1-45
7.2.12 Optical Interconnects II -48
7.2.13 Summary of Packaging Performance Considerations and
Evolution II-48
7.2.13.1 Power-Time Product Theory I1-48
7.2.13.2 Reduction of UsReB by LSI Technology II-50
7.2.13.2 System LSI in Chips and Packages II-53
7.2.13.3 Media Delay Factor II-58
7.3 PACKAGING TECHNOLOGIES I1-62
7.3.1 Packaging Evolution II-63
7.3.1.1 Packaging in the Past II-67
7.3.1.2 Packaging in the Future II-71
7.4 CHIP-LEVEL INTERCONNECTIONS I1-71
7.4.1 Wirebond I1-72
7.4.2 TAB I1-74
7.4.3 Flip Chip II-76
7.5 FIRST-LEVEL PACKAGES II-78
7.5.1 Single-Chip Packages II-78
7.5.2 Chip-Scale or Chip-Size Packages II-82
7.5.3 Single-Chip Package Markets I1-86
7.5.4 Plastic Packages I1-88
7.5.4.1 Process Controls I1-89
7.5.5 Plastic PGA Technology I1-90
7.5.6 Lead-Frame Fabrication I1-90
7.5.7 Logic and Memory Applications II-91
7.5.8 Ceramic Packaging Technologies II-91
7.5.9 BGA Packages II-93
7.5.10 Package Trade-offs II-96
7.5.11 First-Level Multichip Packaging 11-99
7.5.11.1 Definition II-IOO
7.5.11.2 Functions of Multichip Package 11-100
7.5.11.3 Packaging Efficiency I1-10l
7.5.11.4 Electrical Performance II-I02
7.5.11.5 Reliability 11-103
7.5.11.6 Leverages II-104
7.5.11.7 Types of Multichips II-104
7.5.11.8 Cost II-104
7.5.11.9 Applications I1-106
7.6 PACKAGE-TO-BOARD INTERCONNECTIONS II-106
7.7 SECOND-LEVEL PACKAGES II-I 14
7.8 PACKAGING COOLING II-119
7.9 PACKAGE SEALING AND ENCAPSULATION II-123
7.10 BOOK ORGANIZATION AND SCOPE I1-125
7.11 REFERENCES II-127
TABLE OF CONTENTS II-xiii

CHAPTER 8. CHIP-TO-PACKAGE INTERCONNECTIONS II-129


8.1 INTRODUCTION II-129
8.2 CHIP-LEVEL INTERCONNECTION EVOLUTION II-132
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS II-I36
8.3.1 C4 History II-137
8.3.2 Materials II-I44
8.3.3 Design Factors II-146
8.3.4 Bump Fabrication Processes and Tools 11-148
8.3.4.1 Metal Mask Evaporation Technology 11-148
8.3.4.2 Electroplated Bumps 11-152
8.3.5 AssemblylRework 11-154
8.3.5.1 Self-Alignment 11-156
8.3.5.2 Conventional Flux-Assisted Joining II-I60
8.3.5.3 Fluxless Joining with Plasma-Assisted Dry
Soldering 11-161
8.3.5.4 Rework of Multichip Assemblies 11-165
8.3.6 C4 Reliability 11-170
8.3.7 Thermal Mismatch Reliability: Extensions 11-172
8.3.7.1 Design and Geometry 11-173
8.3.7.2 Solder Composition 11-174
8.3.7.3 Substrate Materials 11-176
8.3.7.3.1 Direct Chip Attach (DCA) to PWB with
Underfill 11-178
8.3.7.4 Other Reliability Issues 11-179
8.3.7.5 Alpha-Particle Emission 11-179
8.3.7.6 Burn-In/Known Good Die for Flip
Chip 11-182
8.3.7.7 Heat Dissipation with C4 11-183
8.3.8 C4-Future Trends II-I 85
8.4 WIREBONDING 11-186
8.4.1 Joining Technology 11-187
8.4.2 Wirebonding Joining Mechanism 11-188
8.4.2.1 Ultrasonic Welding II-190
8.4.2.2 Thermosonic Welding 11-193
8.4.2.3 Thermocompression Welding 11-195
8.4.3 Wirebonding History II-196
8.4.4 Wire-Bond Applications II-198
8.4.5 Materials 11-203
8.4.5.1 Bonding Wires II-203
8.4.5.2 Aluminum Wire with I % Silicon 11-203
8.4.5.3 Aluminum-Magnesium Wire II-204
8.4.5.4 Gold Bonding Wire II-205
8.4.5.5 Gold Wire Substitutes II-206
8.4.5.6 Die Attach Materials II-206
8.4.5.6.1 Solders II-206
8.4.5.6.2 Organic Adhesives 11-206
8.4.5.6.3 Glass Adhesives II-207
8.4.5.7 Bonding Tools: Capillaries and Wedges 11-207
II-xiv TABLE OF CONTENTS

8.4.6 Process Automation 11-209


8.4.7 Process Choice Decisions/Guidelines 11-210
8.4.7.1 Wire Type 11-210
8.4.7.2 Ball-Bonding Guidelines 11-211
8.4.7.3 Wedge-Bonding Guidelines 11-212
8.4.8 Evaluating/Optimizing the Process 11-213
8.4.8.1 Engineering Evaluation Methods (Visual) 11-213
8.4.8.2 Engineering Evaluation Methods (Mechanical) 11-215
8.4.8.3 Production Evaluation Methods 11-215
8.4.9 Wirebonding 2000: The Future 11-216
8.5 TAPE AUTOMATED BONDING 11-217
8.5.1 Introduction 11-217
8.5.2 TAB Tape 11-222
8.5.2.1 One-Layer Tape 11-222
8.5.2.2 Two-Layer Tape 11-222
8.5.2.3 Three-Layer Tape 11-224
8.5.2.4 Bumped Tape 11-226
8.5.2.5 Transferred Bump TAB 11-232
8.5.2.6 Two-Conductor Tape 11-233
8.5.2.7 Electrostatic Discharge Protection 11-236
8.5.3 Wafer Bumping 11-238
8.5.4 Inner Lead Bonding 11-240
8.5.4.1 Thermocompression Bonding 11-241
8.5.4.2 Gang Bonding Process Flow 11-242
8.5.4.3 Tape and Die Presentation Options 11-242
8.5.4.4 Single-Point Bonding 11-243
8.5.4.5 Laser Bonding 11-244
8.5.4.6 Factors Affecting ILB 11-246
8.5.4.7 ILB Evaluation 11-250
8.5.5 Outer Lead Bonding 11-250
8.5.5.1 Hot Bar OLB 11-252
8.5.5.2 Anisotropic Conductive Film 11-255
8.5.5.3 ILB and OLB Pitch and Lead-Count Trends 11-256
8.5.5.4 Effect of High 110 and Pitch on ILB and OLB 11-257
8.5.6 TAB Package Applications 11-258
8.5.6.1 Tape Ball-Grid Array 11-258
8.5.6.2 TapePak® 11-259
8.5.6.3 Pentium® TCP 11-260
8.5.6.4 ETA Supercomputer 11-263
8.5.7 Future Trends 11-265
8.6 PRESSURE CONNECTS 11-265
8.7 ADHESIVE BONDING 11-266
8.7.1 Bumps and Nonconducting Resins 11-266
8.7.2 Anisotropic Conductive Adhesives 11-268
8.8 ELECTRICAL PARAMETERS OF INTERCONNECTIONS 11-268
8.9 DENSITY OF CONNECTIONS 11-269
8.10 SUMMARY 11-270
8.11 REFERENCES 11-273
TABLE OF CONTENTS II-xv

CHAPTER 9. CERAMIC PACKAGING II-284


9.1 INTRODUCTION II-284
9.1.1 Unique Attributes of Ceramic Packages II-285
9.1.1.1 Unique Properties II-285
9.1.1.2 Highest Wiring Density II-287
9.1.1.3 Best Reliability and Hermeticity II-288
9.1.2 Drawbacks of Ceramics II-290
9.2 EARLY CERAMIC PACKAGING II-291
9.2.1 Ceramic Solid Logic Technology II-292
9.2.2 Ceramic Dual-in-Line Packages II-292
9.2.3 Advanced SLT II-294
9.2.4 Monolithic Systems Technology II-295
9.2.5 Thick-Film (Crossover) Technology II-295
9.2.6 Ceramic Packages with Thin Films 11-297
9.2.7 Earlier Multilayer Ceramic Technology II-297
9.2.8 Ceramic Chip Carriers, Flat Packs, and Hybrid Packages II-300
9.3 ALUMINA CERAMIC PACKAGING II-305
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND
APPLICATIONS II-308
9.4.1 Ceramic II-309
9.4.2 Greensheet II-311
9.4.3 Binder and Slurry II-312
9.4.4 Via Punching II-313
9.4.5 Screening II-316
9.4.6 Sintering II-319
9.4.7 Plating II-321
9.4.8 Chip Joining II-323
9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING II-324
9.5.1 Aluminum Nitride II-326
9.5.1.1 AlN Powder and Sintering II-327
9.5.1.2 Low-Temperature Firing II-327
9.5.1.3 AlN Properties II-327
9.5.1.4 Metallization of AlN II-329
9.5.1.5 A1N Applications II-329
9.5.2 Beryllium Oxide II-333
9.5.3 Mullite II-333
9.6 LOW-TEMPERATURE CERAMIC OR GLASS + CERAMIC
PACKAGING II-335
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC
PACKAGING II-340
9.7.1 IBM's Glass-Ceramic Substrate II-340
9.7.1.1 Glass-Ceramic Material II -341
9.7.1.2 Copper Powder and Paste Technology II-345
9.7.1.2.1 Copper Thick Film II-346
9.7.1.3 Personalization, Stacking, and Lamination
Technology II-348
9.7.1.3.1 Punching II-348
9.7.1.3.2 Screening II-350
II-xvi TABLE OF CONTENTS

9.7.1.3.3 Inspection II-350


9.7.1.3.4 Stacking and Lamination II-350
9.7.1.4 Glass-Ceramic/Copper Sintering Technology II-35 I
9.7.1.4.1 Binder Removal and Carbon
Oxidation II-35 1
9.7.1.4.2 Densification II-353
9.7.1.4.3 Dimensional Control II-353
9.7.1.5 Substrate Machining and Finishing Technology II-355
9.7.1.5.1 Postsinter Sizing II-356
9.7.1.5.2 Planarization II-357
9.7.1.5.3 Seal Flange Grind II-357
9.7.1.5.4 Polishing II-358
9.7.2 Fujitsu's Glass-Ceramic Substrate II-358
9.7.2.1 Glass-Cordierite System II-358
9.7.2.2 Glass-Alumina System II-359
9.7.2.3 Mainframe Application of Glass-Ceramic II-36 1
9.7.3 NEC's Lead Borosilicate-Alumina Glass + Ceramic II-362
9.7.4 Panasonic's Glass-Ceramic II-363
9.7.5 David Sarnoffs Glass-Ceramic on Metal (LTCC-M) II-365
9.7.6 Other Ceramic Packages II-366
9.8 CHIP ATTACHMENT AND THERMAL DISSIPATION OF CERAMIC
SUBSTRATES II-369
9.9 CERAMIC PACKAGE RELIABILITY II-37 1
9.10 FUTURE CERAMIC PACKAGING II-372
9.10.1 Future Low-Dielectric-Constant Ceramic Materials 11-373
9.10.1.1 Borosilicate Glass + Silica II-373
9.10.1.2 Porous Ceramics and Glasses II-376
9.10.2 Low-Cost Based on Large-Area Processing II-380
9.10.3 Fine-Line and Via Technologies II-382
9.10.4 Ceramic Ball-Grid Array and Chip-Scale Packages II-383
9.10.5 Integrated Ceramic Packaging with Passives II-384
9.10.5.1 Requirements for Wireless Communications II-385
9.10.5.2 Integrated Ceramic Size Requirements II-387
9.10.5.3 Some Examples of Passive Integration II-387
9.11 REFERENCES II-388

CHAPTER 10. PLASTIC PACKAGING II-394


10.1 INTRODUCTION II-394
10.1.1 Through-Hole-Mounted Devices II-398
10.1.2 Surface-Mounted Devices II-399
10.1.3 Plastic Versus Ceramic II-399
10.1.3.1 Size and Weight II-40l
10.1.3.2 Performance II-40l
10.1.3.3 Cost II-40l
10.1.3.4 Reliability II-403
10.1.3.5 Availability II-405
10.1.4 Summary II-406
TABLE OF CONTENTS II-xvii

10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-407


10.2.1 Resins 11-407
10.2.1.1 Fillers and Coupling Agents 11-410
10.2.1.2 Curing Agents and Accelerators 11-415
10.2.1.3 Stress-Release Additives 11-416
10.2.1.4 Flame Retardants 11-418
10.2.1.5 Mold-Release Agents 11-419
10.2.1.6 Ion-Getter Additives 11-420
10.2.1.7 Coloring Agents 11-420
10.2.2 "Low-Stress" Molding Compounds 11-420
10.2.3 Molding Formulation Process 11-423
10.2.4 Lead Frame Design, Materials, and Processes 11-425
10.2.4.1 Materials 11-426
10.2.4.2 Strength 11-427
10.2.4.3 Thermal Conductivity 11-427
10.2.4.4 Thermal Expansion 11-427
10.2.4.5 Mechanical Design 11-427
10.2.4.6 Lead Frame Fabrication 11-435
10.2.4.7 Plating 11-435
10.3 CHARACTERIZATION OF MOLDING-COMPOUND
PROPERTIES 11-437
10.3.1 Shrinkage Stresses 11-437
10.3.2 Coefficient of Thermal Expansion and
Glass-Transition Temperature 11-439
10.3.3 Mechanical Properties 11-440
10.3.4 Spiral Flow Length 11-443
10.3.5 Molding Compound Rheology from Molding Trial 11-443
10.3.6 Characterization of Molding-Compound Hardening 11-446
10.3.7 Gel Time 11-446
10.3.8 Adhesion 11-447
10.3.9 Polymerization Rate 11-447
10.3.10 Hardening 11-448
10.3.11 Postcure 11-449
10.3.12 Moisture Effects on the Epoxy Hardening Process 11-449
10.3.13 Thermal Conductivity 11-449
10.3.14 Electrical Properties 11-449
10.3.15 Chemical Properties 11-451
10.3.16 Flammability and Oxygen Index 11-451
10.3.17 Moisture Ingress 11-451
10.4 THE TRANSFER-MOLDING PROCESS 11-452
10.4.1 Molding Equipment 11-453
10.4.2 Transfer-Molding Process 11-454
10.4.3 Simulation 11-459
10.5 PACKING AND HANDLING 11-460
10.5.1 Considerations in Packing and Handling 11-460
10.5.1.1 Prevention of Moisture 11-461
10.5.1.2 Lead Damage and Solderability 11-463
10.5.1.3 Electrostatic Discharge Protection 11-464
II-xviii TABLE OF CONTENTS

10.5.2 Hierarchy of Packing 11-465


10.5.2.1 First Level Packing 11-465
10.5.2.2 Second Level Packing 11-467
10.5.2.3 Third Level 11-469
10.5.3 Handling of Dry Packages 11-470
10.5.3.1 Classification 11-471
10.5.3.2 Board-Level Rework of Moisture-Sensitive
Components 11-472
10.5.4 Environmental Considerations 11-473
10.6 QUALITY AND RELIABILITY 11-474
10.6.1 Classification of Failure Mechanisms 11-474
10.6.2 Analysis of Failures 11-476
10.6.2.1 Die Fracture 11-476
10.6.2.2 Loss of Chip Passivation Integrity 11-477
10.6.2.3 Die Metallization Corrosion 11-478
10.6.2.4 Metallization Deformation 11-479
10.6.2.5 Wire Sweep 11-479
10.6.2.5.1 Viscosity 11-480
10.6.2.5.2 High Flow Velocity 11-480
10.6.2.5.3 Unbalanced Flows in the Cavities 11-480
10.6.2.5.4 Void Transport 11-480
10.6.2.5.5 Late Packing 11-481
10.6.2.5.6 Filler Collision 11-481
10.6.2.6 Cratering of a Wire-Bond Pad 11-481
10.6.2.7 Loss of Wire Bond 11-482
10.6.2.8 Wire-Bond Fracture and Lift-Off 11-482
10.6.2.9 Wire Bond and Bond-Pad Corrosion 11-483
10.6.2.10 Lead Frame Corrosion 11-483
10.6.2.11 Lead-Frame Deadhesion and Delamination 11-484
10.6.2.12 Encapsulant Rupture 11-485
10.6.2.13 Encapsulant Fatigue Fracture 11-485
10.6.2.14 Package Cracking ("Popcoming") 11-485
10.6.2.15 Electrical Overstress and Electrostatic
Discharge 11-486
10.6.2.16 Soft Errors 11-487
10.6.2.17 Solder-Joint Fatigue 11-487
10.6.3 Ranking of Potential Failures 11-487
10.6.4 Comparison of Plastic and Ceramic Package
Failure Modes 11-489
10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES 11-490
10.7.1 Trends in Materials, Design, and Fabrication 11-490
10.7.1.1 Encapsulation Material Trends 11-490
10.7.1.2 Package-Design Trends 11-494
10.7.1.3 Package Fabrication Trends 11-495
10.7.2 Trends and Challenges in Circuit Card Assemblies 11-498
10.7.3 Standards and Requirements 11-498
10.8 HEAT TRANSFER IN PLASTIC PACKAGES 11-499
10.9 DIE ADHESIVES 11-501
10.10 REFERENCES 11-502
TABLE OF CONTENTS II-xix

CHAPTER 11. POLYMERS IN PACKAGING 11-509


11.1 INTRODUCTION 11-509
11.2 HISTORICAL PERSPECTIVE 11-510
11.3 POLYMER THIN-FILM REQUffiEMENTS FOR PACKAGING AND
SEMICONDUCTORS 11-512
11.3.1 Functional Compatibility and Reliability 11-513
11.4 POLYMERIC DIELECTRICS 11-515
ll.5 POLYIMIDES 11-516
11.5.1 Poly(amic alkyl ester) Development for Thin-Film
Application 11-518
11.5.1.1 Background 11-520
11.5.1.2 M-PAETE Synthesis 11-520
11.5.1.3 M-PAETE Solution Properties 11-521
11.5.2 Characterization Techniques Used for Evaluation of Thin-Film
Polymeric Materials 11-521
11.5.2.1 Mechanical Properties 11-522
11.5.2.2 Tensile Characterization 11-523
11.5.2.3 Process Stressing 11-523
11.5.2.4 Adhesion of Thin-Film Interfaces 11-524
11.5.2.4.1 Polyimide-to-Polyimide Adhesion 11-524
11.5.2.4.2 Polyimide-to-Glass-Cerarnic
Adhesion 11-524
11.5.2.4.3 Polyimide-to-Metal Adhesion 11-525
11.5.2.4.4 Metal-to-Polyimide Adhesion 11-525
11.5.2.5 Dynamic Stress Measurement 11-526
11.5.2.6 Dynamic Mechanical Properties 11-527
11.6 COMMERCIAL PREIMIDIZED POLYIMIDE FILMS 11-528
11.6.1 Desired Properties for Microelectronic Substrates 11-529
11.6.2 Manufacturing Methods 11-530
11.6.3 Commercial Sources, Types, and Chemical Compositions of
Preimidized Polyimide Films 11-531
11.6.3.1 Kapton® Polyimide Films 11-531
11.6.3.2 Apical® Polyimide Films 11-533
11.6.3.3 Upilex® Polyimide Films 11-533
11.6.3.4 Properties of Commercial Polyimide Films 11-533
11.7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS 11-539
11.7.1 Benzocyclobutenes 11-539
11.7.1.1 Electrical Properties 11-539
11.7.1.2 Mechanical Properties 11-540
11.7.1.3 Optical Properties 11-540
11.7.1.4 Water Absorption 11-540
11.7.1.5 Thermal and Thermo-oxidative Stability 11-541
11.7.1.6 Planarization 11-541
11.7.1.7 Adhesion 11-543
11.7.1.8 Thin Film Stress 11-543
11.7.1.9 Metal Migration 11-543
11. 7.1.1 0 Cyclotene 3000 series Processing: 11-543
11.7.1.11 Incorporation of Passive Components 11-546
11.7.1.12 Applications 11-546
II-xx TABLE OF CONTENTS

11.7.2 Polyquinolines 11-546


11.7.3 Olefin Dielectrics 11-548
11.8 PHOTOSENSITIVE POL YIMIDES 11-549
11.8.1 Introduction II-549
11.8.2 Historical Background 11-551
11.8.3 Thin-Film Processing of PSPIs 11-553
11.8.3.1 Polyimide Pattern Fonnation Process using Ester-type
PSPI II-553
11.8.3.2 Comparison with Non-PSPI 11-555
11.8.4 Covalently Bonded Ester-Type PSPI II-556
11.8.4.1 Chemistry and Photopackage: Preparation of
Precursor Polymers II-556
11.8.4.2 Alternate Approaches to the Preparation of
Ester-Type PSPIs 11-557
11.8.4.3 Processing 11-558
11.8.4.3.1 Solution Preparation and Properties II-558
11.8.4.3.2 Photolithographic process 11-559
11.8.4.3.3 Pimel ® Curing II-562
11.8.4.3.4 Pimel Properties II-563
11.8.4.3.5 Functional PSPIs II-565
11.8.4.3.6 Low-Thennal-Expansion PSPIs II-565
11.8.4.3.7 I-Line photoimageable Ester-Type
PSPI II-570
11.8.4.3.8 Base-Catalyzed PSPI II-570
11.9 IONIC BONDED-TYPE PSPIs II-570
11.9.1 Chemistry and Photopackage 11-570
11.9.2 Processing Photoneece (Exposing. Developing. and
Curing) II-572
11.9.3 Lithographic Properties and Characterization 11-574
11.10 PREIMIDIZED PHOTOSENSITIVE POLYIMIDES 11-580
11.10.1 Chemical Characterization 11-580
11.1 0.2 Features 11-581
11.10.3 Origin of Photo-cross-linking 11-583
11.11 POSITIVE PHOTOSENSITIVE POL YIMIDES 11-587
11.11.1 Positive Working PSPI II-587
11.11.2 Properties of Positive Working Photosensitive PI 11-588
11.11.3 Outlook for Positive Photosensitive II-589
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-589
11.12.1 Physical and Chemical Properties II-589
11.12.2 Electrical and Mechanical Properties of Photo BCB 11-590
11.12.3 Processing Photosensitive-BCB II-591
11.12.3.1 Surface Preparation II-591
11.12.3.2 Coating II-592
11.12.3.3 Pre-Exposure Bake 11-592
11.12.3.4 Exposure II-594
11.12.3.5 Solvent Develop 11-595
11.12.3.6 Rework II-600
11.12.3.7 Cure 11-601
11.12.3.8 Descum 11-601
TABLE OF CONTENTS II-xxi

11.12.3.9 Substrate Dicing II-601


11.12.3.10 Applications of Photo BCB II-602
11. I3 PHOTOSENSITIVE EPOXY II-603
11.13.1 Photodefinable Cyanate-esters and Epoxies II-603
11.13.2 Intrinsically photodefinable epoxy II-604
11.14 MICROELECTRONIC APPLICATIONS OF PSPIs II-605
11.14.1 Chip Coating Applications II-605
11.14.1.1 Alpha-Particle Protection Layers II-605
11.14.1.2 Buffer Coating for Stress Relief II-605
11.14.1.3 Thin-Film Interconnect II-606
11.14.2 Packaging Interlayer Applications II-606
11.14.2.1 Interlayer Applications in High-Density Multichip
Modules II-606
11.14.2.2 Applications Using Covalently Bonded Ester-Type
PSPIs II-606
11.14.2.3 Applications Using Ionic Bonded Type PSPIs II-608
11.14.2.4 Application Using Nonshrinking Preimidized
PSPI II-608
11.14.2.5 Optical Waveguides II-61 I
11.14.2.6 Other Applications II-611
11.15 RESEARCH DEVELOPMENTSIFUTURE POLYMERS II-611
11.15.1 Ultrathin PSPIs (LB Membrane) II-61 I
11.16 SUMMARY AND FUTURE CHALLENGES II-612
11.17 REFERENCES II-613

CHAPTER 12. THIN-FILM PACKAGING II-624


12.1 INTRODUCTION II-624
12.1.1 Thin-Film Package Wiring II-630
12.1.2 Thin-Film Module I/O Connections II-634
12.1.3 Power Supply and Heat Removal II-635
12.2 ELECTRICAL PERFORMANCE II-637
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES II-641
12.3.1 Size of Thin-Film and Thick-Film Packages II-641
12.3.2 Cost of Thin-Film and Thick-Film Packages II-645
12.3.3 Single-Layer Thin-Film Packages II-646
12.3.4 Types of Multilayer Thin-Film Structures II-648
12.3.4.1 Planar stacked studs II-649
12.3.4.2 Nonplanar staggered vias II-65 I
12.3.4.3 Electrical Trade-offs II-653
12.4 THIN-FILM MATERIALS AND PROCESSES II-656
12.4.1 Thin-Film Substrates II-656
12.4.2 Dielectric Materials II-659
12.4.2.1 Organic Materials and Their Properties II-660
12.4.2.2 Inorganic Materials and Their Properties II-664
12.4.3 Thin-Film Processes II-666
12.4.3.1 Polymer Processing II-67 I
12.4.3.1.1 Deposition of Polymer Thin Films II-671
12.4.3.1.2 Via Generation Approaches II-677
II-xxii TABLE OF CONTENTS

12.4.3.2 Metallization Processes 11-693


12.4.3.2.1 Wiring-Level Metallization 11-695
12.4.3.2.2 Via Metallization 11-698
12.4.3.2.3 Terminal Metals and Their
Processing 11-699
12.4.3.3 Planarization 11-703
12.4.3.4 Relative Cost Comparison of Thin-Film
Processes 11-707
12.5 YIELD/COST CONSIDERATIONS 11-712
12.5.1 Repair of Thin-Film Packaging 11-712
12.5.2 Leverage of Repair Technologies 11-713
12.5.3 Repair Methodologies 11-714
12.5.3.1 Defect Detection and Repair 11-715
12.5.3.2 Repair of Shorts 11-716
12.5.3.3 Repair of Opens 11-718
12.5.4 Alternative Repair Methodologies 11-723
12.5.4.1 Redundancy 11-723
12.5.4.2 Rerouting 11-724
12.5.5 Performance and Reliability 11-724
12.6 RELIABILITY 11-725
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-726
12.7.1 IBM 11-726
12.7.1.1 Thin-Film Packaging Evolution 11-726
12.7.1.2 Multilayer Thin-Film Implementation 11-728
12.7.1.3 Single-Layer and Multilayer Single-Chip
Packages 11-728
12.7.1.4 Thin-Film Structure and Process for Mainframe
Computers 11-730
12.7.1.4.1 Aluminum-Polyimide Multilevel Thin-Film
on Silicon 11-730
12.7.1.4.2 Stacked Stud Copper-Polyimide 11-733
12.7.1.4.3 Conformal Via Copper-Polyimide 11-734
12.7.2 Siemens: Microwiring Thin-Film Technology for
Mainframes 11-736
12.7.2.1 Microwiring Technology 11-737
12.7.2.2 Sequential Multilayer Thin-Film Process 11-737
12.7.2.2.1 Reinforcement Substrate 11-739
12.7.2.2.2 Signal, Power, and Redistribution
Layers 11-740
12.7.2.2.3 Chip-Mounting Layer 11-741
12.7.3 Hitachi: Thin-Film Packaging for Mainframe Computers 11-742
12.7.4 nCHlP: Packaging with AI/Si0 2 on Silicon Substrates 11-745
12.7.4.1 The Si02 Dielectric 11-746
12.7.4.1.1 Thermal Conductivity 11-747
12.7.4.1.2 Process Cost 11-748
12.7.4.1.3 Inertness and Impermeability 11-748
12.7.4.1.4 Mechanical Properties 11-749
12.7.4.1.5 Consistency and Control 11-751
12.7.4.1.6 SiOl Ruggedness 11-751
TABLE OF CONTENTS II-xxiii

12.7.4.2 Key Electrical and Thermal Characteristics ll-752


12.7.4.3 nClOoo SICB Feature Sizes 11-752
12.7.4.4 MCM Package Assembly 11-753
12.7.5 Digital: VAX. 9000 Multichip Packaging ll-753
12.7.5.1 The HDSC Design and Process 11-754
12.7.5.1.1 The HDSC Design 11-754
12.7.5.1.2 The HDSC Core Processes 11-755
12.7.5.1.3 The HDSC Assembly Process 11-756
12.7.5.2 Evolution of High-Performance Computer
Packaging 11-757
12.7.5.3 Evolution of High-Performance Packaging
Technology 11-760
12.7.5.3.1 MCU-A 11-761
12.7.5.3.2 MCU-B 11-761
12.7.5.3.3 MCU-C 11-762
12.7.5.4 Technology Summary 11-763
12.7.6 NEC: Thin-Film Packaging for Mainframes and
Supercomputers 11-764
12.7.6.1 Polyimide-Ceramic Substrate for High-Performance
System 11-764
12.7.6.2 Wafer Scale Substrate for GaAs Packaging 11-766
12.7.7 MicroModule Systems: Thin-Film Packaging Compatible with IC
Processing 11-767
12.7.8 Other Commercial Schemes 11-768
12.8 EMERGING COMMERCIAL TECHNOLOGIES IN THIN-FILM
PACKAGING 11-769
12.8.1 IBM's Surface Laminar Circuitry 11-769
12.8.2 Sheldahl's Roll-to-Roll Process 11-770
12.8.3 GE's Conformal Multichip-Flex 11-772
12.8.4 DYCOstrate Thin-Film Technology 11-773
12.8.5 IBM's Integrated Rigid-Flex Thin-Film Package 11-774
12.8.6 IBM's Thin-Film Transfer Technology 11-777
12.9 INTEGRATED PASSIVES IN THIN-FILM PACKAGE 11-780
12.9.1 Current Development and Applications of
Integrated Passives 11-782
12.9.1.1 Resistors 11-782
12.9.1.2 Capacitors 11-784
12.9.1.3 Inductors 11-786
12.9.2 Future Directions of Integrated Passives in
Thin-Film Package 11-788
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND
PROCESSES 11-788
12.10.1 Polymeric Materials 11-793
12.10.2 Large-Area Polymer Deposition 11-795
12.10.3 Via Formation 11-797
12.10.3.1 Laser Ablation 11-797
12.10.3.2 Photosensitive Polymers 11-799
12.1 0.4 Planarization 11-800
12.10.5 Electroless Copper Plating 11-800
II-xxiv TABLE OF CONTENTS

12.11 ACKNOWLEDGMENT II-804


12.12 REFERENCES II-804

CHAPTER 13. PACKAGE ELECTRICAL TESTING II-814


13.1 INTRODUCTION II-814
13.2 SUBSTRATE TEST II-817
13.3 SUBSTRATE TEST METHODS II-82 I
13.3.1 Capacitance Testing 11-821
13.3.1.1 Test Implementation II-823
13.3.1.2 Limitations II-825
13.3.2 Resistance and Continuity Testing 11-825
13.3.3 Electron Beam Testing II-829
13.3.3.1 Test Implementation II-829
13.3.3.2 Limitations 11-832
13.3.4 Electrical Module Test II-833
13.3.4.1 Test Implementation 11-834
13.3.4.2 Limitation 11-836
13.4 COMPARISON OF TEST METHODS II-836
13.5 CONTACTS AND PROBES II-838
13.6 OPENS AND SHORTS TEST IMPLEMENTATION II-846
13.7 FUTURE ISSUES AND CHALLENGES IN SUBSTRATE
TESTING II-854
13.8 FUNCTIONAL TESTS 11-855
13.8.1 Boundary Scan Test 11-857
13.8.2 Built-In Self-Test 11-859
13.8.3 Functional Testing of MCMs II-860
13.8.4 A Structured Testability Approach II-861
13.8.5 Assembled-MCM Test and Diagnosis Procedure 11-862
13.8.5.1 Integrity and Identity Check 11-862
13.8.5.2 Substrate Test 11-862
13.8.5.3 Functional At-Speed Test II-863
13.8.5.4 Performance Test II-863
13.8.5.5 Parametric Test 11-864
13.9 FUTURE ISSUES AND CHALLENGES IN FUNCTIONAL
TESTING II-864
13.10 RECENT AND FUTURE DEVELOPMENTS 11-865
13.10.1 Substrate Testing 11-865
13.10.2 Functional Testing II-867
13.11 REFERENCES 11-867

CHAPTER 14. PACKAGE SEALING AND ENCAPSULATION 11-873


14.1 INTRODUCTION II-873
14.2 HERMETIC VERSUS NONHERMETIC PACKAGING 11-876
14.2.1 Hermetic Packaging 11-877
14.2.2 Nonhermetic Packaging 11-878
14.3 IC FAILURE MECHANISMS 11-880
14.3.1 Sources of Moisture 11-884
TABLE OF CONTENTS II-xxv

14.4 PACKAGE SEALING AND ENCAPSULATION 11-887


14.4.l Package Sealing and Encapsulations Material
Requirements 11-888
14.5 TYPES OF HERMETIC PACKAGES 11-890
14.5.l Metal Packages 11-890
14.5.2 Ceramic Packages 11-891
14.5.2.1 Ceramic-Dua1-in-Line Package 11-891
14.5.2.2 Hard Glass Package 11-892
14.5.2.3 Other Ceramic Packages 11-893
14.6 TYPES OF HERMETIC SEALS 11-894
14.6.1 Fused Metal Seals 11-894
14.6.1.1 Soldering 11-894
14.6.1.2 Brazing 11-898
14.6.1.3 Welding 11-898
14.6.1.4 Glass Sealing 11-901
14.6.l.5 Glass-Sealing Methods 11-906
14.6.2 Gasket Seals 11-907
14.7 TESTING OF HERMETIC PACKAGES 11-910
14.7.l Leak Testing 11-911
14.7.1.1 Gross Leak Testing 11-911
14.7.l.2 Fine Leak Testing 11-912
14.7.2 Moisture Measurement and Monitoring II-912
14.8 RELIABILITY TESTING 11-916
14.8.1 Correlation of 85°C/85% rh to HAST 11-916
14.9 RECENT ADVANCES IN SEALING AND ENCAPSULATION II-921
14.10 FUTURE DEVELOPMENTS II-926
14.11 REFERENCES 11-927

GLOSSARY AND SYMBOLS II-931


AUTHORS' BIOGRAPHIES 11-977
INDEX II-999
CONVERSION FACTORS

Length
1 m = 1010 A I A = 10- 10 m
1 m = 109 nm 1 nm = 10-9 m
1 m = 1()6 J,Lm 1 J,Lm = 10- 6 m
1 m = 1()3 mm I mm = 10- 3 m
1 m = 102 em I em = 10- 2 m
1 mm = 0.0394 in. I in. = 25.4 mm
1 em = 0.394 in. 1 in. = 2.54 em
1 m = 3.28 ft 1 ft = 0.3048 m

Area
1 m2 = 104 em2 1 em 2 = 10-4 m2
1 mm 2 = 10- 2 em2 1 em 2 = 102 mm 2
1 m2 = 10.76 fe I ft2 = 0.093 m2
1 cm 2 = 0.1550 in. 2 1 in. 2 = 6.452 cm 2

Volume
1 mJ = 106 em l I em 3 = 10- 6 mJ
1 mm J = 10- 1 em J 1 em l = 103 mm l
1 m l = 35.32 ft l 1 ftl = 0.0283 ml
1 eml = 0.0610 in.J 1 in.l = 16.39 em3

Mass
1 Mg = J()3 kg I kg = 10- 3 Mg
1 kg = J()3 g Ig = 10- 1 kg
1 kg = 2.205 Ibm 1 Ibm = 0.4536 kg
I g = 2.205 X 10- 3 Ibm I Ibm = 453.6 g

Density
1 kglml = 10- 3 glem J I glem 3 = 103 kglm 3
I Mglm 3 = 1 glem 3 I glem 3 = I Mglm 3
I kglm 3 = 0.0624 Ib m/ft3 I Ib m/ft3 = 16.02 kglm 3
1 glcml = 62.4 Ib m/ft 3 I Ib m/ft 3 = 1.602 x 10- 2 glcm3
1 glem J = 0.0361 Ibm/in. J 1 Ibm/in. l = 27.7 gleml
II-xxviii CONVERSION FACTORS

Unit conversion factors begin on previous page.

force
I N = lOS dynes 1 dyne = 10- 5 N
1 N = 0.2248 Ibr 1 Ibr = 4.448 N

Stress
I MPa = 145 psi 1 psi = 6.90 x 10-3 MPa
I MPa = 0.102 kglmm 2 I kglmm 2 = 9.806 MPa
I Pa = 10 dynes/cm2 1 dyne/cm 2 = 0.10 Pa
1 kglmm 2 = 1422 psi I psi = 7.03 X 10- 4 kglmm 2

fracture Toughness
1 psi '\lID. = 1.099 X 10- 3 MPaYm 1 MPa Ym = 910 psi '\lID.
(nergy
1J = Wergs 1 erg = 10- 7 J
I J = 6.24 X to'8 eV 1 eV = 1.602 x 10- 19 J
1J = 0.239 cal I cal = 4.184 J
1J = 9.48 X 10- 4 Btu 1 Btu = 1054 J
IJ = 0.738 ft-Ib r 1 ft-Ib r = 1.356 J
I eV = 3.83 x 10- 20 cal I cal = 2.61 x 10 19 eV
1 cal = 3.97 x 10- 3 Btu I Btu = 252.0 cal

Power
1 W = 0.239 calls 1 calls = 4.184 W
I W = 3.414 Btu/h 1 Btu/h = 0.293 W
1 calls = 14.29 Btu/h 1 Btulh = 0.070 calls

Viscosity
I Pa-s = 10 P I P = 0.1 Pa-s

Temperature, T
T(K) = 273 + T(°C) T(°C) = T(K) - 273
T(K) = i[TeF) - 32] + 273 T(OF) = I[T(K) - 273] + 32
T(°C) = i[T(°F) - 32] T(°F) = I[T(°C)] + 32

Specific Heat
I J/kg-K = 2.39 x 10- 4 cal/g-K I cal/g_OC = 4184 J/kg-K
I J/kg-K = 2.39 x 10-4 BtulIbm-oF 1 BtulIbm-oF = 4184 J/kg-K
I cal/g-OC = 1.0 BtulIbm-oF 1 Btu/lbm-oF = 1.0 cal/g-K

Thermal Conductivity
I W/m-K = 2.39 x 10- 3 callcm-s-K 1 callcm-s-K = 418.4 W/m-K
1 W/m-K = 0.578 Btuift-h-oF 1 Btuift-h_oF = 1.730 W/m-K
I cal/cm-s-K = 241.8 Btuift-h-oF 1 Btuift-h-oF = 4.136 X 10- 3 cal/cm-s-K
CONVERSION FACTORS II-xxix

Unit Abbrevladons

A = ampere in. = inch N = newton


A = angstrom J = joule nm = nanometer
Btu = British thermal unit K = degrees Kelvin P = poise
C = Coulomb kg = kilogram Pa = pascal
°C = degrees Celsius Ibr = pound force s = second
cal = calorie (gram) Ibm = pound mass T = temperature
cm = centimeter m = meter ~m = micrometer
eV = electron volt Mg = megagram (micron)
of = degrees Fahrenheit mm = millimeter W = watt
ft = foot mol = mole psi = pounds per square
g = gram MPa = megapascal inch

SI Muldple and Submultiple


Prefixes

Factor by Which
Multiplied Prefix Symbol

109 giga G
106 mega M
IQ3 kilo k
10- 2 centiU c
IO- J milIi m
10- 6 micro ~
10- 9 nano n
10- 12 pico p
• Avoided when possible.
Parts 1, 2, and 3

SUMMARY OF CONTENTS
PART 1. MICROELECTRONICS PACKAGING HANDBOOK:
TECHNOLOGY DRIVERS
CHAPTER 1. MICROELECTRONICS PACKAGING-AN OVERVIEW
TECHNOLOGY DRIVERS
PACKAGING TECHNOLOGIES
CHIP-LEVEL INTERCONNECTIONS
FIRST-LEVEL PACKAGES
PACKAGE-TO-BOARD INTERCONNECTIONS
SECOND-LEVEL PACKAGES
PACKAGING COOLING
PACKAGE SEALING AND ENCAPSULATION
BOOK ORGANIZATION AND SCOPE

CHAPTER 2. PACKAGE WIRING AND TERMINALS


THE WIRING MODEL AND ITS VARIABLES
RESULTS OF CHIP CIRCUIT PLACEMENT AND WIRING VERSUS MODEL
PREDICTIONS
WIREABILITY OF HIGHER-LEVEL PACKAGES
CONCLUSIONS AND CHALLENGES FOR FUTURE WORK

CHAPTER 3. PACKAGE ELECTRICAL DESIGN


CIRCUIT ATTRIBUTES
SIGNAL DISTRIBUTION
POWER DISTRIBUTION
NOISE CONTAINMENT
LOW-LOSS TRANSMISSION SYSTEMS
PACKAGE MODELING
THE DESIGN SPACE
SYSTEM IMPACT
PACKAGE IMPROVEMENTS
APPENDIX: CALCULATION OF PEAK AVERAGE CURRENT SLEW RATE

CHAPTER 4. HEAT TRANSFER IN ELECTRONIC PACKAGES


HEAT-TRANSFER FUNDAMENTALS
EXPERIMENTAL TECHNIQUES AND INSTRUMENTATION
II-xxxii SUMMARY OF CONTENTS

TYPICAL COOLING DESIGNS


RECENT PACKAGING HEAT TRANSFER DEVELOPMENTS

CHAPTER 5. PACKAGE RELIABILITY


RELIABILITY METROLOGY
CONTACT RESISTANCE
ENVIRONMENTAL INTERACTIONS
THERMAL MISMATCH AND THERMAL FATIGUE
MECHANICAL LOADING
TRIBOLOGICAL DEGRADATION
HEAT-TRANSFER DEGRADATION
APPENDIX-PROBABILITY DENSITY TABLE

CHAPTER 6. PACKAGE MANUFACTURE


QUALITY IN PRODUCT AND PROCESS DESIGN
PRODUCT QUALIFICATION
MANUFACTURING LINE MODELING AND SIMULATION
PROCESS CONTROL
WORLDWIDE QUALITY STANDARD-ISO 9000
APPENDIX: THE 20 ELEMENTS OF ISO 9001

PART 2. MICROELECTRONICS PACKAGING HANDBOOK:


SEMICONDUCTOR PACKAGING

CHAPTER 7. MICROELECTRONICS PACKAGING-AN OVERVIEW


TECHNOLOGY DRIVERS
PACKAGING TECHNOLOGIES
CHIP-LEVEL INTERCONNECTIONS
FIRST-LEVEL PACKAGES
PACKAGE-TO-BOARD INTERCONNECTIONS
SECOND-LEVEL PACKAGES
PACKAGING COOLING
PACKAGE SEALING AND ENCAPSULATION
BOOK ORGANIZATION AND SCOPE

CHAPTER 8. CHIP-TO-PACKAGE INTERCONNECTIONS


CHIP-LEVEL INTERCONNECTION EVOLUTION
FLIP-CHIP SOLDER-BUMP CONNECTIONS
WIREBONDING
TAPE AUTOMATED BONDING
PRESSURE CONNECTS
ADHESIVE BONDING
ELECTRICAL PARAMETERS OF INTERCONNECTIONS
DENSITY OF CONNECTIONS

CHAPTER 9. CERAMIC PACKAGING


EARLY CERAMIC PACKAGING
ALUMINA CERAMIC PACKAGING
SUMMARY OF CONTENTS II-xxxiii

STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS


RECENT DEVELOPMENTS IN CERAMIC PACKAGING
LOW-TEMPERATURE CERAMIC OR GLASS + CERAMIC PACKAGING
STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING
CHIP ATTACHMENT AND THERMAL DISSIPATION OF CERAMIC
SUBSTRATES
CERAMIC PACKAGE RELIABILITY
FUTURE CERAMIC PACKAGING

CHAPTER 10. PLASTIC PACKAGING


MOLDING COMPOUNDS AND LEAD FRAME MATERIALS
CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES
THE TRANSFER-MOLDING PROCESS
PACKING AND HANDLING
QUALITY AND RELIABILITY
FUTURE DEVELOPMENTS IN PLASTIC PACKAGES
HEAT TRANSFER IN PLASTIC PACKAGES
DIE ADHESIVES

CHAPTER 11. POLYMERS IN PACKAGING


HISTORICAL PERSPECTIVE
POLYMER THIN-FILM REQUIREMENTS FOR PACKAGING AND
SEMICONDUCTORS
POLYMERIC DIELECTRICS
POLYIMIDES
COMMERCIAL PREIMIDIZED POLYIMIDE FILMS
HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS
PHOTOSENSITIVE POL YIMIDES
IONIC BONDED-TYPE PSPIs
PREIMIDIZED NONSHRINKING PHOTOSENSITIVE POL YIMIDES
POSITIVE PHOTOSENSITIVE POL YIMIDES
PHOTOSENSITIVE BENZOCYCLOBUTENE
PHOTOSENSITIVE EPOXY
MICROELECTRONIC APPLICATIONS OF PSPIs
RESEARCH DEVELOPMENTSIFUTURE POLYMERS

CHAPTER 12. THIN-FILM PACKAGING


ELECTRICAL PERFORMANCE
THIN-FILM VERSUS THICK-FILM PACKAGES
THIN-FILM MATERIALS AND PROCESSES
YIELD/COST CONSIDERATIONS
RELIABILITY
COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING
EMERGING COMMERCIAL TECHNOLOGIES IN THIN-FILM PACKAGING
INTEGRATED PASSIVES IN THIN-FILM PACKAGING
FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES

CHAPTER 13. PACKAGE ELECTRICAL TESTING


SUBSTRATE TEST
SUBSTRATE TEST METHODS
II-xxxiv SUMMARY OF CONTENTS

COMPARISON OF TEST METHODS


CONTACTS AND PROBES
OPENS AND SHORTS TEST IMPLEMENTATION
FUTURE ISSUES AND CHALLENGES IN SUBSTRATE TESTING
FUNCTIONAL TESTS
FUTURE ISSUES AND CHALLENGES IN FUNCTIONAL TESTING
RECENT AND FUTURE DEVELOPMENTS

CHAPTER 14. PACKAGE SEALING AND ENCAPSULATION


HERMETIC VERSUS NONHERMETIC PACKAGING
IC FAILURE MECHANISMS
PACKAGE SEALING AND ENCAPSULATION
TYPES OF HERMETIC PACKAGES
TYPES OF HERMETIC SEALS
TESTING OF HERMETIC PACKAGES
RELIABILITY TESTING
RECENT ADVANCES IN SEALING AND ENCAPSULATION
FUTURE DEVELOPMENTS

PART 3. MICROELECTRONICS PACKAGING HANDBOOK:


SUBSYSTEM PACKAGING

CHAPTER 15. MICROELECTRONICS PACKAGING-AN OVERVIEW


TECHNOLOGY DRIVERS
PACKAGING TECHNOLOGIES
CHIP-LEVEL INTERCONNECTIONS
FIRST-LEVEL PACKAGES
PACKAGE-TO-BOARD INTERCONNECTIONS
SECOND-LEVEL PACKAGES
PACKAGING COOLING
PACKAGE SEALING AND ENCAPSULATION
BOOK ORGANIZATION AND SCOPE

CHAPTER 16. PACKAGE-TO-BOARD INTERCONNECTIONS


SURFACE-MOUNTED PACKAGES
SOLDER AND SOLDER JOINING
FINE-PITCH TECHNOLOGY
SOLDER PASTE AND ITS DEPOSITION
RELIABILITY
SURFACE-MOUNT ARRAY CONNECTIONS
LEAD-FREE SOLDERS
CONDUCTIVE ADHESIVES
FUTURE DEVELOPMENT

CHAPTER 17. PRINTED-WIRING BOARD PACKAGING


PRINTED-CIRCUIT STRUCTURES AND PROCESSES
INTERCONNECTION TECHNOLOGY CONSIDERATIONS
SUMMARY OF CONTENTS II-xxxv

MATERIALS AND PROCESSES


RELIABILITY AND CONTROLS
RECENT DEVELOPMENTS IN PRINTED-CIRCUIT BOARDS
FUTURE TRENDS IN PRINTED-CIRCUIT BOARDS

CHAPTER 18. COATED-METAL PACKAGING


GENERAL CONSIDERATIONS
METALS AND INSULATOR SUBSTRATES
INSULATORS
METALLIZATION METHODS
SUBSTRATE FABRICATION AND ASSEMBLY PROCESSES
PROPERTIES OF COATINGS
DESIGN RULES
NEW OPPORTUNITIES

CHAPTER 19. CONNECTOR AND CABLE PACKAGING


TYPES OF CONNECTORS
CONNECTOR REQUIREMENTS
APPROACHES TO MCM LEVEL-2 INTERCONNECTIONS
OPTICAL-FIBER CONNECTORS
CONNECTOR RELIABILITY
CONNECTOR STANDARDS ACTIVITIES
CABLES
CABLE SIGNAL THEORY
FUTURE CONNECTORS AND CABLES

CHAPTER 20. PACKAGING OF OPTOELECTRONICS WITH ELECTRONICS


OPTOELECTRONIC DEVICES AND INTERCONNECTION MEDIA
TECHNIQUES FOR INTEGRATING OPTOELECTRONICS WITH ELECTRONICS
SYSTEM APPLICATIONS FOR OPTOELECTRONICS INTEGRATED WITH
ELECTRONICS
Part 2
MICROELECTRONICS
PACKAGING HANDBOOK:
SEMICON DUCTOR
PACKAGING

RAO R. TUMMALA-GEORGIA TECH


EUGENE J. RYMASZEWSKI-RENSSELAER
ALAN G. KLOPFENSTEIN-AGK ENTERPRISES
7
MICROELECTRONICS
PACKAGING-AN OVERVIEW

EUGENE J. RYMASZEWSKI-Rensse/aer
RAO R. TUMMALA-Georgia Tech
TOSHIHIKO WATARI-NEC

7.1 INTRODUCTION

Packaged electronics is the embodiment of all electronic equip-


ment-calculators, personal computers (pes), mainframe computers, tele-
phones, television, and so forth. It consists of the active components, such
as integrated circuit (IC) chips, flat-panel and cathode-ray-tube displays,
loudspeakers, and so on. The active components are interconnected, sup-
plied with electric power and housed in packaging. Success in a very
competitive marketplace hinges on superior performance and price.
Performance is a broad term which encompasses the ability to per-
form a multitude of functions-word processing with embedded graphics
or spreadsheet calculations, airline reservations, banking transactions, and
so on-within a certain time, preferably without imposing an unacceptable
11-4 MICROELECTRONICS PACKAGING-AN OVERVIEW

$/MIPS

10M

CD
u 1M
c
as
...0E
't: 100k

--.,
CD
Do pes Adjusted

0
(,)
10k ,,
,,
,,
1k ..

100------~------~~~--~~----~----~
1960 1970 1980 1990 2000 2010
Year
Figure 7-1. Trends of the System Price-to-Performance Ratio

wait on the users. It is task dependent, so only the general trends will be
discussed throughout this book. In the last several decades, the price (in
$) for perfonnance (measured in millions instructions per second-MIPS)
decreased dramatically as shown in Figure 7-1 which is derived from
Ref. 1.
This gain with PC was in improving the perfonnance at relatively
constant cost and price, thus increasing their value every 2-3 years. These
humble PCs, priced at a (very) few thousand dollars, easily outperfonn
(with certain tasks) the yesteryear's mainframes which had been priced
in millions of dollars. And they are much easier to use.
The mainframe improvements now (the second half of 1990s) mainly
result from the cost reductions enabled by higher levels of integration on
the ICs and in the packaging, thus reducing the size and number of parts,
and from simplifying the packaging complexity (e.g., switch from the
water to air cooling).
Figure 7-2 illustrates the interdependency between processor perfor-
mance and its clock frequency. Another strong variable is the number of
cycles per instruction [2] which depends on the computer design in general
7.1 INTRODUCTION 11-5

1009. 1 CII

10 CII
100

100 CII

1 10 100 1000
Clock Frequency [MHz]

Figure 7-2. System Performance Dependency on Cycle Time and Cycles Per
Instruction, and the Input-Output Data Rate

and on the number of its circuits in particular, as is discussed in Section


7.2.11. The high performance (number of operations per unit of time,
usually 1 s) requires processing and storing of data within short time
intervals, on the order of nanoseconds (ns or 10.9 s). Such high-performance
equipment must be of small physical size because of the finite time
a signal takes to travel through an interconnection-1 ns through an
interconnection 10-30 cm (4-12 in.) long. The evolved, and still evolving,
semiconductor and packaging technologies have enabled this progress to
occur and to continue.
The above scenario is grossly incomplete without highlighting an-
other, perhaps even more significant, paradigm shift-the emergence of
11-6 MICROELECTRONICS PACKAGING-AN OVERVIEW

technologies for the portable systems (Fig. 7-3) with wired (via modem
and telephone lines) and wireless (like cellular telephones) connections
to a rapidly expanding network of interconnected computers, already
providing near-unlimited amounts of information and vast opportunities
for person-to-person as well as group communications. The World Wide
Web name is not an exaggeration.
As already indicated in Figure 7-2 and discussed in more detail in
Section 7.2.11, the communications between the computers, and even
within a powerful single computer, demand very high data rates, on the
order of 10 megabytes (MB) per second (1 byte = 8 bits-binary units)
transmitted in and out of the processor per 1 MIPS; see Figure 7-18. The
past, present, and future mainframes maintain this relationship to balance
their performance and communications capabilities because of the heavy
communications demand of all tasks they perform. The seemingly imbal-
anced ever-increasing performance ofPCs without a correspondingly high
communication bandwidth constitutes the differentiating property between
the two. The PC performance is used to continually enhance the user
friendliness and to perform computation-intensive, but not communica-
tion-intensive, tasks such as pattern manipulation (e.g., maps, images,
voice recognition, and so forth).
Customer expectations are also very high indeed in terms of quality
(work perfectly out of the box), reliability (no failures during useful life),
and rapid and inexpensive service in the unlikely event it is needed. Two
other main branches of electronics, consumer and communications, have
almost completed their transitions from analog to digital circuits wherever
practical. Thus, the long-predicted convergence of these three main
branches of electronics (computer, consumer, communications) is well
underway. The packaging continues to provide strong competitive lever-
ages as it has in the past, in the context of totally different IC chips and
much broader and still widening applications.
Thus, the second edition of this book endeavors to preserve and
build on the timeless fundamentals presented in the first edition and to
update its timely content by splicing in the various advances of the state
of the art. It must be noted, however, that designs for various applications
undergo much more rapid changes and suffer obsolescence much sooner
than at the time of work on the first edition. Consequently, the applications
chapters are no longer included. The interested readers are referred to
many good periodicals, conferences, and workshops. The changing world
of electronics required a different approach to this second edition. First, its
chapter authors represent the broad international professional community.
Second, the book is now being divided into three parts to make it more
complete and comprehensive without approaching a cubical form factor
and to enhance its use as a study book.
The three parts are basically the original book, but now they concen-
.,,~
~ ."Fa
CommunicatiOD8 ." c.ou.r PIoon. ."MuoS-
." T _ ." Os*:
.£~~ ."Coo......,.
."F_
."T.....m..

Handwriline
ReOO(!llitioo
.. Voice RecowUti"..
."F_
.,,~ "PI8iDa
.,,~ .. Na.iphOll
.....
." Piooac .. FuModem
." P..-.
,. Speech To Tat

Fron! End BoIlDi.:s]


"'Ho,..~
0.....
."
." v_............
.,,~

."No........
~
."R..io
"''IV
.,,(;.1..... _
."r..
V.rude Om Link ." Tolioc
." Pooiaa

Figure 7·3. Portable Electronics (Courtesy of NEMI)


11-8 MICROELECTRONICS PACKAGING-AN OVERVIEW

trate on technology drivers and technologies. The first part, "Technology


Drivers," begins with an overview followed by an updated in-depth treat-
ment of the key packaging design aspects-topological, electrical, thermal,
reliability, and manufacturability. The second part, the "Semiconductor
Packaging," focuses on the extremely important and very diverse technol-
ogy options (materials and processes) to package semiconductor chips
in single-chip (SCM) and multichip (MCM) modules. The third part,
"Subsystem Packaging," completes the coverage of the technology options
available to the designer to form a system or a major subsystem. New
chapters were added to address manufacturing considerations, electrical
testing, polymers used in packaging, and optical packaging.
Another aspect of packaging is microelectronics versus general elec-
tronics, in which a great deal of components are heavy and/or bulky, such
as power-supply transformers, filter capacitors, cathode-ray-display tubes,
and flat-panel displays. This book, as its title indicates, concerns itself
mainly with packaging of the wide range of microelectronic components.
The next subsection of this chapter, "Packaged Electronics," high-
lights the two pivotal roles of the electronic packaging-the ability to
support and preserve the on-chip performance, and the main limiting
factors to such an ability. It also raises many questions triggered by
these initial highlights to introduce the subsequent section on packaging
definition, to describe its functions and hierarchy, and to give a brief
overview of these books. The detailed chapter-by-chapter description is
given in Section 7.10, "Book Organization and Scope."

7.1.1 Packaged Electronics

As already mentioned in Section 7.1, the electronics packaging must


support the original design objective/intent of its IC chips, yet it may fail
to do so to a greater or lesser degree, either by limiting the realization
of the intended IC performance potential or by "diluting" the on-chip
performance, or by both. To illustrate these points for digital applications,
let us consider an assembly (e.g., in a workstation) of 750,000 computing
elements, all of which having the same value of the "speed-power product"
(actually energy), assumed to be 0.1 pJ (0.1 x 10. 12 J).
Let us select three operating points (signal delay at an average power
dissipation) roughly representative of performance (but not the power)
of yesterday's, today's, and tomorrow's technologies: 10 ns at 10 ~W, 1
ns at 100 ~W, 0.1 ns at 1 mW, respectively. During each system cycle,
the signal being processed has to traverse 20 of these computing elements.
Furthermore, there are three packaging options, each assumed to be a
square with a 1-cm edge typical of an IC chip, 6-cm edge typical of a
multichip module, and a 36-cm edge typical of a printed wiring board
(all these terms will soon be discussed). Insulating material of the signal
7.1 INTRODUCTION 11-9

Table 7-1. Attributes of Nine Workstation Implementations

IC Chip Packaged Chips


Multichip Module Printed Wiring Board

Edge dimension lem 6em 36 em


Manhattan td 0.13 ns 0.8 ns 4.8 ns
Area I cm2 36 cm2 1296 cm2

"Yesterday's designs," 10 ns
Wire delay adder -0% 0.4% 2.5%
IOIlW Clock -5 MHz 4.98 MHz 4.88 MHz
7.5 W Power density 7.5 W/cm2 0.21 W/cm2 0.006 W/cm2

"Today's designs," 1 ns
Wire delay adder 0.5% 3.8% 19.3%
100 IlW Clock 49.7 MHz 48.1 MHz 40.3 MHz
75 W Power density 75 W/cm2 2.1 W/cm2 0.057 W/cm2

''Tomorrow's designs," 0.1 ns


Wire delay adder 6.1% 28% 67%
ImW Clock 463 MHz 357 MHz 147 MHz
750W Power density 750W/cm2 20.8 W/cm2 0.58 W/cm2

interconnections has a dielectric constant of 4, an approximate value of


many actual materials.
In a fairly representative situation, the signal will originate by the
circuits located in one comer of the square package and will have to
travel ultimately to the diagonally opposite comer via paths mostly parallel
to the edges, commonly called "Manhattan delay" (from the streets and
avenues of the borough of Manhattan of New York City). In such an
intentionally oversimplified design, the cycle time is determined by the
sum total of the two signal propagation times:

Through the 20 computing elements, connected in series (on-chip


delay)
Manhattan (along two packaging edges) wire delay for Er =4, which
is Vl5 nstcm.

Thus, there are nine technology design options presented in detail


in Table 7-1. The three columns correspond to the designs with each of
the three edge dimensions (a single IC chip and two packaging structures).
The three rows reflect the designs with each of the three performance
levels representative of the three time frames. The individual cells of
Table 7-1 list the resulting clock frequency and power density, and the
% adder of the wire delay. The individual cells of Table 7-1 present a
11-10 MICROELECTRONICS PACKAGING-AN OVERVIEW

more complete picture by listing the total signal delay of the on-chip
computing elements (mostly logic circuits), the cycle path delay (and the
% adder of the wire delay), the corresponding clock frequency, and the
power density.
Contemplating the key system attributes of this table, we note minus-
cule effects ofthe wire delay on the cycle time for alllO-ns ("yesterday's")
designs, regardless of the edge dimension. The added delay is 2.5% or
less of the total, and the power densities are easily managed, as will
become more apparent in Section 7.2.4 (note that this subject is fully
treated in Chapter 4). Section 7.2.3 will show another aspect of minimal
packaging effects.
The 1-ns ("today's") designs begin to show noticeable Manhattan
wire delay adder (- 20% ) for the 36-cm-edge case and a rather challenging
power density (75 W/cm2) for the 1-cm edge. Again, Section 7.2.3 will
show similar situation.
The O.l-ns ("tomorrow's") designs are most severely affected by
the packaging: up to 67% added wire delay (3.4x performance degradation)
on the 36-cm-edge package and, on the 1-cm-edge package, a 750-
W/cm2 power density-on the order of the power density of a nuclear
blast 1 mile from its center, as shown in Figure 7-12. Such a power
density cannot be supported by any practical packaging design. Either a
circuit power reduction (and correspondingly greater signal delay) or an
increased edge dimension (greater Manhattan delay), or both, are required.
Consequently, they preclude any practical full utilization of the theoreti-
cally possible 2-ns cycle time, or 500-MHz clock, and leave one to hope
that the future ICs will achieve the same O.l-ns performance at the circuit
power levels lower than 1 mW (i.e., will have a speed-power product
less than 0.1 pJ). The power distribution for these tomorrow's ICs will
also need significant design innovations.
In these examples, only the power density and wire delay were
considered, which are, in fact, the primary variables. Further considera-
tions raise some additional intriguing questions.
The 1-cm edge is most likely to be that of an IC chip itself. The
larger edges imply a somewhat uniform distribution of the circuits over
the larger area, which is possible only with partitioning of the total circuit
content into several IC chips. How many of them? Just a few or quite a
few (a high number)? What are the consequences of such partitioning in
terms of chip-to-package interface on the number and arrangements of
the signal, ground and power terminals, the on-packaging interconnections
between them, and the thermal interfaces/paths to the equipment exterior?
What are the topological aspects, the physical dimensions, and the
electromagnetic characteristics of the interconnections? How does one
deal with the chip-to-package interface regions? What are the materials,
in what shapes and dimensions, and how to process them during manufac-
turing?
7.1 INTRODUCTION 11-11

What are the values of stray electric and magnetic fields, and the
resultant capacitances and inductances, and what impact do they have on
the system performance and/or reliability of its operation (intermittent
failures-the most onerous kind)? How well can it all be manufactured
and what reliability (hard failure rates and lifetimes) to expect?
How to integrate this portion of a larger data processing system with
the rest of it in a competitively advantageous manner? What performance
leverage, or penalty, results from employing either less or more than the
750,000 computing elements exemplified above?
This book endeavors to deal with all of these questions, and raise
and answer a few more. This chapter sets the stage for the entire book
(all three parts) by introducing and highlighting the topics of all subsequent
chapters. This concludes with the packaging functions and structural hier-
archy, and presents the evolving trends of semiconductors and packaging.
The next major section, Section 7.2, deals with the two sets of
driving forces-the design considerations (topological, electrical, thermal,
reliability, manufacturability, and testability) and the applications (memo-
ries, pes and workstations, portable electronics, mainframes, optical inter-
connects) which optimally integrate the various technologies. Section
7.2.13 concludes the description of driving forces, gives a bird's-eye view
of evolution of the entire range of digital-electronics-based products, and
sets the stage for the next six sections to deal with the various packaging
technologies (structures, materials, and processing) and their trends and ap-
plications.
Section 7.3 introduces the packaging technology and describes its
trends. Section 7.4 describes the various technologies for chip-to-package
connections. Section 7.5 focuses on the single-chip and multichip mod-
ules-the first-level packages. Section 7.6 covers the package-to-board
interconnections. Section 7.7 deals with the second-level packaging tech-
nologies. The packaging cooling is the subject of Section 7.8. Section 7.9
handles the packaging sealing and encapsulation.
Section 7.10 summarizes this book with a schematic overview and
a detailed chapter-by-chapter description of all three parts. It is followed
by listing of references.

7.1.2 Packaging Functions and Hierarchy

Electronic packages contain many electrical circuit components-


up to several millions or even tens of millions-mainly transistors assem-
bled in integrated circuit (IC) chips, but also resistors, diodes, capacitors,
and other components. To form circuits, these components need intercon-
nections. Individual circuits must also be connected with each other to
form functional entities. Many of the intercircuit connections are steadily
migrating into the Ie chips as the levels of integration continue to increase.
Mechanical support and environmental protection are required for the Ies
11-12 MICROELECTRONICS PACKAGING-AN OVERVIEW

and their interconnections. To function, electrical circuits must be supplied


with the electrical energy which is consumed and converted into the
thermal energy (heat). Because all circuits operate best within a limited
temperature range, packaging must provide an adequate means for removal
of heat.
Thus, the packaging has four major functions:
• Signal distribution, involving mainly topological and electromag-
netic consideration
• Power distribution, involving electromagnetic, structural, and materi-
als aspects
• Heat dissipation (cooling), involving structural and materials consid-
eration
• Protection (mechanical, chemical, electromagnetic) of components
and interconnections

as illustrated in Figure 7-4. They are discussed throughout this series


of books.
Arguably, the thin-film wires (interconnections) within semiconduc-
tor IC chips are an important packaging component. Increased integration
had migrated onto the IC chip at a steadily increasing percentage of the
wires previously provided by the packaging structures outside the chips.
Packaged components must be manufactured and assembled at af-
fordable costs. Repairs must be few, quick, and inexpensive. In many
applications, it is also desirable/mandatory to enable the addition or re-
arrangement of functional features (e.g., expansion of storage capacity,
addition of a coprocessor or a modem, or even a processor upgrade).
All interconnections on an IC chip (even though some ofthem might
have been part of packaging in previous-lower level of integration-
technology) or all internal structures of any other component (e.g., a
crystal or those of discrete capacitors and inductors) are not considered
packaging in the common parlance. Thus, the packaging begins at the
interfaces to these ICs and other components: mechanical, to provide
support and, if needed, an enhanced thermal path to facilitate heat transfer;
and electrical, for power and signal connections, whose electromagnetic
properties also play an important, but not always desirable, role.
These interfaces may be the exterior surfaces of an IC chip (die)
itself-the common term being a bare chip-often in the context of a
"direct chip attach." Or, as it is still often the case, the IC chip has an
enclosure, commonly referred to as a single-chip module (SCM), or a
chip carrier, or a header. In either case, it is considered a first-level
package. This is an extremely important and very diversified level of
packaging which fills the entire Part 2, Semiconductor Packaging. This
book also deals with a direct attach of several unencapsulated (bare) chips
to a substrate of various materials produces a multichip module (MCM)
7.1 INTRODUCTION 11-13

Signal Distribution
__ -=-_-=x

~ ~~~Signal

Power Distribution
~~~

Heat Dissipation

Package Protection

Figure 7-4. Four Major Functions of a Package

or multichip packaging (MCP). MCMs are often assembled on PWB along


with SCMs and other discrete components.
The groups of chips (packaged in SCMs), along with other compo-
nents such as capacitors, resistors, inductors, filters, switches, and optical
and RF components, are assembled on a second-level packaging-cov-
ered in Part 3-usually copper-clad sheets of epoxy-glass laminates.
After patterning of the metallic sheet followed by mechanical drilling and
lamination to form interconnections, such structures are commonly called
"printed wiring boards" (PWB) or card. See Figure 7-5.
The next level packaging (3rd or higher) may be the outer shell
of a small piece of equipment (e.g., a hand-held calculator, a portable
phone, or a pluggable unit to supplement or to enhance functions of a
11-14 MICROELECTRONICS PACKAGING-AN OVERVIEW

W1rebond Rlp-Chlp (C4) TAB


Heat

Lead
Removal
1 Heat
Removal
Frame 'Io--fop.-fl elI!:l!!~~Lead

tpower

Figure 7-5. Packaging Levels, Interconnections, and Technologies

larger equipment-a pluggable PC card with a hard disk or with a wireless


modem). In a larger equipment (desktop or floor standing), several cards
are plugged into a PWB. Such cards are sometimes referred to a "daughter
cards" and the PWB as "mother boards." There may be several boards
within a box/frame oflarge equipment such as a workstation or, especially,
a mainframe. As already mentioned, the continuing increases in the number
of circuits and/or bits per IC chip tend to simplify the packaging hierarchy.
In addition to the relatively small and light IC chips, many electronic
applications require bulkier and heavier components such as power trans-
formers, large electrolytic capacitors, disk drives, display tubes, and so
on. They are mounted separately and are electrically connected to the rest
of the electronics via various cables. Cables also are used to connect
boards either to each other or to the rest of the equipment, including its
external terminals (e.g., a PC to its monitor, keyboard, mouse, printer,
and, via modem, to the house telephone line).

7.1.3 Evolving Trends


7.1.3.1 Product Applications
The main driver is the price for performance shown in Fig. 7-1 for
the digital data processing. The general trends of the attributes of packaged
electronics are shown in Figure 7-6. Note how lower prices enhance
production volume.
7.1 INTRODUCTION 11-15

.".
U
U
·c
A.

$3
·r····r··········· ..···········....··
Consumer Electronia

1970 1980 1990 2000


Years

Figure 7-6. Attributes of Next-Generation Electronics

The mainframe improvements were progressing at a slow rate until


emergence of the microprocessors based on the complementary-metal-
on-silicon (CMOS) technology. This technology was initially too slow to
challenge the bipolar-based emitter (driver output) complementary logic
(ECL), as shown in Figure 7-7. The advent of better (smaller dimensions)
photolithography greatly benefited the CMOS-based logic-circuit perfor-
mance, at a better speed-power product. It had also enabled higher circuit
density leading to lower circuit costs. These higher levels of integration
on the ICs lead to higher packaged circuit density, thus reducing the size
and number of parts. Lower power densities have simplified the packaging
complexity, mainly by the switch from water cooling to air cooling.
The clock frequencies of the CMOS-based microprocessors (used

1000 ..------r------,~---.....,

---3>.
1990 1995 2000

Year

Figure 7-7. Performance Trends of CMOS and Bipolar Logic Technologies


11-16 MICROELECTRONICS PACKAGING-AN OVERVIEW

mainly in PCs) have caught up with the mainframes and are chasing
the supercomputers (in the range of 200-350 MHz). The CMOS-based
mainframes are expected to do so before the year 2000. It must be noted,
however, that it is extremely difficult to compare PC performance (essen-
tially a single-user system) with the mainframes, which have a totally
different instruction set and serve a vast number of users. In Figure 7-1,
1 mainframe MIPS was assumed to equal approximately 10-100 PC
MIPS. The main differences between the two are in the instruction sets
(tasks) and in the communications bandwidth-the number of bits per
second transmitted into and out of the processor for each MIPS, as will
be presented in detail in Figure 7-18 of Section 7.2.11.
The zero power consumption of the CMOS circuits when in the
standby mode (no data processing) contrasts sharply with the near-constant
power consumption of the ECL technology. It has enabled portable, bat-
tery-powered, equipment (e.g., PCs, cellular phones) with reasonably long
operating times. The trend here is toward smaller dimensions, lower
weight, more functionality, and very attractive prices, leading to high
volumes.
Low technology costs and the better understanding and control of
the failure mechanisms resulted in highly reliable products (see, for exam-
ple, Fig. 7-13), with the new applications, such as automotive electronics.

7.1.3.2 Semiconductors

Until the mid-fifties, most of the electronic circuits were produced


from individual discrete components: vacuum tubes, resistors, capacitors,
and, notably, discrete wires. The first step toward integration occurred
with the emergence of the printed wiring board (PWB). An insulating
layer, a fraction of 1 in. thick, was clad with metal layers on one or
both sides. Interconnections were photolithographically patterned. Wire
terminals of the still discrete components were inserted into prepunched
holes in the PWB and soldered, all at once.
It all began to change rapidly in late sixties, with the advent of
integrated circuits (IC) which contained in a little semiconductor (mostly
silicon) "chip" an ever-increasing number of planar devices (mainly,
transistors, field-effect-FET, or bipolars) isolated from each other, with
their terminals on the surface of the chip. These devices were intercon-
nected by "wires" photolithographically formed from one, and later more,
layers of metallic thin film, predominantly aluminum. Enormous density
improvements (Fig. 7-8) resulted from advances in the photolithography,
for instance more than a fourfold increase in storage cells per chip every
3 years or approximately a 100-fold increase every decade. This mass
production of electronic circuits dramatically reduced (and continues to
reduce) their cost. Table 7-2 highlights the evolution of the key IC parame-
7.1 INTRODUCTION 11-17

1 Memory
p ackage
.-
nn •. V J'\'.

u Single Level
c.
g>
.-
= .. .. Integrated Modul

..
:2 i!' ..." (SLIM)
(J 0 oJ Memory increase 1.5lyear 100'"" '
.. E ' 1n ,
GO GO
c.:E E
N =
MPU increase 1.35/year ./ P""./' ...-
I!! ' .!!
ON I!!

...
- E 0
CJ 'U
.!! 17 ./
:!!"ii
as:t::
.. m
I- . . . .
·iii
t:
CO 1M
1M i/ ./ ' /
... 0 I-
o ....
.. GO 0
GO .a .. / /'
.a E GO
E ::I .a ~ .,/

~
:i Z § 100
Z

/ ~ ./

1nl
~ V

1 ..,. I~ ==Microprocessor/Logic
1K V '" Memory (DRAM)
1975 1980 1985 1990 1995 2000 2005
Year

Figure 7-8. Microelectronics Density Trends (Adapted from Ref. 22.)

ters and attributes from 1992 to 2007. Figure 7-9 offers comparisons of
the major ICs-bipolars (emitter-current logic, ECL), complementary
metal on silicon (CMOS), BICMOS, GaAs-in terms of their major attri-
butes.
These semiconductor advancements have put, and are still putting,
an ever-increasing burden on the packaging-in terms of interconnecting,
powering, cooling (Fig. 7-10), and protection-to keep pace with and
provide sufficient support to the ICs as it is highlighted in Table 7-1 and
will be discussed in more detail in Section 7.2. An added demand to
reduce cost is becoming stronger and more pervasive since the shift in
the application thrusts from the large (and still expensive, but getting less
so) mainframes and supercomputers to desktop PCs and workstations
(shown in Fig. 7-1) and portable electronics. Portable electronics also
demand very energy-efficient designs to prolong operating time on one
battery charge.

7,1.3.3 Future Packaged-Electronics Considerations


The quest for the further cost reduction opportunities now focuses
on those attributes/elements of technology whose contributions to the
total cost are becoming disproportionaly large (e.g., pins, individual chip
carriers, individual discrete components, extensive rework to replace faulty
11-18 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7·2. Semiconductor Evolution

1992 1995 1998 2001 2004 2007

Feature size (11m) 0.5 0.35 0.25 0.18 0.12 0.10


Gates/chip 300K 800K 2M 5M 10M 20M
Bits/chip
DRAM 16M 64M 256M 1G 4G 16G
SRAM 4M 16M 64M 256M 1G 4G
Wafer processing cost ($/cm') $4.00 $3.90 $3.80 $3.70 $3.60 $3.50
Chip size (mm')
Logic/unprocessor 250 400 600 800 1000 1250
DRAM 132 200 320 500 700 1000
Wafer diameter (mm) 200 200 200-400 200-400 200-400 200-400
Defect density (defects/cm') 0.1 0.05 0.03 0.01 0.004 0.002
No. of interconnect levels-logic 3 4-5 5 5--6 6 6-7
Maximum power (W/die)
High performance 10 15 30 40 40-120 40-200
Portable 3 4 4 4 4 4
Power supply voltage (V)
Desktop 5 3.3 2.2 2.2 1.5 1.5
Portable 3.3 2.2 2.2 1.5 1.5 1.5
No. IIOs 500 750 1500 2000 3500 5000
Performance (MHz)
Off chip 60 100 175 250 350 500
On chip 120 200 350 500 700 1000

components). These cost-reducing efforts may disrupt the continuity of


the technology evolution; they may even traumatize the established tech-
nology-flow and business-flow patterns; for example, the replacement of
the surface-mount technology (SMT) for discrete components (mainly
resistors and capacitors) by embedding them into the substrate to enable
higher density, better performance, and lower costs.
Equally traumatic are transitions from one technological approach
to another which is vastly different. Examples from the past are replace-
ments of the ferrite memory-core arrays with the integrated-circuit (IC)
dynamic and static random-access memories (DRAMs and SRAMs), or
advent of the microprocessors and application-specific integrated circuits
(ASICs). In the examples just cited, the cost and performance improve-
ments went approximately hand in hand.
Introduction of the dense liquid-cooled multichip modules (MCM),
with well over 100 bipolar IC chips directly attached (without benefit of
individual packages-SCMs) to multilayer ceramic substrates, provided
very significant performance leverages, but it was not a low-cost approach.
Yet, it lasted for several generations of IBM mainframes and was emulated
by mainframe and supercomputer producing competitors; see Figures 7-32
7.1 INTRODUCTION 11-19

6 ~CMOS
I:LZJ BICMOS
~GaAa
~ 5
'a _ECL
:J
1loll 4
~
~
0 3
Z

GATES POWER SPEED PERFORMANCE COST


pg CHIP UNIPROCESSOR CRIT. PA11f UNIPROCESSOR UNIPROCESSOR

Figure 7-9. Semiconductor Technology Comparisons

and 7-33. The direct attach of bare chips to MCM substrates is still resisted
by some suppliers of electronics technology, at least in part due to the
business separation between the makers of the ICs, the makers of the
PWBs, and the makers of the end products or subproducts.
The traditional single-chip-module approach, when extended to giga-
scale integration (GSI, with large IC edge dimensions-25 mm and be-
yond) and, especially, combined with decreasing cross-sectional geome-
tries of interconnections (down to 0.1 /-lm wide and thick), is likely to be
competitively inferior to the alternative implementation in MCMs or MCPs
on two most significant counts: cost and performance. The following is
a generalized qualitative examination of both aspects.

Manufacturing costs: Inscribing large chips within a circular wafer


produces a disproportionately smaller number of them than a
smaller chip-size alternative. Given identical defect density on
wafers, a higher percentage of larger chips will be defective.
Consequently, the cost of good chips will suffer disproportion-
ately. The chip yield problems will be further aggravated by the
need for additional metal levels. The cost differential is likely to
be large enough to offset the additional MCMIMCP costs.
Performance: Smaller device geometries necessitate smaller contact
areas and narrower interconnections. To avoid "degenerate" aspect
ratios, interconnections also have to become thinner. These two
factors will reduce the cross-sectional area at a faster rate than
11-20 MICROELECTRONICS PACKAGING-AN OVERVIEW

the device size and interconnection lengths reductions. The smaller


area increases the resistance and, therefore, reduces bandwidth,
unless the interconnection length is also reduced by the same
factor. In fact, the interconnection lengths become greater with
the increasing IC chip size, further aggravating the performance
degradation. An attempt to widen the interconnections will lead
to an even higher number of metal layers to render the chip
wireable-hardly a desired by-product because of the added com-
plexity and increased number of defects.

Note that many GSI chips consist of "functional islands," and busses:
16-bit or 32-bit (or more) sets of interconnections between them. The
wiring within functional islands is at least partially random. A properly
partitioned GSI chip may be readily subdivided into smaller chips. The
interconnections between the subchips will have increased performance
when their cross-sectional geometries are increased: an easy task, if done
on a physically separate chip carrier (MCM or MCP). The chip-to-carrier
interface must present minimum electromagnetic loading. Indeed, it can
be done with the advanced flip-chip technologies. Performance of such
nets will be vastly superior to those implemented on a single chip of
reasonably complexity-with a limited number of thin-film metal layers.
As most of the web-surfers already know, the wait inflicted by the
interconnection's bandwidth is usually much longer and, therefore, less
tolerable than that caused by the signal round-trip delay. This bottleneck
is very typical for distributed computing (actually data processing). Ulti-
mately, it is also gated by the affordable price/cost. To continue with this
example, the service provider accepts a regular telephone line with its
severe data-rate limitation, but will be happy to provide a data link of
(much) greater bandwidth at a correspondingly higher price.
In high-speed data processing, both the round-trip delay and the
bandwidth are important. The shorter round-trip delay is achieved by
denser (usually a smaller area for the same electronic content) technology.
Denser interconnections have smaller cross-sectional areas and, conse-
quently, reduced bandwidth simply because the cross-sectional area scales
down with the square of the scaling factor while the distance reduces
linearly. Such a dimensional scaling-down process has been continuing
since the early seventies, as already indicated. The bandwidth reductions
were barely noticeable, if at all, until recently,when detailed design and
performance simulation work on multi-GHz clock systems required ex-
tremely wide busses (on the order of 1000 lines) between the processor
and its temporary (cache) memory sections; see Figure 7-24 and Refs. 3
and 4. For a single processor configuration, a multilayer planar packaging
structure is manufacturable. A multiprocessor configuration will probably
choke on a submarginal data-transfer rate between the processors and
7.2 TECHNOLOGY DRIVERS 11-21

their caches. Optical connections are coming to the rescue, with their
much greater bandwidths.
As already mentioned in Section 7.1, a great deal of digital data
processing in the mid-nineties is microprocessor based. The microproces-
sors are often contained within a single chip (characteristically, such
was the case in the seventies, eighties, and early nineties) or the higher
performers are partitioned into a small group of chips, typically between
two (Intel Pentium-Pro) to about nine (eight in the newer IBM workstations
[5] and Fig. 7-20).
This represents two competing schools of thought-and ways of
doing business: single-chip design, with chips projected to grow rather
large; and multichip design, with several smaller chips tightly packaged
on an MCM or MCP.
Although the MCM approach had been the technology of choice for
the mainframes, its use in PCs is still emerging (in 1996). A study reported
in Ref. 6 substantiates the above conclusions with a quantitative analysis.
Such studies confirm the general applicability of the insights gained from
R&D of the high-end mainframe and supercomputer packaged electronics,
as discussed in detail in Section 7.2.13.
A multichip approach seems the only practical choice when heteroge-
neous semiconductors are a necessity. The most obvious case is with
wireless communications, with the silicon-based ICs handling the digital
signal processing and the compound semiconductors-based (e.g., gallium
arsenide) devices handling the radio-frequency (RF) transmitting and re-
ceiving tasks [7]. The SCM carriers on a PWB for such ICs are, in fact,
inferior solutions to the MCMs in terms of higher packaging volume
(premium for portable equipment) and poorer perfonnance. Thus, there
is a strong drive for dense, low-cost interconnections capable of providing
high degree of electromagnetic separation between the digital and RF
signal and power interconnections. The detailed packaging technology
trends are given in the introduction to Section 7.3.
Yet another set of challenges stems from distributed computing
(actually data processing) with its need for rapid and faultless transmission
of enormously large data/information volumes. Each challenge involves
several key parameters or characteristics which often counteract each
other-improvements in one often cause deterioration of another. Further-
more, the current technologies are approaching numerous fundamental
limits set by the properties of materials and by the structural geometries
(form factors and dimensions) of packaging.

7.2 TECHNOLOGY DRIVERS


A successful package design will satisfy all given application require-
ments at an acceptable design, manufacturing, and operating expense. As
11-22 MICROELECTRONICS PACKAGING-AN OVERVIEW

a rule, application requirements prescribe the number of logic circuits


and bits of storage (possibly along with other components, such as
switches, lights, displays, etc.) which must be packaged-interconnected,
supplied with appropriate electric power, kept within proper temperature
range, mechanically supported, and protected against the environment.
Ever-increasing integration at the chip level toughens demands on the
number and density of signal terminals and interconnections. The resulting
smaller physical dimensions tax both the properties of materials, and the
production assembly and test techniques. These challenges are further
compounded by the need to preserve or even to enhance the thermal and
electromagnetic characteristics while maintaining or improving the quality
and reliability.
There are two major sets of technology drivers-one governs the
technology selection for a particular application (e.g., mainframe proces-
sor, its main memory, PC), the other set is derived from the key design
considerations: topological (wiring, number of terminals), electrical (sig-
nal and power interconnections), thermal, reliability, manufacturability,
and testing. The technology selection for a particular application integrates
and optimizes these design considerations and specific parameter values-
power density, electrical characteristics, number and density of terminals
and interconnections, and so forth. In this section, the design considerations
are reviewed first, following the subject of Chapters 2-6 (the first part).
Test design consideration are covered in Part 2, Chapter 13 and design
considerations for optical are covered in Part 3, Chapter 20. Sections
7.2.8-7.2.12 highlight the various generic application considerations-
memories, PCs and workstations, portable equipment, mainframes, and
optical interconnects. Section 7.2.13 gives a bird's-eye view of the evolu-
tion of the entire range of products with regard to several key parameters.
7.2.1 Wireability, Number of Terminals, and Rent's Rule
The total number of terminals and wires at packaging interfaces
(module pins, card connectors) is a major cost factor. As will be seen
later, the electrical characteristics of the interconnections are also the key
performance determinators, along with the limits in the ability to supply
power and remove heat. They can also pose serious reliability challenges.
Essentially all of the electronic equipment components and subas-
semblies have two distinct sets of terminals (individual electrical connec-
tion)-signals and power. Common to both is usually the "ground" termi-
nal(s), a carryover from the old radios which had to be connected to an
areal (antenna) and to the ground, literally a spike driven into the ground
or other suitable artifact such as a water pipe. House electrical wiring,
telephones, and so forth still have such a connection to prevent an electrical
shock to a person in contact with an electrically conducting part of an
appliance. However, in electronic applications, the "ground" designates
7.2 TECHNOLOGY DRIVERS 11-23

a neutral reference voltage to be considered with a much greater care


than frequently done (for more details see Section 7.2.3).
Signal interconnections and terminals constitute the majority of con-
ducting elements, especially in low-cost packaging. Other conductors
supply power and provide ground or other reference voltages. Packaging
designed for the high-performance circuits must contain more elaborate
power and ground connections to ensure acceptably low levels of electrical
interference (noise). This significantly complicates the design and, there-
fore, raises the cost.
The number of terminals supporting a group of circuits is strongly
dependent on the function of this group. The smallest number goes with
the memories (storage of binary coded information) because the stream
of information can be as limited as a single bit, requiring only one data
line. The address data depend on the storage capacity. For example, a 4K
bit (4096 = 212) random access memory (RAM) will require 12 address
lines (terminals) and 16M RAM bits will be satisfied with 24 address
terminals. Controls (e.g., read, write, clock) require a few more lines.
Thus, the packaging of memories is usually simple and inexpensive.
Exactly the opposite is the case with the groups of logical circuits
which result from random partitioning of a computer. Repeated studies
of the relationship between the number of circuits within an enclosure
(module, card, board, gate of Figure 7-10) and the number of terminals
required have reconfirmed findings of Rent [8,9,12], as discussed in depth
in Chapter 2. The relationship is known as Rent's rule [8] which can be
generally expressed as 1:

N=KMP [7-1]

where N is the number of required input-output terminals, K is the average


number of terminals used for an individual logic circuit, M is the total
number of circuits, and p (0 ~ p ~ 1) is a constant, called Rent's constant.
A detailed discussion is given in Section 2.2.2.
Of main concern here is the number of signal terminals in digital
applications, especially digital data processing as opposed to data storage.
The more circuits (gates) the data processing IC (or other packages) have,
the more terminals will be required at their interface. Although the general
dependency follows the same rule, different classes of applications demand
somewhat smaller or higher number of terminals for the same number

This expression stresses the fact that the average number of wired circuits depends on
the pinout provided by the package. Maximum available circuits must exceed the average
by approximately a factor of 2. Other limitations are wiring channels, power supply,
and cooling.
11-24 MICROELECTRONICS PACKAGING-AN OVERVIEW

Mainframes
Bipolar gate arrays
CMOS gate arrays
• Microprocessors
c Static RAM
o Dynamic RAM

Number of Circuits or Bits

Figure 7-10. Number of Terminals and Number of Circuits (Adapted from


Ref. 8.)

of circuits (Fig. 7-10). The highest number of terminals with randomly


partitioned logic is needed for the highest-performance applications such
as supercomputers or mainframes.
Generally, the "dictate" of Rent's rule applies to information being
conveyed to the assembly of circuits, not necessarily to the number of
actual lines or terminals. Information can be coded on several lines,
somewhat analogous to memory address, to trade the total number of
terminals against the need to code-decode and to buffer data, with the
resultant performance penalty. Such a penalty is often small when many
logic (and memory) blocks are encountered by the data between entering
and exiting the enclosure (chip, card, board), as it occurs in a microproces-
sor. Therefore, significant pinout reduction can be utilized with an accept-
able performance degradation. In the case of high-performance logic, only
very few logic blocks are present between the input and output terminals.
Consequently, to utilize most circuits and to avoid severe performance
degradation, Rent's rule of Equation [7-1] has to be followed.

7.2.2 Electrical Design Considerations


Objectives of the electrical package design are to assure reliable and
predictable signal transmission through the on-chip and off-chip intercon-
7.2 TECHNOLOGY DRIVERS 11-25

nections. It is accomplished by numerous design considerations, all of


which have to be met as follows.

• Assure reliable functioning of the interchip signal interconnec-


tions (nets):
By preservation of the transmitted signal (within acceptable toler-
ances): DC levels, length of signal transitions, and distortions
caused by reflections
By containment of the noises (unwanted signals) injected through
various paths, such as interline coupling (controlled by the lines'
cross sections and shared lengths) and the common path imped-
ances
• Assure reliable on-chip nets performance:
By ascertaining integrity of power distribution
By providing integrity of the chip-to-package interface regions:
self-inductances and mutual inductances and capacitances
• Optimize the packaging cross-sectional geometries and, if possi-
ble, materials:
By avoiding insufficient/excessive interline separation and line
cross sections
By providing adequate ground and power planes
• Devise and verify "wiring rules":
For the "legal" signal nets configurations-and to grant expections
To facilitate implementation of the noise containment constraints
• Enable performance prediction, or at least assessment

These objectives were pioneered by the developers of packaged


electronics for the mainframes and supercomputers; now they are becom-
ing increasingly pervasive for the entire range of digital processors, be-
cause their clock frequencies are going well beyond the lOO-MHz mark.
As was mentioned in Section 7.1 and will be shown in Section 7.2.11,
the demand for higher performance had been driving down the overall
dimensions of a computer while demanding more circuits and more bits
of storage. The higher level of IC integration enables such size reductions.
In addition to the impact on physical dimensions, two other key considera-
tions need further discussion-the electrical considerations, to be dealt
with in this section, and the thermal considerations, which will be covered
in the Section 7.2.4. The in-depth coverage of these topics will be given
in Chapters 3 and 4, respectively.

7.2.2.1 General
Digital information processing is based on assigning a discrete (bi-
nary) value to a small range of an electrical signal and another value to
11-26 MICROELECTRONICS PACKAGING-AN OVERVIEW

another range. This value can the be transmitted through a wire to another
location, stored, or applied to a logic circuit as its input signal and, as a
rule, combined with the other input signal(s) to form an output signal,
the value of which is governed by the desirable logical function (e.g.,
INVERT, OR, AND, etc.). New information is processed by changing
the input signals and using the output signal as an input to the subse-
quent operation(s).
Changing information then requires a change in the signal level from
one value to the other. The transition to a new signal value cannot occur
instantaneously. The time it takes is called the signal transition time. Short
transmission times facilitate better performances but also cause a variety
of problems.
Unlike the liquid or heat flow, or the propagation of light and other
electromagnetic radiation, transmission of electrical signals requires the
use of two conductors, such as the wires to a light bulb or to a loudspeaker.
In the majority of data-processing applications, the one conductor is well
defined, frequently used terms being a wire, a connection, or a signal
line, whereas the other conductor is shared by many other signal and
power circuits. From the early days of electronics, that second conductor
has been usually provided by the metallic packaging structures, such as
the chassis of an amplifier or a TV set. The presence of the second
conductor is often only implied in the schematic diagrams, based on the
assumption that no special attention is required to its role in conveying
signals.
Similarly, properties of the connecting conductor are also frequently
assumed to be ideal, and of no particular consequence to the task of
connecting two circuit elements or two complete circuits. Again, no further
attention is paid to any of its properties. Such lack of attention may, and
in the case of high-performance electronic circuits definitely will, cause the
appearance of several phenomena that are undesirable and often harmful to
the intended function.
A particularly lucid treatment of this subject-in fact, an excellent
supplement to Chapter 3-is accomplished by Ron K. Poon in Section
8 of his very comprehensive book published in late 1995 [10].

7.2.2.2 Electromagnetic Properties of Signal Wire

Three electrical parameters-resistance, capacitance, inductance-


are always present2 and cause signal delays and many forms of signal
distortions, including the unwanted appearance of signals on supposedly
quiescent (not switched) lines. It is these unwanted noise signals that are

2 Except zero DC resistance in a very special case of superconductivity.


7.2 TECHNOLOGY DRIVERS 11-27

the biggest troublemakers, because they are hard to eliminate from a


finished design; they must be addressed during the initial design phase.
All signal lines and also the input electrodes of switching devices
(e.g., gate, base) have a certain capacitance; they store an electric charge
when a voltage is applied. To change the voltage, the amount of charge
must be changed by the flow of electric current. This flow causes the
buildup of a magnetic field that induces voltage into the current-carrying
line (self-inductance), resisting the current flow, and into the other lines
in the proximity of the line being switched (mutual inductance), causing
a current flow there. Furthermore, the source resistance of the switching
circuit and the signal line resistance also limit the amount of current flow
and, thus, control the time during which the signal transits from its old
value to the new. The transition time increases with the increase in capaci-
tance.
The design of faster devices and circuits concerns itself, among other
things, with the faster switching speeds of devices, with the reduction of
the device/circuit input capacitance, and with the optimization (reduction)
of the driving (source) impedance. The packaging design (geometries and
materials) can control the signal line capacitance to the reference plane,
the line resistance, its self-inductance, and interline capacitances and
inductances. Consequently, it can control the signal distortions and the
appearance of unwanted interferences. In spite of best efforts, it may even
limit performance of the high-speed computers.
The signal line capacitance is directly proportional to the line length,
whereas the capacitance per unit length is a function of cross-sectional
geometry and the value of the dielectric constant of chosen materials.
Note that the scaling of geometry, preserving all ratios of dimensions,
preserves the capacitance per unit length.
7.2.2.3 Signal Degradation
There are two internal causes of degradation (omitting externally
induced interferences) as the signal propagates through the package-
reflections and the effects of line resistance. Resistance mainly causes
voltage drops and increases in the signal transition time. Reflections, in
addition to causing increases in the transition time, may split the signal
into two or more pulses, with the potential of causing erroneous switching
in the subsequent circuits and, thus, malfunction of the system. As will
be shown in detail in Chapter 3, the reduction in signal level due to the
resistive voltage drop will decrease immunity to the signal noise which
has several causes; mainly cross-coupling noise, switching (M) noise, and
signal reflections.

7.2.2.4 Cross-Coupling Noise


The cause of cross-coupling noise is illustrated in Figure 7-11, along
with the causes of switching and reflected noises. The structure shown
11-28 MICROELECTRONICS PACKAGING-AN OVERVIEW

Figure 7-11. Causes of Reflected, Coupled, and Switching (M) Noises

represents two short sections of typical wiring. Each signal line, in addition
to its own resistance, capacitance to the reference (ground) plane, and
self-inductance, has mutual inductance L12 and interline capacitance 12 e
to its neighbor.
When driver 01 sends a changing signal voltage and current through
the first signal line toward its terminating resistor Tl, line 2 between 02
and T2 is supposed to be in a steady state, not switched. Voltage change
on line 1 induces current Ie through e)2 which splits and flows toward
both ends of line 2. The current change on line 1 induces a voltage on
line 2, which, in tum, causes the current h to flow in the direction opposite
to the signal current on line 1.
In many cases, induced currents Ij2 and h, flowing toward the far
end of line 2, cancel each other. By contrast, the currents flowing toward
02 (near end) add up, are reflected by the low impedance of 02, propagate
toward T2, and contribute to the signal noise.

7.2.2.5 Switching (AI) Noise


Another serious, and potentially the worst, noise originates by the
driver switching current drawn from the power distribution system P,
through the inductance Leff common to some, many, or all drivers of the
same chip. A voltage spike across Leff will usually propagate through all
drivers (even the quiescent 02 in this example) and, because of the additive
effect of simultaneously switching drivers, can cause severe problems.
The remedy may be either in reducing the value of Leff by changing
7.2 TECHNOLOGY DRIVERS 11-29

packaging geometries or in restricting the total switched current by either


slowing down the driver speed, by limiting the number of drivers allowed
to switch at the same time, or by incorporating decoupling capacitors,
which shunt some of the switching current, effectively reducing M through
Leff and the noise generated by it.

7.2.2.6 Signal Reflections


A potentially very serious signal degradation occurs at the junction
between the signal line and the stub of Figure 7-11, providing connection
to the input of a circuit. If the stub delay is less than one-fourth of the
signal transition time, then it acts as a capacitive load, causing a dip in
the signal transition waveform and sending a negative reflection toward the
driver. This by itself is undesirable but unavoidable. It may be intolerable if
the stub capacitance in combination with receiver Rl input capacitance
generates a large reflection that can be magnified by several other closely
spaced (less than one-fourth the transition time) stubs and receivers. The
resultant degradation of the signal transition waveforms may also be
unacceptable.
These problems are further magnified by the long(er) or multiple
stubs. When the signal transition arrives at the stub location, it splits. One
part, with the reduced amplitude, travels along the signal line and the
remainder travels along the stub toward the receiver input. A negative
(reflected) signal is sent on the signal line toward the driver. After charging
the receiver input capacitance, it is reflected back, traveling on the stub
toward the signal line. Upon arrival there, it restores the signal to its
original value. In the meantime, the negative reflection may affect receivers
located between the driver and this location. Other receivers, located
between the stub and the line terminator, may not switch until the restored
signal amplitude arrives at their respective locations.
Shorter signal transition times improve performance. They also cause
increased noise and restrict the maximum stub length and number of stubs.
Their generation usually requires higher power, just as the achievement of
higher circuit performance does.

7.2.3 Electrical Performance-Power Distribution


• Leverages of a "quiet" power distribution:
Reduced noise due to circuits switching activity
Enables either wider processing tolerances or better peiformance
Permits lower power and signal voltages ~ better speed-
power product
• How to achieve a "quiet" power distribution:
Use high-speed voltage regulator and enough decoupling capac-
itors
11-30 MICROELECTRONICS PACKAGING-AN OVERVIEW

"Hierarchical" decoupling ~ control of resonance peaks, as shown


in Figures 3-12 and 3-13
Need to contain deleterious effects of interconnection inductances
Employ distributed decoupling with low inductance

7.2.3.1 Rigorous Analysis


Following is the rigorous analysis of the noise generated in any
power distribution system, derived from Ref. 11. Let us designate:

v, = signal voltage, as a function of time


Vp = voltage variation at the power terminals of a circuit or a circuit group
L = inductance of the decoupling capacitor and its connections to the cir-
cuit(s)
ip = current drawn by ckt(s) from the power distribution
Cp = decoupling (power distribution) capacitance
Rc = intrinsic resistance of the decoupling capacitor
Vci = voltage change across the decoupling capacitor caused by the dielec-
tric roll off

The rigorous equation for Vp is

V() -L~
p t - dt + J.-f·d
Cp lp t + R'
c1p + V ci [7-2]

For unterminated lines, with pure capacitive load C"

. _ C dVs
lp - s dt [7-3]

Sustitution yields

rPVs C dV,
Vp(t ) = L Cs df + CpVs(t) + KC dt + Vci [7-4]

The ratio of noise-to-signal voltage is

Vit) _ _1_ rPVs C _1_ dVs ~


Vs(t) - LCs V,(t) df + Cp + RcC V,(t) dt + V,(t) [7-5]

This is a precise but rather complex depiction. Note the significance of


the four terms:
7.2 TECHNOLOGY DRIVERS 11-31

LCs indicates the importance of keeping small the value of L-the


inductance of the "last" power-distribution link between the
decoupling capacitor and the circuits being decoupled-for
high-performance applications with high values of dV/dt and
tfV/dt2. It will be rather crudely illustrated on the following
pages in which tfV/dt2 is replaced with (dV/dtf
C/Cp indicates the fraction of the charge the decoupling capacitor(s)
are to lose when supplying current to (re)charge the signal
lines through the driver circuits. Thus, it determines the mini-
mum value of Cp for a given Cs.
ReCs is the intrinsic time constant of the decoupling capacitor(s),
assuming no dielectric rolloff. It indicates the need for high-
quality capacitors (low Rs) for high-performance (high dV/
dt) applications.
Vc/V.(t) indicates the additional voltage drop caused by the dielectric
constant rolloff at high frequencies, a common occurrence with
the high-dielectric-constant materials which can negate the high
values of Er occurring at low frequencies-typical for very high
values of Er• As a rule, the dielectrics with the higher values
of Er begin to drop it at lower frequencies and rolloff more
than those with the lower values. The breakdown field (Vfcm)
is also lower for the dielectric materials with higher values of Er•

7.2.3.2 Very Rough Estimate of Acceptable Lead Inductance L = Lmax

Note that rigorous calculations involve second-order nonlinear dif-


ferential equations, as shown above.
One circuit is driving a capacitive load Csignal = Cs with the signal
amplitude Vs. During the signal transition time !:It, there will be a surge
of current M through the signal and ground/power leads.

M = c.(Vs / ~t) [7-6]

(With more outputs n, simultaneously switching, the total current surge


is M = nMsingle). This current surge M during ~ts will produce a voltage
spike ~V at the corresponding parasitic inductances L of signal, ground,
and power leads.

[7-7]

Substituting equation [7-6] for M in equation [7-7]:

~V = V.LCs[l / (!:lt~t.)] [7-8]


11-32 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-3. Values of Maximum Power Lead Inductances Lmax

For ti V/V,;g"l = 0.1


and n = I e, = 10 pF e, = I pF e, = 0.1 PF

tit = 10 ns = 10,000 ps
(&)2 = 108 Lm" = I I1H Lm" = 10 I1H Lm" = IOOI1H
"Yesterday's designs"

tit = I ns = 1000 ps
(tit)' = 106 Lmn = 10 nH Lmn = 100 nH Lm" = I I1H
"Today's designs"
tit = 0.1 ns = 100 ps
(tit)' = I (J4 Lmn = 100 pH Lmn=lnH Lmn = 10 nH
"Tomorrow's designs"

ill = 10 ps
(&)'= 100 Lm" = I pH Lm" = 10 pH Lmn = 100 pH
"Future designs"

By rearranging,

~V / Vsignal = LC,I / (M~tJ "" LCs(1 / ~t)2 for ~tv "" ~t [7-9]

Note that if L is in pH, C s is in pF, and ~t is in ps, no multiplier is required.


A typical interconnecting wire in a PWB is a transmission line
with the characteristic impedance Zo "" 80 Q and Cr "" 4.5 [12]. Such lines
have L = 550 pH/mm = 5.5 nHlcm and C = 85 fF/mm = 0.085 pF/mm =
0.85 pF/cm.
Contemplating Table 7-3, we notice that the typical PWB wire
lengths of "yesterday'S designs" were on the order of 1-50 cm. The
corresponding C, were then in the 1-50-pF range, roughly represented
in the first two columns of Table 7-3. The long signal rise times permitted
the power interconnectin inductances in the 0.2-1O-flH range, or > 40
cm-an easily manageable length of a signal wire serving as a power-
distribution connection.
By contract, the "today's designs," which may well have become
those of yesterday when you read this, have similar wire lengths, but the
signal transition times shrunk to about 1 ns, forcing the power distribution
wires into the several-centimeter range-still an easy to meet requirement.
The today's (ca. 1996) mainframe designs, with -0.1 ns signal transi-
tion times, will be almost all designs around the year 2000. The signal
transmission lines become shorter, with capacitances in the 0.1-1 + pF.
These designs can tolerate the power-distribution interconnection lengths
7.2 TECHNOLOGY DRIVERS 11-33

of only a few millimeters as, for example, was done for the IBM ESI
9000 mainframe MCM packaging [11,13].
Very advanced designs, still in the research phase in the late nineties,
aim at clock frequencies of 1 GHz and higher. This drives the signal
transition times into the lO-ps range. The signal-line capacitances are still
in the O.l-I-pF range. Consequently, the affordable power-connection
inductances must be in the few pH range, requiring the interconnection
lengths of a fraction of Imm! This is a significant paradigm shift, which
amounts to replacement of discrete decoupling capacitors, such as in Refs.
11 and 13, with the low-impedance power-distribution systems, such as
demonstrated with thin high-dielectric films sandwiched between the
power and ground planes [14,15].

7.2.4 Thermal Design Considerations

A somewhat humorous way to describe electronic data processing


might be by saying that its prime purpose is the conversion of the electrical
energy into thermal-generation of heat. Some circuits, notably those
employing bipolar devices, generate most of the heat even if they idle
(i.e., process no information at all). Others, notably CMOS, generate heat
only when they switch and do so in direct proportion to their operating
(clock) frequency. High concentration of circuits within a small area of
an IC, leads to high power density, measured in W/cm2• The power
density is actually another term for a heat flux-the number of calls (1
W = 0.239 calls) to be transported from the location of their generation
to the location of their removal. Such transport is associated with a
temperature difference between the two locations. Thermal resistance
quantifies the temperature rise for a given flux and is expressed in °CIW
or KIW.
The quest for higher and higher levels of integration drives technolog-
ies to produce ever-smaller devices, interconnections, and terminals. As
already illustrated in Table 7-1, the demands on heat-removal capability
can easily snowball. Consider, for example, Figure 7-12 which relates
the power density, in W/cm 2, demands of lO-mm x lO-mm IC chips with
various numbers of circuits and operating at various levels of circuit
power. It also benchmarks temperatures and power densities of a light
bulb and the sun surface. A nuclear blast, a mile away, fits in between.
So will chips with a circuit count range around 1990 (lO,OOO-lOO,OOO)
if each were to consume l-lO mW. But the chip temperature must be
below 100°C to assure proper electrical performance and to contain the
propensity to fail.
Figure 7-12 clearly demonstrates the need to enhance heat removal
and to contain the power hunger of the logic circuits. It also indicates
that the power density, and not the more commonly used power-per-chip
11-34 MICROELECTRONICS PACKAGING-AN OVERVIEW

1()6

...o
Off-Chip Driver
Power Neglected

10 1()2 1()3
Wattslcm2

Sun Surface

Nuclear Blast
(1 Mile Away)

Light Bulb

High Performance
Computer Chips

10~ ______- L________ ~ ______ ~ ______ ~

1 10
Wattslcm 2

Figure 7-12. Power Density Requirements

figure, is the measure of the heat-removal capability of a package. This


capability is particularly important when large assemblies of closely spaced
high-density high-performance chips must be employed, as in a power-
ful mainframe.
Figure 7-13 presents an evolution summary of the heat-flux demands
for a wide range of applications, from the hand-held equipment to high-
performance mainframes and supercomputers. It also indicates the pro-
jected increases in the chip (die) dimensions. The conventional package
consisting of single-chip modules and card on board is capable of about
an order-of-magnitude lower power density than the state-of-the-art multi-
layer ceramic technology with liquid cooling. However, significant prog-
ress has been made, and continues to be made, in increasing the manageable
power density, as shown in Figure 7-14.
7.2 TECHNOLOGY DRIVERS 11-35

<ii> Ole Size mm2


1lOW

100 H\gh performa


nce
---®-- ...e
lOW
....®- 15W/ern'

®-
law/em'
t:"'
e 55W

I
..I
>C 10 6W1ern'
:::s Automotive 3-6W/cm2
i!

Commodity 1-2W/cm2

Hand-Held 0.5-1.0W/cm2
1

1995 2000 2005 2010

Year

Figure 7-13. Evolution of the Thermal-Dissipation Requirements

Thermal expansion caused by heating up the packaging structure is


not uniform-it varies in accordance with the temperature gradient at
any point in time and with the mismatches in the thermal coefficient of
expansion. Mechanical stresses result from these differences and are one
of the contributors to the finite lifetime and the failure rate of any packag-
ing structure.

7.2.5 Reliability

There are two fundamentally opposite approaches to the reliability


aspect of a design. The first one is to build the hardware of a given design
and either just hope for the best or subject it to some sort of stress test(s)
to assess the potential for performing at acceptable failure rates within
the anticipated/specified lifetime and under expected/specified operat-
ing conditions.
When the risk of failing rules out such an approach (e.g., in the case
of nuclear plants, aircraft, or electronics for critical applications), then
the second alternative is the only choice. It involves understanding the
failure mechanisms, systematic elimination of their causes during the
design phase, and, finally, passing a set of qualification tests. Such tests
11-36 MICROELECTRONICS PACKAGING-AN OVERVIEW

l00W/cm>

~\b
"e 80 ~o: .
~
"
~
~
!::
20
Air-Cooled Technology
« lOW/em> ,
~j9()()()(50
3;),fC W/cm»

i
...
l:

:>-
10

...
~
:i:
u 5
5-18 Watts/cm 1
3
IBM 438.
2 IBM IBM
acuum 360 3033
oneywell DPS-88
1955 1960 1970 1980 1990 2000
Yeelf Source: IEEE Proc. Dec. 1992

Figure 7·14. Evolution of AirlLiquid Cooling Capabilities. (From IEEE Proc.


Dec. 1992.)

(or subsets of them) may also be repeated at prescribed intervals to


ascertain continued stability of the product manufacturing process.
In the case of microelectronics, the failure to function may have
two basically different causes-error signals occurring in the otherwise
fully functional product and a compromise to the product integrity, such
as opens and shorts in the signal lines. Containment of error signals has
already been highlighted in Section 7.2.2 and is covered in great detail
in Chapter 3.
Product integrity disruptions are caused either by the intrinsic wear-
out mechanisms or by hostile environmental factors, such as corroding
agents penetrating the package interior and attacking sensitive parts or
their interfaces. The formation of cracks in the solder joints caused by
mechanical stresses which result from nonuniform thermal expansions
and contractions is an example of the wear-out mechanism. Corrosion of
connector contacts may result either from a process residue not fully
removed or from the environmental contaminants. Chapter 5 deals with
the determination of reliability parameters and Chapter 10 describes tech-
nologies specifically aimed at protection against the environment. Chapters
7-13 include discussions of the reliability aspects.
The failure-rate reductions in leading-edge electronic products were
quite phenomenal-about four orders of magnitude in a quarter of century,
or a twofold reduction every 2 years-as presented in Figure 7-15.
7.2 TECHNOLOGY DRIVERS 11-37

10000

1000

100

10

0.1

0.01 +----t---+---+----II----f----+--~
1970 1975 1980 1985 1990 1995 2000 2005
Year

Figure 7-15. Intrinsic Failure-Rate Improvement Trends for IBM Logic-Cir-


cuit Reliability

7.2.6 Manufacturability and Quality


For the majority of the packaged electronic products, the emphasis
has shifted from performance-the major preoccupation at the time of
writing the first edition-to cost. Performance is still very important as,
for example, manifested by the continuing new PC announcements with
improved price per performance ($IMIPS) shown in Figure 7-1. But the
drive for a constant cost-performance factor has been replaced by the
drive to maintain or reduce the cost or raise the performance without
sacrifice in quality or reliability. There are two major enabling factors to
accomplish these goals:

• Establishment and maintenance of manufacturing process tolerance


windows necessary and sufficient for the product to pass its tests,
actually with ever decreasing incidence of fails as illustrated in
Figure 7 -16. Elimination of causes of "killer defects," such as sloppy
handling of parts, is also an essential part of this cost-reduction effort.
• Elimination (or a drastic reduction) of buffer inventories by a bal-
anced manufacturing throughput capability and an uninterrupted flow
of product through various manufacturing steps. Increased through-
put lowers manufacturing costs in proportion.

In recognition of the pivotal role these factors play in product compet-


itiveness and success in the marketplace, a new Chapter 6, "Package
11-38 MICROELECTRONICS PACKAGING-AN OVERVIEW

1,000,000 IRS Tax Advice

10,000 Restaurant Billing


Payroll Processing
23 Airline Baggage Handling
.= 100 3.

IBM's TCM:
0.01
5 Billion Component 0.018
Hours Without Fail
0.0001 Courtesy: IBM
2 4 6 8
Sigma

Figure 7-16. Defects Versus Statistical-Distribution Limits, Trend Is To-


ward ~6cr

Manufacturing" has been added in this second edition. It reviews these


five key principles:

• Quality in product and process designs


• Product qualification
• Manufacturing line modeling and simulation
• Process control
• Worldwide quality standard-ISO-9000-for quality systems certi-
fication and verification procedures. In many cases, it has already
become a necessary element when participating in trade within and
between certain countries.

7.2.7 Testability
Implicit in reliability considerations is the assumption of a flawless
product function after its initial assembly-a zero-defect manufacturing.
Even though feasible in principle, it is rarely practiced because of the
high costs and possible loss of competitive edge due to conservative
dimensions, tolerances, materials, and process choices. The commonly
practiced alternative is to balance the above variables against loss of
defective product, which must be screened out by testing.
At the dawn of electronics, discrete components were individually
tested prior to assembly. Such tests were often simple; for example,
the value of a resistor or capacitor. Completed cards were tested for
functionality. Again, such tests involved a small number of steps necessary
to ascertain freedom of assembly defects. With the introduction of statisti-
cal designs in the late fifties, additional tests were required to weed out
the small percentage of circuits (cards) whose components contained
7.2 TECHNOLOGY DRIVERS 11-39

extreme values of their parameters and caused at least one important


circuit parameter to fall outside of specified tolerances. Such cards were
commonly reworked by identifying, desoldering, removing, and replacing
the likely culprit-an item still within its own tolerance band but not
blending well with other components on that particular card. Additional
sophistication was required in devising and performing tests and in analyz-
ing the test results. Similar methods were employed in determining a
faulty component, one which either slipped through its testing screen or
was damaged during the card assembly. These techniques also worked
well with cards that failed in the field, which were found, replaced, and
returned for repair and reuse.
Finding a faulty card was facilitated by connector pins protruding
out of the other side of the board and offering convenient test points for
a voltmeter or an oscilloscope. "Swapping" two cards of the same design
(part number) often could cure a problem arising from the statistical design
of the off-card signal nets.
None of these techniques work with the large-scale integration (LSI)
and very large-scale integration (VLSI) chips. Circuit terminals are buried
in the chip wiring, and of course, nothing on the chip can be removed
and replaced. Whereas at lower levels of integration, some external test
points (terminals) would permit access to internal nodes, the higher level
of integration and the resultant pinout demands severely limit such an
option. Two different solutions have evolved, each suitable for different
designs. Ultimately, a combination of these solutions, and other ideas
already on the horizon, will provide an adequate level of testability.
Storage arrays lend themselves to testing through their regular termi-
nals by writing in and reading out patterns that determine the ability of
each individual cell to store ones and zeros and to be immune to the
switching activity of other cells or groups of cells. By contrast, random
logic requires strict design discipline, sophisticated test pattern generation,
and involved analysis to verify proper functionality and to uniquely iden-
tify causes of malfunction [9].
Multichip modules (MCMs) or multichip packages (MCPs) popu-
lated with VLSI chips and/or with specialized microwave transmitter/
receiver chips have become high-value products because of the expensive
chips and because of the relatively complex, high-cost substrates. Avoid-
ance of non shippable modules is an economic must. It is accomplished by
the use of so-called "known good die" (KGD)-pretested chips, including
burn-in-and thorough pretesting ofthe substrates. Doing both minimizes
the risk of manufacturing defective MCMs but does not reduce it to zero.
Therefore, testing of populated MCMs is also required. This testing has
two major facets. One is an extension of the design for testability methodol-
ogies and techniques developed for the KGD. The second is akin to
functional testing of subsystems which, in fact, many MCMs are.
The subsystem testing is highly application dependent. It is still
11-40 MICROELECTRONICS PACKAGING-AN OVERVIEW

beyond the scope of this book. Testing of the substrates and KGD-type
testing (functional and diagnostic) of populated MCMs has been added
to this second edition in Chapter 13, "Electrical Testing." It covers various
techniques (both, currently practiced and still in the R&D phase) for opens
and shorts testing of the wiring nets, contacting the substrates with single-
terminal and multiterminal probes, the boundary-scan and built-in self-
tests of the populated modules/packages.

7.2.8 Memory Packaging


The requirements for the main memory packaging-as opposed to
various buffers and caches which are often integral with logic-were least
affected by the performance race, primarily because the speed of the
dynamic random-access memory (DRAM) chips is improving at a very low
rate. The cost improvements are the main driving force. This improvements
were leveraged by several key factors. One, very consistent of them, is
the steadily increasing number of bits per chip and the bits density on
the chip, a planar-two-dimensional-density increase of approximately
2 times every 3 years, or 10 times every 10 years due to improvements
in the photolithography. In the past, there had been an approximately
sixfold gain in the on-chip density produced by the circuit inventions
which replaced a six-transistor storage cell with a one-transistor plus
capacitor cell in DRAMs. This maps into a prorated annual density im-
provement of 9.4% (1.094x), or 2.45x in 10 years. Thus, the on-chip
advancements contribute to approximately 25-fold density improvements
every 10 years.
Figure 7-17 shows an interesting long-range trend of memory density

~
MAGNEnC
~ CORE MEMORIES

II... 10'
1K
~ 10'
Q
4~6K
64K
w 256K
~ 1 1M

8
4M
16M
w 10 MOS 84M
:I DRAMs ~
::l
-' 10 HUMAN
g BRAIN

1950 1960 1970 1980 1990 2010


YEAR

Figure 7-17. Trend in the Volume Required for 1 MB of Storage (Courtesy of


SRC, reproduced with pennission.)
7.2 TECHNOLOGY DRIVERS 11-41

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

Figure 7-18. Microprocessor Clock Frequencies (Courtesy of TechSearch Interna-


tional; from Ref. 26.)

improvements over a 60-year span. The continuous trend is approximately


200x every 10 years, even though the ferrite cores of the 1950-1970s
were replaced by the semiconductor chips. With the on-chip improvements
contributing -25x, the balance of 8x in 10 years must be, and in fact is,
contributed by the packaging advancements such as increased planar and
volumetric silicon efficiency.
Finally, also note that in the first decade of the next century, we
anticipate the volumetric bit density of the electronic memories to match
that of the human brain.
The requirements for memory packaging are somewhat different
from those for the packaging of logic discussed in almost all the chapters.
At the systems level, the demand for memory in terms of numbers of
active elements per chip has outpaced all the other packaging requirements
(Figs. 7-8 and 7-17). Today's supercomputers, for example, require as
much as 2-10 gigabytes of main memory, which must be packaged in
such a way that the access time to it, as well as retrieval from it, is fast,
often no more than a few cycles of the machine. All of this clearly requires
that the memory package be as dense as possible and either be part of
the logic or be very close to it. Also see Section 7.2.11 and Figure 7-24.
Memory chips generally have very low-power and I/O requirements, as
pointed out earlier in the chapter, and thus lend themselves to the low-
cost, high-density packaging.
The cardoon-board technology is the most common technology used
for packaging memory to date. However, a departure from this, particularly
11-42 MICROELECTRONICS PACKAGING-AN OVERVIEW

1000

Pentium. 1993
en 100

-:E
Il.

r 1486.1989
~
c 10
as
1386.1985
E
.g 80286. 1982
CD
Il. 8086. 19n

0.1
II I
10 100 1000

Clock Speed (MHz)

Figure 7-19. Intel Microprocessor Clock Frequency Versus Performance


(From Ref. 26.)

for high-end mainframes and top-of-the-line supercomputers that can pro-


cess information in a cycle of a few nanoseconds, is expected. In the
early 1970s DIPs with 16 leads were capable of handling 4K dynamic
random access memory (DRAM), Single inline packages (SIPs) and DIPs
handled up to 256K DRAM's in the late 1970s. The megabit DRAMs
that were beginning to be available in the mid-1980s were expected to
be packaged in PLCC or quad packs. Each card (250 x 270 mm) is capable
of holding about 160 chips, along with logic support chips, to store 1
megabyte with 64K DRAMs, 4 megabytes with 256K DRAMS, and 16
megabytes with I-megabyte chips. Both sides mounting with six layers
of interconnecting lines within the card could be used to double the
memory capacity to 32 megabytes. Half a gigabyte can be achieved by
edge connecting 16 of these cards in the width of an 18-layer board.

7.2.9 Personal Computers and Workstations


At first it would seem that personal computers should be among the
easier products to design; that is, with the large systems driving perfor-
mance technology development and the consumer products driving cost,
designing a personal computer should simply mean taking the best from
both disciplines. This would indeed be the case if it were not for the
many contending user and business requirements for entry systems which
include, but are not limited to, the following.
First and foremost, the machine must be cost-competitive and offer
7.2 TECHNOLOGY DRIVERS 11-43

Figure 7-20. Packaging of a Workstation (IBM RS/6000) (From Ref. 26.)

on the order of a factor of 2 improvement in value in comparison to the


products marketed just 2 or 3 years before. In the short history of personal
computers this has been the compound rate of cost-performance improve-
ment as already indicated in Figure 7-1. Today's machines must provide
performance in the range of mainframe processors of just 7 years ago.
The reliability of the machine means that it must work right out of the
box just as any other modem electrical appliance, and for a comparable
period of time. Also it should be easily assembled and operated by the
customers who usually do not possess a high level of technical skill.
The personal computer, more than any other data-processing ma-
chine, serves a broad range of applications and therefore must offer "expan-
sion" cards for configuration to the problems of that specific user. As
personal computers are typically customer-operated systems, provisions
for automatic-fault detection and location are required to enable the owner
to make repairs with a reasonable level of confidence. The "footprint" of
the machine should be as small as possible, so that the machine will
consume the minimum of desktop work space and be an unobtrusive part
of the office environment. It should be of reasonable weight for portability,
whether for office rearrangement or for relocation to where the work is
to be done.
11-44 MICROELECTRONICS PACKAGING-AN OVERVIEW

The display capabilities expected by the customers are far more than
the text-quality graphics provided by the TV-like screens of the first-
generation personal computers. Current product generations approach
CAD/CAM-like system's display capability requiring high speed and
high bandwidth for the manipulation of these data. Video-processing
requirements and the hardware content for this aspect can easily surpass
the complexity of the "central processor" hardware of the machine. The
personal computer must not only be sparing in its power requirements
but also should be tolerant of electrically noisy power environments. Its
electromagnetic radiation must achieve legislated standards which have
been tightened worldwide in the recent past. Finally, the personal computer
must "look good," be "human engineered," and also "user friendly."
The mid-range data-processing systems require a blend of perfor-
mance and low cost optimized for the lowest possible cost for a given
performance. The typical performance range of these is between about
one MIPs (million instructions per second) or less, to almost the bottom
end of mainframes, which in the mid-nineties is around 100 MIPS per
uniprocessor. This range is generally achieved by a single processor,
but sometimes with multiple processors, particularly in the top range
of performance.
The semiconductor technology capable of providing the function
necessary to achieve these performances has been primarily ECL and
TTL, but CMOS is emerging as the leading candidates offering much
higher integration levels with lower power requirements [5,8]. Single-
card central-processing units are marketed with those silicon technologies
and improved system's performance. BICMOS is a strongly emerging
technology taking the best of bipolar and CMOS transistors on the same
die, providing an alternative to both those technologies in many applica-
tions.
The internal CMOS circuits offer very low-power and low-current
operation sufficient for the short on-chip circuit-to-circuit communications
and the bipolar circuits are used primarily for high-speed off-chip drivers.
In comparison, this technology provides a lower speed-power product
than can be achieved by either of the semiconductor technologies.

7.2.10 Portable and Mixed-Signal Equipment

All portable equipment, from hand-held calculator, to lap-top PCs,


to various communications equipment (e.g., pagers, portable telephones,
geo-positioning systems, etc.) must consume very little power to prolong
the operating time on one battery charge. As a by-product, the thermal
management is simplified. However, the operating environment, especially
automotive and marine, is considerably more hostile than a typical office
environment. The ever-smaller size and mass ofthe equipment (Fig. 7-21)
7.2 TECHNOLOGY DRIVERS 11-45

•
•
Il!! •
.9
~
~ 100
'Eas
ECD
en
N

10
85 86 87 88 89 90 91 92 93 94 95 96 97
Year

Figure 7-21. Trends in Mass and Size of Cellular Phones (From Ref. 26.)

leads to small dimensions of the components and interconnections and


adds to the challenge of design for ruggedness.
The mixed-signal applications environment has its own set of design
challenges: good electromagnetic isolation between the digital and the
analog, usually radio-frequency (RF), circuits to avoid malfunctions due
to injections (either way) of potentially strong noise signals; and well-
controlled low-loss properties of interconnections which, in fact, may
often constitute various circuits and circuit elements-resonance circuits
and filters, impedance transformers, and even receiving and transmitting
antennas [7]. Once again, the cost-competitiveness is a very important
design and manufacturing parameter.

7.2.11 High-Performance Processors

The general trends and key dependencies were highlighted in Section


7.1. Figure 7-22 quantifies the performance trends as determined by the
two key parameters: the cycle time (reciprocal of the clock frequency)
and the average number of cycles the processor needs to execute an
instructions-cycles per instruction or CII. The short cycle times require
correspondingly short interconnections-signal propagation time through
the interconnections must be less than the cycle time. The scale on the
right side of Figure 7-22 indicates the propagation times though the
interconnections whose insulating materials have dielectric constants of
1.6-an extremely small number, not yet available-and lO-a number
11-46 MICROELECTRONICS PACKAGING-AN OVERVIEW

Er = 1.6 Er = 10
Domain of 100
Parallel
Processing
100

--
10

-
Ii)
c::
Q)
E
10
E
Q)
(,)
c::
i= aI
Q)
g, 0.5 1ii
is
0
0.5
Domain 0.1
1 of 0.2
Uni Processors
0.1

0.1 0.01
0.1 1 10 100 1000
MIPS

1960 1970 1980 1990 2000


Years

Figure 7-22. System Performance Dependency on Cycle Time and Cycles


per Instruction

somewhat higher than present with alumina-based ceramics. These two


numbers, therefore, bracket all practical materials.
Practically all packaging technologies have predominantly planar
interconnections. Thus, the distances of Figure 7-22 apply to the maximum
Manhattan length-the sum of the two packaging edges. All circuits and
associated bits of storage (e.g., buffers and first-level caches) must be
contained within these dimensions. Therefore, the maximum available
power density and the total packaged electronics area determine the total
power available for all circuits. The total number of circuits depends
greatly on the intended number of cycles per instruction as shown in
Figure 7-23. Obviously, the lower number of ell may lead to reduced
circuit performance because of the lower power per circuit, if the design
calls for higher than available power density.
The mainframes have provisions for high data-rate transmission into
7.2 TECHNOLOGY DRIVERS 11-47

-
til

~ 100k
(3

10k

1k~ __~~~~~~__~~~~~
1 10
Cycles per Instruction

Figure 7-23. Circuit Count Dependency on the Number of Cycles per In-
struction

and out of the processor. This interdependency is quantified in Figure


7-24 and amounts to ~ 8 MBIMIPS. Note the three data points obtained
by independent researchers (Refs. 3 and 4 and Chapter 20) in their work
on very advanced system applications. They confirm continuation of the
same numerical relationship to at least 1000 MIPS.

10000 Research
Designs

1000

~
III 100
:!. It.

i C
A A ttHJ.
Il '#S¢lMfIf>
~ 10 ~ ~ Actual
x Mainframes

0.1
0.1 10 100 1000
Performance [MIPS]

Figure 7-24. Input-Output Data-Rate Dependency on Performance


11-48 MICROELECTRONICS PACKAGING-AN OVERVIEW

Section 7.2.13 offers an in-depth study of the simplistic examples


of Section 7.1.1 and shows the evolution trajectories of various semicon-
ductor technologies in a wide range of applications.

7.2.12 Optical Interconnects


All electrical interconnections have a limited bandwidth-the highest
transmittable number of pulses per unit time, such as megabits per second.
The bandwidth with a given cross-sectional geometry and materials is
reciprocal to the distance (cable length). The interconnection bandwidth
must keep up with the increasing performance, whereas higher densities
and reduced geometries do the opposite-decrease the bandwidth per unit
length. A very practical alternative for higher data rates is modulation
of light, especially coherent light, and its channeling from the point of
origination (transmitter) to the destination (receiver). A description of
such technologies in Chapter 20, "Optical Packaging," concludes Part 3
of this book.

7.2.13 Summary of Packaging Performance Considerations


and Evolution
In developing digital technologies, various parameters have been
applied to circuit-performance evaluations. The power-time product of
a gate is one of the well-known parameters. The advances in digital
circuits have resulted in remarkable reduction of this equivalent energy.
Energy levels of electromagnetic relays (early logic blocks) were several
millijoules (mJ) and the levels of magnetic switching cores for memories
had approached the microjoule (flJ) level. Silicon transistor gates had
opened the way to the nanojoule (nJ) level, and large-scale integration
(LSI) technology have been developed with picojoule (pJ) and sub-pJ en-
ergies.
This progress indicates that the equivalent energy driven by power
is one of the fundamentals of understanding the characteristics of digital
circuit design. The performance of a digital functional block is quantita-
tively measured by their equivalent energy values and thermal resistance
to the ultimate heat sink. The hardware design concept of processors
(packaged electronics) is described with both chips and packaging techno-
logies. Energy reduction of chips and packaging and appropriate thermal
resistance are essential for appropriate processor performance. The follow-
ing subsection discusses the Power-Time Product Theory proposed by
H. Kanai in 1967 [19].
7.2.13.1 Power-Time Product Theory
Given a digital functional block B, its equivalent energy UB can be
defined as the product of its average power PB and signal delay time tB
of performing the function B:
7.2 TECHNOLOGY DRIVERS 11-49

[7-10]

Another important factor in the design of the functional block B is the


thermal characteristic, represented by

[7-11]

where SB is the temperature rise in the functional block B and ReB is the
equivalent thermal resistance defined by Equation 7-11. The failure rate
AB of the block B can be represented by

[7-12]

where ~ and a are given constants.


The major target of digital system design is to provide satisfactory
service to users not only in terms of high-speed data processing but also
in terms of high reliability. The performance (PF) of a digital functional
block B in terms of speed and reliability can be represented as

PF = (I-bAB) / TiLogical function of block B) [7-13]

where b is a given constant. By introducting Equations 7-10 through 7-12


into Equation 7-13,

PF = 8B[1-b~ exp(a8B)]
[7-14]
UaReB / Logical function

In the case of digital processors, the logical function of generalized


blocks correspond to instructions of a processor and TB corresponds to
instruction execution time:

PF = 8B[l-b~ exp(a8B)] [7-15]


UBR eB / Instruction

The upper part of Equation 7-15, 8 B [1 - b~ exp(a8B)], describes the


reliability of the logic-function block B. This value also indicates the
long-term performance of the system because the long-term performance,
MTBF (mean time between failure) can be represented by a reciprocal
of the failure rate. The lower part of Equation 7-15, UBReB/Instruction,
shows the short-term performance-the processing time of a unit instruc-
tion. Thus, this equation summarizes the total system performance, the
circuit speed, and circuit reliability.
II-50 MICROELECTRONICS PACKAGING-AN OVERVIEW

Equation 7-14 or 7-15 gives us ways to maximize the performance


of either generalized function blocks or processors:

1. Reduction of UBRsBlfunction (or instruction)


(a) Reduction of energy UB
(b) Reduction of thermal resistance ReB
(c) Minimization of UoReB by optimization of packaging structural
dimensions in functional blocks
2. Reduction of constants a,~, and b by passivation of devices, reduc-
tion of failure rates, and high-availability design of blocks in system
3. e e
Optimized B to maximize B [1 - b~ exp(aeB)]; design value of
PB given by (Optimized eB) I ReB'

From Equations 7-10 and 7-12, a nominator of Equation 7-14 can be


represented as

Uo(Instruction) = tBeB / Logic function [7-16]

e
As the temperature rise B is obtained by maximization of the numerator
of Equation 7-14, UBR eB is directly related to the function time tB'

7.2.13.2 Reduction of VeRee by LSI Technology


The progress in LSI technology has had a major effect on increasing
the performance and reducing the cost of digital systems. The design
concept of digital processors represented by Equation 7-15 and items 1-3
enables us to comprehend this effect. The remarkable reduction of logic-
block energy UB resulted from the highly integrated fine patterns of transis-
tor circuits and their on-chip interconnections. Figure 7-25a shows a
typical model of sequentially connected logic circuits (gates), called
switches. Each circuit is supposed to have the same power supply voltage
Vo and resistor Ro. It stores energy in its two parts: the switching energy
and the connecting energy. Therefore, to reduce the equivalent stored
energy UB represented by Equation 7-10, both parts should be reduced.
Even if the stored switching energy could be reduced, the total reduction
is insufficient unless the stored interconnection energy is also reduced.
If the switching energy in Figure 7 -25a is proportional to the current
10 = VolRo to be switched, then the interdependence between circuit power
and delay time are shown in Figure 7-25b. In this circuit model, the
reduction in interconnection size is effective in decreasing the circuit
energy, and the reduction switch (device) size moves the reduced energy
area into a higher-speed domain.
The LSI technology is one of the best approaches available to achieve
these objectives because of the microscopic dimensions of its devices and
7.2 TECHNOLOGY DRIVERS II-51

Yo Yo

FANOUT

>- >- CJ >-


CJ IE: CJ
IE: CJ W IE:
W IE: Z W
Z w w zw
w zw
CJ
z i CJ
i § z
:c
j:
uw :c w
z
z
g z
z
~
i
0
0
U ~ U tn

- - - - - -
(a)

",\
REDUCTION OF CONNECTION SIZE

w
~
tn
:J: C
• U
b'·· ~
.! c'

~
:::;)
C
W
IE:

GATE POWER (LOG)


(b)

Figure 7-25. Typical Model of a Sequentially Connected Logic Circuits (a)


and Its Power-Delay-Time Characteristics (b). Reduction of geometrical dimensions of
connections and switches of ICs is effective in reducing energy.

interconnections arranged in a two-dimensional plane. The advances in


silicon LSI technology have enabled a remarkable reduction of the energy
UB of the digital-function blocks and, thus, contributed to increases in
processor performance as indicated in Equation 7-15.
In addition to the LSI advancements, the reduction in equivalent
thermal resistance ReB of LSI MCMs is pivotal in the minimization of
II-52 MICROELECTRONICS PACKAGING-AN OVERVIEW

IC-CHIP IC-CHIPS

0 0 0 0
0 000
MINIMIZING U•• RIB BY

o
MULTI-CHIPPACKAGING

c::::::> 0000
(REDUCTION OF
\INCREASE OF U.
R..) 0 0 0 0
(8) (b)

Figure 7-26. Comparison Between Single-Chip and Multichip Packages. The


value of UaR-SB of a processor can be minimized by optimizing design relative to the
number of chips, chip size chip locations, and intrachip connections for multichip structure.
(a) Single-chip package for small processor and (b) multichip package for large processor.

UsReB, essential for high-performance processors. Figure 7-26 shows a


comparison of SCM and MCM structures. A processor is considered for
integration on a single silicon chip in Figure 7-26. The value of UsRSB
of a processor in an SCM is adequate for a low-performance design,
because of the value of PF corresponding to a low UB associated with
the UBR eB of an LSI chip on an SCM.
However, the small value of UsReB required for a high-performance
processor cannot be realized with an SCM. Further chip improvements
may sufficiently reduce UB, but the effective approach for a small UsReB
is to use an MCM. The chip of Figure 7-26a can be subdivided into m
chips and packaged using thin-film interchip connections as shown in
Figure 7-26b. With this MCM structure, the equivalent thermal resistance
UsReB can approach the level of UBReB/m of an SCM structure. On the
other hand, the value of UB of the processor increases slightly due to the
additional energy of the chip interconnections. Therefore, the value of
UsReB can be minimized in MCM designs by optimization of the number
of chips, chip size, chip location, and interchip connections. Thus, a
sufficiently low level of UsReB for high-performance processors is achiev-
able with MCMs.
In the case of high-performance processors, large numbers of chips
are to be interconnected not only on MCMs but also on cards andlor
boards which support MCMs. The same design method for minimization
of UsReB can be employed for optimization of the processor performance.
These processor design concepts based on energy and thermal resistance
7.2 TECHNOLOGY DRIVERS II-53

0.1 "--_ _ _-'<-_ _ _- " -_ _ _-"'-_ _ _-"'-_ _ _-"'-_ _ _--'-_ __

10 100 1.000 10.000 100.000 1000.000


N: NUMBER OF GATES

Figure 7-27. Correlation Between Energy Level of Gates on Silicon chips and
Processor Design. The three coordinates are N: number of gates, tpd: average gate delay
time, and Ug : required energy level of gates on a chip when maximum power of a chip
is 1 W. Design points of three types of processors, I, II, and III, are illustrated in the case
of lO-pJ energy level of gates. Processor II fits in one chip structure. Processor III requires
multichip packagings. Processor I can be designed with additional functions on one chip.

are illustrated in Figure 7-27. This graph has three coordinates: N is the
number of circuits (gates), tpd is the average circuit delay time, and Ug is
the average energy level per circuit (gate) on a chip. In Figure 7-27, the
thermal resistance of a semiconductor chip is assumed to be constant,
and the maximum power of each chip is supposed to be 1 W to give the
optimized temperature rise 8 B for processor performance. If a circuit
(gate) should be designed for a chip containing 100 circuits, each is
allowed to consume 10 mW. Therefore, in the case of tpd = 100 ns, the
average energy level of the circuit Ug = 1000 pJ. In the case of tpd = 10
ns, Ug = 100 pJ. In the case of tpd = 1 ns, Ug = 10 pJ, and so forth. If the
circuit should be designed for a chip of 1000 circuits, each of them can
have I mW. Then, in case where tpd = 100 ns, Ug = 100 pJ; tpd = 10 ns,
Ug = 10 pJ, and so forth. Therefore, the coordinate Ug can be indicated
diagonally on the tpd-N plane as shown in Figure 7-27.

7.2.13.2 System LSI in Chips and Packages


As is well known, digital processors enabled significant advance-
ments in computer and communication systems consisting of hardware
II-54 MICROELECTRONICS PACKAGING-AN OVERVIEW

and software. At the dawn of the integrated circuit (IC) development,


small-scale integrated (SSI) chips had limited functionality as basic logic
gates, flip-flops, and so on. The SSI-MSI-LSI-VLSI-ULSI evolution
expanded the LSI functional and performance capabilities due to the
remarkable reduction of the circuit energy Ug • Likewise, the semiconductor
and packaging improvements have contributed to significant enhance-
ments of the processor figure of merit UBR eB as illustrated in Figure 7-
28 which has the same coordinates N, tpd' and Ug as Figure 7-27. The
initial Ug is valid only for I W maximum chip power. In the case of the
chip power being nW, the value of Ug will be lin times the value from
Figure 7-28. In both figures, seven various processor domains (very large,
large, medium, small, mini, micro, and calculators) are approximated by
their respective tpd (circuit delay time) and N (number of gates) plane.
Figure 7-28a shows the integrated structure of processors of early
IC development (SSI) at microjoule energy levels. Each SSI chip had a
small number of MaS, TTL, or ECL gates and was packaged in an SCM.
These SCMs were soldered or socketed on printed wiring boards; and
the processors had many cards plugged into boards assembled in cabinets.
Figure 7-28b shows the technology evolution at its second stage. The
gate energy level had dropped to approximately 100 pJ, with a MSI or
even LSI level of integration. During this stage, low-performance proces-
sors could be implemented within a single MaS LSI chip. Production
and marketing of the calculator LSI and four-bit microprocessors had a
remarkable effect on the growth of the electronic industry. At the high
level of processor hierarchy, large processors had been implemented with
the high-speed bipolar MSI chips in SCMs. Packaging hierarchy of these
high-performance processors was similar to the preceding SSI implemen-
tations of Figure 7-28a.
Figure 7-28c shows the packaging structures of the third-generation
ICs, when the gate energy dropped to the several picojoule level and the
integration levels reached the advanced LSI, the functionality of which
was custom-tailored to a particular system design. For lower-level proces-
sors, the functional capability of a (MaS) LSI chip exceeded that required
for Processor I, as shown in Figure 7-27. This excess capability was
utilized to incorporate software elements onto the chip via inclusion of
memory circuits. Thus, a single-chip (low-level) processor had emerged,
capable of performing various functions such as calculations, conversions,
controls, and so forth. This high functional ability also provided for great
architectural design flexibility of processors on a chip. For medium-class
processors, the number of chips and the total processor size have been
greatly reduced, enabling more compact lower-cost higher-performing
machines than the second stage of Figure 7-28b.
At the high end, large or very large processors were implementable
with the third-generation LSI chips having more than 100 gates of less
7.2 TECHNOLOGY DRIVERS II-55

(a)

.
.s
W
100
::;:
~
>-
:5
w
0
W
f- 10
~
'C
C.
f-

0.1 " -_ _ _-"'-_ _ _ _" -_ _ _--IL _ _ _ _IL-_ _ _--IL_ _ _ _L -_ __

10 100 1,000 10,000 100,000


N : NUMBER OF GATES

(b)

~ 100
W
::;:
~

~
W
0
W
f- 10
«(!)
'C
C.
f-

0.1 IL_ _ _ ~~ _ _ _JL_ _ _---l<:_ _ _____IIL-_ _ _--IL_ _ ____'L-_ __


10 100 1,000 10,000 100,000
N . NUMBER OF GATES

Figure 7-28. Correlation Between Energy Levels on IC Chips and LSI Design
of Processors from Calculators to Very Large Mainframes. Coordinates are
the same as shown in Figure 7-27. The progress of LSI technology can be viewed by energy-
level reduction of gates on chips. LSI chip technology and LSI packaging technology are
closely related to design processors in each level of progression.
II-56 MICROELECTRONICS PACKAGING-AN OVERVIEW

(c)

"'
S
W
::;
100

1=
~
w
Cl
W
10

~
"0
Co
t-

0.1 ~ ________ ~ __ ~ ____- L________ ~~ ______ ~L- ________ ~ ________ ~ _________

10 100 1,000 10,000 100,000


N : NUMBER OF GATES

(d)

1,000

S'" 100
W
::;
i=

~
w
Cl
W
t- 10
«
(!)

"0
Co
t-

0.1 ~ ________ ~ ________- L____ ~ __ ~~ ______ ~~ ________ ~ ________- L_________

10 100 1,000 10,000 100,000 1000,000


N . NUMBER OF GATES

Figure 7-28. (Continued)


7.2 TECHNOLOGY DRIVERS II-57

(e)

R
f-

01L-_ _ _ _ ~~ ____ ~ ______ ~ ____ _4~~ __ ~ ______ ~ _ _ _ __ _

10 100 1.000 10.000 100,000 1000,000


N : NUMBER OF GATES

Figure 7·28. (Continued)

than 1 ns delay each, Figure 7-28c. To maximize performance by minimiz-


ing UBReB/instruction of Equation 7-15; these large processors require
packaging technology as Processor III of Figure 7-27, with a special
consideration toward dense chip interconnections (linewidths and pitch)
and low thermal resistance between device junctions and cooling air. To
achieve this, one of the best methods is to use MCMs on cards and boards,
as indicated in Figure 7-28c. Without the effort to reduce the energy
stored in the chip interconnections, it is difficult to reduce the equivalent
energy level of large high-speed processors, even if low-energy chips can
be developed with low-energy bipolar CML gates.
Figure 7-28d shows the integration technology for processors at its
fourth step of evolution. The gate energies are in the sub-picojoule range.
The level of integration has reached VLSI. At this stage, the energy
reduction provided by the technology has more effect on the data process-
ing system than during the third stage. For low-class processors, the
utilization of the increased/excess chip functional capabilities due to fur-
ther gate-energy reductions gives more intelligence to the chip by further
expansion of the built-in software. This evolutionary direction will con-
tinue because an almost infinite demand for intelligence improvement
exists in the data-processing market. Other applications are to extend
the processing capability from 16 to 32 bits at high speed or to have
II-58 MICROELECTRONICS PACKAGING-AN OVERVIEW

Logic Circuits
Signal In - - -........ Si nal Out
Flip Flip
Flop Flop

Clock

Figure 7-29. Basic Chain of Logic Blocks Operating in One Clock Cycle

multiprocessors on the chip. (Also consider arguments for smaller chips


made in Section 7.1.3.3).
Medium to large processors can be produced with better performance
and lower cost, in one chip or one-card packaging by application of bipolar
or high-speed MOS VLSI chips with an integration range of more than
several thousand gates per chip. Very large computers will use VLSI
MCM packages, with 100-SOO-ps gates. These large computers also will
require very fine and highly multilayered PWB technology for interconnec-
tions between MCMs.

7.2.13.3 Media Delay Factor

In digital data processing, the information to be processed is stored


in a register (flip-flop) as shown in Figure 7-29. At the arrival of the first
clock pulse it is sent for processing through a chain of logic gates (circuits)
and has to arrive at the receiving register (flip-flop) ahead of the second
clock pulse to enable its error-free storage and further subsequent process-
ing. Therefore, the clock cycle time tclock is driven by the sum-total delay
of the logic circuits between the registers. It consists of the sum of delays
within chips (on-chip circuits) tchip and the chip interconnections delay
tmedia' The values of these two time intervals are similar. Therefore, to
minimize the clock time, the reduction in the interconnections delay tmedia
is an effective as the use of higher-speed VLSI chips. Figure 7-30 shows
the relationship between media delay and gate density. In an attempt to
cut the media delay in half, the packaging edge L which contains chips
must become U2, assuming tmedia proportionality with the packaging edge.
Consequently, the gate density in the half-edge package becomes fourfold,
7.2 TECHNOLOGY DRIVERS II-59

Substrate

Substrate
G gates

L S =L2 G gates
U2 S = (1/4)L2
112 tm nS
tm nS
U2
L
Media Delay 1 ) 1/2
Gate Density 1 ) 4

Figure 7·30. Relationship Between Media Delay and Circuit Density

because the new package area is one-fourth of the original. The media
delay tm is, therefore, reciprocal to the square root of the gate density
GIS, where G is the number of gates within the package and S is the area.
The unit signal delay 't in a medium of dielectric constant E is

't=~
-IE,
[7-17]
c

where c is the speed of light in vacuum. The media delay factor can,
therefore, be expressed as

[7-18]

To evaluate the packaging performance in the digital systems, the


tm value is one of the figures of merit; the smaller it is, the better. Figure
7-31 shows tm values for various substrates. Earlier substrates, PWBs,
were made from organic materials such as FR-4 and polyirnide. Those
PWBs, however, had wiring-density limitations because of plated-through-
holes (PTH) which required mechanical drilling. Even though the dielectric
constant of PWB is small, the packaging density was not high, caused
by limited PTH density. By contrast, the ceramic cofired substrates, such
as alumina, have much denser vias and more closely spaced chips but
have a higher dielectric constant. The tm values of the ceramic substrates,
therefore, could not be dramatically improved versus the PWBs. Several
improvements were made by using glass-ceramic substrates, which have
11-60 MICROELECTRONICS PACKAGING-AN OVERVIEW

10 \~

l::. PWB Packages


A PWB Packages (Double-Sided Mount)
o Multilayer Ceramic Packages
o Ceramic+Polyimide Packages

0.1

1980 1990
Year
Figure 7-31. Prefonnance-Factor tm Trends in Various Packages

lower dielectric constants of Er = 5 to 7 and polyimide-ceramic substrates


(Er = 3.5) for the application of high-end computers and supercomputers,
as shown in Figure 7-32. By using those high-speed ('t) and high-density
(GIS) substrate techniques, the tm value has been dramatically improved,
as shown in Figure 7-31. In order to improve the gate density on the
substrate, chip mounting and its cooling techniques are important as well
as dense wirings. C4 technology (IBM) is the best way for highest chip
mounting when the power dissipation of the LSI chips are optimized to
match cooling methods. In case of up-to-date high-end computers, such
as supercomputers, the LSI chip design is rather optimized for the highest-
speed and high-density integration, and results in a higher power density
by depending on lower-thermal-resistance chip-mounting techniques and
cooling techniques. Flipped TAB carrier (FfC) (NEC, Fujitsu) and com-
pact chip carrier (Hitachi) technologies are widely used as the high-density
chip mountings for high-performance computers, as shown in Figure 7-33.
CD ® @
Thin Film
Multl-Layar Ceramic Wlrln. Layer
PWB Substrate
Multi-Layer
PWB

-Coolin. -Wlrln.
Limitations Limitations @

PWB {]
-Wlrln. Air
Limitations water Water
~ Multi-Chip Packa.e Multi-Layer
Glees Ceramic Substrate

Air

-Coolin.
Limitations @
-Wlrln.
Limitations
- Raductlon of
Multi-Layer Er and
Ceramic Resistance
substrate

Water Water

Figure 7-32. MCM Designs with Significantly Improved Values of tm


11-62 MICROELECTRONICS PACKAGING-AN OVERVIEW

Termination Raalstor Modula

Figure 7-33. Cross-Section of SX-3 Flipped-TAB Carrier

Those chip carriers have the excellent features of high pin counts and
lower thermal resistance. As shown in Figure 7-33, 110 pads are arranged
in a pad-grid array on the rear surface of the substrate which includes
many wiring layers inside for the interconnection between the LSI chip
and the pads. The cap of the chip carrier acts as a heat spreader which
can conduct heat generated on the LSI chips to the cooling mechanism ef-
fectively.

7.3 PACKAGING TECHNOLOGIES


As pointed out earlier and throughout the book dramatic changes are
underway in the computer, telecommunication, automotive, and consumer
electronics industries. In computers, the cost per unit of computing has
come down from $4MIMIPS (millions of instructions per second) with
mainframes to less than $10 with PCs and some workstations, as indicated
in Figure 7-1. In addition to increased performance, computing is increas-
ingly being performed in hand-held next-generation computers. The over-
all trend during the last three decades has been from centralized systems
such as mainframe to distributed systems such as PCs to portable, wireless,
and network systems. There is a similar trend in the telecommunication
industry where the user is employing high-performance, multifunctional,
portable units. In the consumer industry, multimedia products with voice,
image, video, text, speech recognition, speech to text, and other functions
as illustrated in Figure 7-3 in wireless applications are expected to be
commonplace within the next decade. Major changes in the automotive
industry, both in the amount of electronics in automobiles and the functions
needed to support the intelligent vehicle highway systems (IVHS) are
occurring in this decade. The common and pervasive requirements in all
of these electronics are (1) ultra-low-cost, (2) thin, light, and portable,
(3) very high performance, (4) diverse functions involving a variety of
7.3 PACKAGING TECHNOLOGIES 11-63

semiconductor chips, and (5) user friendliness. Figure 7-6 reflects this
trend not only for consumer products but most of the electronics of the
next century.

7.3.1 Packaging Evolution


Tables 7-4A through 7-4B project the packaging trend in consumer,
automotive, and high-performance systems in lIOs, lIO density, power,
and acceptable cost for board and wiring dimensions. The packaging
technologies summarized in this chapter as well as the details of the rest
of the technologies required to meet these needs is what the book is all
about. All these technologies from chip-level connections to the first-level
package and to the printed wiring board is referred to as the packaging
hierarchy, as illustrated in Figure 7-5. The evolution of this hierarchy is

Table 7-4A. Packaging Trends in Consumer Products

1996 2000 2004

110 per chip 100 200 375


110 density/in. 2 100 200 375
Area array pitch (mm) 0.4 0.25 0.20
Board lines/spaces (mil) 5 4 2
Power/chip (W/chip) 1
Board cost ($/in. 2) 1.0 0.6 0.4

Source: Courtesy of NEMI.

Table 7-4B. Packaging Trend in Automotive Electronics

1996 2000 2004

110 per chip 150 200 500


110 density/in.2 30 100 150
Area array pitch (mm) 0.4 0.25 0.20
Board lines/spaces (mil) 6 3 2
Power/chip (W/chip) 6 12 15
Max. board temp. (DC) 125 155 170
Board cost ($/in. 2 ) 1.0 0.70 0.40
(6 layers)

Source: Courtesy of NEMI.


11-64 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-4C. Packaging Trend in High-Performance Systems

1996 2000 2004

110 per chip 600 1,500 3,000


110 per MCM 3,000 6,000 10,000
110 density/in.2 150 300 400
Board lines/spaces 3 2 I
Power/chip (W/chip) 40 60 80
Board cost ($/in. 2) 10 5 2.5
(30 layer)

Source: Courtesy of NEMI.

depicted in Figures 7-34 and 7-35 from dual-in-line packaging (DIPS) in


the 1970s to quad flat package (QFP) in the 1980s to ball-grid arrays in
the 1990s and to a single-level integrated module referred to as SLIM in
the next century. In contrast to all the current and previous packages
which not only have a complex hierarchy involving two or more levels
and the interconnections between them, but also 5-10 components such
as resistors, capacitors, inductors, filters, switches, RF and optoelectronic
and so forth, the SLIM package under development at the Georgia Institute
of Technology is a single hierarchy-level and single component system-
level package. Table 7-5 illustrates this trend in packaging hierarchy,
number of levels required to complete a high-performance system, and
the ultimate measure of packaging referred to as silicon efficiency, defined
as the total area of all silicon chips divided by the area of the board. The
silicon efficiency of 2% achieved in the 1970s is both due to the large
size of the package as well as the area occupied by discrete components.
The BOAs and chip-scale packages currently in development will increase

-00

SLIM
19705 19805 19905 2000 <=C>

Figure 7·34. Packaging Evolution


7.3 PACKAGING TECHNOLOGIES 11-65

__________~S~m~o~lI~
er~S~ e_________
iz~ I.
11_"
U'- .0"---. SOP --
s-sop
I SOJ
QFJ
" (PLCC)

..
u
...
~
a
r
~
~ PIH 1000

I
" CSP

Year 1970.
DIP/PGA

1980. 1990.
~ SLIM
2000-2010

Figure 7-35. Single-Chip Package Evolution Leading to Multichip Packaging

Table 7-5. Packaging Evolution

Past Current Future


1970s 1980s 1990s 2000 2005

Chip connection Wire bond Wire bond Wire bond Flip chip Low-cost flip chip
Package DIP PQFP P/C-BGA None
Package assembly PTH SMT BGA-SMT None
Passives C-discretes C-discretes C-discretes C-discretes Integrated
Board Organic Organic Organic DCA to board SLIM
No. of levels 3 3 3 1
No. of types of
components 5-10 5-10 5-10 5-10
Si efficiency (%) 2 7 10 25 >75

Source: Courtesy of Georgia Tech.


11-66 MICROELECTRONICS PACKAGING-AN OVERVIEW

~ J, /. ,
R _R /! \ / i\
,JU"r-ellinn-in-ation Resistor • < Assembly
W Ii :r.---Integrated R
«l ....
Optical Wtye,puide ••--........_
~ Integrated 0
----y---- - ...,.......-Integrated L
Dec. Cae ~I§~ --oiiiiiio....
["""! r-integrated C
...--GND

Ceramic or PWB

Power 110

Figure 7-36. Single-Level Integrated Module (SLIM) (Courtesy of Georgia Tech.)

the efficiency to about 25% because discretes are still on the board. The
SLIM package illustrated in Figure 7-36 integrates all or most of the
discretes as well as RF and optical components into a single-level board
to which the chip is attached in area-array fashion. This package, conse-
quently, provides packaging efficiency in excess of 75%, perhaps close
to 90%, thus maintaining almost the on-chip transistor density. This is
referred to as "Moore's Law of Packaging" as illustrated in Figure 7-8
in both Si efficiency or in transistor density on board.
Table 7-6 illustrates a more detailed evolution for each of the major
packaging hierarchy technologies. Interconnections at the chip level were
and are being achieved by wire bonding to plastic or ceramic single-chip
packages which are then bonded to printed wiring boards using surface-
mount technology (SMT). The trend is toward flip chip and ball-grid array
(BGA) in the short term, and to direct-chip attach to the board in the long
run. The board will be fabricated using fine-line photolith via as contrast
to drilled via in laminates and to photolith wiring on greensheets in
ceramic boards.
There appears to be a paradigm shift from inorganic to organic
packaging; however:

• Move to nonhermatic packaging


• Direct-chip attach to organic board
• BGA-SMT reliability of ceramic packages in contrast to plastic
BGAs
• Cost and size requirements driving to a large organic board with
thin film fabrication
7.3 PACKAGING TECHNOLOGIES 11-67

Table 7-6. Packaging Technology Evolution

Chip interconnection
Wire bond ..... TAB ..... Flip chip
Packaging interconnection
PTH ..... SMT ..... Fine-Pitch SMT ..... BGA SMT ..... CSP SMT
Single chip
DIP ..... QFP ..... TTQFP ..... BGA ..... CSP
Multichip
Ceramic ..... Thin film ..... Thin film on PWB/Ceramic ..... SLIM

Source: Courtesy of Georgia Tech.

7.3.1.1 Packaging in the Past

The packaging technologies that span the microelectronics industry


from consumer electronics to low-end systems to high-performance sys-
tems are very diverse, as illustrated in Table 7-7. Several important obser-
vations can be made from this table. The number of chips needed to form
a system in the past increased from a few, in consumer electronics, to
several thousands, in supercomputers. Given this, the packaging hierarchi-
cal technologies necessary to interconnect all these chips became complex;
consumer electronics typically required a card or flexible-circuit carrier,
and supercomputers required several boards, each containing several cards
or multichip modules. In addition, these boards needed to be intercon-
nected by high-speed cables and connectors. NEC systems, for example,
used as many as four levels of packaging hierarchy to complete the system.
They are, referring to Table 7-7, and Figs. 7-32 and 7-33, flip-TAB carrier
(FTC, first level) bonded to 100-mm multilayer ceramic substrate (MLC-
MCM, second level) plugged into polyimide-glass board (third level)
interconnected by cables and connectors (fourth level) [20], as illustrated
in Figs. 7-32 and 7-33.
The power and cooling capabilities of these systems range from no
special provisions for cooling in consumer electronics to very complex
cooling technologies in mainframes and supercomputers. The driving
forces for the latter stem from the fact that the chips need to be placed
and bonded very close to each other in order to minimize signal delay in
addition to the use of high-speed high-power chips.
The main emphasis in consumer electronics is cost. Given the need
for smaller functions to be packaged, chips that are wire-bonded or tape-
automated-bonded (TAB) to such low-cost plastic carriers as plastic leaded
chip carriers (PLCC) and plastic dual-in-line (DIPs) packages on cards
with minimum internal planes generally suffice. Space limitations in these
11-68 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-7. Packaging Technologies in Some Previous Electronic Products

1st 1st to 2nd 2nd 2nd to 3rd 3rd Max.


Chip Level Level Level Level Level Chip Chipsl
Conn. Package Conn. Package Conn. Package Cooling Systems

Sony CD WB PSCM SMTI Card <10


PTH
TFfY WBITAB PSCM SMTI Card <10
PTH
Apple PC WB PSCM SMTI Card Conn. Board lO's
PTH
HP Work- Wire- P-SCM SMT Board Conn. Air lO's
station bond
IBM 390 C4 C-MCM PGA FR-4 Conn. Cable Water 1000's
Mainframe
NEC SX-4 TAB FfC SMT LCM PTH P-G Water 1000's
Supercomputer Board

Key:
C-MCM: Ceramic multichip module PSCM: Plastic single-chip module
Conn.: Connector PTH: Pin-through-hole
FR-4 Board: Epoxy-glass board SMT: Surface mount technology
PC: Personal computer TAB: Tape automated bonding
PGA: Pin-grid array TFfY: Thin-film television
P-G Board: Poly imide-glass board WB: Wirebond

systems have brought about the need to package these components with
surface-mount technology with lead spacing as low as 0.4 mm. In some
circumstances, a flexible-circuit carrier is used that utilizes fully the three-
dimensional space available in small systems, such as the computer termi-
nal illustrated in Figure 7-37.
The low-end systems, such as personal computers and printers, need
additional functions, thus requiring increased sophistication in package
technologies. Surface-mounted as well as pinned components containing
wire-bonded or TAB chips are assembled on cards which are then plugged
into back panels or boards. Power requirements are generally low, not
requiring special provisions for cooling of components, in addition to
heat dissipation by convection and by conduction through the structural
members. In applications that require higher-power dissipation, as in
power supplies or motors to drive typewriter hammers, chips can either
be backbonded with adhesives or metal eutectics to a thermally enhanced
epoxy-glass or metal core substrates or, alternatively, to TAB chips with
heat sinks on the back of the chips. As in consumer electronics, the use
7.3 PACKAGING TECHNOLOGIES 11-69

Figure 7-37. Flexible Circuit Carrier (After R.K. Hayes, IBM Raleigh.)

of a flexible-circuit carrier to provide design freedom and to minimize


cables and connectors is being explored (Figure 7-37).
Midrange systems and workstations that contain a few dozen chips
required single-chip or small multichip packages with 1I0s in excess
of 200. Surface-mounted chip carriers or pinned modules are generally
mounted on cards which are then plugged into boards. The use of single-
chip packages, particularly single-chip ceramic packages, with closely
spaced 110 connections is often extended to high-performance systems
such as mainframes and supercomputers. The large computers of Fujitsu
and Hitachi, for example, use single-chip ceramic modules surface moun-
ted to very sophisticated boards. In contrast, IBM and NEC computers
use very sophisticated multichip pin-grid modules that are plugged into
large multilayer boards. The single-chip and small-number multichip-
modules approach generally allows air cooling to achieve the desired
power dissipation (except the Fujitsu M780 which uses water), whereas
the large-number multichip-module approach, with its very close chip-
to-chip spacing, demands power densities beyond the capabilities provided
by air cooling. The closer chip pitch, as in IBM's Thermal Conduction
Technology with a center-to-center chip spacing of about 9 mm, accounts
11-70 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-8. Packaging Technologies and Processes

Typical
Technology Technology Typical Typical Process
Function Options Materials Process Temp. {oq

Connection to chip Wirebond Gold, aluminum Wirebond 225


Solder bond (C4) Pb-Sn Reftow 360
TAB Copper, gold, Thermo- 550
aluminum, compression
polyimide
First-level package High temperature Ah03' SiC, BeO Sintering 1500-2000
Low-Temp. Ceramic Glass-ceramic Sintering 900
Plastic Epoxy Molding 200
TAB Cu on Kapton* Adhesive- 200
Bond
Molding
First-to-second- Surface mount Pb-Sn Reftow 220
level Connection Solder Kovar, Pb/Sn Reftow 220
Pin-in-hole solder Kovar, AU/Sn Braze 400
pin-braze
Second-level Card Epoxy-glass Cure 200
package Metal carrier Glass on steel, Invar Fuse 1000
Hex Cu on Kapton Adhesive- 200
Inj. mold. card Resin Bond 200
Molding
Third-level Board Epoxy-glass Cure 175
package Pol yimide, glass Cure 200
Second-to-third Connector Polymer, BeCu Cure 200
level connection Cable Polymer, copper Cure 200

*Kapton is a trademark of Dupont Company.

for its very high performance while handling high-power densities. Thus,
power densities and performance generally go hand in hand.
The packaging technologies required to form a logic function or
gate described earlier, involving three levels of package as well as an
interconnection between these and the chip, are illustrated in Table 7-8.
These technologies, together with thermal technology for cooling the chip,
and sealing and encapsulation technologies required to protect the package
and device circuitry, form the basis of packaging-materials science and
technology. The basic requirements, the status of, and the future direction
for each of these technologies are summarized in this section. Detailed
discussions and additional references can be found for each technology
in the appropriate chapter.
Because each packaging level is manufactured independently of the
7.4 CHlp·LEVEL INTERCONNECTIONS 11-71

other, the selection of materials and processes, including the process


temperatures for fabrication of the package, is determined for that level.
However, the packaging hierarchy does impose thermal-hierarchy consid-
eration in the assembly of the total package system. For example, if the
first level requires a brazed pin in forming a pin-grid array module, the
maximum temperature at which the chip attachment (to the first-level
package) can take place is the lowest temperature at which the braze is
dimensionally stable. Assuming the chip joining is done by the use of
conventional Pb-Sn solders, thermal-assembly-hierarchy considerations
require that the connection between the first and second levels is done at
yet a lower temperature than the temperature at which the chip solder is
stable. If a third level exists in the function, further thermal hierarchy is
extended. Rework of each of the packages and chips requires similar
considerations as well. Although in principle the thermal-assembly hierar-
chy could be set up in either direction of the packaging hierarchy, it is
determined by the maximum temperature a particular level of packages
can withstand. Because most conventional second- and third-level pack-
ages are made of organic cards and boards that are limited to about a
250°C joining temperature, the assembly of the package system involving
these is done as the last and lowest temperature operation.

7.3.1.2 Packaging in the Future

As indicated above, the future systems for the most part are either
portable, wireless, network or other systems with the same packaging
attributes as these. The system-level attributes include weight, cost, and
system functions as indicated in Figure 7-6.
The second important change driving packaging technology is the
type of semiconductor used to form the systems of the future. In the past,
bipolar was the primary engine of high-performance systems, and CMOS
met all other requirements. As illustrated in Figures 7-7 and 7-9, CMOS
provides the lowest cost and highest integration while it approaches the
performance of a bipolar but at much lower cost at the system level [22].
As a result, CMOS becomes the most perversive technology from PCs
to workstations to mainframes. The packaging in the future, therefore,
must be consistent with this evolution.

7.4 CHIP-LEVEL INTERCONNECTIONS

Connections between chip and package are commonly performed


by one of three technologies: wirebond, solder, or controlled collapse
chip connection (C4), also called "flip chip (FC)", or "solder bump," and
tape automated bonding (TAB), depending on the number and spacing
of 110 connections on the chip and the substrate as well as permissible
cost. The possible number of connections per chip with each of the
/1-72 MICROELECTRONICS PACKAGING-AN OVERVIEW

C-4

f
.c
~
TAB
Eo<

Wirebond

10 100 1000 10000 100000


Connections/Chip

Figure 7-38. Chip-Connection Technology Status

technologies is illustrated in Figure 7-38, where the arrow indicates the


current technology in manufacturing (1995) and the higher number indi-
cates the technology feasibility.

7.4.1 Wirebond
Wirebonding is the most common chip-bonding technology, span-
ning the needs from consumer electronics to mainframes. Its widespread
use is primarily based on the fact that the maximum number of chip
connections in products in use can be accomplished with this technology
in addition to providing the lowest cost per connection. It should be
pointed out that wirebonding concurrently provides for thermal dissipation
by backbonding or diebonding the chip to the substrate. The semiconductor
trend for the next decade as indicated in Table 7-2 illustrates the leading-
ledge 1I0s, lithographic dimensions, chip and wafer size, and power
requirements for portable and high-performance semiconductor chips.
The maximum number of lIOs in the leading-edge high-performance
semiconductors in Table 7-2 calls for as many as 5000 lIOs on a single
chip, which is best accomplished by area-array connections such as flip
chip using solder or conductive adhesives. There are several reasons for
this. First, referring to Table 7-1, the connection density (lIO/chip size
in mm) increases by a factor of 2 (from 750/400) to 4 (500011250). Thus,
if leading-edge wirebonding is to meet today's requirement, it needs to
be enhanced by a factor of 4 in pitch-a major challenge. Second, the
peripheral pad connections on the edge of the chip, particularly in large
7.4 CHIP-LEVEL INTERCONNECTIONS 11-73

Table 7-9. Materials for Chip Connections

Flip Chip TAB Wirebond

Metal Solder, conductive adhesive Copper on polymide Gold, AI-Si


Process Melt/rellow Thermo and Compression Thermo/ultrasonic
Chip metallurgy Cr/CulAu TiIW/Au AI, TiIW/Au
Substrate metallurgy NilAu Pb-SnlAu NilAu

Source: From Ref. 23.

chips, do not provide optimal power distribution to all the transistors on


that chip. Perhaps, the most important limitation of wirebond has to do
with the economics of chip usage itself. Because wirebonding cannot be
performed easily in manufacturing over the active area of chip compared
to area-array solder flip-chip connections, the size of the chip [21] is
larger by about one-third with the wirebonded chip. This difference will
account for tens to hundreds of dollars, as it directly relates to the number
of chips per wafer. In addition to the above reasons, the reliability of flip
chip in Table 7-10 has been shown to be three orders of magnitude
better than wirebond. Because the leading-edge products constitute a
small fraction of all semiconductors, all three chip-level interconnection
technologies will be used and enhanced. The basic materials and processes
that are currently in use in these technologies are indicated in Table 7-9
[23], whereas the electrical and other parameters, including reliability,
are illustrated in Table 7-10 [23].

Table 7-10. Chip-Level Connection Parameters

Wirebonding TAB Flip Chip


Conn. Metallurgy Al Au Cu Pb/Sn

Resistance (0) 0.035 0.03 0.Q2 0.002


Inductance (nH) 0.65 0.65 2.10 0.200
Capacitance (pF) 0.006 0.006 0.04 0.001
110 density (cm-2) 400 400 400 1600
Rework Poor Poor Poor Good
Failure rate (%/1000 h) Ixlo-5 IxlO-5 N/A <lxlO-3

Source: From Ref. 23.


11-74 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-11. Chip-Level Interconnection Development Trend for Wirebond


(WB), TAB, and Flip Chip (FC).

1990 1995 2000 2005

WB LSI chip 15 x 15/700 20 x 20 11000 20 x 20 11200 25 x 25


size (mm) 110 count
Wire pitch 105 70 60 50
ball bond (~m) (line) (staggered) (staggered)
Wedge bond (~) 100 65 55 50
TAB LSI chip 15 x 151500 20 x 20/1000
size (mm) 110 count
Inner lead pitch (~m) 100 75 50 30
Outer pitch bond ~m) (wire bump) (Plating bump)
100 75 50 30
Fe LSI chip 15 x 151900 15 x 151 1000 20 x 20 25 x 25
size (mm) 110 count
Bump pitch (~m) 250 250 200 150
Bump diameter (~m) 130 125 100 75

Table 7-11 demonstrates the chip-level developments underway in


all three technologies-wirebond (WB), tape automated bonding (TAB),
and flip chip (FC). The development goals for wire bond by ball-bond
and wedge-bond technologies at a pitch of 50 Jlm and for TAB at a pitch
of 75 Jlm are capable of providing 1200 1I0s on a 20 x 20-mm chip using
a staggered pin configuration.
In addition to the thermal-hierarchy considerations discussed pre-
viously, an important factor in deciding the chip-connection technology
is the lead inductance, resistance, chip rework, and failure rate as indicated
in Table 7-10 that a particular technology provides; wirebond having the
highest value and C4 the lowest. Detailed discussions on this and other
aspects of chip-level interconnections can be found in Chapter 8, "Chip-
Level Interconnections."

7.4.2 TAB
The progress and trend in TAB listed in Table 7-12 show two or
more layers with an inner lead bonding pitch of 60 Jlm and an outer lead
bonding pitch of 90 Jlm. On a 28 mm size chip, these leads provide in
excess of 1100 1I0s. TAB has been a major accomplishment in Japan
but not in the United States. Various advancements in TAB technology
are being pursued by such Japanese companies as Shinko-Denshi, NEC,
Fujimitso, Mitsui-Kinzoku, Oki, and Nitto Denko. One particular enhance-
ment of TAB being pursued by Nitto Denko is illustrated in Figure
7.4 CHIP-LEVEL INTERCONNECTIONS 11-75

Table 7-12. TAB Trend

1990 1995 2000

TAB Tape Signal layer 1 layer 2 layer 2 layer


Max. tape width 35 mm 70mm 70mm
Cu thickness min. plating 35 11m 18 11m 18 11m
Min. pattern Inner lead pitch 100 11m 60 11m 60 11m
Outer lead pitch 200 11m 90 11m 90 11m
Pattern pitch 100 11m 60 11m 60 11m
Max. height 1.0mm 0.7 mm 0.6mm
28 mm Max. Pin count 520 pins 1152 pins 1152 pins

Source: From Ref. 24; courtesy of Oki.

7-39a, comparing the new two-layer direct copper bonding process with
a conventional three-layer process. The new process coats polyimide onto
copper, the opposite of the 3M process in the United States that coats
copper on Kapton or other polyimides by electroplating. The advantages
claimed for this new process, shown in Table 7-13, include high heat

Table 7-13. Advantages of the Nitto Process

• High Heat Resistance ~ No Adhesive Layer


• No Curling ~ CTE = 16 ppm (0--3OO°C)

Kapton-H

CTE (ppm) 16 20 8 16
. . 24 hours water
• Low MOIsture AbsorptIon ~ 1% immersion at 230C

Nitto Kapton-V Kapton-E

(%) 1.0 3.1 2.4 1.3

• Strong Peel Strength ~ 6.7 lbs/in (1.2 kg/cm)

Nitto 3M
Peel Strength 6.7 4.0 (lbs/in)
Minimum Value 1.2 0.7 (kg/cm)
1I~76 MICROELECTRONICS PACKAGING-AN OVERVIEW

(a)
Nitto Process Others

2 layer substrate 3 layer substrate

I Polyimide I Polyimide
I Copper I
r

Unique features:
• 2 layer substrate - no adhesive layer
• Polyimide-synthesized by Nitto

• Polyimide onto Cu by coating (Nitto)


• Cu onto polyimide by electroplating (US)

(b) I. Polyimide coating onto Cu foil


IE;'Zll a' 2. Cu pattern making by etching
i} J
'iti/"1l11li\!1 'M! M' 3. Fine hole making into poly imide layer

IM""I'M' '. .' 4. Metal filling into hole on Cu pattern

lel!II!!.' 1& &' 5. Bump making: Ni,Au, Solder, Cu, etc.


i}

Copper lead

Figure 7-39. Nitto Process (a) TAB and (b) Bump (Courtesy of JTEC.)

resistance, low moisture absorption by proper selection of polyimide, and


better adhesion. There are other advantages to using this new process (I)
a very high aspect ratio-25-/lm-diameter holes in a 50-/lm-thick film;
(2) large area processing (300 x 300 mm); and (3) complete wiring patterns
(both vias and lines). Nitto is also applying this technology for burn-in
and electrical testing, as illustrated in Figure 7-39b.

7.4.3 Flip Chip


Flip-chip technology is being extensively studied by almost all major
companies and is already used in products by Hitachi, IBM, and Delco.
7.4 CHIP-LEVEL INTERCONNECTIONS 11-77

Soldered

Solder bump
(minimum)
250 mm pitch)
seal
Thin film layers
(5 wiring layers,
Mullite ceramic base board 1 resistor layer)
(7 conductor layers)

Solder bump
(450 )lm pitch)

Figure 7-40. Microprocessor Carrier (BGA) for LSI (Courtesy of Hitachi.)

The Hitachi flip chip, together with microcarrier BGA assembly to next-
level (mullite glass-ceramic) package, is illustrated in Figure 7-40 [25].
The microcarrier, which is only larger than the chip itself by about 2 mm,
and henced can be called a chip scale package, is a single-chip carrier
fabricated with seven layers of mullite ceramic and first levels of polyim-
ide-aluminum thin-film technology. Flip-chip enhancements being pur-
sued in the industry generally consist of one of two approaches-solder
bonding (including Pb-Sn, Pb-In) and conductive adhesive bonding. Bump
technology itself, like Fujitsu's bump integration technology (BIT), is
generating considerable interest. One example is illustrated in Figure 7-
41 using thin-film and electroplating processes. In contrast, the Germans
are pursuing electroless-plate bumping, and the British are trying gold-
ball bumping by wire-bonding tools. Bumping, because of its high
cost has received considerable attention leading to various processes
as illustrated in Figure 7-42. These include IBM's original process
based on high lead solder, followed by high lead solder on chip and
eutectic on board, elimination of high lead with the substitution of
eutectic on chip and conductive adhesives on substrate. Low cost
bumping processes, as illustrated in Figure 7-43 are also being pursued.
The biggest breakthrough in flip-chip bonding is the technology IBM
(Japan) pioneered as an extension of IBM (U.S.) flip-chip technology
developed three decades ago [25], It involves direct bonding of a bumped
chip to a PWB by the use of low-temperature solder that is hot-injection
deposited onto PWB through a mask. The challenge here is to develop
a thermally compatible encapsulant to reduce the strain on the solder joint
arising from the great mismatch in thermal expansions between the PWB
(17 ppm) and the chip (3 ppm). This invention together with fine-line
thin-film technology is capable of revolutionizing packaging technology
11-78 MICROELECTRONICS PACKAGING-AN OVERVIEW

/AuPad
I
-Wafer
--------------------------------

",
I I
",

l j. . ---Bum p

I I

Figure 7-41. Bump-Fabrication Process (Sharp)

for minimizing the number of package levels and reducing size, weight,
and cost.
Figure 7-44 illustrates a tenfold strain reduction when the encapsulant
is used between the PWB and the chip. This discovery has major implica-
tions for the Japanese packaging industry, particularly for consumer elec-
tronics, as it allows Japan to continue to use its existing infrastructure
for PWB. Figure 7-45 shows the eutectic solder to be more effective than
high-Pb solder (95/5) in achieving the desired fatigue life.

7.5 FIRST-LEVEL PACKAGES


7.5.1 Single-Chip Packages
A summary of all first-level packages, both plastic (including TAB)
and ceramic, is given in Table 7-14, wherein each package and its accepted
abbreviation is described in terms of the insulator material it is made of
and the maximum number of 110 connections and spacings that were
available in 1995 as well as year 2000 projections. These packages as
illustrated in Figure 7-46 and as illustrated in Table 7-14 fall into six
[ - - - - ---~

C 1 ~~__~~\~utectic -]
High U U Hi/Au o 0 High Pb-Sn ~«R 0
..... Pb-Sn '-Pad
o
co o 0 Eutectic or
Bump 1 = = 1
'---------'1 Solder '-------~I~S~R \CWJ GDi
IBM'. Original Proces. IBM Japan'. Process Latest Proces.es
Figure 7·42. Advances in Flip-Chip Bumping
11-80 MICROELECTRONICS PACKAGING-AN OVERVIEW

(b) Solder Bolls Forming

".--- LSI or Package


(c) Heat-Pressing ..

',"
"
-, '
..._--
(d) Solder Bolls Tronscripfing

~
.......
, Solder Bump

"; I. •

Figure 7-43. Low-Cost Bumping (Courtesy of Fujitsu.)

types: (1) those with pins often referred to as pin-grid array (PGA) for
pin-through-hole or butt surface mounting, (2) those with lead-frame leads
that are meant for surface mounting, (3) those with pads or balls on area-
array pads that are called ball-grid arrays (BGA), (4) solder or other
columns on an area array called solder column packages, (5) TAB bonded
to the board typically with solder, and (6) chip-scale packages that are
bonded to the board by one of the above connection technologies, The
substrate materials these are made of can be plastic, ceramic, printed
wiring board, or thin film, The development of all of these first-level
packages over the last three decades was aimed at two categories: those
that contain one chip, namely single-chip modules (SCM), and those that
can support more than one, called multichip module (MCM), Multichip
modules sometimes support up to and in excess of 100 chips, as discussed
in Chapter 9, "Ceramic Packaging," and in Chapter 12, "Thin-Film Packag-
ing'" These latter ones are referred to as thermal-conduction modules
(TCM) or liquid-cooled modules (LCM) in this book. The chronological
7.5 FIRST-LEVEL PACKAGES 11-81

10~--------------------------------------'

Without Encapsulation
.....................
......................
--
S"

............
'ffl. 6"
................................
c
l!
- 4"
tJ)

2" With Encapsulation

o • • • • • • • • • • • • •
o 40 SO 120 160 200 220
Joint Height (micron)

Figure 7-44. Effects of Encapsulation on Strain in Solder (Courtesy of IBM


Japan.)

12~--------------------------------------,

...........
10' Without encapsulation 5195 ..•••.•.••...

..............................................
and 63/37 solders ••••••••••••

-'l!
~8

-
C 6'

t J) 4"

0%
. 25% 50%
• •
75%
•
100%
63137 63/37 63137 63/37 63137
Solder Composition

Figure 7-45. Effects of Encapsulation and Solder Composition on Strain in


Solder (Courtesy of IBM Japan.)
11-82 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-14. First-Level Single-Chip Packages and Their Characteristics

Minimum No.
of Interconnections
Available Future 110 Spacing
Package Package Materials 1995 2000 {mm}

Dual-in-line (DIP) Alumina ceramic, 64 2.54


plastic 64 2.54
Shrink DIP Plastic 64 1.77
Skinny DIP Plastic 64 2.54
Single-in-line (SIP) Plastic 21 2.54
Leadless chip carrier (LLCC) Ceramic 132 1.27
300 400 0.63
Plastic 180 1.00
Small outline package (SOP) Plastic 40 1.27
Leaded chip carrier (LCC) Plastic 84 1.27
144 0.63
Quad flat pack (QFP) Plastic 130 1.00
160 500 0.65
Ceramic 180 0.40
200 0.63
Very small periphery array Plastic 300 600 0.40
500 0.40
Pin-grid array (PGA) Alumina (single 312 2.54
chip)
>1000 1.27
Ceramic 2177 >5000 2.54
(multichip)
Plastic 240 2.54
>500 2.54
Tape automated bonding Plastic 300 0.50
(TAB) >1000 0.25
Ball grid array Plastic 300 >500 0.50
Ceramic 604 >1000 0.40
Chip scale packages Thin film 300-1000 >1000 0.5
Ceramic 300-1000 >1000 0.5
SLIM Thin film >5000 0.25

development of those packages as a function of single and multichip


1I0s is illustrated in Figure 7-47.

7.5.2 Chip-Scale or Chip-Size Packages


An improvement in packaging efficiency beyond BGAs is being
achieved by so-called chip-scale or chip-size packages (CSP) which are
hardly bigger than the chip itself. Similar to QFP and BGA, CSP offers
7.5 FIRST-LEVEL PACKAGES 11-83

Through Hole Package Surface Mounted Package

so Of sOP
DIP (OUel In-llne
a g (Smell Out-llne
Peckage)
Peckage)

QFP
SH-DIP
h (Qued Ret
b (Shrink DIP)
Peckage)

C
SK-DlP, SL-DIP
(Sldnny DIP,
Slim DIP)
0 LCC
(Ludl...
Chip c.rrter)

PLCC, SOJ
SIP (PI_tic LNded
d (Single In-llne Chip Center with
Pecuge) Butt Leeds)

BGA
ZIP k (Bell Grid Arrey)
(Zlg-ug In-lln.
e P.cuge)

PGA (Pin Grid


f Arrey) Of Column
P.ckage m

Figure 7·46, Types of First·Level Packages

bum-in and testability of ICs prior to joining to the printed-wiring board.


There are a number of types of these packages (Fig. 7-48) that include
such rigid substrates as ceramics with straight-through vias, flex or TAB-
like connections with bumps or balls on either end and wafer-level assem-
bly such as redistribution of peripheral IIOs to area array IIOs on chip
and lead-on-chip (LOC). Two of the most popular of these packages are
by Tessera and by Hitachi. The heat transfer characteristics of these
packages range from 0.5 to 10°C/watt.
First-level packages, in addition to providing the required number
of contacts for power and signal transmission, need to have the desired
number of wiring layers discussed previously, provide thermal expansion
compatibility with the chip, provide a thermal path for heat removal from
the chip, and keep signal transmission delay and electrical noise to a
minimum. Package sealing, to provide protection of both the package
metallurgy and chip metallurgy, is yet another important function of the
first-level package. Minimum electrical delay requires a low dielectric
constant. Low electrical noise requires low self-inductances and low inter-
line capacitances and inductances. Maximum heat removal requires high
11-84 MICROELECTRONICS PACKAGING-AN OVERVIEW

10000

II 2000
"3 1800 (90MMjO
-0
::J
l: 1000

.
~
I: 400
Multi Chip
iL

j 300
E
::J
Z 200 PGA-50 Mil Grid
TAB (169)
QFP
100 SLIM: Single Level Integrated Module
PLCC (84) SCIM: Single Chip Integrated Module

0'64 70 '75 '85 '90 '95


Yeor

Figure 7-47. Chronological Development of Single-Chip and Multichip


Modules

Example CSPType Companies

General Electric
Flex circuit interposer NEe
Nitto Denko
Tessera

IBM, Kyocera
Rigid substrate Matsushita
interposer Motorola
Toshiba

Fujitsu
Hitachi
Custom lead frame LG Semicon
Mltsublshi Electric
Rohm

ChipScale
Wafer-level assembly ShellCase
Sandia

Figure 7-48. Chip-Scale Packages


7.5 FIRST-LEVEL PACKAGES 11-85

Table 7-15. Properties of Package Insnlator Materials.

Thermal Approximate
Dielectric Expansion Thermal Processing
Constant Coefficient Conductivit:t TemQerature
1O-7/oC W/mK °C

Non-organics

92% Alumina 9.2 60 18 1500


86% Alumina 9.4 66 20 1600
Si/V, 7 23 30 1600
SiC 42 37 270 2000
AIN 8.8 33 230 1900
BeG 6.8 68 240 2000
BN 6.5 37 600 >2000
Diamond
High pressure 5.7 23 2000 >2000
Plasma CVD 3.5 23 400 ",1000
Glass-Ceramics 4-8 30-50 5.0 1000
Copper Clad lnvar 30 100 800
(10% Copper)/
(Glass Coated)
Glass Coated Steel 6 100 50 1000
Organics
Epoxy-Kevlar (x-y) 3.6 60 0.2 200
(60%)
Polyimide-Quartz 4.0 118 0.35 200
(x-axis)
Fr-4(x-y plane) 4.7 158 0.2 175
Polyimide 3.5 500 0.2 350
Benzocyclobutene 2.6 300-600 0.2 240
Tefton®* 2.2 200 0.1 400

*Tefton is a trademark of Dupont Company.

thermal conductivity. Maximum power distribution requires high electrical


conductivity of package metallization, and high reliability requires a close
thermal-expansion match between the chip and substrate. The thermal
and electrical properties of substrate materials in use in 1995 are shown
in Table 7-15 for insulating materials and in Table 7-16 for conductors.
The trend in the 1990s is toward glass±ceramic and organic materials
with low dielectric constants. Thin-film multilevel wiring embedded in
thin-film dielectrics of these materials and other low-dielectric-constant
materials are important in the minimization of transmission delays that
will be required in the 21 st century.
11--86 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-16. Properties of Package Conductor Materials

Thennal
Electrical Expansion Thennal
Metal Melting Point Resistivity Coefficient Conductivity
°C l~n·cm 1O-7/°C W/m·K

Copper 1083 1.7 170 393


Silver 960 1.6 197 418
Gold 1063 2.2 142 297
Tungsten 3415 5.5 45 200
Molybdenum 2625 5.2 50 146
Platinum 1774 10.6 90 71
Palladium 1552 10.8 110 70
Nickel 1455 6.8 133 92
Chromium 1900 20 63 66
Invar 1500 46 15 11
Kovar 1450 50 53 17
Silver-palladium 1145 20 140 150
Gold-platinum 1350 30 100 130
Aluminum 660 4.3 230 240
Au-20% Sn 280 16 159 57
Pb--5% Sn 310 19 290 63
CU-W[20%Cu) 1083 2.5 70 248
Cu-Mo(20%Cu) 1083 2.4 72 197

7.5.3 Single-Chip Package Markets


The worldwide package demand for the single chip packages dis-
cussed above is indicated in Table 7-17, which includes the application
for diodes, transistors, optoelectronics, and discretes, both for memory
and logic. Ceramic PGAs are used primarily for military and for high-
reliability commercial applications (PCs to mainframes) that require either
hermeticity or pluggability. Of the total ceramic PGA usage of about 150
million units, Intel alone uses about 50 million units for its 486, Pentium,
P6 and others, and IBM uses about 30 million units for PC, workstation,
midrange, and mainframe computers.
Plastic quad flat package is the most common package in use today,
exceeding 10 billion units worldwide in 1995 and doubling by the year
2000. A variation of this package as indicated in Figure 7-46 is used in
much smaller size with less pins [hence, on only two sides called small
outline package (SOP) or small outline J-Iead package (SOJ)], is actually
manufactured in higher volume, almost by a factor of 3, than PQFP.
Because Japan produces most of these packages, Figure 7-49 that lists
the Japanese production of these packages is of great use. It should be
7.5 FIRST-LEVEL PACKAGES 11-87

Table 7-17. Worldwide Package Demand by Package Type


(in Millions of Units)

PackagelYear 1994 1995 1996 1997 1998

CPGA* 151 166 168 181 203


PPGA 10 10 15 15 15
CQFP* 7.8 7.7 7.8 8.1 8.7
PQFP* 10,122 10,875 12,721 15,214 18,897
PBGA 14 17 32 41 80
CBGA <1 2 4-10 4-10 4-10
TBGA <1 2-8 2-8 4-10
TCP 0.5 1.5-2 2-4 2-4

Source: TechSearch International, Inc.


*VLSI Research, Inc.

noted that a number of these packages, particularly PGA, QFP, DIP as well
as BGA, are available in ceramics from Kyocera, NTK, Sumitomo, IBM.
Ball-grid arrays (BGA) are the natural outgrowth of flip-chip technol-
ogy and pin-grid-array connections, providing area-array connections on
a smaller grid. The need for this type of package arose for mainframe

10 20 30 40 50
DIP .1

Through-hole

-~~~
sOP/SOJ ....
••.•
17.0
TSaP
Surface Mount
QFPF=~·
TQFP

PLCC

LCC
MoIdodTAB

[[3
TapoTAB
Bar. o..p 1992

IGA
o 1995
o
Other

Figure 7-49. Japanese Packaging Production (Percent) (From Ref. 27.)


11-88 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-18. BGA Market Trends (Packages in Millions)

BGA Type/Year 1995 1998 2000

PBGA 17 100 200


CBGA 2 4-10 10--15
TBGA 4-10 10--15

Source: TechSearch International, Inc.

and supercomputer applications by IBM, Hitachi, and NEe. NEC applied


a modification of this package called flip TAB carrier (FfC) with solder
ball connections in 1985 and Hitachi has been using, since 1990, a mi-
crochip carrier (MCC) with 7 layers of thin film, and resistor wiring,
providing in excess of 600 balls or 1I0s. As discussed in the trade-off
section (Section 7.5.8) and as indicated in the chip-level interconnections,
the BGA strategy is clearly the current wave in single-chip packages,
providing cost-effective packaging solution at almost one-fourth in size
from current ceramic PGAs for microprocessor PC applications. Of the
dozens of companies currently evaluating the plastic or ceramic BGAs,
IBM, Motorola, Sun Microsystems, COMPAQ, Ericsson, and BULL,
among others, are using these in their products. Table 7-18 lists the market
trend that is expected of these packages.

7.5.4 Plastic Packages


Plastic package developments currently underway include the fol-
lowing

1. New molding materials and processes


2. Process controls and quality assurance
3. Fine-pitch interconnections
4. New lead frame materials
5. Multilayer plastic PGA packages up to 2000 1I0s

Table 7-19 lists the historical and projected development in molding


compounds by Nitto. Future developments are aimed at improving pro-
cesses for moldabi1ity into thinner packages, antidelamination, higher-
temperature solder resistance, and low stress.
In addition, the molding compound developments include composi-
7.5 FIRST-LEVEL PACKAGES 11-89

Table 7-19. Molding Compound Development in Japan

Discrete MP-2000 Modability

1970s Small DIP MP-3000 Moisture resistance, moldability


Middle DIP HC-IO-2 Moisture resistance, moldability
Large DIP, SOP MP 150S0 Low stress
1980s PLCC MP-180 Low stress
QFP MP-190 Low stress, processability
SSOP,TSOP MP-7000 Soldering resistance
SQFP, TQFP MP-7000 Soldering resistance
I 990s SOP, DIP MP-80 Antidelamination, processability
PLCC,QFP MP-8000 Floor life for soldering, low stress, high adhesion
Large package Target I Soldering resistance, low stress, moldability
Ultrathin package Target 2 Moldability, soldering resistance
Specific package Target 3 Adaptability to components

Source: Nitto Denko

tion and formulation, moldability, and temperature and humidity bias


reliability. These developments are summarized in Figure 7-50.

7.5.4.1 Process Controls

The increasing use of analytical techniques and micro sensors to fine-


tune materials and control process-induced defects falls into the category
of enhanced process control and quality assurance. X-ray radiography
and C-mode scanning acoustic microscopy-tomography are finding in-
creased use as nondestructive techniques for evaluating a number of
processes-plastic delamination; die metal, die attach, and bonding wire
deformation; die metal and wire-bond voiding; lead frame, die passivation,
die match, wire, wire bond, and case brittle fracture; and dendritic growth
under bias. Mercury porosimetry has also been used successfully to find the
number and size of epoxy and epoxy-metal pores in epoxy-encapsulated
packages. Epoxy pores are less than 0.2 /lm in diameter, epoxy-lead
frame voids are about 1 /lm in diameter, and surface pores range from 5
to 500 /lm in diameter. Piezoresistive strain gauges integrated into test
chips will continue to be used to directly measure the mechanical stress
in a plastic encapsulated module (PEM}-caused by encapsulation, die
bonding, and other factors, either during fabrication or under environmen-
tal stress testing. Solid-state moisture microsensors will also be used to
measure the moisture content at any specific location inside a PEM.
11-90 MICROELECTRONICS PACKAGING-AN OVERVIEW

Epoxy Molding Compound


I Research and Development 1
Reliability Moldability
I
11 Composition
1 (temperature, humidity bias) and formulation 1 1 (viscosity control)

Low stress, humidity resistance Resin flexibilation, hardener,


(megabit memories) coupling agent, etc.

Thermal conductivity for


moldability and reliability
(high heat dissipation devices)
1 Filler type and composition

Non post-Qlfe type with high glass


transition temperature (integrated Catalyst
circuit, large scale integration)

Non cleaning type for demoldability


and reliability (integrated circuit, Basic resin release agent
large scale integration)

Molding compound for multiplier use Low viscosity compound,


with high fluidity and optimized cure curing agent, additives
(integrated circuit, large scale integration)

Figure 7-50. Development Trends of Epoxy-Molded Compounds (From Ref. 24.)

7.5.5 Plastic PGA Technology

Even though plastic PGA is not used as extensively as QFP or SOP,


it is clearly the leading-edge technology providing very high IIOs with
multilayer structures. As shown in Table 7-20, Ibiden's plan for future
plastic PGA technology is expected to increase the pin count from the
current 500-600 pins to 2000 pins by the year 2000. This increase in pin
count is possible with multilayer plastic packaging technology that in-
cludes new materials with lower dielectric constant and high glass-transi-
tion temperature, finer-line wiring, down to 25 11m lines to 100 11m pitch,
and layer counts up to 15. Ibiden expects land-grid arrays (LGAs) and
ball-grid arrays (BGAs) to approach 600 pins before the end ofthe decade.

7.5.6 Lead-Frame Fabrication

Trends in lead-frame technology have important implications for the


future of molded plastic packaging. Japanese companies are replacing
copper alloys with 42Fe/58Ni (Alloy 42) as the lead-frame material,
7.5 FIRST-LEVEL PACKAGES 11-91

Table 7-20. Plastic PGA Technology Trends (lbiden) (From Ref. 24)

1992 1995 1998

Pin count 300-400 pin 1000 pin 2000 pin


Pin pitch 2.54 mm 1.5 mm (LOA) 1.27 mm
1.27 mm (POA)
B'g pitch 180 11m -120 11m -100 11m
Linewidth 100 11m -50 11m -25 11m
Trace thickness 18 11m -10 11m -10 11m
Multilayer 6 layers 10 layers IS layers
Via hole size (diarn.) 0.3 mm 0.1 mm 0.1 mm
Insulation-layer thickness 0.1 mm 0.06 mm 0.06 mm
Dielectric constant 4.4 3.S 3.5
Impedance control ±IS% ±1O% ±5%
TCE. (ppm) 14 10 10
Tg Point 180°C 200°C up 200°C up
Material BT New BT New BT

Source: Courtesy of Ibiden.

especially for moderate to high heat-dissipating devices such as processors.


For thin (:~O.15 mm) lead frames with close lead tips for very high lead-
count packages, chemical etching has replaced mechanical punching.

7.5.7 Logic and Memory Applications

The technology developments discussed above are expected to lead to


enhancing both memory and logic chips. A typical representative roadmap
from Oki for both these are included in Tables 7-21 and 7-22. The memory
roadmap indicates reduction in mounting and wire loop heights, leadframe
thickness and reliability improvements as a result of new materials. The
logic roadmap, in addition to indicating the role of plastic QFP, also
indicates high I/O achievability with ceramic PGAs and thin-film
packages.

7.5.8 Ceramic Packaging Technologies

The chronological development in ceramic packaging is indicated


in Figure 7-51, which illustrates the recent materials development from
alumina to AIN, mullite and glass-ceramic and future developments in
composite ceramics. The properties of the substrate materials currently
in use are indicated in Table 7-23. The process advances in cofiring in
11-92 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-21. Logic LSI Package Roadmap

Year {CY} 1990 1993 1996 1999 2002

LSI chip Gate density 200K 300K 500K 1M 2M


trend Max. chip size 155 mm 15mm 17.5 mm 20 mrn 25 mm
Max. power 1.5 W 2W 3W 5W lOW
Max. speed 50 MHz 150 MHz 300 MHz 500 MHz 500 MHz
Max. pin count 250 350 500 750 1000

Package QFP size (mm) 40 40 28 40


technology QFP pitch (mm) 0.65 0.5 0.4 0.3
C-PGA 110 401 526 750 1000
Size (mm) 40 32 36 40
Pitch (mm) 1.778 1.27 1.27 1.27
Thin package
110 1.44 216 344 2966
Size (mm) 20 24 28 24
Height (mrn) 1.2 0.5 0.4 0.3 0.3
Cooling Cu lead- QFP with QFP with QFP with
frame Fn heat cool
spreader module

Source: Courtesy of Oki.

terms of line and via dimensions and the number of layers that are cofired
are indicated in Figure 7-51. The state of the art in ceramic packaging
involves 50-100 Jlm lines and vias placed on 225-450 Jlm apart and the
substrate cofired with dimensional control better than to.l % in 63 layers
of metal and ceramic. It appears that these dimensions, dimensional con-
trols, and the number of layers can be improved technically by about a
factor of 2 within the next decade. Whereas ceramic packaging was used
primarily for either high performance or higher-reliability applications in
the past, Panasonic, Kyocera, and others have begun to apply to con-
sumer products.
Table 7-23 shows the variety of materials being used as ceramic
substrates. These materials include A1 20 3, AIN, mullite, and a variety of
glass-ceramics that include both glass added to alumina and crystalizable
glasses. Whereas most of these low-temperature ceramics are metalized
with Ag, Ag-Pd, or Au as fired in air, a few firms are beginning to cofire
with copper using special binders, or special atmosphere cycles, to remove
organics from greensheets. Table 7-23 also illustrates the properties of
some of the glass-ceramics being pursued by Kyocera, Panasonic, Oki,
Fujitsu, NEe, and NTK.
7.5 FIRST-LEVEL PACKAGES 11-93

Table 7-22. Memory Package (TSOP) Technology Roadmap

1990 1995 2001

Device Configuration Density 4M 64M 10


Die size (mm) 90 190 400
Package mounting height (mm) 1.2 1.2 0.8
Customer usage condition Storage I day usage
Rellow IR (220°C) IR (240°C)
Condition
Ratio of die size and package size (%) 64 79 80
Spacing between package edge and die 1.0 0.4
edge (mm)
Packaging technology Minimization Conventional LOC
Low profile 200 ~m max. 120 ~m max
wire loop
height
Lead-frame 0.15 ~m 0.15 !lm
thickness
Reliability (rellow resistance) Conventional Biphenyl Imid modified
molding epoxy resin (EOCN) epoxy (high Tg)

Source: Courtesy of Oki.

7.5.9 BGA Packages


There are several BGAs including micro-BGAs (/l--BGA) that are
in the horizon, as illustrated in Figure 7-52. These are packages with
reduced I/O pitches but still compatible with the leading-edge printed
wiring board technologies. These technologies and their I/O spacings are

400
'[
-;: 300 30
s
'iii ~
i 200 20 ~
E
o
g'100 10
~
0~----~19~8~0----~1~9~90~----~2000
Year

Figure 7-51. Ceramic Packaging Trends (Courtesy of Kyocera.)


11-94 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-23. Ceramic Substrate Materials

Glass-Ceramic
AIN SiO, +
Alumina Mullite (lbiden) Kyocera Pansonic IBM Fujitsu NEC B,O,

Thermal 18 5 180 2 2 2 2 2 2
conductivity
(W/mk, RT)
TCE. (1~1 7.0 4.4 4.5 4.0 3.0 4.5 3.5 2.5
0c)
40-400°C
Dielectric 10.0 6.8 8.9 5.0 7.4 5 5.6 4.4 4.0
constant (1
MHz, RT)
Tan Ii 24 10 20
(I X 10-4)
Conductor W,Mo W,Mo W Cu Ag,Cu Cu Cu Au Ag-Pd
metal
Sheet 10 10 15 3 2-5 2 3 3 2-5
resistance
(mQ/sq)

PBGA CPBGA

,Lui!uuuuuu2Vu:)
Plastic Ball Grid Array Cavity Plastic Ball Grid Array

TBGA
Flip Chip

Tape Ball Grid Array

CBGA Wire Bond CBGA


Flip Chip - Cap

fubuuuu3uuuulJiuuuuJ
Ceramic Ball Grid Array Ceramic Ball Grid Array

Figure 7-52. Various Types of BGA Packages


7.5 FIRST-LEVEL PACKAGES 11-95

Table 7-24. BGA Packages

Technology (Company) I/O Configuration I/O Pitch (urn)

Micro SMT (Mplus Microwave) Peripheral 100-150


m-BGA (Sandia) Part area array 500-750
~BGA (Tessera) Peripheral 500-1000
DCAM (IBM) Part area array 200-500
Slice (Motorola) Area array 825-1500
MC2M (Valtronic USA) Peripheral 1000-1500

indicated in Table 7-24. Some of these packages are also referred to as


chip scale packages, particularly in harsh environments_ While the BOA
trend continues, there are certain reliability issues to which BOAs are
prone_ Quad flat pack, although low in cost, is too large in size and too
low in thermal performance. VSPA (very small peripheral array) illustrated
in Figure 7-53 seems to offer the best of both worlds ofQFP and BOA [281-
The trend in the past has been from pin-through-hole to SMT as
electronic products shifted from mainframes to desktop. The most recent
trend, however, is from SMT to BOA, consistent with the product trend
from desktop to laptop to notebooks.

QFP 304 leads


BGA 313 leads VSPATI. 320 leads

36mmX36mm
3SmmX3Smm 27 X 27 mm (including leads)
Plastic QFP
Peripheral BGA Cavity Down VSPA
O.S mm pitch .S mm/.8 mm pitch (SO/SO%)
1.2 mm pitch
Easily Rewarkable Reworkable
Reworkable
4-6 layers required in PCB 4-6 layers required in PCB
6-8 layers required in PCB
Visual Inspection OK
No Visual Inspection Visual Inspectian OK
Normal SMT
Not "narmal" SMT NormalSMT
Thermal via array
Back side card OK Back side card OK
Back side NOT OK
$O.OII/pin $O.OI/pin
$0.02/pin

Figure 7-53. QFP, BGA, and VSPA™Package Comparison (From Ref. 28.)
11-96 MICROELECTRONICS PACKAGING-AN OVERVIEW

The advantages of BGA are as follows:

• Assembled with same SMT equipment as QFP


• Smaller area board and yet, larger VO Pitch
• Self alignment during soldering process as with flip chip
• Low assembly cost
• Repairable
• Component cost reduction
• Inspection unnecessary compared to QFP

7.5.10 Package Trade-offs


Figure 7-54 plots the cost of various single-chip packages as a
function of VOs. Given the cost-sensitive nature of electronic packaging,
plastic quad flat pack and single-layer TAB are the leading candidates,
compared to ceramic and plastic PGAs. This is already reflected in the
huge volume of these packages as described in Section 7.5.1, "Single-
Chip Packages." As pointed out earlier, the QFP is being improved upon
for both increased VOs by improving on lead-frame pitch from the current
0.4 mm to 0.15 mm and for thinness by reduction of molded or ground
thickness. In spite of these improvements, the problems with QFP are its
huge size and poor thermal and electrical performance. The popularity
of BGAs and their exponential growth discussed above can be attributed
to this deficiency. In addition, a new revolutionary, simple, and elegant
package known as very small peripheral array (VSPA) is beginning to
emerge as the best overall medium VO package from the Panda Project

~
011
U
.;:
iCI.

..
~
o
';:
011
o
011
01
o
.:.I.
U
o
iCI.

600 700 800 4000

Pin Count

Figure 7-54. Single-Chip Packaging Cost (Courtesy of Otsuka.)


[18]. As compared in Figure 7-53, this package is smaller in size than
QFP and BGA and yet provides far superior thermal performance, due
to its metallic heat sink to which the die is bonded. Because the heat sink
can be copper, thermal conductivity improvement of the order of 20 over
ceramic and 500-1000 over plastic packages can be achieved. The cost
per I/O is expected to be the same as QFP. Because of these advantages
and better electrical performance, this package may emerge as the most
competitive package to QFP.
Figure 7-55 and 7-56 compare the package weight and mounting
height on PWB for a number of single-chip packages as a function of
pin count. Only TAB and TSOP, besides QFP, BGA and VSPA, can meet
future competitive requirements for thinness and lightweight in packages.
As shown in Figure 7-57, a number of companies such as Hitachi
has selected QFP and TAB, as their front up packages for low to medium
1I0s. The only exception seems to be for very high 1I0s, where PGA and
BGA will be their strategic packages. There is general agreement that the
best alternative to QFP in low-pin-count consumer products and high-
pin-count computer products will be BGA or surface-mount PGA, the
new VSPA package. BGA is seen as a high-speed, high-pin-count package
that also provides a compact solution. Figure 7-58 shows how BGA
provides a smaller footprint at a 1-mm pitch than the ultimate 0.15-mm-
pitch QFP beyond 600 1I0s. This is one of the primary reasons for the
emergence of BGA.
11-98 MICROELECTRONICS PACKAGING-AN OVERVIEW

(mm)

~
IP
OS.DIP ZIP

...
.J:
DIP
PGA
01
·iii QFP
.J: SOP
01
C
:;:::
C
:::J
o
~ 1.0

o @)
0.3 0.5 0.8 1.0 1.5 2.0 3.0(mm)
, ,
o 40 50 70 100 (mil)

Lead Pitch

Figure 7-56. Lead Pitch and Mounting Height

In contrast to the inspection needs of fine-pitch QFP, manufacturers


like Hitachi and IBM, do not believe there is a need to inspect the BGA
joint, even though X-ray inspection may be possible. This is so because
of the large pitch that BGA provides. Hitachi also claims great flexibility
in circuit design using BGAs, allowing Vee and Vdd connection everywhere,

1000
t
PGA
800
TAB
t &
BGA
....C I I
:J
0 Range of applicability
u 600

11
c
.Q.
E
:J 400 QFP
E
.5(
&
CI
~ VSPA
200
+
0
'88 '90 '92 '94 '96 '98
Year

Figure 7-57. High-Pin-Count Packages (Courtesy of Hitachi.)


7.5 FIRST-LEVEL PACKAGES 11-99

50 0.3 mm Japan
0.4 mm U.S.
0.4 mm Eurape
1996
40
0.3mm QFP

0.2mm QFP
~
o 30
:l
0-
.. ' ....
'" O. ISmmQFP
.sE 20~--Y
.....

10 ........................
............................
O~~ __~____~__~__~____~~
100 200 300 400 500 600
Pin count

Figure 7-58. Relative Package Areas: BGA Versus QFP, Illustrating Superior-
ity of BGA

in addition to providing power and ground for each group of output


buffers, reducing the simultaneous switching noise. The QFP approach
does not provide this flexibility.
In parallel to surface-mount options, the electronics industry is pursu-
ing PGA options in limited fashion both in ceramics and plastics. Toshiba,
for example, is already pursuing an 820-pin ceramic PGA on a 60-mm
square ceramic substrate. Kyocera supplies a lOOO-pin surface-mount
PGA in ceramics with its improved PGA that compares favorably with
QFP, as illustrated in Figure 7-59. The pin-grid pitch in these packages
is 1.27 mm (50 mils). Toshiba chose TAB connection to the 20-mm chip
using gang inner lead bonding and single-point outer lead bonding. The
plastic PGA trend providing in excess of 1000 1I0s was discussed earlier
(Section 7.5.3).

7.5.11 First-Level Multichip Packaging

Multichip packaging is one of the fastest growing segments of all


the packaging technologies. It came into existence for mainframes and
supercomputer applications that required the most number of circuits to
be packaged in the least amount of space. These attributes are now expected
to be expanded into workstations, consumer electronics, medical, aero-
space, automotive and telecommunication technology. This subject is
11-100 MICROELECTRONICS PACKAGING-AN OVERVIEW

70
60 .........HM.H.l...

~
50
40
30
20
10
0
100 200 300 400 500 600 700 800 900 1000 1100

---....;). I/O's

Figure 7-59. Package Body Size Versus Terminal Count; SMT PGA Same
Size as QFP

discussed in more detail in Part 2, Chapter 9, Ceramic Packaging, Part


2, Chapter 12, Thin Film Packaging, and elsewhere [29].

7.5.11.1 Definition
Multichip packaging requires the definitions of multichip and pack-
aging. Packaging has been defined elsewhere as interconnecting, power-
ing, cooling, and protecting the semiconductor chips. Because single-chip
package is defined as interconnecting bare chips onto a single plastic,
ceramic, or some other carrier, the multichip package has been defined
as providing the same function with more than one bare chip on to a
single carrier or substrate. With this definition, however, there are very
few current manufacturers, such as IBM, that qualify as the multichip
fabricators, as there are very few companies that bond the bare chips
directly to the multichip substrate.
The benefits of multichip packaging, however, can be realized, al-
though not as fully as in direct bare-chip attachment, by bonding prepack-
aged chips in their single-chip carriers to the multi chip substrate. This is
the most common multichip packaging that is currently in use. Both of
the above types are illustrated in Figure 7-60. The multichip carrier on
which the chips are bonded can be ceramic, printed wiring board, or thin
films deposited on any substrate, including metals like silicon, copper,
and aluminum, and ceramics like AIN, diamond, and alumina, and organics
like polyimide-glass, composite organics-inorganics.

7.5.11.2 Functions of Multichip Package


Consistent with the above definitions of multichip and package, the
multichip package must perform the following functions:
7.5 FIRST-LEVEL PACKAGES 11-101

Multichip Multichip

a) 2 or More Bare Chips b) 2 or More Prepackaged


Directly on Multichip Chips on Multichip

Figure 7-6f). Multichip Module Defined as ~ 40% Silicon Efficiency

1. Remove heat effectively from the chips


2. Provide interconnections between all the chips with as many circuits
and as high a circuit performance as available to the multichip sub-
strate
3. Provide wiring density with high-conductivity metal to interconnect
all the chips with minimum chip-to-chip spacing
4. Provide multichip substrate connections for signal and power distri-
butions
5. Provide protection to all the chips and the multichip substrate itself.

The schematic in Figure 7-61 illustrates these functions.

7.5.11.3 Packaging Efficiency


High packaging efficiency results as a result of placing a large
number of chips closely spaced on an individual multichip package to
meet a certain function (e.g., processor on a module).

Close Chip Pitch

Figure 7-61. Functions of Multichip Packages


11-102 MICROELECTRONICS PACKAGING-AN OVERVIEW

300-
PWB
250
'[
'-' 200
'"
s::
.g Multichips

5'" 150 ~40% Si


.§
Q
Current
01)
s:: 100
:5 / Co-Fired Ceramic
~
50

1."'--- 50---MCM---·.lOO
Si Packaging Efficiency

Figure 7-62. Packaging Efficiency of Various Packages

Packaging efficiency (Perr) is defined by

P = Total active silicon area [7 19]


eff Total active multichip packaging area -

Figure 7-62 illustrates packaging efficiency of various packages including


multichip with various line widths that are available in various packages,
illustrating clearly the leverage of multichip module in packing more
chips in least space than any other packaging including wafer scale integra-
tion (WSI). This advantage of multichip is expected to provide a wide
array of applications in aerospace, medical, consumer, telecommunication,
and computer industries.

7.5.11.4 Electrical Performance


Electrical performance is the primary reason for the use of multichip
packages in computers. There are two parameters that contribute to the
system-level performance, which is often quoted by a throughput parame-
ter, MIPS (millions of instructions per second). These are cycle time and
cycles per instruction and are related to the MIPS by the following equa-
tion, as discussed earlier in this chapter:

lQl
Number of MIPS = . .. [7 20]
(Cycle ttme)(Cycles per mstructlOn) -
7.5 FIRST-LEVEL PACKAGES 11-103

4E3 .'..
3E3
• .... 2E3
.'...
• Long Term Failure Rate
' • Infant Mortality Goals
2E3
• ....
1E3

-.
II···. 800
..... ..
.. .. 600 41

41
0 • / .... ..... 400
0
a::
:t
a::
...41 .. 300
•........
400 200 ~0
~
c 300 ~
u.. ....c
200
• •
..
..... 100
.... 60
.Ec

40
• 40

30
30
20
20
10
72 76 80 84 88 92 96 2000
Source: Intel
Year

Figure 7-63. Reliability of Semiconductors and Multichip Packages

Leading-edge computer performance in MIPS requires leading-edge cycle


time and leading-edge cycle per instruction. The latter parameter is often
ignored, and as can be calculated from Equation 7-20, it will have a great
effect on the MIPS. The multichip package influences both of these
parameters, as both of these are influenced by packaging efficiency of cir-
cuits.

7.5.11.5 Reliability

Reliability enhancement is an essential function of any multichip


package, particularly when it carries so many leading-edge and expensive
chips, making it prohibitively expensive to discard due to any field failure.
Any multichip package must inherently be more reliable than the semicon-
ductors it packages, the reliability of which continues to be improved
upon according to Figure 7-63, which plots both the infant mortality and
long-term failure rate in parts per million as a function of time [30].
Fortunately, multichip packaging provides the fundamental basis for
11-104 MICROELECTRONICS PACKAGING-AN OVERVIEW

improvement of reliability, as illustrated in Figure 7-45, wherein the total


number of interconnections and wiring lengths in forming an equivalent
processor involving chips and various packaging levels is illustrated.
Reliability of multi chip packages must be treated at three different levels
to ensure its success:

1. Design for reliability upfront


2. Build-in reliability by process controls during manufacturing
3. Accelerate test of final product to guarantee the design and the
process

7.5.11.6 Leverages
There are four major leverages of multichip packaging:

1. Packaging efficiency
2. Electrical performance
3. Reliability
4. Cost

7.5.11.7 Types of Multichips


Consistent with the definition of multichip packaging technology in
this chapter, one can visualize three different multichip technologies (Fig.
7-64): (1) those formed with laminated, organic boardlike technology
referred to as MCM-L, to which chips can be bonded by TAB or wire
bond, (2) those formed with screen-printed ceramic thick film, either
sequentially formed by print and fire process, or parallel formed by lamina-
tion of screened green sheets and subsequently cosintered, referred to as
MCM-C, to which chips are attached directly by flip-chip or indirectly
in prepackaged carriers, and (3) those formed with deposited thin-film
technology using thin films of polymers and metals, referred to as MCM-D,
to which the chips are attached. Variations of these can also be considered
multichips. For example, a variation of laminated MCM using plastic
packages, involving molding compounds and lead frames, so-called multi-
chip plastic quad packs (MCM-P), containing four chips became available.

7.5.11.8 Cost
The cost of multichip modules varies with the type of multichip and
wiring density it provides.
Within various multi chip options, it is controversial as to which type
is cheaper. Figure 7-65 illustrates the relative costs of various multichip
technologies as a function of wiring density. It is generally agreed that
7.5 FIRST-LEVEL PACKAGES 11-105

(A) (B) (C)


Co-Fired MLC Thin Film on Si Printed Wiring
MCM-C MCM-D MCM-L
F=:::::::!C=!::::::j~pcrFoil

Metallize Pattem Etching


I
Greensheet

III III
Apply Resist
1= = 1
Via's Punched Etch Metal
Remove Remove
C::J
Form Plating Mask
1
C' ,.....,

III III
Via's Metallized
Lines Screened
Spin Coat
Polyimide c:Drilling
~ Permanent
d ~~Film

1
r
Apply Resist

ExposelDevelop
=I bl[d
Seeding

@ = I !
bl!~
Laminated

r-o
Etch Polyimide

=I f\
In: til Remove Resist
Cure
Cata1izer
Electroless Copper
p::q ~ ~:::'I."
lee =1 I:a::::r:J b::::sI Plating Film
Co-Sintered p

Metallize
Via/Conductor

Figure 7-64. Three Types of MCMs and Their Processes

the laminated technology, although it is cheapest at low wiring densities,


becomes very expensive at medium to high wiring densities. The contro-
versy is mainly in the multilayer ceramic versus thin-film multichip pack-
ages. Whereas the former is based on a parallel process involving screened
greensheets and a final high-yield cosintering process that assembles all
the layers, the latter is a sequential and an expensive ultraclean-room
process. Thus, even if all operations have better than 95% yield in the
sequential thin-film process, the final yield can be significantly lower than
desired. IBM's experience with multilayer ceramic during the parallel
process approach has given yields in excess of 90%. Because of these
and other reasons, the cost of multichip technologies, as illustrated in
Figure 7-65 has been determined [29].
There is a need yet to have another category, based on deposition
of thin films (MCM-D) on ceramic multichip modules. The wiring in this
type of multichip is shared in cofired ceramic thick film and polymer-metal
11-106 MICROELECTRONICS PACKAGING-AN OVERVIEW

.5 15.00 MCM-D
0- (projected)
~'"
8'" 10.00
5.00

200 400 600 800 1000


Interconnect Density (in./sq. in.)
Ref. NEPCON, "Multichip Module Fundamentals'
R. Wayne Johnson, Auburn University

Figure 7-65. Interconnect Density Versus Cost for Various MCMs

thin film, offering the best of both technologies in cost and function. This
type of multichip is expected to be the most predominant of all multichips
and has already formed the basis of two multichips at two major companies,
NEC [31] and IBM [32].
The multichip package that is expected to be dominant, however,
in the next decade is thin film on laminate MCM-D on MCM-L. A good
example of this is the IBM SLC surface laminar circuitry discussed in
Section 7.7, "Second-Level Packages."
Table 7-25 illustrates typical package parameters for each of the
three multichip types. The materials and their properties are illustrated in
Table 7-26 and in Figure 7-66.

7.5.11.9 Applications
Because of high electrical performance, reduced space and weight,
and improved reliability over single-chip modules, as discussed above,
to form the required system-level function, multichip packaging is ex-
pected to have a broad spectrum of applications. Some of these are shown
in Table 7-27.

7.6 PACKAGE-TO-BOARD INTERCONNECTIONS


As pointed out earlier, the system needs such as cellular phones or
disk drives, as illustrated in Figure 7-67, drives the technologies toward
small sizes and low weight but with high functionality and at low consumer
7.6 PACKAGE-TO-BOARD INTERCONNECTIONS 11-107

Table 7-25. Typical Package Parameters of the Three Multichip


Module Types

Characteristic MCM-L MCM-C MCM-D

Description High-density Cofired low- Thin film on


laminated dielectric- silicon
printed circuit constant ceramic
board substrate
Max. wiring density 300 800 250-750
(cmlcm2)
Min. linewidth 60-100 75-100 8-25
(11m)
Line space (11m) 625-2250 125-450 25-75
Max. substrate size (mm) 700 245 50-225
Dielectric constant 3.7-4.5 5-5.9 3.5
Pinout grid (mm) Array 2.54 Array 2.54 Peripheral
(staggered) 0.63 mm
Max. No. of 46 63 4
wiring layers
Via grid (11m) 1250 225-450 25-75
Via diameter (11m) 300-500 100 8-25

prices. These requirements translate directly into component density-


number of components per unit area of board. The components include
active plastic or ceramic packages containing logic and memory chips as
well as nonactive components like capacitors, resistors, inductors, and so
forth. Figure 7-68 illustrates the trend in component density which is
currently at about 20-25/cm2 and which is expected to be 50/cm2 by the
years 2000-2005. This figure also illustrates, to some extent, how this
component density will be achieved by reducing the size of each compo-
nent, both active and passive while providing more pins or connections.
The active component reduction is achieved for most part by reducing
the 110 pitch or by area-array connections or both. The passive component
improvements are made by reducing the sizes, as also illustrated in Figure
7-68 from 3.2 x 1.6 mm to 0.8 x 0.4 mm.
There are three basic types of connections between the first- and
second-level packages: those with pins, requiring pin-through-holes
(PTH), and others with lead frames meant for surface-mounting the device
(SMD) by the use of surface-mounting technology (SMT). The third
category is BGA, using balls in area-array fashion in contrast to peripheral
lead frames. Package assembly includes, in addition to, an active plastic
or ceramic components containing logic and memory chips, nonactive
components like capacitors, resistors, and inductors. All the available
11-108 MICROElECTRONICS PACKAGING-AN OVERVIEW

Table 7-26. Properties of Multichip Substrate Materials

Thermal Reuse!
Dielectric Expansion Fabr.l Thermal
MCM Constant Coefficient Temp. Conductivity
Material (I MHz) (l0"7/°q rn (W/mk)

Alumina 9.4 66 1600 20


Aluminum
nitride 8.8 35 1600 20

Mullite + glass 5.9 35 1200 6


Borosilicate +
alumina 5.6 40 900 6
Lead borosilicate
+ alumina 7.8 42 900 5

Glass--ceramic 5.0 30 950 7


Composite Si02
+ glass 3.9 19 900 4

Epoxy Resin 4.7 1400 200 0.1


Polyimide resin 4.7 750 300 0.1
MS resin 3.7 680 250 0.1

Polyimide
PMDA-ODA 3.5 250-400 400 0.2
BPDA-ODA 3.0 20-60 400 0.2
Benzocyclobutene 2.7 650 350 0.2
Polyphenyl
quinoxaline 2.7 350 450 0.2

packages were previously classified in Figure 7-64 into one of these cate-
gories.
The surface-mount technology was the fastest growing technology
in small systems and consumer electronics during the 1980s. It presents
a number of advantages. The packages for SMT can be developed with
pads or connections out of the package at the tightest spacings, thus
allowing more IIOs per unit area of the package. Second, the technology
can be used to increase the packaging density by requiring less board
area and, thus, lowering the cost of packaging the total system. As a
result, surface mounting, which accounted for roughly 20% of all consumer
and low-end packages in the late 1980s, is already at about 80%. This
technology will, however, pave the way for BGAs in the late 1990s.
The overall trend in package-to-board assembly is illustrated in
Figure 7-69 showing the expected trend to fine-pitch quad flat pack (from
7.6 PACKAGE-TO-BOARD INTERCONNECTIONS 11-109

11

10
~I: 9
>.
0
u 8
c
I:
0
-..:0 7
01
0

e
Q.
6
Q.

Dielectric Constant

Figure 7-66. MCM Substrate Materials and Their Dielectric Constants

Table 7-27. Potential Multichip Package Applications and


Their Characteristics

Application Characteristics

Computers High electrical perfonnance


Less space, high reliability
Military Less space and weight
Telecommunications High bandwidth perfonnance
Consumers electronics Less space and weight
Medical Less space
Automotive Less space

0.4 mm pitch in 1995 to 0.15 mm in 2000) as well as to ball-grid array,


chip scale package, and direct flip-chip attach to organic board. The QFP
to BGA change appears in Figure 7-70 around 400 1I0s, but as pointed
out earlier, a number of companies, particularly Japanese with massive
investments in peripheral SMT, will continue to push beyond 400 1I0s.
The lead-pitch improvement consistent with this trend is indicated in
Figure 7-71 for a number of actual Japanese packages.
Consumer products require thin and lightweight packaging. Plastic
packages such as QFP that are surface-mounted onto PWB have effectively
11-110 MICROELECTRONICS PACKAGING-AN OVERVIEW

70
60
:;-
50
~
c'II 40
~
30
~ Disk Drives
g 20
ID
10
Source: BPA
0
80 85 90 95

700
600
1v 500
-; 400
] 300
0

•
> 200
100 Source: JTEC TZ-804B
0
80 85 90 95

Figure 7-67. Portable Need Drives Size Reduction

!
I/J
CD
:::l

~ Component
"0
CD Density
.~
(ij
E
.... 0.4
0
Z
0.2

o Source: Hitachi

80 85 90 95 2000

Year

Figure 7-68. Portable Component Density (Courtesy of JTEC.)


7.6 PACKAGE-TO-BOARD INTERCONNECTIONS 11-111

~
'iij
c:
CD
o
g Fine
Pitch
Surface ........... SMT
Throug h ., Mount
Hole RiI';;'.'t'.""
1980 1985 1990 1995 2000 2005
Figure 7-69. Package-to-Board Interconnection Density Trends

1000

800

..
e 600
~
0
u
c:
i:L 400

200

o~----------------
1990 1992 1994
__ __________ ____..
~
1996 1998
~
2000 2002

Year

Figure 7-70. Pin-Count Evolution


11-112 MICROELECTRONICS PACKAGING-AN OVERVIEW

60r--------------------------------
124
50~----~._----------------------

40~------~.------------------------

30 I----~:_:__....;

20~-----------------~~~~~~---
•
172
• •
10~----------------~-------------

O~ __ ~ __ ~ __ ~ __ ~~~~ __ ~ __ ~

50 100 150 200 250 300 350 400 450


Lead Count

Figure 7-71. QFP Lead Pitch Versus Lead Count

met consumer product requirements. The industry vision of next-genera-


tion products requires packages that are smaller and cheaper than in the
past, roughly 50% smaller for each new generation. Given industry's past
investments in PWB and SMT technologies, and given increased global
cost-competitive pressures, the industry is expected to pursue the use of
plastic packages to the ultimate limit. The ultimate limit accepted by
Japan currently is 0.15 mm lead-frame pitch, giving rise to 800 pins in
30-mm sizes and 1000 pins in 38-mm sizes. The United States seems to
follow the path of BGA.
It should be noted that while the QFP and BGA improvements as
discussed above are taking place, the pin-grid-array technology is also
capable of being enhanced for those limited applications that require
insertable components for upgradeability, for example. Both ceramic and
plastic PGAs are capable of being enhanced to 1000 1I0s at a pin pitch
of 1.27 mm, or 50 mils. It should also be noted that these enhanced
PGAs can be surface-mountable as well using standard or modified SMT
equipment. Figure 7-59 illustrated the advantage of this approach-that
the size of this new PGA is no bigger than the current QFP at 0.4 mm
and that a 1000 110 PGA is readily available.
The continued use ofP-QFP beyond the current O.4-mm pitch toward
the 0.15-mm pitch, however, requires major enhancements in SMT pick
and placement tools, solder deposition technologies, reflow tools and
7.6 PACKAGE-TO-BOARD INTERCONNECTIONS 11-113

technologies, inspection, solder repair for opens and shorts, and electromi-
gration resistance of both the plastic package and the printed wiring board.
Contrary to what one might expect, the industry will incrementally enhance
each of these to a level that will guarantee high yield and high reliability.
This conclusion is supported by (1) Sony's advancements in factory auto-
mation and (2) Oki's single-ppm-defect soldering systems. These systems
and processes have lowered assembly defects to less than 20 ppm, as
shown in Figure 7-72. Sony's precision robots have improved placement
repeatability to 0.01 mm from 0.05 mm during the last 6 years. Matsushi-
ta's new SMT machine has 11 placement heads with 0.01 mm repeatability.
Toshiba's advanced TAB equipment can place 0.2-mm-pitch parts using
CCD vision, because pitch size has reached the limits of human vision.
With increased miniaturization, soldering technologies continue to
evolve. For example, Oki's single-ppm-defect technology includes devel-
opments in the following:

• New wave soldering machine for zero defects


• Nitrogen flow soldering process technology
• Rheology and printability of solder paste
• Inspection technique for solder paste printability and printing param-
eter optimization
• Development of an automatic solder-joint inspection system

Figure 7-73 illustrates the general trend in soldering techniques that the
Japanese microelectronics industry is expected to follow, shifting from
reflow to local soldering techniques in order to meet ultrafine pitch assem-
bly requirements.
Table 7-28 summarizes the overall packaging that can be expected
in terms of chip and package connections, board substrate wiring, chip

Failure rate per 1.5


PWB(%}

1.0- 0.63%
Solder bridge
0.5 - 0.57 ~ 0.02%
"-
Soldering atmosphere Air (1992.7) N (1992.9)
Totol PWBs 18,506 PWBs 21,002 PWBs
Defect PWBs 117 PWBs 4PWBs

Figure 7-72. Soldering Defects Improvement Achieved at Old (From Ref. 24.)
11-114 MICROELECTRONICS PACKAGING-AN OVERVIEW

Wave
1.27 soldering 1----=--=-----..

1972 1977 1982 1987 1992 1997 2002

Figure 7-73. Soldering Technology Trend in Japan (Courtesy of Hitachi.)

placement accuracy and cost for leading-edge consumer products. Similar


parameters were reviewed earlier in Tables 7-4A, 7-4B, and 7-4C for
consumer, automotive, and high-performance systems, respectively.

7.7 SECOND-LEVEL PACKAGES


Four technologies capable of meeting the requirements of the second-
level package are as follows:

Table 7-28. Package Assembly Trend for Consumer Electronics

1995 1998 2000

Cost
Component assembly (Conversion) cost (¢ per I/O) 2 0.4
Substrate cost (6-layer H.D.~ 4 layer) ($ per in. 2) 0.5 0.25
IC package cost (¢ per I/O) 1.5 0.4 0.2
Package connections
Component I/O density (I/O per in. 2) 100 400 600
Component I/O package pitch-peripheral (mm) 0.4 0.25 0.15
Component I/O Package pitch-area array (mm) 0.4 0.25 0.25
Chip connections
Component I/O perimeter flip-chip pitch (mm) 0.15 0.05 0.05
Component I/O area chip pitch (mm) 0.30 0.20 0.15
Substrate wiring
Substrate lines and spaces (mils) 5 2
Substrate pad diameter (mils) 20 2
Equipment
Chip placement accuracy (mils) 2 0.5 0.25

Source: Courtesy of NEMI.


7.7 SECOND-LEVEL PACKAGES 11-115

1. Organic cards and boards


2. Polymer- or glass-coated metal carrier
3. Flex film carrier
4. Injection-molded card
5. Ceramic boards

These are schematically illustrated in Figure 7-74 and their characteristics


described in Table 7-29. Additional details can be found in Chapter
17, "Printed-Circuit Board Packaging," and Chapter 18, "Coated-Metal
Packaging." Additionally, ceramic boards, described in Chapter 9, "Ce-
ramic Packaging," can be considered second level as well.
Card and board technologies based on epoxy-glass, and enhance-
ments in these by the use of such materials as polyimide, Telfton, benzo-
cyclobutene organics, and quartz glass to provide better electrical charac-
teristics and packaging densities, form the most common second-level
package. The technology development is summarized in Figure 7-75 in
terms of such package parameters as linewidths, intervia diameter, plated-
through-hole diameter, aspect ratio, and number of layers. Printed-circuit
boards with up to 42 layers capable of supporting as many as 336 chips
on an individual board of 600 x 700 mm have been reported [33]. Very

Flexible
Insulator Circuit
Adhesive Carrier
(Flex)

~
~////////////////////ff/1!

~~ Card
~//////////////////////~
~

Copper
Chip _ __

Solder Insulator
Conductor
Metal Carrier

Metal Core

Figure 7-74. Second-Level Packaging Technologies


11-116 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7-29. Second-Level Package Characteristics

Organic Inorganic Inorganic Coated Aexible Inj.-Molded


Card Board Board Metal Carrier Card

Insulator material Epoxy- Glass- alumina Glass or Polyimide Resin


glass ceramic polyimide
Conductor Cu Cu MoorW Ag-Pd or Cu Cu
materials Cu
Number of layers 8 63 >33 4 2 2
Coefficient of 150 30 66 30-70 200 300
thermal
expansion
(RT*-IOQ°C)
x ]Q-7/0C
Dielectric 4.0 5.0 9.5 3.5-6.0 3.5 4
constant
Thermal 0.2 4 20 150 5 0.1
conductivity
(W/m K)

*Room temperature.

high-performance boards with materials of dielectric constant around 2.2


with five or more lines per channel are under development for potential
applications in the next decade. Injection-molded cards (IMC) , as the
name implies, involve molding plastics into two- and three-dimensional
shapes, and subsequently metallizing with seeding and plating technolog-
ies. Only surface wiring, with no internal planes (NIP), has been accom-
plished to date. A 50% cost savings, primarily associated with the reduced
number of wet process steps, compared with the epoxy-glass card process,
have been reported by a number of companies. Accordingly, injection-
molding card technology may find widespread use in low-end systems.
The flexible carrier, which consists of two surface layers of thin-
film copper wiring on each side of polyimide or other polymeric film on
which protective epoxy could be masked and screened to form terminals
for bonding the first-level package is a growing second-level package.
This kind of second level, which is most commonly used in consumer
electronics, can eliminate the need for a number of cables and connectors
that are otherwise required to interconnect packages and can thus provide
significant cost savings. Additionally, this type of package can be consid-
ered a three-dimensional package, as the flex film with mounted compo-
nents could be bent in any dimension, as conceptually illustrated earlier
in Figure 7-37 for display applications.
7.7 SECOND-LEVEL PACKAGES 11-117

100

50
40

30

20

10

4
3

Mechanical Drill PhotolLaser


1.0 Plated 1:
hill /[,
OJeD'
0.5 lill1:Jeter C
11ll1l)
0.4

0.3

0.2

0.1

0.05

0
75 80 85 90 95 2000
Year of 1st Use

Figure 7-75. Evolution of Second-Level Packaging Parameters

The coated metal (sometimes referred to as metal carrier or metal


substrate), with a coating of glass, glass-ceramic, or polymer on such
metal carriers as low-carbon steel (as in porcelain enamels for dishwashers,
sinks, and so forth, or aluminum, stainless steel, or copper-invar-copper)
can form excellent second-level packages with a built-in ground plane,
mechanical stiffness, and thermal-expansion tailorability. This technology,
discussed in Chapter 18, "Coated-Metal Packaging," can be applied for
removal of heat without heat sinks on the back of the chip by the direct
attachment of the chip backside to the metal carrier. Alternatively, solder
bonding of chips (C4) can be accomplished with copper-invar-copper
substrates, the thermal expansion of which can be made to approximate
that of silicon.
A revolutionary approach to organic board enhancements has been
advanced by IBM Japan [26] that preserve low-cost attributes, such as
large-area processing, low-cost wet processing, but enhance the wiring
11-118 MICROELECTRONICS PACKAGING-AN OVERVIEW

r---
-,
Build-up

-- - - - - Lamination

- - - - - multilayer

"T -c::::..J Build-up

Figure 7-76. Additive Process Enhancement (Courtesy of IBM Japan, Ibiden.)

density by photolithographic via and line processes in contrast to drilled


via and laminate technologies. This is referred to as thin film on printed
wiring laminates and also as surface laminar circuitry (SLC). A typical
structure is illustrated in Figure 7-76. These enhancements fall into sev-
eral categories:

• Thin-line and fine-line conventional subtractive etching process.


• Low-cost, fine-line, thin-film additive process. Shown in Figure 7-77
is a new sequential process involving deposition of a photosensitive
polymer or epoxy, formation of via holes by large-area photo expo-
sure, and subsequent chemical etching and metalization by catalytic
chemical seeding and electroless plating. The advantages of this
process are many, including pattern shape, pattern thickness control,
pattern width control, mounting reliability, and, most importantly,
small via size. Ibiden compares this additive process with subtractive
processes in Figure 7-78 [24].
• Minimization of solder bridge by using dry film.
• New materials. Examples include aramid-based laminates with low
thermal expansion, good electromigration resistance, high glass-
transition temperature, and excellent processability. Another material
is ceracom, a combination of porous ceramic laminated with glass
and epoxy resin to form very low-TCE boards suitable for direct
chip bonding. Chapter 9, "Ceramic Packaging."
• Direct bonding of chip. This may be by COB (wire bonding), tape
on board (TAB), and flip chip on board to the printed wiring board
with appropriate low-stress encapsulants.

Japan has invested in a variety of PWB materials that include FR-


4, polyimide-glass, maleimide styryl, BT resin, and a new aramid-based
laminate consisting of aramid-based paper as a reinforcement in a matrix
of a new epoxy resin by Teijin Limited. The superior properties of this
7.8 PACKAGING COOLING 11-119

I- - Formation of internal circuit

1- - 1 Coating of interlayer insulator

Formation of via-hole

Chemical etching & drilling

Catolyst seeding & masking

Electroless copper pattern plating

Figure 7-77. Low-Cost, Fine-Line, Thin-Film Process (Courtesy of mM Japan,


Ibiden.)

aramid-based board for potential MCM applications include low TCE


(6-16 ppm/QC), very high electrical resistance, very low impurities in the
aramid fiber, and processability with fine via holes.
The shape and accuracy of conductor patterns, as well as the mount-
ing reliability of solder bridging, as illustrated in Figure 7-78, compares
with the standard substractive process. Table 7-30 indicates the dielectric
and metal ground rules, as well as the drilled and photolith dimensions
resulting in fine-line, thin-film structures on both sides of a PWB. The
structure of the additive process as practiced by IBM (Japan) and Ibiden
is illustrated in Figure 7-76, using two layers on each side of the PWB.

7.8 PACKAGING COOLING


Removal of heat from today's highly integrated ICs remains a major
bottleneck. The dissipative needs of microprocessors are in the range of
11-120 MICROELECTRONICS PACKAGING-AN OVERVIEW

Item New Process Subtractive


Etch1ns..Raiot
Pattern
Shape
"
wp/
Conductor

S_ _
1"""""""""""""",,1

~ Pattern
C
u
Thickness
R
A
C
Pattern
Y Width
Ability mass pro. mass pro.

Mifi
Patterning -100/150
Ability

--
Sold., Dry-film
Mounting
Reliability
Conductor

Figure 7-78. Shape and Accuracy of Conductor Pattern by Additive Process


(Courtesy of Ibiden.)

Table 7-30. Characteristics of Additive-Plated PWB

Number of layers 6
Thickness 0.8 mm
Insular thickness 50 11m
Conductor thickness 15 11m
Minimum wiring width/space/pitch 50 I1rn150 I1rn1lO 11m
Minimum via <I> 100 11m

Drill <I>

I. Inner via (drill) (mm) <I> 0.2 <I> 0.4


2. Blind TH (photo) (mm) <I> 0.3 <I> 0.5
3. Through hole (drill) (mm) <I> 0.5 <I> 0.7

Source: From Ref. 14.


7.8 PACKAGING COOLING 11-121

Conduction and Silicone-Rubber


Convection =::..3-5 watts/cm2
(Air)
~~~~~~-Frame
Sealing
Glass
Logic LSI
Chip
"'ea p
(M llite)
u

Conduction and
Convection
(Uquid) =::..20 watts/cm2

Convection Boiling =::..50 watts/cm2


(Boiling Uquid) Uquid

Figure 7-79. Cooling Technology Options

5-20 WIchip, whereas those in leading-edge ECL and CMOS technologies


are about 20-100 WIchip.
Even though the power per circuit has been steadily decreasing with
time because of improvements in the device and circuit technologies, the
growing total number of circuits per chip and the growing total number
of circuits per system are continually increasing the power and power
density requirements of logic functions. Figure 7-13 illustrated the package
power cooling requirements in watts per chip area.
The low-end systems have used and will continue to use forced-air
cooling of single-chip modules, surface mounted onto cards or boards.
The future use of chips with very high integration levels is still expected
to use this kind of cooling technology enhanced by heat sinks mounted
on the backside of chips or, alternatively, chips back-bonded to either
thermal cards based on coated-metal substrates or to high-conductivity
ceramics (AIN) or metallic sinks (Cu-W) joined to the ceramic packages
by solder glasses or braze alloys, as discussed in Chapter 9, "Ceramic Pack-
aging."
Cooling technologies used to provide these capabilities generally
fall into three categories, as shown in Figure 7 -79, and are further discussed
in Chapter 4, "Heat Transfer in Electronic Packages."
11-122 MICROELECTRONICS PACKAGING-AN OVERVIEW

t! 60
3
E
8-
...e 40
l-
e.
:..a 20
u
0
0.1 W/cm2 10

Power Dissipation
Effective Cooling Area

Figure 7-80. Cooling Regimes by Air, Liquid, and Boiling Liquid

1. Those that take advantage of the higher thermal conductivity of the


substrate, such as aluminum nitride ceramics, with thermal conduc-
tivity approximately 10 times better than alumina
2. Those that do not require the high thermal conductivity of the first-
level package, and thus depend on external cooling by means of
heat sinks bonded on the backside of the chip or package
3. Those that involve direct immersion of the chip or package in inert
liquids such as fluorocarbons

Figure 7-80 illustrates the cooling capabilities provided by air, water,


and boiling liquids as a function of both chip temperature and power
dissipation density. This figure clearly indicates the cooling challenges
being faced by PCs and workstations that require high heat flux removal,
which must be accomplished by air cooling.
In most of the systems shipped in the eighties and early nineties, a
combination of conduction, convection, and sometimes radiation has been
used with liquid or air. In the future, air cooling is the predominant choice,
and only if it cannot be achieved, liquid cooling will be pursued. A major
breakthrough in air cooling as indicated in Figure 7-80 appears to be the
microjet technology being pursued by the Packaging Research Center at
Georgia Tech [34] that has the potential to enhance the air-cooling regime
to in excess of 10 W/cm 2 from about 3 W/cm2 currently. It involves
7.9 PACKAGE SEALING AND ENCAPSULATION 11-123

Permeability (g/cm • Torr)

10-6 10-8 10- 10 10-12 10- 14 10- 16


10 1,.........,...--..,....--..,....--_--.,.....--,...-__,

Min H Day Mo Year

Figure 7-81. Permeability of Water Through Organic and Inorganic Mate-


rials

microjet air cooling. It focuses on the development of zero net mass flux
synthetic jet technology for integrated cooling of single-chip and multichip
modules in open and closed flow systems. An experimental technology
capability of about 800 W/cm 2 has been demonstrated [35] by direct liquid
cooling of silicon with fine groves directly etched in silicon as discussed
in Chapter 4, "Heat Transfer in Electronic Packages."

7.9 PACKAGE SEALING AND ENCAPSULATION


Package sealing is performed to protect the device and package
metallization from corroding environments and from mechanical damage
due to handling. Moisture is one of the major sources of corrosion of
the semiconductor devices. Electro-oxidation and metal migration are
associated with the presence of moisture. The diffusion rate of moisture
also depends on the encapsulant material, as shown in Figure 7-81, which
compares the permeability of water for various organic and inorganic
materials. In spite of the poor permeability of organic materials in the
figure, a number of encapsulants, particularly silicones, certain epoxies,
polyimides, and imide-siloxanes, have been shown to have great promise
[36]. The characteristics of these materials are listed in Table 7-31. In
plastic packages for consumer electronics and low-end systems, as dis-
cussed in greater detail in Chapter 8, "Plastic Packaging," a variety of
plastic encapsulants have been developed to minimize the diffusion of
11-124 MICROELECTRONICS PACKAGING-AN OVERVIEW

Table 7·31. Characteristics of Potential Encapsulants

Materials Application Advantages Disadvantages

(I) Epoxies • Normal dispensing • Good solvent • Nonrepairable


or molding resistance • High stress
• Excellent • Marginal
mechanical strength electrical
performance
(2) Polyimides • Normal dispensing • Good solvent • High-
(spin-coat) resistance (-500°C) temperature cure
• High stress
(3) Polyxylylene • Thermal deposition • Good solvent • Thin film only
(Parylene® Trademark (reactor) resistance
from Union Carbide) • Conformal coating
(4) Silicone-polyimides • Normal dispensing • Less stress vs. • Higher thermal
polyimide expansion
• Better solvent • Thin-film only
resistance vs.
polyimide resistance
vs. silicone
(5) Silicons • Normal dispensing • Good temperature • Weak solvent
(RTV, Gel) cycling resistance
• Good electrical • Low mechanical
properties strength
• Very low modulus

Source: After Ref. 38.

water. This results in significant improvements in reliability of plastic-


encapsulated packages. The diffusion coefficients of these materials range
from 1 x 10-8 cm2/s for epoxy resins, 12 x 10-8 cm2/s for typical silicone
305, to 0.1 x 10-8 cm2/s for NYSOL transfer molding material. The develop-
ments with silicone rubber have recently led to further improvements with
silicone gel [37], which is soft and can be deposited to thinner dimensions
with minimal stress problems.
High-performance systems, however, have depended on hermetic
packages. Hermeticity of the package requires a He leak rate of better than
Ix 10-8 cm2/s. The usual methods used in hermetic sealing are as follows:

• Welding
• Brazing
• Soldering
• Glass sealing
7.10 BOOK ORGANIZATION AND SCOPE 11-125

Three basic welding techniques are used: resistance welding, seam weld-
ing, and cold welding. Brazing is performed by melting metal or an alloy
to join package metals of higher melting points. The common brazes are
gold-tin eutectic with a melting point of 280°C, In-Cu-Sil with a melting
point of 680°C, and Cu-Sil with a melting point of 750°C. Solder sealing
is accomplished in the same way, except by the use of solders, the most
common being Pb-Sn eutectic at 163°C. Both soldering and brazing require
that the packages and lids (covers) to be sealed be metallized by thick-
film, thin-film, or plating technologies. Glass sealing, which requires an
excellent thermal expansion match between the composite package to be
sealed and the glass used for sealing, requires glass of a slightly lower
thermal expansion coefficient so as to put the seal in slight compression.
A number of solder glasses have been developed in the PbO-ZnO-AbOr
B20 r Si0 2 systems to meet a variety of sealing requirements [29].

7.10 BOOK ORGANIZATION AND SCOPE

For the readers convenience, an extensive glossary of terms and


symbols are included at the end of each part and are available at our
publishers-Chapman and Hall-internet address: ftp://ftp.thomson.com/
chapman&hall/tummalalglossary. As mentioned in early sections of this
chapter, this second edition is partitioned into three parts. Figure 7-82
depicts how the 20 chapters are organized. The introductory chapters of
each part-I, 7, 15-are identical. This was done intentionally for those
that do not have access to the complete set.
Part 1, "Technology Drivers", deals with several driving forces,
design considerations and technology options available. Chapter 2, "Pack-
aging Wiring and Terminals," appropriately begins this part by addressing
the topological (layout, placements, and wiring) considerations. The elec-
tromagnetic design considerations are followed in Chapter 3, "Package
Electrical Design." The thermal design considerations, driven to a great
degree by the electrical design, are presented in Chapter 4, "Heat Transfer
in Electronic Packages." The reliability considerations are in Chapter 5,
"Package Reliability." A new Chapter 6, "Package Manufacturing," has
been added in this edition for the competitively pivotal manufacturability
considerations.
Previously, it was noted that all interconnections on an IC chip or
all internal structures of any other component (e.g., a crystal or those of
discrete capacitors and inductors) are not considered packaging in the
common parlance. Thus, the packaging begins at the interfaces to intercon-
nect, power, cool, and protect all ICs.
Parts 2 and 3 deal with packaging technologies; Part 2 relating to
so-called semiconductor packaging and Part 3 relating to subsystem or
board-level packaging. Semiconductor packaging relates all technologies
involved in the packaging of semiconductors ready to be assembled onto
11-126 MICROELECTRONICS PACKAGING-AN OVERVIEW

<,.to"" I - li,ro<l,,'rooia PI,I<.o,I., - An Ovt.vi.,. J. J>


~.nJ,· P RT I - TECn OLOGY DRIVE ~.4'" liz
~~~ ~
""tI"!:

,."l' - "'1t!","lt

~.~~
, -;~
_---------p;;;;~~:::-::D ::::::~::~:_~__________ ~-,~
v~ ~

~' r------ 14· Staling .ad En<lP "I:;~~-l '\


i~ I
I
I e : "\
I 1
:r~ : t ~
III . Polymer. i. TbiD Film
iI Il - TbiD film
13.
10 _ fluti. PII( ...,in,
9 - Corami, PICI<.oC;'" I
iI
"

I & I

r-===========-----
:
-----========================~============::~-
PART 3 - SYSTEM-LEVEL PACKAGING I
I 20 a Oplottec1ronic
I • Prinled Wlrin, Board
I
I 18 - C•• led ldal Board
I
I
I
I
I
I
I
I
t
t ___________________________________________________________________ J
t

Figure 7-82. Book Organization. Numbers refer to chapter numbers in this three part
book. Chapter One is repeated in each part as Chapters 7 and 15.

board. Accordingly, it includes all chip-to-package connections such as


TAB, flip chip, and wire bond in Chapter 8 and all first-level packages
such as ceramic and plastic as in Chapters 9 and 10. Because some of
these packages, particularly the most advanced ones, contain thin-films
of polymers and metals, they are covered in Chapters 11 and 12. All those
packages can be single-chip or multichip and these chapters therefore
review both of these. Some of these are very complex and involve IIOs
in excess of 1000 per IC and wiring density in excess of 500 cmlcm 2•
To assure "known-good substrate" before assembled onto board, electrical
testing is performed. This technology is covered in Chapter 13. The
semiconductor packages is ready to be shipped for assembly provided it
is sealed or encapsulated to guarantee reliability with or without hermetic-
ity. This final technology operation is covered in Chapter 14.
Part 3 deals with all technologies required to assemble into subsys-
tems or total systems depending on the need for systems from consumer
electronics to mainframes. The first-level packages described in Part 2
are typically assembled as pin-through-hole (PTH) in the board, surface-
mount (SMT) onto the board and more recently ball-grid array or chip-
scale package connection to the board. These package-to-board connec-
tions are reviewed in Chapter 16. The board technology itself is undergoing
7.11 REFERENCES 11-127

major changes to accommodate the high I/O IC demand. Some of these


changes include alternatives to drilling, alternatives to epoxy-glass dielec-
tric and photolithography. All these printed wiring board technologies are
covered in Chapter 17. There are alternative technologies to printed wiring
board that include metal or ceramic boards. The ceramic boards are covered
in the chapter "Ceramic Packaging" in Part 2 and, therefore, it is not
repeated here. The metal board, however, referred to as coated-metal that
involves polymer, glass and ceramic coatings is covered in Chapter 18.
For very high performance applications such as mainframe and supercom-
puters, several boards are required to meet their semiconductor needs.
The interconnections between these are referred as connector and cable
packaging. Optoelectronic packaging is a fast-moving technology because
of its advantages both for long-distance communications replacing copper
and for short interconnections because of its bandwidth and data rate.
This technology is covered in Chapter 20.
7.11 REFERENCES
1. S. Sherman. "The New Computer Revolution," Fortune, 14 June 1993, pp. 56-80.
2. J. L. Hennesy and D. A. Patterson. Computer Architecture: A Quantitative Approach,
2nd ed. Morgan Kaufmann Publishers, San Matao, CA, 1966, pp. 1-59.
3. H. J. Greub, J. F. McDonald, and T. Creedon. "FRISC-E: a 250-MIPS Hybrid Micro-
processor," Circuits & Devices, 6(3): pp. 16-25, 1990.
4. R. Philhower. "Spartan RISC Architecture for Yield-Limited Technology," Ph.D.
Dissertation, Rensselaer Polytechnic Institute, 1993.
5. R. F. Sechler and G. F. Grohoski. "Design at the System level with VLSI and CMOS,"
IBM 1. Res. Devel., 39(112): pp. 5-22, 1995.
6. P. D. Franzon, A. Stanaski, Y. Tekmen, and S. Banerjia. "System Design Optimization
for MCM-DlFlip-Chip," IEEE Trans. Components, Packaging, and Manu! Technol.,
Part B: Advanced Packaging, CPMT-18(4), 1995.
7. Mixed Signal Design, Short Course, GaAs IC Symposium co-sponsored by IEEE
Electron Devices Society and IEEE Microwave Theory and Techniques Society, 16
October 1994.
8. H. B. Bakoglu, Circuits, Interconnections and Packagingfor VLSI, Addison-Wesley,
Reading, MA, 1990, Fig. 9.12.
9. See B. 1. Landman and R. L. Russo. "Pin vs. Block Relationship for Partitions of
Logic Graphs," IEEE Trans. Computers, C-20(12): pp. 1469-1479, 1971.
10. R. K. Poon. Computer Circuits Electrical Design, Prentice-Hall, Englewood Cliffs,
NJ., 1995.
11. J. N. Humenik, 1. M. Oberschmidt, L. L. Wu, and S. G. Paul. "Low-Inductance
Decoupling Capacitor for the Thermal Conduction Modules of the IBM Enterprise
Systeml9000 Processors," IBM 1. Res. Devel., 36(5): pp. 935-942, 1992.
12. D. P. Seraphim, R. Lasky, and c.- Y. Li (eds.). Principles of Electronic Packaging,
McGraw-Hili, New York, 1989.
13. E. E. Davidson, P. W. Hardin, G. A. Katopis, M. G. Nealon, and L. L. Wu. "Physical
and Electrical Features of the IBM Enterprise Systeml9000 Circuit Module," IBM
1. Res. Devel., 36(5): pp. 877-877, 1992.
14. W.-T. Liu, S. Cochrane, S. T. Lakshmikumar, D. B. Knorr, E. J. Rymaszewski, J. M.
Borrego, and T.-M. Lu. "Low Temperature Fabrication of Amorphous BaTiO J Thin
Film Bypass Capacitors," IEEE Electron. Device Lett. 14(7): p. 320, 1993.
11-128 MICROELECTRONICS PACKAGING-AN OVERVIEW

15. P. K. Singh, R. S. Cochrane, 1. M. Borrego, E. J. Rymaszewski, T.-M. Lu, and K.


Chen. "High Frequency Measurement of Dielectric Thin Films," Proceedings of
Microwave Theory and Techniques Conference, 1994.
16. R. Crowley, E. J. Vardaman, and I. Yee. "Worldwide Multichip Module Market
Analysis," A Multi-Client Study, TechSearch International, Austin, TX, July 1993.
17. D. A. Doane and P. D. Franzon (eds.). Multichip Module Technologies and Alterna-
tives: The Basics, Van Nostrand Reinhold, New York, 1993.
18. E. L. Boyd, Guest Lecture Notes for Electronic Packaging course, Rensselaer Poly-
technic Institute, 1993.
19. Hisao Kanai, "Low Energy LSI and Packaging for System Performance," IEEE
Trans. on CHMT, CHMT-4, No.2, pp. 173-180, June 1981.
20. T. Watari and H. Murano. "Packaging Technology for the NEC SX Supercomputer,"
Proceedings of Electronic Components Conference, 1985, pp. 192-198.
21. The National Technology Roadmap for Semiconductors, Semiconductor Industry
Association, San Jose, CA, 1994.
22. R. Hillemann. "Bipolar-based Mainframes vs. CMOS-based Mainframes," Computer
Packaging Workshop, 1992.
23. R. R. Tummala. "Next Generation of Packaging beyond BGA, MCM and Flipchip,"
Proceedings of IMC, 1996.
24. R. R. Tummala and M. Pechi. "Japan's Electronic Packaging Technologies," JTEC
Panel Report, pp. 59-96, 1995.
25. F. Kobayashi, Y. Watanabe, M. Yamamotu, A. Anzai, A. Takahashi, T. Daikoku,
and T. Fujita. "Hardware Technology for Hitachi M-880 Processor Group," Proceed-
ings of 42nd IEEE-ECTC, pp. 693-703, 1992.
26. Y. Tsukada, "Surface Laminar Circuit Packaging," Proceedings of 42nd IEEE-
CHMT, pp. 22-27, 1992.
27. J. Vardaman, Tech Search International, Austin, Texas, private communication.
28. R. R. Tummala. Proceedings of Emerging Microelectric and Information Technolog-
ies, pp. 1-8, 1996.
29. R. R. Tummala. "Multichip Packaging-A Tutorial," Proceedings of IEEE, Vol. 80,
No. 12, pp. 1-17, 1992.
30. D. L. Crook. "Evolution of VLSI Reliability Engineering Meeting," Proceedings
2nd European Symposium on Reliability of Electronic Devices, pp. 293-312, 1991.
31. T. Watari. "Computer Packaging Technology for System Performance," NEC Res.
Devel., No. 98, pp. 45-59, 1990.
32. R. R. Tummala and B. T. Clark. "Multichip Packaging Technologies in IBM," Pro-
ceedings of 42nd IEEE-CHMT, pp. 1-9, 1992.
33. M. Nishihaka, K. Kurosawa, and T. Sakamura. "Facom M-780," Fujitsu J., 32(2):
pp. 2-8, 1986.
34. A. Glezer and M. Thompson. "High Heat Flux Air Cooling Using Synthetic Micro-
jets," Semicon West, 1995.
35. Y. Tsukada. "Surface Laminar Circuit Packaging," Proceedings of 42nd IEEE-
CHMT, pp. 22-27, 1992.
36. J. Vardaman, Tech Search International, Austin, Texas, private communication.
37. R. R. Tummala, Proceedings of Emerging Microelectronic and Information Techno-
logies, pp. 1-8, 1996.
38. C. P. Wong. "High Performance Silicone Gel as IC Device Chip Protection," Mat.
Res. Symp. (108): pp. 175-187, 1988.
8
CHIP-TO-PACKAGE
INTERCONNECTIONS

PAUL A. TOTTA-18M
SUBASH KHADPE-Semiconductor Technology Center
NICHOLAS G. KOOPMAN-Microelectronic Center of HC
TIMOTHY C. REILEY-18M
MICHAEL J. SHEAFFER-Kulicke & Soffa

8.1 INTRODUCTION
Integration of circuits to semiconductor devices, driving the need
for improvements in packaging, has been discussed in Chapter 7, "Micro-
electronics Packaging-An Overview." This is further illustrated in Figure
8-1, wherein the cost of interconnecting on silicon is compared with
interconnecting on ceramic substrates and on organic boards, clearly show-
ing the lower cost of interconnecting on silicon [1]. Although the trend
is toward total integration on Si there is, however, a practical, growing
limit to the number of circuits which can be made on a single piece of
silicon, which is currently at about 1.6 million circuits for CMOS logic,
40,000 circuits for bipolar logic, and 64 megabits for memory. The highest
integrated transistor counts are approximately 5 million on advanced
microprocessors. Therefore, because most current information systems
11-130 CHIP-TO-PACKAGE INTERCONNECTIONS

~
~I~r----r--~r---~--~~--~--~
&l
z
z
o
~ t02r---t~
~
~
ffi
A-
10'
I-
eI)
o
um°r---t----~~--~~=r~~SILICO~)
e
i
N

10-1......-"---'---L..----I"----I---.;:s;;;;;::w
G: 10- 1 100 10 1 10 1 101 104
i INTERCONNECTION DENSITY (clII-l,

Figure 8-1. Cost of Interconnections. The relative cost of interconnection for DSR
(double-sided rigid printed wiring board), MLB (multilayer printed wiring board), ceramic
hybrids, gate array, and custom silicon. The curves nearly show the advantage of silicon
integration in decreasing interconnection costs. (From Ref. 1.)

require a greater number of circuits and interconnections, a number of


chips still need to be interconnected on organic or ceramic first-level
packages. The electrical connections between the chip and the package,
referred to here as chip-level interconnections, are the subject matter of
this chapter. Because, for the systems considered here, no first-level pack-
age can usually accommodate all the required chips, a second-level pack-
age interconnecting the first levels is often required. These interconnec-
tions, referred to as package-to-board interconnections, are reviewed in
Chapter 16. A recent deviation from this packaging pattern, also to be
discussed, is the direct surface mounting of flip chips on FR4 cards or
flexible circuits which is referred to as direct chip attach (DCA) or chip-
on-board (COB).
An abstract of the SIA national roadmap for semiconductors and
packages for various applications is presented in Table 8-1 [2]. Very
clearly the trend for input-output (110) on chips is upward with advanced
application-specific integrated circuits (ASICs) requiring in excess of 1000
110 by the tum of the century. For other applications, the growth of 110
is not as rapid, with automotive and commodity applications requiring
only modest increases.
There are three principal chip-level interconnection technologies
currently in use:
8.1 INTRODUCTION 11-131

Table 8·1. SIA Roadmap for chip interconnections. Extraction of estimated I/O
Requirements on chips in different applications for each step in feature size reduction.

1995 1988 2001 2004 2007 2010


Years 1998 2000 2003 2006 2009 2012
Smallest Feature 0.35 11m 0.25 11m 0.18 11m 0.13 11m 0.10 11m 0.07 11m

Chip Pad Count: (chip-to-package)


Commodity 208 256 324 420 550 700
Hand held 300 450 675 880 1140 1500
Cost/performance 540 810 1200 1600 2000 2600
High-performance 900 1350 2000 2600 3600 4800
Automotive 132 200 300 400 500 700

Pad/pin-count ratios: mem, comm, port, and auto = 1 : I; performance classes = 1 : 1.2
Source: From Ref. 2.

• Wirebonding (WB)
• Tape automated bonding (TAB)
• Flip-chip solder connection (C4)

Less common, but also important, are some chip-to-package inter-


connections such as conductive adhesives, elastomeric connections, and
combinations of these. The latter connections dominate the flat panel,
liquid crystal display applications.
This chapter reviews primarily WB, TAB, and C4, indicating the
advances made over the last two to three decades in materials, design
factors, fabrication processes, tools, assembly, rework, and reliability.
Future trends in this very important area are also discussed.
The most basic function of the chip-level interconnections is to
provide electrical paths to and from the substrate for power and signal
distribution. Electrical parameters such as resistance, inductance, and
capacitance need to be quantified for each interconnection design, as each
of these will affect the total system's performance. In addition to the
electrical functions, the C4 and TAB designs provide mechanical support
for the chip, which is augmented by the encapsulants used to seal the
chip metallization or to modify strain in the interconnections. Because
almost any electrical conductor is also a good thermal conductor, some
interconnections together with package materials are designed for removal
11-132 CHIP-TO-PACKAGE INTERCONNECTIONS

of heat from the chip. Thus, each interconnection provides electrical,


mechanical, and thermal functions which are discussed in the chapter.

8.2 CHIP-LEVEL INTERCONNECTION EVOLUTION

The silicon integrated-circuit chip was invented in the late 1950s


and was in manufacturing by the mid-1960s. Prior to its existence, chips
or die of silicon or germanium were single, discrete transistors or diodes
for a number of years. These devices were initially packaged in hermeti-
cally sealed metal cans and interconnected with passive components such
as resistors and capacitors on printed-circuit boards. The first transistorized
generation of computers and entertainment systems were built this way.
During the early, discrete transistor era (1955-1960) there was only
one kind of chip-level interconnection, the thermocompression wirebond.
This technology came from Bell Laboratories simultaneously with the
germanium and silicon semiconductor technology [3]. The implementation
of thermocompression (TC) wirebonding at that time led to many problems
with reliability, manufacturability, and cost. Simple TC bonding of gold
wires on aluminum chip metallization was a hot process, typically in the
350-400°C region, which resulted in excessive intermetallic formation
between Au and Al (observed as "purple plague") and caused bonding
weakness and field-failure problems. The manual nature of the bonding
also resulted in great bond variability and operator dependence. The cost
was therefore very high and the supply of transistors limited. For all these
reasons there was a quest in industry for a better way.
At Bell Laboratories and at IBM, the chip was taken out of its
hermetic enclosure and made into a smaller, less vulnerable component.
The approach at Bell Laboratories was to use a combination of silicon
nitride on the chip to protect junctions, and surface gold wiring with
"beam leads" to keep the interconnections free of corrosion [4]. At IBM,
a thin layer of glass passivation sealed the chip surface and its aluminum-
based wiring [5,6]. Solder-bump interconnections joined the chip to its
package [7,8]. Both the Bell and the IBM chips were usually mounted
face down in the "flip-chip" configuration on ceramic substrates that had
thin-film or thick-film wiring and passive components. The chips were then
encapsulated in silicone gel, which prevented the formation of continuous
water films and electrolytic metal migration between exposed electrodes.
Thus began the nonhermetic, hybrid packaging era, which was to accom-
pany second-generation transistors in computer and telecommunications
applications.
The first integrated circuits became available in manufacturing quan-
tities in the mid-1960s. As with their hybrid module predecessors, there
were few circuits on each chip of silicon. Aluminum-based or gold-based
thin-film wires were used to "integrate" the active and passive devices
8.2 CHIP-LEVEL INTERCONNECTION EVOLUTION 11-133

embedded in the silicon. The exponential growth of the circuit count per
die during the following decades has been phenomenal and was faster
than expected. At the start there were only one to five bipolar logic circuits
on a chip, and the first bipolar memory chips had a modest 16 bits in a
scratch pad memory application. In the early 1970s, the bipolar logic
count grew to about 100 circuits and monolithic memory to 128 bits for
the first commercial, bipolar main memory replacing ferrite cores, as, for
example, in the IBM 370 system. Today, the number of logic circuits has
grown to about 40,000 per chip (bipolar) or 1.6 million circuits (CMOS)
and up to 64-megabit memory arrays with single FET (field-effect transis-
tor) cells totally replacing multi-transistor main memory bits. Microproces-
sors, essentially computers-on-a-chip, are commonplace in personal com-
puters, workstations, and highly parallel supercomputers. Very-large-scale
integration (VLSI) has turned into ultralarge-scale integration (ULSI).
The integration and densification process in integrated circuits has
caused the continuous migration of intercircuit wiring and connections
from boards, cards, and modules to the chip itself. The surface of the
chip, with its multilayer wiring, has become a microcosm of the conductor
and insulator configurations that were common on previous multilayer
printed-circuit boards and on multilayer ceramic packages (Figs. 8-2 and
8-3). As many as four to six levels of wiring have been created on the
chip. Advances in interlevel vias (etched, tapered holes) to vertical wires
or studs and significant planarization of wiring layers has led to the ability
to have 1.6 million circuits on a 15-20-mm chip with five levels of wiring.
Even with this progress, the wiring capability on the chip continues to
lag the potential density of silicon devices. Therefore, wiring pitches are
rapidly dropping into the low-submicron dimensions. It is anticipated that

Figure 8-2. Chip Multilayer Wiring. Contrast of conventional etched via, three-layer
wiring with excessive topography (left) with studded, planarized wiring for four-level
structures (right). (From Refs. 9 and 10.)
11-134 CHIp· TO·PACKAGE INTERCONNECTIONS

Figure 8-3. Metallographic Cross Section of Four-Level Wiring on IBM Bipo-


lar Hip, Illustrating Damascene Studs and Chem-mech Polishing Planariza-
tion. (From Ref. 10.)

0.25 J..lm wiring will be commonplace by the end of the decade and chip
sizes up to 30 mm are envisioned for the logic circuits or microprocessors
of the future.
The technology of chip surface wiring is truly a part of the packaging
technology, but it is too complex and different from typical board and
substrate technology to be treated adequately in this chapter. It is a fact,
however, that all the physics and engineering of high-speed transmission
line theory applies in the chip-level wiring as well. The role of multilevel
wiring in reducing on-chip delay is shown in Figure 8-4. Compared are
the distributions and magnitudes of on-chip delays attributable to wiring
versus Si for three generations of IBM logic chips for mainframe comput-
ers. The importance of effective wiring is illustrated. In the oldest, first
family shown, only 10% of the delay was attributed to the three-level
chip wiring. This grew to 27% of the delay in the third family with four-
level wiring and would have been 45% if the progress from three- to
four-level wiring had not been made. The gains in circuit performance for
five and six levels are not illustrated but might be anticipated by projection.
The progress in integrated circuits has led not only to enormous
densification of circuits on a chip but also to the total integration of a
computer on a chip. Modern microprocessor chips in hand-held calculators
have the computing power of second-generation large-scale computers of
the mid-sixties. Powerful personal computers or workstations in parallel
8.2 CHIP-LEVEL INTERCONNECTION EVOLUTION 11-135

4SO
M3 11% II Wile delay

400
II Other delays

350

300
-;
..e 2SO
~
22S
~ 200 M4 22% r-,
195
M31 :44%
150
ISO
M4 27%
100

SO

0
3080 3090-S ESI9000
System

Figure 8-4. Relative Contributions of Wiring and Silicon to the On-Chip


Propagation Delays in Three Generations of Bipolar Switched Computers.
The third bar contrasts difference in delays between the three and four wiring levels for
the same chip. (From Ref. 10.)

are so capable that they are beginning to consume some of the market
share that was the domain of supercomputers and mainframes.
The advanced VLSI era has put great demands on the functionality
and reliability of ever-increasing numbers of input-output (110) connec-
tions. An empirical relationship between 110 and the number of circuits
to be wired (Rent's Rule) appears to be holding well for mid-to-Iarge
systems. This means that the logic 110 demand will expand from about
500 today to 5000 1I0s by the year 2010 [2].
Heretofore, serial wirebonding of one or two rows of 1I0s around
the perimeter of the chip has satisfied the needs of ceramic or plastic
packages for logic packaging. Automated wirebonding today is very fast,
efficient, and reliable compared to the manual bonding of the early sixties.
Wirebonding has been displaced with TAB bonding in instances where
11-136 CHIP-TO-PACKAGE INTERCONNECTIONS

the perimeter pitch must be finer to keep the chip size smaller, yet accom-
modate 500-600 connections. On the other hand, the solder-bump periph-
eral counterpart evolved differently into an area-array C4 configuration
in which the entire surface of the chip is covered with C4s for the highest
possible I/O counts, as high as 2000-3000 pads. Unlike wirebonding, C4
and usually TAB demand bump formation on the surface of the chip
when the chip is in wafer form. Bumping the chip has been and is a
constraint to the widespread adoption ofC4 or TAB in commercial devices.
It is an added expense and a commitment to fixed bonding patterns, but,
in the VLSI era, perhaps a necessary one. Typically, for bumping a wafer,
a layer of silicon oxide, silicon nitride, or polyimide passivation must be
used on the final wiring level of the chip. But this has become a common-
place precaution anyway to protect the fine chip wiring from corrosion
and mechanical damage, even in advanced wirebonded chips. The details
of the required bump fabrication will be discussed later in this chapter.
The die then has evolved to become a total "minipackage" of its
own. In a sense, it is the world's smallest electronic package. The role
of future packaging will be that of protecting the chip, getting power in
and heat and signal out, which will, of course, become more challenging
as microprocessors grow larger and more powerful. The chip-level connec-
tions toward achieving this will be discussed in the following sections.

8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS

The solder-bump flip-chip interconnection was initiated in the early


1960s to eliminate the expense, unreliability, and low productivity of
manual wirebonding. Whereas the initial, low-complexity chips typically
had peripheral contacts similar to wirebonded chips, this technology has
allowed considerable extendibility in I/O density as it progressed to full-
popUlation area arrays. The so-called controlled-collapse-chip connection
(C 4 or C4) utilizes solder bumps deposited on wettable metal terminals
on the chip and joined to a matching footprint of solder wettable terminals
on the substrate [11]. The upside-down chip (flip chip) is aligned to the
substrate, and all joints are made simultaneously by reflowing the solder
(Fig. 8-5).
The C4 joining process has been described in the literature with
numerous acronyms. In the industry, it has also been called CCB-
controlled-collapse bonding-and flip-chip joining, recognizing the fact
that it is opposite to the traditional backside-down method of bonding,
in which the active side of the chip, facing up, is wirebonded. The terms
C4 or flip chip are used interchangeably in this book.
Two other acronyms are also used in this section: BLM and TSM.
These refer to the terminal-connecting metallurgies at the chip and sub-
strate, respectively. BLM stands for ball-limiting metallurgy and refers
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-137

Substrate

Figure 8-5. ControUed CoUapse Chip Connection (C4). The upside-down chip
(flip chip) is aligned to the substrate and all joints are made simultaneously by reflowing
the solder. (From Refs 9 and 12.)

to the region of terminal metallurgy on the top surface of the chip that
is wettable by the solder. Alternate references to BLM are PLM (pad-
limiting metallurgy) and UBM (under-bump metallurgy). TSM stands for
top-surface metallurgy and refers to the terminal metallurgy on the sub-
strate to which the chip and its associated solder balls are joined.

8.3.1 C4 History
The solder-bump interconnection of flip chips, the face-down solder-
ing of silicon devices to alumina substrates, has been practiced for approxi-
mately 30 years. First introduced in 1964 with the solid logic technology
(SLT) hybrid modules of IBM's Systeml360, it was part of a design to
eliminate the expense, unreliability, and low productivity of the early
manual wirebonding [5]. The solder bump was also an integral part of a
chip-level hermetic sealing system created by the glass passivation film
on the wafer [6]. Most semiconductor devices of that era were, in contrast,
protected by expensive, hermetically sealed can enclosures. The terminal
bump design was created to hermetically reseal the access or "via" hole
through the glass protection layer as well as to provide a means for testing
and joining the chip (Fig. 8-6).
Initially, for the discrete transistors or diodes of the hybrid SLT,
copper ball, positive standoffs, embedded in the solder bumps, were used
to keep the unpassivated silicon edges of the chips from electrically
shorting to solder-coated thick-film lands [7]. Later, in the early integrated-
circuit era, the controlled collapse chip connection was devised. In this
technique, a pure solder bump was restrained (controlled) from collapsing
or flowing out over the electrode land by using thick-film glass dams, or
stop-offs [11], which limited device solder-bump flow to the tip of the
substrate metallization (Fig. 8-7).
11-138 CHIP-TO-PACKAGE INTERCONNECTIONS

coppcrb3110
i and Au plated) ,',- .

Figure 8-6. Terminal Metallurgy Design. The original SLT flip chip (0,69 mm
square) with glass passivation, BLM sealing of via holes, and Cu ball solder bumps,
(From Ref. 7.)

As with SLT, the solder flow on the chip surface is limited by a


ball-limiting metallurgy (BLM) pad, which is a circular pad of sequentially
evaporated thin-film metals. Chromium, copper, and gold, in that order,
provide the sealing of the via as well as a solderable, conductive base
for the solder bump. A very thick deposit (100-125 /lm) of evaporated
97 Pb/3 Sn solder acts as the primary conduction and joining material
between chip and substrate [7,8].
The earliest integrated-circuit chips typically had peripheral C4 I/O
pads like their wirebonded counterparts. The pads were 125-150 /lm in
diameter, on 300-375-/lm centers. The spacing (pitch) of connections
was compatible with the screening resolution and pitch capability of thick-
film-AuPt or AgPd-electrodes on the ceramic substrates.
Occasionally it was convenient to have an "in-board" power pad or
two in the thick-film technology, but larger numbers of inside I/O pads
could not be used until thin film metallized ceramic (MC) technology
became available in the mid-1970s, as discussed in Chapter 9, "Ceramic
Packaging," and Chapter 12, "Thin-Film Packaging." The narrower lines
and spaces made possible with etched thin-film Cr-Cu-Cr on ceramic,
as fine as 20-/lm lines with 60-/lm pitches, allowed wiring escape for
double rows of I/O pads and many internal connections [13] . Later, a
"depopulated" grid of bumps allowed the interconnection of 700 circuit
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-139

(a)

"'.1'E---- Device
~t:-_--Conlacl to active region
Cu
solderable ~*--- Device passivation
layer ~--- adherence layer
Au flash

Pb: Snpad

(b)

Figure 8-7. Controlled Collapse Chip Connection. (a) Side view of a device or
chip; (b) Side view of a device or chip on a substrate (dam method). (From Ref. 11.)

logic chips. The fully populated area-grid array, in which every grid point
is occupied by a solder bump, required the complexity of multilayered,
cofired ceramic packages. In these packages, the distribution of 110 wiring
could be accommodated by via "microsockets" and multiple buried redis-
tribution layers of wiring as opposed to single-level wiring, where the
"escape" of wires is geometrically restricted by the maximum number of
lines per channel between 110 connections [14]. The progression of C4
"footprint" geometry on chips is shown in Figure 8-8.
An example of an early full area-array C4 configuration is shown
in Figure 8-9. The 110 count was only 120 in an efficient square grid
array, which is 11 C4 pads long by 11 pads wide on 250-llm (lO-mil)
centers. A 125-llm (5-mil) solder bump is located at every intersection
in the grid except one, which is displaced for orientation purposes [16].
Some packages, such as the cofired alumina multilayer ceramic (MLC)
(Figure 8-10) used 9-133 area-array chip sites per package to attain high
bipolar circuit densities in IBM's 4300 and 3081 series computers. Logic
11-140 CHIP-TO-PACKAGE INTERCONNECTIONS

(d)

Figure 8-8. Progression of C4 area array from (a) essentially peripheral 110 to (b) staggered
double row to (c) depopulated array to (d) full-area arrays. (From Ref. 15; reprinted with
permission of Solid State Technology.)
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-141

Figure 8-9. Area Array C4 Configuration. (a) An 11 x 11 full area array of solder
bumps on a 700-circuit logic chip for use with multilayered ceramic. (b) SEM view
of solder bumps. After Goldmann, Ref. [6], 1983, reprinted with permission of Solid
State Technology.

and memory chips were mixed as required. As many as 25,000 logic


circuits or 300,000 memory bits were packaged on a single TCM substrate
in this technology [14,17-20). By adding thin films to a glass ceramic
MLC base, IBM was able to achieve 70,000 C4's on a single module for
its 9000 series computers [21] . These modules hold 121 integrated-circuit
(IC) chips with 648 pads each, and 144 special low-inductance C4 MLC
capacitors with 16 pads each (Fig_ 8-11)_ A highly populated logic flip
chip of earlier times was a "computer-on-a-chip" which had 762 C4 solder
bumps in a 29 x 29 area array (Fig. 8-12) [24]. Four levels of metal wiring
in this chip were used, compared to the two or three levels used previously.
More recently, the tremendous improvement in lithography with
minimum dimensions in fabs at 0_5 /lm (0.35 just starting) and advanced
chip wiring rapidly advancing with chemical vapor deposition stud forma-
tion and chemical-mechanical polishing to planarize wiring (five and six
levels), great strides have occurred in device density, chip size, and UO
count. Leading-edge ASICs have 40,000 bipolar circuits (Hitachi Skyline
machine) with approximately 2250 UOs which is well ahead of the SIA
roadmap. Power PC microprocessor flip chips from IBM and Motorola
have area arrays of approximately 650 UO.
11-142 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8·10. Cofired Ceramic Modules. Multilayered ceramic packages: The 9-chip
multichip module (MCM) in alumina/molybdenum technology (a), and the l21-chip
thermal-conduction module (TCM) in glass-ceramic/Cu technology (b).

Over time, the C4 technology has been extended to other applications.


C4s are used on thin-film resistor and capacitor chips in hybrid module
applications [25]. Solder pads for some applications can be very large-
750 ~m in diameter. At the other extreme, Schmid and Me1choir [26]
used C4's for precision registration and alignment in the joining of a
GaAs waveguide. The C4's in this case were only 25 ~m wide and high.
The most dense area-array application reported has been a 128 x 128
array of 25-~m bumps on 60-~m centers, resulting in 16,000 pads [27].
The photolith process for forming these is discussed later. C4' s or solder
connections on substrates similar to C4's and referred to as ball-grid
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-143

Figure 8-11. Flip-Chip Decoupling Capacitor Made with Pt Metallized


BaTi03 Platelets Terminated with 16 C4's. (From Ref. 22.)

Figure 8-12. Computer-on-a-Chip. An II,OOO-circuit "computer-on-a-chip" with 762


solder bumps in a 29 x 29 area array. (From Ref. 23.)
11-144 CHIP-TO-PACKAGE INTERCONNECTIONS

arrays (BGA's) are used for attaching chip carriers to boards and have
become part of the surface-mount revolution discussed in Chapter 7,
"Microelectronics Packaging-An Overview." New applications of this
technology are being explored continually.

8.3.2 Materials
Melting point has been a prime consideration in the choice of solder
alloys for C4's. High-lead solders, especially 95 Pbl5 Sn and 97 Pb 3 Sn,
have been most widely used with alumina ceramic substrates because of
their high melting point, approximately 315°C. Their use for the chip
connection allows other, lower-melting-point solders to be used at the
module-to-card or card-to-board packaging level without remelting the
chip C4's.
A reverse order of assembly (e.g., modules-to-board, then chips-
to-module) would require a reverse order of melting point. Josephson
superconducting devices have been joined in such a fashion, using an
alloy of 51 Inl32.5 Bil16.5 Sn (having a melting point of 60°C) for the
chip C4's, whereas a higher-melting-point eutectic alloy, 52 Inl48 Sn
(having a melting point of 117°C), was used for pins and for orthogonal
connections to the chip carrier [28-30].
Joining to advanced organic carriers such as polyimide-Kevlar@l
[16] or ordinary FR-4 printed-circuit boards [31-33] also requires lower
processing temperatures. Here, intermediate-me1ting-point solders, such
as eutectic 63 Snl37 Pb (melting point 183°C), and PbIn alloys, such as
50 Pbl50 In (melting point of approximately 220°C), have been used.
Joining with mixed solders has also been demonstrated. IBM has joined
chips with high-melting-point (315°C) PbSn solder balls to printed-circuit
boards with low-melting-point (l83°C) eutectic PbSn solder [34]. In this
case, the processing temperature is intermediate (250°C) and the high-
lead solder ball does not melt but is wetted by the low-melt solder on
the board. A listing of solder alloy compositions and melting points is
shown in Table 8-2. Some phase diagrams relevant to C4 solder joints
are shown in Figure 8-13.
The choice of terminal metals, which is described in detail later,
will depend on the choice of solder. For example, silver and gold are
poor terminal metals to be used with the SnPb alloy. In only a few seconds,
gold completely dissolves into the liquid solder. In these cases, another
solder alloy could be used, such as indium, which has a much lower
solubility for gold; or one of the other lower-solubility metals could be
used for the terminal. Thus, Cu, Pd, Pt, and Ni are very commonly used

Trademark of Dupont Corp.


8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-145

Table 8-2. Selection of Low-Melting Solder Alloys.


Melting Point ComQQsition in Mass Percent Other
m Sn Pb Bi In Cd Elements

16 24 76 Ga
20 8 92 Ga
25 95 Ga; 5 Zn
29.8 100 Ga
46.5 lO.8 22.4 40.6 18 8.2
47.2 8.3 22.6 44.7 19.1 5.3
58 12 18 49 21
61 16 33 51
70 13.1 27.3 49.5 10.1
70-74 12.5 25 50 12.5
72.4 34 66
79 17 57 26
91.5 40.2 51.7 8.1
93 42 44 14
95 18.7 31.3 50
96 16 32 52
96-98 25 25 50
103.0 26 53.5 20.5
96-110 22 28 50
117 48 52
125 43.5 56.5
127.7 75 25
139 43 57
144 62 38
145 49.8 32 18.2
156.4 100
170 57 43T1
176 67 33
178 62.5 36 1.5 Ag
180 63 34 3
183 61.9 38.l
183 62 38
198 91 9 Zn
215 85 15 Au
221 96.5 3.5 Ag
232 100
248 82.6 17.4
251 89 11 Sb
266 82.6 17.4 Zn
271 100
280 20 80 Au
288 97.2 2.8 As
304 97.5 2.5 Ag
304-312 5 95
318 99.5 0.5 Zn
321 100
327 100
356 88 Au; 12 Ge
370 97 Au; 3 Si
420 100 Zn

Source: From Ref. 35; reprinted with permission of Electrochemical Publications Ltd., Ayr,
Scotland.
11-146 CHIP-TO-PACKAGE INTERCONNECTIONS

1000

500 ~~~
,:aL:]
·""---'---'--'---t.Sn Pb 5n
1000

500

~~------~Bi A~U--~~~

1000' -

500~~\~
Cu Pb Pb Sb b Sn
1000 400

500
~~'98
Sn Zn Sn

Figure 8·13. Some Phase Diagrams Relevant to Soldering. Temperature, along


the vertical axes, is given in °C; concentration, along the horizontal axes, is given in mass
percent (Hansen and Smithells give concentrations in atomic percentages). (From Ref.
35; reprinted with permission of Electrochemical Publications Ltd., Ayr, Scotland.)

for both BLM and TSM. All of these metals form intermetallics with Sn,
which limits the reaction rates with PbSn solders. On the chip side, this
terminal metal is normally sandwiched between an adhesion metal layer
of Cr or Ti, and a passivation metal layer, usually of thin gold. The copper,
palladium, or nickel thin films on the substrate are typically protected
with gold [17-20] or are tinned with solder. In the latter case, some of
the solder for the C4 joint is supplied by the substrate [13,36]. MLC
substrates usually use a flash of gold on nickel [17-20,25,37-39]. Thick-
film substrates have the palladium or platinum alloyed with gold or silver
and are dip-soldered prior to the joining operation. AuPt, AgPd, AgPdAu,
and AgPt have been reported [8,11,40-44] as thick-film TSM pads.

8.3.3 Design Factors


Some of the factors affecting the material choices for the terminal
and solder have already been discussed, but other variables must also be
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-147

considered in C4 design. The joints must be high enough to compensate


for substrate nonplanarity, especially for the older version of thick-film
substrates. Because solder surface tension "holds up the chip," a sufficient
number of pads must be provided to support the weight of the chip.
Typically, this only becomes a cause for concern with very low I/O
devices, such as memory chips or chip carriers, which are very bulky.
Numerous studies have been published showing the interrelationship be-
tween BLM and TSM size, solder volume, chip weight, and C4 height.
Figure 8-14 shows these relationships over a very wide range [25].
Extra "dummy" pads, added to supplement those needed for simple
electrical connection, have often been used to enhance the mechanical
behavior, reliability, or thermal performance of the assembly [45]. Pad
location is important from the standpoints of both electrical design and
reliability. The effect of distance to neutral point (dnp) is discussed later
in this chapter as it relates to thermal-cycle fatigue.
The original solder bumps were placed over inactive Si in SLT, or
even over diode-isolated regions in early integrated circuits, just as wire-
bond or TAB pads are placed over inactive Si today. But it was soon
discovered in the early 1970s that C4 solder bumps could be safely placed

800

500

E
":-400
E'
i
is
.., 300
..'0•
.I:

l:ell 200
"i
:J:

100

Joint Solder Volume, VJ (mm 3)

Figure 8-14. Interrelationship Between BLM and TSM Size. Height of the joint
(h) vs. solder volume of the joint (~) as a function of chip weight (W in grams); rb is
radius of joint. (From Ref. 25.)
11-148 CHIP-TO-PACKAGE INTERCONNECTIONS

over active silicon devices and multilayer wiring. This unique capability
has made area-array bumping a very powerful packaging attribute and
has provided the designer with much freedom in wiring complex ULSI
devices. Automatically, area-array flip chips are smaller than their periph-
eral counterparts, meaning more chips per wafer and lower chip cost.
As VLSI logic chips become more and more dense, higher 110 counts
will drive full area arrays of terminals. In this case, the pad size and
location are fixed by the chip size and available real estate per pad allocated
by a fully populated area array.
The number of C4 pads as a function of chip size and pad geometries
is shown in Table 8-3 wherein the opportunity for 155,000 pads on a 20-
mm chip is indicated. Figure 8-15 shows the pronounced density advantage
of area array versus a single-perimeter row, as pad sizes and spacings de-
crease.

8.3.4 Bump Fabrication Processes and Tools


Evaporation through a metal stencil mask [7,8,15,25,29,41,42,46,47]
is still the most widely used technology for C4 terminal fabrication. BLM
and solder are both evaporated through holes in the metal mask and
deposited as an array of pads onto the wafer surface. This has proven to
be a practical and reliable batch process, simultaneously processing many
chips per wafer and many wafers per evaporation. However, it is not the
lowest-cost process nor will it accommodate finer bumps and pitches,
larger wafers, more precise overlays, or low-melting solders rich in Sn
which has a lower vapor pressure than Pb. Therefore, a number of innova-
tive and alternative fabrication technologies for bumping wafers have
begun to appear. Photolithographic mask fabrication with thin or thick
polymer films, screening of solder creams, electroplating of solders
through a photomask where the seed layer for plating becomes the BLM,
deposition of blanket sputtered BLM films followed by subtractive etching,
or combinations of these process designs are all beginning to appear to
bump wafers for various applications.

8.3.4.1 Metal Mask Evaporation Technology


IBM's C4 evaporated bump technology has been practiced for about
30 years and would be typical of this methodology. A passivated wafer
with via holes opened is aligned and assembled under a thin Mo stencil
mask (Fig. 8-16). The spring-loaded mask assembly is clamped tightly
against the wafer to avoid shifting during handling and deposition.
Argon ion sputter cleaning or etching of the via hole and the passiv-
ation surface around it is routinely used to remove aluminum oxide from
the last metal film on the chip and to remove any photoresist residues
from the passivation. This guarantees low contact resistance (typically
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 1\-149

Table 8-3. Pad Density Comparisons.

Chip Size Pad Diam. Pad Pitch # Pads # Pads


(mm) ll!:!!!l ll!:!!!l Perimeter Array

5 762 1524 8 9
635 1270 12 16
508 1016 16 25
381 762 24 49
305 610 28 64
254 508 36 100
203 406 44 144
152 305 60 256
127 254 76 400
102 203 96 625
76 152 128 1,089
51 102 192 2,401
25 51 388 9,604
10 762 1524 24 49
635 1270 28 64
508 1016 36 100
381 762 48 169
305 610 60 256
254 508 76 400
203 406 96 625
152 305 128 1,089
127 254 152 1,521
102 203 192 2,401
76 152 260 4,356
51 102 388 9,604
25 51 784 38,809
20 762 1524 48 169
635 1270 60 256
508 1016 76 400
381 762 100 676
305 610 128 1,089
254 508 152 1,521
203 406 192 2,401
152 305 260 4,356
127 254 312 6,241
102 203 388 9,604
76 152 520 17,161
51 102 784 38,809
25 51 1,572 155,236

Source: H. Nye, personal communication, 1986.


11-150 CHIP-TO-PACKAGE INTERCONNECTIONS

1.DIDI~

20

....a..
D 5
......

....~ 20

I
• 5
CHIP SIZE

PERItETER TERMINALS

1
1 10
TERMINAL SPACIN& (MILS )

Figure 8-15. Number of Pads Versus Pad Separation for Different Chip Sizes.
Input/output terminal trends. (Courtesy of H. Nye, 1986.)

less than 10 mn) and good adhesion to the SiOz or polyimide at this
interface. Originally, the plasma was created by a high DC voltage but has
migrated to an RF/AC process due to charge accumulations at unconnected
mechanical or thermal pads and because FET devices are more sensitive
to damage or parameter shifts.
The multilayered metals of the BLM are then deposited in the follow-
ing way. A typical evaporator would have numerous metal sources with
thermal energy supplied by either resistance, induction, or electron beam
(e-gun). Cr is sublimated first to provide adhesion to the passivation layer,
as well as to form a solder reaction barrier to the aluminum. A "phased"
or gradually mixed layer of Cr and Cu is coevaporated next, to provide
resistance to separation during mUltiple reflows. This is followed by a
pure Cu layer to form the basic solderable metallurgy. Finally, a flash of
gold is provided as an antioxidation protection layer and to promote
wettability. This is desirable because the wafers are normally exposed to
air before going on to the next step which is the solder evaporation. Solder
deposition requires a "thick" (on the order of 100 /lm) stencil mask.
Although lead and tin are usually in the same charge (single alloy, molten
pool), the higher vapor pressure component, Pb, evaporates and deposits
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-151

Chrome/Copper/Gold

(b)

(c)

<8>

Figure 8-16. Metal Mask Technology. (a) Tooling for alignment of mask to wafer;
(b) masking and evaporation of chromium/copper/gold; (c) masking and evaporation of
lead/tin; (d) reflowed solder bump. (From Ref. 42; reprinted with permission of ISHM.)

first, followed by tin on top of the lead. Reflow in a subsequent hydrogen


ambient furnace at about 350°C melts and homogenizes the two-layer
solder pad and brings it to a spherical shape. In addition, H2 promotes
some reduction of oxides of the Pb and avoids excessive oxidation of
the Sn.
11-152 CHIP-TO-PACKAGE INTERCONNECTIONS

Photolithographic processes and combinations of photolith and metal


mask are becoming more and more popular for fabricating terminals.
Most common is a sequence of blanket deposition of the BLM, application
of photoresist, development of a pattern in the resist followed by electro-
deposition of the solder; then removal of the resist and subetching of the
BLM, using the plated solder bumps as a "mask" [37,48-51]. An alterna-
tive sequence is to blanket-deposit BLM by sputtering, apply photoresist,
subetch the BLM through the resist, then deposit solder by a variety of
techniques including solder dip [52], solder ball placement [53], or stencil
mask evaporation [54]. The single-masking processes ("unimask") are
significantly simpler and cheaper than multiple-mask techniques, although
they are not as flexible in providing varying amounts of solder. In some
applications, higher volumes of solder are able to decrease strain, as
discussed in Section 8.3.6, "C4 Reliability."

8.3.4.2 Electroplated Bumps

It has been pointed out that extendability issues may place restrictions
on metal mask evaporation. As semiconductor wafers have increased in
size, it is becoming more and more difficult to hold height and volume
uniformity of the bumps across a wafer. Very small bumps, typically
desired for optoelectronic applications, are difficult to make with thick
metal masks, and alignment tolerances limit extendability for high-density
arrays. In addition, the drive to direct chip attach (DCA) on organic boards
is driving more users to demand lower-melting-point solders to be used
for the bump, especially SnPb eutectic with 63% Sn. These very high tin
contents are extremely difficult to evaporate because the tin vapor pressure
is so low as to require hard driving of the sources which can result in
meltback of the evaporated solder on the wafers during the evaporation. For
these reasons, photolithographic patterning, coupled with electroplating, is
gradually displacing the metal mask evaporation.
Photolithographic processes use photoresist layers directly on the
wafer for the terminal or solder bump definition and do not have the
problems of extendability that metal masks have. Smaller bumps are in
fact easier to make with resist. Electroplating various compositions can
be accomplished by adjusting the plating bath composition or by the
sequential plating of the individual components. Various combinations
of these methodologies have been reported [51,55,56]. Most common is
a sequence of blanket deposition of the "seed layer" (to be the BLM),
application of photoresist, development of a pattern in the resist, followed
by electrodeposition of the solder, then removal of the resist and subetching
of the BLM using the plated solder bumps as a "mask." A typical flowsheet
for the MCNC electroplating process is shown in Figure 8-17.
Hitachi [48] and Honeywell-Bull [57] have also created solder bumps
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-153

Evaporate UBM

-MIlD _
Pattern Solder Dam
<i

,
Pattern Plating Template

,
Electroplate Solder

,
Strip Photoresist

,
Reftow

,
Etch Solder Dam & UBM

Figure 8-17. Flow Sheet of a Regular Plating Process for Low-Cost Solder
Bumps at Microelectronic Center of North Carolina (MCNC). (From Ref. 51 .)

in the past by electroplating solder on a seed layer. Hitachi used its bumped
chips for early hybrid circuits [25]. Honeywell-Bull replaced conventional
TAB Au bumps with solder bumps to gang bond inner lead bonds by
soldering instead of using AuSn eutectic attachment. The similar Hitachi
and Honeywell-Bull structures are shown in Figure 8-18.
A unique extension of the single-mask concept has been demon-
strated at Microelectronic Center of North Carolina (MCNC), where not
only the solder and terminal are patterned simultaneously but also the
redistribution layer [B]. When a wafer has been designed for wirebonding
11-154 CHIP-TO-PACKAGE INTERCONNECTIONS

Solder Bump
after "wetback"

I
, ..
I
J
Nickel--.[=="
€L.B-~-~-~~~~~~~~~paSSivation
Film
TiW ~,aI!~i)-r;;;:;=-'I/Ji~:L..

AI-Si
Si

Figure 8-18. Electroplated Bump Structures from Hitachi and Honeywell-


Bull (Left and Right). (From Refs. 48 and 57.)

and one wishes to convert it to solder-bump flip-chip without extensive


redesign of chip surface wiring, this system combines inward redistribution
of the wiring simultaneous with bump formation to have a cost-effective,
dual solution. The process is referred to as single-mask redistribution
(SMR). The uniformly plated solder flows during wafer reflow to distribute
the solder primarily to the bump region as shown in Figure 8-19. The
resultant structure contains a very high-conductivity redistribution line
lightly coated with solder, connected to a larger-volume solder bump, all
fabricated from the same metallization/patterning steps.
Examples of MCNC electroplated bumps are shown in Figure 8-20.
A high-density array of as-plated bumps is shown in Figure 8-20(a). The
uniformity of solder deposit on this chip with 1679 bumps is evident.
After reflow, the bumps appear spherical as shown in Figure 8-20(b). The
process has been exercised on bumps as small as 40 )lm and as large as
600 )lm in diameter.
The final operation before dicing of the wafer into individual chips
is the electrical testing of each chip. Mechanical probes are used to contact
the soft solder bumps fabricated earlier. Area-array bumps are tested with
special buckling beam assemblies of wires [58].

8.3.5 Assembly/Rework
The formation of wettable-surface contacts on the substrate (provid-
ing a mirror image or "footprint" to the chip contacts) is achieved by
thick- or thin-film technologies. Thin-film contact technology is similar
to the BLM described previously, but thick-film technology involves the
development of wettable surfaces by plating nickel and gold over generally
nonwettable surfaces such as Mo or W (conductors usually used within
the ceramic substrate). Solder flow may be restricted by the use of glass
or chromium dams where necessary. Various thin- and thick-film processes
which are typically used are illustrated in Figure 8-21.
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-155

'-01 J
Plating Template
Top VIew

Solder Oeposlt
Sid. VIew

Ouling Reflow

Ca)

(c)

Figure 8-19. Flow Diagram for SMR Template Formation [B]; (b) Plated
SMR Template [B]; (c) Completed SMR Structure, Including the Redistribu-
tion Trace and Solder Bump [B]. (From Ref. 56.)
11-156 CHIP-TO-PACKAGE INTERCONNECTIONS

(a)

(b)

Figure 8-20. Electroplated C4 Bumps in the as Deposited (a) and Reflowed


Conditions (b). (From Ref. 56.)

8.3.5.1 Self-Alignment

One of the finest attributes of the C4 process is its self-alignment


capability arising from the high-surface-tension forces of solders [47].
Chip pads and their counterparts on the substrate may be separated by as
much as three times the average bump radius, but if the mating surfaces
touch and are reasonably wettable, self-alignment will occur (Fig. 8-22).
Optoelectronic package assemblies use this phenomenon extensively [60-
64] and allows precision ± 1 J..lm self-alignment of optical devices to
waveguides and optical fibers. The productivity of the reflow joining
process when done in an oven is extremely high. As many as a million
C4 bonds can be made in 1 h using automated tools.
The self-alignment dynamics have been modeled as in Ref. 65.
Figure 8-23a shows the restoring force versus misalignment for a chip
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-157

Screen & Fire Thick Film


Thick Film Conductor

Plated Au

Plated NI

Thick Film MLC


Mo
(Multilayer Ceramic)

crRemOVed\

Thin Film

__ PO_I_yl_m_ld_e_....,~~----\
\_ /_
rr--
Copper MuHllayer
Thin Film
Polylmlde

Copper

Figure 8-21. Various Substrate Pad Structures (TSM) for Thick- and Thin-
Film Substrates.

controlled
collapse
so~,\aer
w.r~bl:~:S,\ / EquHibrium

- I /

8
shape
ball

..........
Initial confact

(a) (bl (el (d)

Figure 8-22. Stages of Joining Process for C4: (a) Rough Alignment of Compo-
nents in Initial Placement; (b) Joining; (c) Self-Alignment Begins; and (d)
Bond Complete, Components Aligned. (From Ref. 59.)
11-158 CHIP-TO-PACKAGE INTERCONNECTIONS

(lor pod liz. = 100 mlcronl dla)


0.10

0.08

......
...c 0.08
0

.!.
•.u 0.04
~

0.02

0.00
0 10 20 .30 40 :10 10 70
(a) p(mlcrons)

40~-----------------------------------'

20

..
c:
eu o
'e
'Q:

-20
DEVICE WI 49 BUMPS

-40+---------~--------~--------~----~
o 2 3

(b) .(s.c)

Figure 8-23. Self-Alignment Restoring Forces and Oscillations During Dy-


namics of Solder Wetting. (From Ref. 65.)

with 49 bumps IOO-l1m in diameter. During joining, a dynamic oscillation


is predicted which decreases in amplitude due to the viscous damping
forces of the liquid solder. This is illustrated in Figure 8-23(b). The
oscillation frequency is very high, about 10 times per second. The viscous
damping gradually reduces the amplitude of the oscillation, and reaches
the I-11m range after 3 s. Another chip with only 12 bumps has restoring
and damping forces only one-fourth of those in the first case. These levels
are small compared to the chip inertia and result in the damping being
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-159

not as effective, reaching 10 !lm amplitude after 3 s. Thus, a minimum


time is required for the solder to be in the liquid state to allow the full
damping to occur before the solder freezes. Bache (Fig. 8-24) has used
two solder bump sizes on the same chip to allow the self-alignment to
occur in stages. The larger bumps "catch" first and in self-aligning allow
the smaller bumps to "catch" and finish the precision alignment. A better
than l-!lm accuracy is attained with the process.
Inspection of the joined chips to assess the degree of self-alignment
is difficult due to the joint being "hidden." Optical techniques can be used
only if the devices are transparent to the optical wavelengths-such as
infrared for silicon devices. Bache [62] has used a vernier made of high-
density solder bars on both chip and substrate coupled with x-ray-
radiography to measure the degree of alignment as shown in Figure 8-
25. The vernier structures consist of a series of aligned bars on each of
the two components. These bars have slightly different spacings on the
two components such that any misalignment will result in the bars of the
two halves of the vernier coinciding away from the central bar. The actual
misalignment of the assembly can then be measured by counting the
number of bars between the central bar and that bar where the registration
occurs, and multiplying this by the difference in spacing of the two sets
of bars. The z dimension (chip-substrate gap) is also extremely important
for certain optical applications. This has been measured [66] by using
Fourier transform infrared spectroscopy (FTIR) and measuring the spacing

LD o
I

I-RA-j

_I RN I.. , I
iO
: I j r.'-
I
I
I
,

) )
Figure 8-24. Alignment of Fine Bumps with the Use of Coarser Bumps. (From
Ref. 62.)
11-160 CHIP-TO-PACKAGE INTERCONNECTIONS

a) 'Observer'

Chip

Solder vernier bars Main solder bump array Substrate

b)

-- 100!1m

r--------- -- -- - -- - - -- - - - -- - -I Vernier B
: I Substrate

1000 ~O&UAtr~i~tr~tr~o 0 0 00:


I Solderable Bars"", ---___ :
IL _ _ _ _
Solder
_ _Applications
_ _ _ _Area Vernier
___ __ Centre
_ _Bar_ _ _ _ _ _ JI

Figure 8-25. Solder Vernier Structure for Monitoring Flip-Chip Component


Alignment. (a) Cross Section, Schematic; (b) Plan View of 2-fLm Resolution
Vernier. (From Ref. 62.)

between the interference peaks (Fig. 8-26). Bond heights to ± 0.5 J.1m
was achieved with nominal 16-J.1m-high joints.

8.3.5.2 Conventional Flux-Assisted Joining


Once the BLM, TSM, and solder are in place, as described previously,
the joining of chips to the substrate using C4 technology is straightforward.
The chip needs to be aligned and placed upside-down on the substrate.
Commercial tools are available to give ± 2-J.1m placement accuracy in
the laboratory, and ± 8-11m in manufacturing [67]. Flux, either water-white
rosin [50,68-70] for high-lead solders or water-soluble flux [28,29,71] for
low-lead and other low-melting solders, is normally placed on the substrate
as a temporary adhesive to hold the chips in place. Fluxless and no-
elean processes are being explored as well and will be described in the
next section.
Once the chip-joining operation is complete, cleaning of flux residues
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-161

~ooo 3700 3~OO 3100 2800 2500 2200


WAVENUf'1BEfI

Figure 8-26. Interference Fringes Obtained by FTIR Spectroscopy for the


Measurement of Solder Bond Height. The height in this example in 15.5 ± 0.2 11m.
(From Ref. 66.)

is accomplished with such solvents as chlorinated solvents or xylene


for rosin flux and water for water-soluble flux. The assembly is then
electrically tested.

8.3.5.3 Fluxless Joining with Plasma-Assisted Dry Soldering


For fluxless plasma-assisted dry soldering (PADS) processing, flux
is replaced with a dry pretreatment performed using a commercial plasma
pretreatment tool (see Fig. 8-27). During this pretreatment, the oxide
which inhibits joining is converted to oxyfluoride. This conversion film
passivates the solder and has the unique property of breaking up when
the solder melts, exposing a free solder surface and allowing reflow and
joining to occur in the absence of a liquid flux or reducing gas.
The chip-joining task involves first flipping the chip over and placing
it precisely on the matching pads of the substrate or carrier. Both manual
[73] and automated tools [67] are available. Manual tools use an optical
head incorporating a microscope and vertical illumination system that
enables the operator to view both the chip and substrate simultaneously.
Coupled with a joy-stick-controlled sample motion with 6 degrees of
freedom, the chip and substrate can be aligned within 2 ~m [73]. For
rough surfaces, the optical probe have been modified by adding two
illuminators and fiber-optic light pipes which introduce the light to the
11-162 CHlp·TO·PACKAGE INTERCONNECTIONS

Figure 8·27. Fluxless Pretreatment PADS Tool. (From Ref. 72.)

substrate at an angle. This arrangement allows any type of substrate to


be used. An example is shown in Figure 8-28, where silicon chips with
high-lead (95 Pb Sn) solder bumps were fluxlessly joined to a eutectic
solder-dipped FR-4 organic substrate made from printed circuit board
stock. The rough surface of the FR-4 is visible in the bottom part of the
lower cross section (Fig. 8-28b). It is also visible on the virgin surfaces
before joining in Figure 8-28c. Automated tools use pattern recognition to
place chips very rapidly on the substrate with accuracies in the ± 5-8-J-lm.
In flux joining, flux acts as an adhesive to hold the chip in place
during the transfer and handling involved in mass reflow operations in
belt furnaces. For fluxless joining there is no "glue." Instead, a tacking
regimen has been employed which utilizes various combinations of pres-
sure, time, andlor temperature to temporarily bond the chip to the substrate
in the aligner bonder during the chip placement operation [731- Both upper
and lower chucks could be heated, and the chip placement control computer
set to apply the required loads/times/temperatures needed for the applica-
tion. Once tacked, the assemblies are transported to the nitrogen reflow
oven where the permanent chip joining is accomplished and self-align-
ment occurs.
The accuracy of the chip placement is most important for small
bumps. In addition, the position of the chip relative to the substrate in the
final assembly is critical for optoelectronic applications [62-64,66,74-76].
Self-alignment takes place when the solder reflows [47] and allows the
final chip to be located ± I-J-lm with precision [73] . The relative positions
of the chip and substrate after placement or after reflow can be observed
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-163

Figure 8-28. Fluxless Joining of 95 Pb 5 Sn C4's to Eutectic Tinned FR4


Terminations by PADS Fluxless Joining. (a) Cross section of chip-on-board; (b)
cross section showing 95 Pb 5 Sn bump soldered to card with 63 Sn 37 Pb eutectic, and
(c) eutectic tinned FR4 terminals and rough FR4 surface. (From Ref. 72.)
11-164 CHIP-TO-PACKAGE INTERCONNECTIONS

and measured optically with white light for glass substrates, or with
infrared (IR) for silicon which is IR transparent. Separate infrared micro-
scopes are available which, when used in conjunction with thin-film orien-
tation patterns and verniers built into the components, allow the relative
chip-to-substrate positions to be measured very accurately. Several thin-
film patterns are shown in Figure 8-29a, whereas Figure 8-29b shows the
vernier patterns imaged with IR [18]. The images are slightly out of
focus because both the chip and substrate surfaces are being viewed
simultaneously with the focal plane halfway between the two surfaces.
They are sufficiently clear to be able to demonstrate better than l-/lm
self-alignment accuracy using photolithographically defined solder-bump
structures with 40-/lm diameters.
The final chip joining is usually performed in a belt furnace with
inert nitrogen. The same furnace can be used with either flux or fluxless
joining. Belt furnaces have extremely high productivities-as many as a
million flip-chip solder bonds can be made in 1 h. For laboratory use,
local heat sources have been employed to bond one chip at a time [51].
An infrared belt furnace was used to produce the fluxlessly joined and

Aligned Misaligned

-.. •. • • • ...,,'.,. I

Figure 8-29. (a) Positional Test Patterns (Thin Film); (b) IR Images.
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-165

self-aligned eutectic solder bump shown in cross section in Figure 8-30.


The substrate used had bare copper terminals. No post-join-cleaning is
required for the fluxless process. However, cleaning of the flux residues
is normally required for flux-based processes. Chlorinated solvents had
been widely used for many years but have been banned due to their
propensity to deplete the ozone layer. Other solvents such as xylene are
now used for rosin fluxes, and water-based formulations are used for water-
soluble fluxes. Once the chip-join operation is complete, the assembly is
electrically tested.

8.3.5.4 Rework of Multichip Assemblies


One of the most important issues in high-end multichip module
products is that of rework. Whereas the chip-join process is an extremely
high-yield process, and defective joints are rarely found in manufacturing,
defective chips are occasionally encountered which require replacement.
Even when "known good die" are used, it is sometimes necessary to
replace a die in a multichip assembly for engineering changes. The rework
process for a flip chip usually requires three steps: (l) chip removal, (2)
site dress, and (3) placement and joining of a new chip.
The chip removal is accomplished mechanically by ultrasonic torqu-
ing of the chip (Fig. 8-31) or (Fig. 8-32). This minimizes any damage
that can occur on the substrate or chip (as can happen, for example, when
a cold mechanical pull, shear, or torque is used). Local heat can be supplied
by IR [69,77] or by conduction through upper and lower chuck heaters
in an alignerlbonder [55] . In order not to melt all the solder joints on
passenger chips-those chips on the assembly not being reworked-the

Figure 8·30. Self-Aligned Fluxless Joint to Copper.


11-166 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8-31. Excess Solder. SEM photograph of residual solder left on a typical
substrate microsocket after mechanical removal of the device. (From Ref. 69; reprinted
with permission from Journal of Electronic Materials.)

chuck holding the substrate is heated to a temperature slightly below the


melting point of the solder while the upper chuck is heated to a temperature
above the melting point of the solder. The pedestal is aligned to the die
and the lower chuck raised until the chip is secured to the upper chuck
by means of the vacuum contact on the pedestal. The chip stays on the
pedestal when the bottom chuck is lowered.
An example of a substrate surface after hot chip removal is shown
in Figure 8-32. The solder is 95/5 Pb/Sn in a 41 x 41 array of l25-mm-
diameter bumps on 250-mm centers. This was performed in air without
flux . To accommodate the solder melting temperature of approximately
315°C, the upper chuck was heated to 350°C while the lower chuck was
heated to 300°C.
After chip removal, the site usually needs to be dressed. This entails
removing most of the solder so the residual is very uniform and does not
interfere with the joining yield of the new device or the reliability goals
of the assembly. Hot-gas tools [69] have been used in the past (Fig. 8-33),
but currently the most common practice is to use a solder wick made out
of a powder metallurgy block of copper [78] (a proprietary development
of IBM corporation). The block is placed on the substrate site either
manually (alignment is not critical) or with the aligner bonder as if it
were a chip to be joined. The assembly is then given a reflow during
which the excess solder is absorbed into the copper block by surface
tension. It can be picked off the site which is now ready for the new chip
to be attached. A dressed site is shown in Figure 8-34.
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-167

..

{' ~ () t) (,,, ('


() (~ (' (", ('to, (~
------- .

,~ /~ (' ~ (~ ('
,
~ . (' /
(' (>tt; I (1 ~

r
-

(~ C' (~ (~ - ~
0016 20KU X70 100Vu H019
Figure 8·32. Substrate Site After Hot Pull

The new chip is joined to the substrate in exactly the same fashion
as the original chips, using a chip placement tool in its normal fashion
to align and place the new chip. Joining reflow (and clean if necessary)
is accomplished also in the normal fashion as with the original chips.
Besides the electrical testing, numerous techniques are available to
characterize the joints. The concern for quality of the joints is often raised

.~""'""-01_,!,"""_~,,, Exit:
Nitrogen from Hot Gas &
the Heaters ,.....--"I""ir-i Solder Debris

/k-J-+--+-NI---Insert

Insert Recess

I I
. - - - - 1IlIo.II......I~......,............&;.;--lnsert

I Exit Port
Residual Solder

Figure 8·33. Hot-Gas Tool. Side-view sketch showing the direction of gas flow within
the probe tip of the hot-gas dress tool. (From Ref. 69; reprinted with permission from
Journal of Electronic Materials.) .
11-168 CHIP-TO-PACKAGE INTERCONNECTIONS

oo_eo-
ee __ ee
I

.'
eoeeee I

e " .e;,e e e
I '

t ,<e e _ e ~ ~
,

; ,899S . 20KU X70 1getJM WD24

Figure 8-34. Substrate Site After Copper Block Dress

as an issue for new users or developing processes because the joints are
hidden and not visible with standard inspection tools. X-radiography
[75,79-81] has been used to supplement the optical techniques described
earlier. The x-rays can detect misshaped joints, voiding, and low-volume
solder pad defects. Acoustic imaging has also been used to detect interfa-
cial debonding of solder bumps to pads [80].
The solder-joint height, usually the chip-to-substrate gap, is also an
important factor for thermal-cycle fatigue reliability and optoelectronic
device alignment (z axis). This has been measured by Fourier transform
infrared spectroscopy [75]. Reflections at the chip and substrate give
rise to interference peaks whose separation is a function of the chip-to-
substrate spacing.
Several application examples of solder flip-chip assemblies are
shown below to illustrate the technology. Chips joined in flip-chip fashion
to FR4 are shown in Figure 8-35.
A cross section of these joints was shown earlier in Figure 8-28.
The chips used in this demonstration were fabricated test vehicles with
1679 C4 bumps. Each chip contains many unique structures, such as
temperature-sensing diodes, Kelvin connections for four-point contact
resistance measurements, daisy chains for yield determinations, piezoresis-
tive strain gauges, and reliability test patterns. These structures allow one
to fully characterize the joining and rework processes as well as the chip
and substrate fabrication processes. A MEMS device (microelectromecha-
nical structure) device assembly is shown in Figure 8-36. The fine movable
structures of the thin films are not disturbed by joining process because
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-169

Figure 8-35. Chips Joined to FR4

Figure 8-36. Fluxless Flip-Chip Joined MEMS Device


11-170 CHIP-TO-PACKAGE INTERCONNECTIONS

no flux was used in the assembly. A cross section of one of the joints
was shown in Figure 8-30 and illustrated the compatibility of the fluxless
process with bare copper metallurgy.
An application of flip chip to flexible circuit attachment is shown
in Figure 8-37. Here, as well as in the case with FR4 organic substrates,
low process temperatures are required to be compatible with the organic
materials involved and low-melting eutectic lead tin alloys are preferred.
The joints in Figure 8-37 are under the flex circuit copper leads,
therefore not visible. However, they are stronger than the leads which
tore from the flex circuit in a pull test evaluation.
The fluxless joining of high-lead alloys in the range 90/1 0 PbSn to
97/3 PbSn is done at higher temperatures, typically 350°C. An example
of a 97/3 PbSn flip chip joined to MoNiAu microsockets on a multilayer
ceramic substrate is shown in Figure 8-38. These are fracture surfaces
after pull testing indicating ductile taffy pulls which indicate excellent
wetting to the MLC microsockets occurred.

8.3.6 C4 Reliability
This subject is covered in detail in Chapter 5, "Package Reliability,"
but because one of the major limitations of this technology is how large
a chip area can be bonded and still remain reliable, a brief discussion is
included here.
A question often raised regarding flip-chip bonding is the ability of
the joint to maintain structural integrity and electrical continuity over a
lifetime of module thermal cycling. A thermal expansivity mismatch
between chip and substrate will cause a shear displacement to be applied
on each terminal. Over the lifetime of a module, this may lead to an
accumulated plastic deformation exceeding 1000% [83]. A quasi-empirical

Figure 8·37. Flip Chip on Flex


8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-171

Figure 8·38. 9713 PbSn Fluxless Flip-Chip Joined to MLC. (From Ref. 82.)

model was developed by Norris and Landzberg [40] that relates the cyclic
lifetime to cyclic deformation parameters. The modified Coffin-Manson

relationship after Norris and Landzberg is NJ = (~r9 f0 3 e-(0.123IkTmaxl. It


is based on the Coffin-Manson relationship [84] between fatigue life and
plastic strain amplitude but with two terms added to account for time-
dependent behavior: a frequency term, where lifetime increases with fre-
quency to a lower power, and a maximum-temperature term, where lifetime
decreases with maximum temperature. With the assumption of a lognormal
failure distribution, a product sample may be tested in an accelerated
thermal cycle; then, based on the statistical lifetime to electrical failure,
the projected field lifetime may be extrapolated. Using this technique, an
interconnection failure rate projection was made for logic chips in System!
370 [40] of no more than 10-7%11000 h per bond at the end of life.
By the end of 1975, 540 billion MST interconnection hours had been
accumulated with no wear-out failures reported [85], yielding a 50%
confidence-level-estimated failure rate of 1.3 x 10-7%11000 h, in good
agreement with the earlier projection. In 1991, IBM reported that there
had been zero failures due to thermal-cycle fatigue wear-out on the MLC
C4 technology over a 12-year period since its introduction in 1980 [86].
This is more than 10 12 Power-on-Hours.
These results demonstrate the inherent capability of solder intercon-
nections to withstand high-strain accumulations; the results also demon-
strate the approximate validity of the projected failure rate. However, the
early MST chip was only about 1 x 1 mm in size and had only 12
peripheral bumps. As discussed in Chapter 5, "Package Reliability," the
mismatch shear deformation is proportional to the distance between a
given pad and the neutral point referred to as dnp (the point on the chip
1/-172 CHIP-TO-PACKAGE INTERCONNECTIONS

that remains stationary relative to the substrate during a thermal excursion).


Because the neutral point is near the chip center, the maximum value of
dnp is roughly proportional to chip size. Moreover, the Norris-Landzberg
model and most subsequent experiments show that lifetime is inversely
proportional to shear deformation raised to a power that approaches 2.
Thus, with the evolution from MST to much larger and denser C4 foot-
prints, thermal wear-out was considered with renewed interest.
The wear-out model subsequently has come under close scrutiny. It
has been suggested, for instance, that a dwell time factor be added [87].
Also recommended is a complete reformulation starting with the constitu-
tive equation for solder that incorporates crack growth and creep deforma-
tion [86]. It has been further suggested that several competing mechanisms
come into play in thermal wear-out, including cavitation. Recent experi-
mentation [88] also suggests that wear-out is much more complex than
implied by the simple equation discussed above. Scanning electron micros-
copy (SEM) of joints at various stages of thermal cycling show that
mismatch between the solder and chip plays a role in the damage. Also,
low and high volumes of solder joints fail by different mechanisms, thus
emphasizing the effect of C4 shape.
A further complicating factor in modeling wear-out is that thermally
induced strains are not uniform within the joint. An initial attempt [89]
was made to incorporate joint shape into failure-rate projections. Despite
the simplicity of the model, reasonable agreement has been obtained in
experiments where joining geometry has been intentionally varied and
the joints mechanically tested by single-cycle or cyclic torquing of the
chip [25,89]. More sophisticated techniques, taking into account time-
dependent and temperature-dependent solder properties, will be required
to understand geometric effects fully. Chip bending also must be included
in a complete analysis [90,91].
In summary, simple models that accompanied the introduction of
C4 joining have, until now, proven adequate to estimate field behavior
and as an aid in product design. This can be attributed not only to their
simplicity and their qualitative rationality but also primarily to the fact
that existing products have experienced wear-out failure rates too low to
be of concern. As chips grow larger and pad counts in the hundreds
become common, new or revised models will be required that reflect a
greater understanding of the wear-out mechanism and better precision in
failure projection.
8.3.7 Thermal Mismatch Reliability: Extensions
Together with a greater understanding must come an extension of
the C4 in its ability to accommodate larger and denser chips without
affecting system reliability. Existing modeling and testing techniques,
however imperfect, have been used to evaluate various extension schemes,
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-173

several of which show promise. They may be subdivided into strain


reduction, geometry or shape improvement, alternate solders, and underfill
encapsulant effects.

8.3.7.1 Design and Geometry


The thermal mismatch displacement across the pad can be kept to
a minimum by arranging the footprint to minimize the dnp: for example, by
deleting comer pads or, in an extreme case, by a quasi-circular array [15].
Joint geometry is dictated by the wetting areas on chip and substrate,
solder volume, and the weight of the chip. Unless the chip is very heavy,
the joint has the shape of a "truncated" sphere [25,89,92], and its height
is uniquely determined by interface radii and volume. The previously
mentioned geometric model [89] claims that wear-out depends on shape
and that ajoint can be geometrically optimized to extend lifetime. Although
verifications have been primarily by mechanical testing, with its inherent
limitations, the following optimization philosophy and sequence are proba-
bly valid:

1. Consistent with other design and process constraints, the interface


areas should generally be as large as possible.
2. There is an optimum ratio of substrate-wetting area to chip-wetting
area, which must be determined experimentally for each particular
material set. For thin-film copper lands on ceramic (discussed in
Chapter 9, "Ceramic Packaging"), the ratio is about 1.2, with the
substrate pad larger. The optimum ratio is characterized by a roughly
even distribution of thermal-cycle fails between chip and substrate
(solder crack near BLM or TSM intermetallics).
3. For fixed interface radii, there is an optimum solder volume. A
model exists, but because verification has been by mechanical testing
only [25], thermal-cycle testing for the optimum is recommended.

The preceding discussion pertains to joints whose shapes are domi-


nated by solder surface tension. Under these conditions, the joint takes
the shape of a doubly "truncated" sphere, truncated at each end by the
contact metallurgies. Normally, optimization of the spherical segment
joint is of limited extendibility value, improving lifetime by less than
50%. Enforced changes in shape away from a spherical segment, on the
other hand, can produce very large effects. Heavy chips that depress
the joint, causing deep notches [25], severely reduce lifetime, whereas
stretched or elongated joints substantially extend lifetime. Mechanical
testing has shown an order-of-magnitude difference in fatigue life between
an hourglass and a barrel-shaped joint, with the fracture location of the
hourglass joint shifted to the center of the joint [93]. Stretched pads have
11-174 CHIP-TO-PACKAGE INTERCONNECTIONS

been fabricated by a number of techniques including using two different


solders on the same chip [93]. Called SST (self-stretching soldering tech-
nology), this technique makes use of the surface-tension forces of larger
bumps of one solder to stretch the lower-volume functional bumps (Fig.
8-39). Two additional concepts being pursued in Japan are illustrated in
Figure 8-40, wherein solder columns are being stacked to achieve im-
proved fatigue life.
In addition, solder has been cast into helical copper coils to form very-
high-aspect-ratio solder columns. It has been applied to joining leadless
ceramic chip carriers (LCCC) to glass-epoxy printed-circuit boards (Fig.
8-41). This structure has not been scaled down to sufficiently fine dimen-
sions to be applicable to integrated-circuit chip interconnections. Free-
standing cast-solder pillars (without the copper helix) have been developed
for package-to-board interconnections [33]. See Chapter 16, "Package-
to-Board Interconnections."

8.3.7.2 Solder Composition


Among solders, which have been evaluated as alternatives to 95
Pb/5 Sn, the Pblln system has shown the most fatigue enhancement.
Thermal-cycle lifetime has been shown to be quite sensitive to composition
with a minimum at 15-20% In [68,97]. There is a 2 times improvement
over 95 Pb/5 Sn at 5% In, 3 times at 50%, and 20 times at 100% In [97].
Pure In is being used for optoelectronic device joining [26,27]. Early
work on integrated circuits emphasized 50 Pb/50 In as a compromise
between ultrahigh thermal-cycle reliability and processing constraints.
Implementation of this alloy was limited by two factors: increased corro-
sion susceptibility in nonhermetic packages [68,97,98], and a substantially
accelerated thermomigration rate over PblSn alloy [99]. By the latter
mechanism, the thermal gradient between chip and substrate causes a
condensation of vacancies at the chip BLM region, leading to premature
high resistance or mechanical wear-out. Later work emphasized low (3-
5%) In alloys [68], which provided less fatigue enhancement but were

r.:::::I t"",., , - 51 Chip


~ \ iWvnmu&«U!J1 -*(& VA (
'" - Control Electrode

)
&4 nep wm?p@mh e#:..-- Solder J~nt •
~ (Tmp-314 C)
'-=~-J
mmnmx Control Solder

..
I
m~ _
~ ms--
I (T
mp-
-318'C)

AJ203 Substrate

mSI Joint Height hi

Figure 8·39. Self·Stretching Soldering Technology. (From Ref. 93.)


8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-175

~ Pad (iJ200/Lm)
b
"Vfl"':@2"-"0""7fi~@>m0""7fi~Wll4k;m~_ Ceramic Substrate
In-Sn (t =20/Lm)
Sn-Pb (t =1oo/L)
Mask (Metal)

.m:- ChiP

Figure 8-40. Some Methods Being Pursued to Extend C4 Life. (a) Stacked solder
bumps using polymide (from Ref. 94); (b) stacked solder bumps using multiple solders
(from Ref. 95).

not as susceptible to the corrosion and thermo migration difficulties of 50


Pb/50 In. Testing 95 Pb/5 In at several thermal-cycle frequencies showed
a similar relationship to that found by Norris and Landzberg for 95 Pbl
5 Sn.
Subsequent optimization of the 95 Pb/5 Sn system showed that a

Figure 8-41. Solder Columns. Closeup view of the solder columns connecting a LCCC
to a glass-epoxy printed-circuit board. (From Ref. 96.)
11-176 CHIP-TO-PACKAGE INTERCONNECTIONS

97 Pb/3 Sn composition also brought nearly a 2x improvement in fatigue


life over 95 Pb/5 Sn without the thermomigration and corrosion concerns
of the Pblln alloy. Therefore, when IBM System 3080 was introduced,
the solder composition was altered to realize the increased reliability
limits. At about the same time, it was appreciated that tinning metallized
ceramic Cr-Cu-Cr C4 pad tips was deleterious to fatigue life because
the solder used had 10% Sn-the wrong Sn direction for long fatigue
life. A process was therefore invented for putting solder balls on pin heads
only, leaving the C4 sites bare [79]. All the C4 solder then came from
the chip only, and both fatigue life and chip joining yields were simultane-
ously improved.

8.3.7.3 Substrate Materials


By far, the largest effort to reduce the strain drastically has been by
matching the substrate thermal expansion to that of silicon, as demon-
strated in Figure 8-42.

Si or Glass Ceramic

10,000

0
I/)
Z
rn
CD
'0
>- 1000
9
.l!!
:::i
CD
:::l
Cl
:;
L1.

100
1/"(2 Coffin-Manson Equation

o 40
Coefficient of Thermal Expansion (10 - 7/ 0 C)

Figure 8-42. Effect of Thermal Expansion Coefficient of Substrate on C4


Fatigue Life. (Modification of Figure from Ref. 16.)
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-177

-
i.-_ _ _ _ _ _ _-.,._ _ _ _ _ _
---------~
ICCHIP

~=~

V. 1
ROUTING
SIGNAL
LEVEL 1
ROUTING
SIGNAL2
LEVEL

~~~~OIHG
POWER
( SOLDER BUMP
CONNECTION
\ \ I PAO ... VIA

/ ~OUNO D~

,
POWER PLANE CONNECTION DIELECTRIC
CAPACITOR
DOPED SIUCON WAFER
GROUND PLANE

Figure 8-43. Silicon Flip-Chip Joined to Silicon Carrier with Thin-Film inter-
connections. (From Ref. 100.)

Pure Si substrates with multilayer Al and poly imide or Si02 insulators


have been fabricated and demonstrated. Figure 8-43 illustrates a typical
Si carrier system [100]. An application of such a system is shown in
Figure 8-44 which is essentially the core of an IBM RiSe 6000 workstation
on a single piece of silicon. There are two weaknesses in such a system:
(1) The lack of vias through the silicon means that all wiring must lead
to the perimeter for wirebonding or TAB bonding, and (2) thin-film Al

Figure 8-44. Workstation (IBM RS 6000) on Silicon Carrier. (Courtesy of IBM).


11-178 CHlp·TO·PACKAGE INTERCONNECTIONS

wires are more resistive and longer than multilevel Cu wiring on ball-
grid-array-type substrates. Therefore, this approach typically falls short
in optimum performance.
Glass-ceramic substrates in which crystallizable glass, cordeirite-like
ceramic has been used as the carrier also matches silicon CTE near perfectly.
The multilayer structure with punched and filled vias and Cu wiring allows
short paths in good conductors to reach a pin-grid-array configuration on
the bottom [21]. This approach works well but is probably too costly for
many applications. Also, the mismatch in the TCE problem is now trans-
ferred to the package-to-board interface and would have to be solved to
avoid premature failure of those joints (e.g. soldered BGA joints).
An early effort using a polyimide-Kevlar® organic substrate [16]
with near-matching TCE has been followed by a number of packaging
efforts that take advantage of the improvement in thermal fatigue lifetime
of the C4 connections. The use of AIN and SiC as first-level packages
(as discussed in Chapter 18, "Coated-Metal Packaging") were actively
pursued for direct chip-attach applications. Gallium arsenide has also been
matched using sapphire [26] or A1 20 3, which matches TCE's well.
Power cycling is complementing thermal cycling to evaluate these
material combinations [91,101,102] because it is more realistic in simulat-
ing temperature differences between the chip and substrate. Process con-
straints, wireability, dielectric constant, and heat dissipation must, of
course, be among other factors considered in selecting an alternative
substrate material.

8.3.7.3.1 Direct Chip Attach (DCA) to PWB with Underfill


Perhaps one of the most innovative developments recently has been
the discovery that certain polymers such as epoxies heavily filled with
particles of Si02, to match the solder TCE and bond chip to substrate,
will increase the fatigue life of the solder by a factor of 10-100 [31,103].
Nakano (Fig. 8-45) showed this in 1987 [103], and the work was applied
by Tsukada [34] in attaching silicon chips directly to thin films on FR4
in a product called surface laminar circuitry (SLC). For the first time, Si
chips with low expansivity could be made reliable in direct flip-chip
attachment to high-expansivity FR4 PWB. Some of the first products to
use such technology commercially are IBM PCMCIA modules for portable
computers (Fig. 8-46) and Motorola pen-sized pagers (Fig. 8-47).
The technology appears to work well because the encapsulant shares
and reduces the all-important solder strain level to 0.10-0.25 of the strain
in joints which are not encapsulated. This has been shown by finite element
analysis calculation (Fig. 8-48) [34] and by Micro-Moire experimental
analysis (Fig. 8-49) [104], The latter shows the strain to be 10% relative
to the solder of the unencapsulated solder and indistinguishable in solder
or filler. Both ceramic and polymeric flip-chip substrates benefit from this
underfill enhancement.
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-179

LSI
~ump (95Pb·5Sn) Ji
Epoxy Resin

~ 0 I:::::~~~~~~::::J
AI 20 3 Substrate
- Size of LSI; 10 x 10 mm
- No. of 1/0; 224 II0s

Conditions -55·C - 150·C

Resin II

_"(~'~<:j"~J
(a,'" "-'7
o 2000 5000
No. of Cycle (1 Cycle = 1 Hour)

Figure 8-45. C4 Life Extension by the Use of Thermal Expansion Matched


(to Solder) Resins. Resin II is assumed to include filler to match solder CTE of 27
ppm. (From Ref. 103.)

Perhaps the only negatives with the underfill process are the difficulty
with reworkability and the slowness of fill and cure. All of these things
are being investigated and developed. Cleavable epoxides have been dem-
onstrated. Thennal degradable encapsulants show promise. Until these
materials become qualified, the option remains practical to apply the
encapsulant only at the end of processing, after test, when it is most
probable that the chip is good.

8.3.7.4 Other Reliability Issues


Among other reliability concerns that have been reported in the
literature, thennomigration has been shown to be a major concern for
applications where high-temperature gradients are coupled with high-
diffusivity, low-melting-point solders [99]. Corrosion has been encoun-
tered in high-humidity testing of PbIn [98]. Solder void defects are ad-
dressed in Ref. 85. Palladium depletion of AgPd thick films by PbSn can
allow Ag corrosion and prompted a change to AgPdAu ternary alloy for
the substrate electrode [105,106].

8.3.7.5 Alpha-Particle Emission


A more fundamental problem receiving increased attention is that
of soft errors in devices caused by alpha-particle emission of trace quanti-
11-180 CHIP-TO-PACKAGE INTERCONNECTIONS

-
Encapsulation

...
--
--
Cu 18p thickness
ThlnFlim /
PTH O_Smm
Epoxy · Glass Core

Figure 8-46. (a) SLC Cross Section Showing Thin-Film Surface Layer Wiring
on FR4 Base and Underfilled Flip Chip. (b) SLC Token Ring Adapter Card
Compared to Regular Surface Mount Technology. (From Ref. 34_)
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-181

I,:, I(M)~ I ,( -
i~."--. ~ ~_ ~ \ .

Figure 8·47. Motorola Pen·Sized Pager

ties of radioactive materials in the packaged assembly [36,107,108]. In


the case of C4, high-lead alloys are mostly used for computer applications,
and lead almost invariably brings with it trace amounts of uranium and
thorium. This problem will demand more attention as VLSI devices be·
come more dense and as the critical charge levels in the device become
smaller. U and Th contaminants have been successfully removed from

~ 4
~
C

-
.~
CJ)
III
3

:
::
;
~
:
;
~
~
:
~
~
:

5: 2
: : : : ; :
~I ~Withb~~Eri~~p~~iat on
C
g1

1 2 3 4

Bump Location No.

Figure 8·48. Finite Element Analysis of Strain on Chip C4's for SLC. (From
Ref. 34.)
11-182 CHIP-TO-PACKAGE INTERCONNECTIONS

No Encapsulation Encapsulation
C4 Bump Chip

~~/.~ICB
o 200
I I

p= 4 52 nmlfringe

Thermal loading of .M =-60°C


Figure 8·49. Micro-Moire Interferometry of Underfilled Chip's Showing
Stain Reduction and Distribution. (From Ref. 104.)

Pb. There is even less of a problem with Sn. However, there are isotopes
of Pb, 214Pb and 2IOPb which are a part of the 238U radioactive decay chain
which themselves decay to Bi and Po on the way to 206Pb, the ultimate,
stable form of Pb. It has been discovered that some naturally occurring
deposits of Pb are very low in the alpha-producing isotopes and therefore
have several orders of magnitude lower alpha radiation [109]. Isotope
separation of other deposits or secondary Pb were thought to be prohibi-
tively expensive, but are still under active consideration [110].

8.3.7.6 Burn-In/Known Good Die for Flip Chip


C4 chips can be given a bum-in cycle on a temporary substrate to
improve chip defect-related reliability [111]. To facilitate the removal of
the chip from the temporary substrate after the bum-in, IBM has used a
reduced radius removal (R3) process. As shown in Figure 8-50, the sub-
strate pad is reduced in size to approximately one-fifth of the area of a
functional terminal. After joining and bum-in, the chip is mechanically
torqued off the temporary substrate, with the fracture being in the solder
adjacent to the small-sized terminal. Because most of the solder remains
with the chip, there is no need to dress or rework the chip except for a
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-183

C4

rl )

C4

Figure 8-50_ R3 Burn-In Methodology for C4 Flip Chips Featuring Solder


Attach to Greatly Reduced Substrate Pads for Burn-In. (From Ref. Ill.)

reballing in a dry H2 reflow furnace. The process has been verified on


10,000 CMOS memory chips in the field, with early life-defect reliability
being improved by more than a factor of 30. It should be understood that
the defects referred to are not C4 defects, but internal defects in the IC
chip induced to fail during the bum-in cycle instead of being allowed to
fail in the field application. The temporary substrate is reusable-a sample
substrate was used repeatedly for over 20 cycles.

8.3.7.7 Heat Dissipation with C4


Heat dissipation is covered in great detail in Chapter 4, "Heat Transfer
in Electronic Packages," and only a brief discussion is included here. Due
to the very short length and large area of contact of a C4 joint, the thermal
path through the interconnection itself is very low and far superior than
for TAB or wirebond. Typical dimensions are a 125-llm diameter and a
60-llm "length" for C4 as compared to a 25-llm diameter and a 2500-
11-184 CHIP-TO-PACKAGE INTERCONNECTIONS

/lm length for wirebond, and a 50-/lm square and a 1750-/lm length for
TAB. The chip-to-substrate resistance is so low for C4's that for many
years it was not necessary to supplement it at all. Early chips with 10-20
peripheral C4 bumps bonded to an alumina substrate could dissipate
approximately 0.5 W [112]. An air-cooled module with six 4.5-mm chips,
each having an 11 x 11 array of solder joints, can dissipate approximately
1.5 W per chip [8]. Somewhat higher power levels have been achieved
with new, high-thermal-conductivity ceramics such as AIN [113] and SiC
[91]. Numerical analysis techniques are required for reasonable perfor-
mance projections because the thermal path for flip chips depends on
device location, size, metallization, number of terminals, and the thermal
resistance of the substrate [112,114].
Today's high-power-level devices, however, require a supplemental
heat path through the back of the chip. Die bonding accomplishes this
for wire or TAB-bonded chips. In the C4 case, because the back of the
chip is free of mechanically or electrically delicate surface features, it is
amenable to direct contact by a wide variety of heat sinks, thermal greases,
or solders, whose thermal conductivity is often better than the plastic or
ceramic package to which they are backbonded. An example is the IBM
multichip module [114,115] where spring-loaded pistons transfer heat
from the back of each chip to a water-cooled plate, augmenting the
traditional solder-joint thermal path. Four watts per chip or 300-400 W
for a 100-chip module can be dissipated. This has been enhanced in the
S/390lES9000 series computers to a 16.7-W/cm2 capability [115].
Full bonding to the back of the chip entails more risk for C4 mounted
chips because the solder joints, being very short, are not as compliant as
wires and the designer must be concerned with possible solder-joint fa-
tigue. Bonding to the back of a C4 mounted chip may impose additional
forces on the C4's and lower the fatigue life (see Chapter 5). Designs
have been proposed which combine a solder backside die bond with C4's
to give extremely low thermal resistances-on the order of O.4°CIW for
chips joined directly to a water-cooled plate in a MCM [41,42,44]. This
would give a power capability of over 100 W per chip. Such power levels
have been projected for the year 2000 [41]. Hitachi has implemented a
slightly less efficient single-chip module version of this design (Fig. 8-51)
in their M880 Processor Group computers [116]. Chip to water thermal
resistance is 2°CIW, not quite as good as the multichip design because
more interfaces and layers are between the chip and the flowing water.
Various conduction cooling designs for high-power C4 mounted chips
have been compared (see review by Darveau [117]).
Finally, liquid immersion cooling [117] and cryogenic applications
[28,29] have also been demonstrated for C4 interconnected structures.
The latter application, for Josephson devices that operate at 4.2 K, was
especially noteworthy, because all of the materials were subjected to a
8.3 FLIP-CHIP SOLDER-BUMP CONNECTIONS 11-185

SoIdefBamp
(minimum
250Jun pitch)

Thin FHm Layers


(5 wiAlling layers.
1 realator layer)

Figure 8-51. Micro Chip Carrier in Hitachi M880 Mainframe; Enhanced


Heat Dissipation Through Back-Side Soldering and High-Density C4 Area
Arrays on Chip and Package. (From Ref. 116.)

very large tlT between room temperature and operating temperature and
because the materials' intrinsic properties, such as resistance to cracking,
are quite degraded at low temperatures. Orthogonal solder connections
are also used to join silicon slices at right angles to each other. For such
a joint, matched expansion materials are required [28-30). Compatibility
of standard 95/5 Pb/Sn C4's in liquid-nitrogen environments for CMOS
applications has been demonstrated on alumina ceramic packages [118].
Parylene coatings have been shown to be effective in lowering the thermal
shock during transfer into and out of the liquid nitrogen [119].

8.3.8 C4-Future Trends


As VLSI proceeds to denser chips, C4 densification will follow.
Area-array structures will dominate over traditional peripheral devices.
Bumps of 25 !lm, resulting in over 10,000 pads per chip in experimental
devices, have already been fabricated [27], ahead of the predictions of
the SIA roadmap. Photolithographic processing will be preferred for such
devices, and with that, area arrays will become more and more common.
Significant activity in matched-expansion substrates, or alternatives that
alleviate the fatigue limitations such as under chip fillers are being actively
applied. High-thermal-conductivity packaging materials and innovative
configurations to improve heat dissipation of the devices are also being
used and extended. More attention will be given to defects, for both yield
and reliability needs. Efforts to purify all packaging materials for low alpha
emission or substitute new ones which are inherently better will continue.
The range of applications for C4-like structures is becoming much
wider, given the more diverse materials used in new electronic devices
and substrates (for example, In C4's on GaAs chips on sapphire substrates
for optoelectronics) [26], chips C4 bonded to flexible circuits [120] and
printed-circuit boards [121]. The latter developments suggest a future,
strong presence in the commodity world as well as high-end computers.
11-186 CHIP-TO-PACKAGE INTERCONNECTIONS

8.4 WIRE BONDING

Wire bonding has totally redeemed itself in reliability, manufactura-


bility, and cost from where it was in the earliest era of semiconductor
devices. It is still the dominant chip-connection technology in the semicon-
ductor industry. Virtually all dynamic random access memory (DRAM)
chips and most commodity chips are still in plastic packages with wire-
bonds. Therefore, because the vast majority of all integrated-circuit (IC)
packages are assembled by wirebonding [122], it is important to understand
this technology. This section explains the basic wirebonding process,
emphasizing the metallurgy, materials, and typical applications. Guide-
lines are provided for selecting the appropriate methods for various pack-
age types and optimizing these methods.
Wirebonding differs from C4 and TAB techniques in that electrical
package connections are created at this assembly stage by attaching a fine
wire between each device 110 and its associated package pin (one at a
time), as opposed to attaching prefabricated patterns of interconnecting
materials. This may not seem advantageous for wirebonding, but it is
actually the basis for one of its biggest strengths: flexibility.
Wiring changes are easily accomplished without the need for expen-
sive tooling and material changes. A new "bond program" can be taught!
saved in a matter of minutes, followed by immediate production of the
new devices, saving time and money.
The speed and cost-effectiveness of changes also applies to the initial
package design. Chip preparation cost is significantly lower than C4 and
TAB devices because the wirebond welds are made directly to the chip
(wafer) metallization, saving substantial tooling and production costs in-
volved in the bumping process. Further tooling savings can be realized
through the use of a common package or substrate which can be used for
a number of devices using various wire configurationsnengths. Therefore,
flexibility and cost are two of the primary reasons for selecting wire-
bonding.
A third reason for the prominence of the wirebonding process is
much improved reliability. In the early 1990s, 1.2-1.4 trillion wire inter-
connections were produced annually (roughly 2.6 trillion welds). Manufac-
turing losses and test failures are extremely low (40-1000 ppm) and
trending downward each year. High yields and reliability rates continue
to be achieved around the world in a large variety of facilities, from
completely automated, mass production factories making millions of pack-
ages per week, to single-room facilities producing low-volume, custom
parts.
Today, a global network of manufacturing people, vendors/suppliers
of equipment, and materials comprise a large wirebonding knowledge
base. The infrastructure is vast and tends to keep its user base unless
there are extremely strong reasons to deviate to a different technology.
8.4 WIREBONDING 11-187

Wirebonding is popular as a result of continuous process improve-


ment achieved through the development of sophisticated, automated equip-
ment. Today's production lines can assemble the latest generation of
packages with the evolutionary derivative of the welding technique used
to manufacture the first transistor. It is the fundamental Au to Al intercon-
nection technology that has continued to grow with the semiconductor
industry. Wirebonding appears to be the method for connecting the major-
ity of packages in the near future, with C4 and TAB providing the tech-
niques needed for specialty packages and high-end, high-pin-count de-
vices.
Today, hundreds of diverse plastic and ceramic packages utilize
wirebonding. Despite this diversity, commonalities do exist. First, the
chip is normally mounted with the back side attached to a metal frame
or substrate, although there is at least one noteworthy exception where
the metal framelleads are attached to the top side of the chip, called lead-
on-chip (LOC) [123]. This mounting process, called die bonding or die
attach, is achieved by a variety of conductive epoxies or solders. The
chip and substrate assembly are then presented to the wirebonder. The
wires are attached to the chip and package substrate, one at a time, using
either a thermosonic or ultrasonic welding process. (A third method,
thermocompression welding, may also be used, but it is not commonly
used.) State-of-the-art bonding technology produces two welds and a
precisely shaped/routed wire loop in 100-125 ms, depending on the pro-
cess used.
In factories throughout the world, thousands of automatic machines
produce wirebonded packages with little, or no human intervention. In
each machine, die attached parts are delicately transported and presented
to the bonding area. Key features of the chip and package carrier are located
by a video camera and recognition system using artificial intelligence
to identify individual targets for the intended wires. As the welds and
interconnecting loops are made, the machine uses closed-loop servo sys-
tems to ensure that each step of the process is being performed as intended.
After all the connections are made, devices are then transported to an
output station where they are staged for the next operation (usually plastic
encapsulation or application of a lid to the hermetic devices).

8.4.1 Joining Technology

Most wirebonding processes combine either thermo sonic (TIS) or


ultrasonic (U/S) welding methods, with two different bonding methods
for applying the wire to the chip and to the package surfaces. The two
basic wirebonding techniques are ball bonding and wedge bonding. Ap-
proximately 93% of all semiconductor packages are manufactured using
ball bonding method, while wedge bonding is used to produce about 5%
of all assembled packages. Both ball bonding and wedge bonding were
11-188 CHIP-TO-PACKAGE INTERCONNECTIONS

once pure thermocompression (TIC) bonding, but because of the high


bonding temperatures required (approximately 350°C) have changed to
lower-temperature (150-200°C) TIS Au ball bonding and room-tempera-
ture Al wedge bonding.
Ball bonding. In this technique (Fig. 8-52), wire is fed vertically
through a tool called a capillary. The wire is heated to a liquid state with
an electronic spark discharge called an electronic flame-off (EFO). The
surface tension of the molten metal forms a spherical shape, or ball, as
the wire material solidifies, hence the process name "ball bonding."
The capillary descends and positions the ball on a bond pad of the
chip using the bonder's video and servo systems. Thermosonic welding
is then performed. Once a weld is completed, the capillary rises and
follows a prescribed trajectory calculated for each wire. A precisely shaped
wire connection called a wire loop is created as the capillary descends
to a target position for the second bond. A second weld, a crescent bond,
is now formed, with the wire extending through the center of the capillary,
under its face, and out to the newly formed loop. The finished weld has
a crescent or fishtail shape made by the imprint of the capillary's outer
geometry. The capillary ascends once again, breaking the wire at its
thinnest point near the bond and, simultaneously, feeding an exact wire
length needed for the EFO to form a new ball to begin bonding the
next wire. Figure 8-53 shows typical ball and crescent bonds, as well as
loop shapes.
Wedge bonding. Named for the shape of the bonding tool used in
the process, this technique feeds wire through a hole in the back of a
bonding wedge at an angle usually 30°-60° from the horizontal chip
surface (Fig. 8-54). As the wedge descends onto the Ie bond pad, it pins
the wire against the surface and performs either an VIS or TIS weld. Next,
the wedge rises and executes a series of calculated motions, creating a
desired loop shape.
In addition to vertical and horizontal motions needed for ball bonding,
a rotational axis (called 8) is required to align the package so that the
wire about to be bonded is on the same axis as the bonding wedge feed
hole. This allows the wire to feed freely through the hole in the wedge
during the loop formation. At the second bond location, the wedge de-
scends to make the second weld. A mechanical clamp is closed behind
the wedge and the clamp is moved to the rear to tear the wire. As the
wedge ascends, the clamped wire is fed under it to begin bonding the
next wire. Figure 8-55 shows examples of first and second wedge bonds.

8.4.2 Wirebonding Joining Mechanism

Like other technologies that join together two materials, some of


the exact metallurgical mechanisms of joining in wirebonding are still
8.4 WIRE BONDING 11-189

~
6

•
1

~
7
a

•
2

~
8

•
3 9

a ~
49 10

~
51
l!
11

Figure 8-52. Ball-Bonding Steps to Complete One Cycle. 1. Ball is formed below
capillary (note that it should not contact the capillary during the formation process). 2.
Capillary descends, capturing/centering the ball so that it is seated in the capillary's inner
chamfer. 3. High velocity is reduced to contact velocity at a programmed distance above
the chip. The point where this velocity transition occurs is called the tool inflection point
(TIP). 4. Bond force and ultrasonic energy is applied to form the ball bond. 5. The capillary
ascends vertically to payout sufficient wire to form the loop of wire between bonds. The
wire clamps are typically closed at the highest point, prior to movements toward second
bond. 6. High-velocity motions form the wire shape as the tool moves over the second
bond site. 7. Vertical motions are slowed to contact velocity at the TIP above the lead.
8. The crescent bond is formed through the application of force and ultrasonic energy.
9. The capillary ascends a prescribed distance to the pay-out wire (called a tail). The
material contained in this wire length will be used to form the correct ball size. The wire
clamps are closed again when this motion is completed. 10. The capillary moves vertically
to break the bottom of the tail (tack welded to the lead). 11. The capillary and the EFO
wand/electrode are brought in close proximity while a spark discharge provides heat to
form a molten ball.
11-190 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8-53. Ball and Crescent Bond Examples. Left: A typical ball bond without
fine pitch restrictions (25-~lm wire with 80-~m bond squash). Note the small amount of
material around the wire that was extruded into the capillary bore. Examination of the
top surface reveals the shape of the tool's inner chamfer and face. Surface roughness is
typical of "secondary bonding" that causes the ball to adhere to the capillary tip while
the primary bond is being formed between the ball and chip. Right: A typical crescent
bond with the familiar fishtail shape. Small irregularities at the bottom edge of the bond
are remnants of the tack weld that temporarily attached the wire tail before it was pulled
away. (Courtesy of K&S.)

subject to discussion. Methods for attaching metals include a range of


processes from adhesive metal to metal bonds to metallurgical mechanisms
(diffusion bonds with intermetallic compound formation) and liquid metal
brazing. Often there is a combination of mechanisms which results in
"bonding." The following section will summarize the most accepted views
about U/S and T/S wirebonding which have emerged.

8.4.2.1 Ultrasonic Welding


This method of welding is most commonly used to join aluminum
wire to a variety of package materials, including the aluminum metalliza-
tion found on the chip, as well as thick- and thin-film package metals
consisting of gold, silver, nickel, or palladium. The bond is made at
room temperature without an applied heat source. It is accomplished
by combining two primary elements: downward force and horizontal
oscillatory motions.
The frequency of these oscillations range from 60 to 120 kHz,
hence the name "ultrasonic." The downward clamping force is continually
applied to the bonding wedge to assure that the ultrasonic motion of the
tool remains in contact with the wire as it plastically deforms/yields to
the applied energy. A weld results from this intimate contact between the
8.4 WIREBONDING 11-191

1 2 3
Oe.cent to 1st Bond 18t Bond Impact, Rise to Loop Height
U.S. Energy Applied

WEDG~~ CLAMP
-
UL TAA.ONICS

18TBO~D ~
~
\\\\\\\\\\\\\\\\,
x-v TABLE
~
4 5 6
Descent to 2nd Bond 2nd Bond Impact, Clamp Plvote Away from
(Programmable' Clamp Clo •• U.S. Energy Applied Tool to Tear at.2nd Bond

-
UL TRASONICS TEA A

7
Clamp Pivots Forward to
Feed for Next 1.t Bond

Figure 8-54. Wedge-Bonding Steps to Complete One Cycle. 1. The wedge de-
scends at high velocity (not shown) and makes transition to contact velocity at TIP. The
wire clamps are closed to assure the wire does not move under the tool. 2. The first bond
is formed by application of bond force and ultrasonic energy. 3. The clamps are opened
and the wedge ascends vertically, then it moves horizontally to a position over the second
bond site. 4. The clamps are closed again and high-speed motions move the wedge
downward until it reaches the TIP distance above the package. 5. The second bond is
formed. 6. Articulated clamps move away from the wedge to break the wire at the back
radius of the tool. 7. The wedge ascends from second bond (not shown) while the clamps
push the wire through the wedge feed hole to provide the wire for the next bond cycle.

wire and the surface in less than 25 ms. The ultrasonic energy aids in the
wire deformation and the breakup of the hard aluminum oxide on the
surface of the bond pad.
Although other wires can be used, aluminum appears best suited for
the VIS technique due to the thin, naturally occurring aluminum oxide
layer on its surface. The ultrasonic movement of this abrasive material
11-192 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8·55. Aluminum Wedge Bonds. Left: Typical first bond with tail; right:
Typical second bond.

removes surface contaminants/oxides on the mating surfaces to expose


pure metal for the joining process.
In 1978, Winchell and Berg published their findings on ultrasonic
bond development [124] that is still a definitive paper for ultrasonic
wirebonding. Briefly summarized, they conclude that an ultrasonic bond
occurs from the material flow associated with several static and dynamic
forces applied by the wedge-bonding machine.
First, the aluminum wire is softened and plastic deformation occurs
when ultrasonic energy is applied horizontally by the bonding wedge. It
is interesting to note that ultrasonic energy needed for the elongation of
aluminum is about 100 times less than the heat energy required for the
same result [125]. In fact, no external heat is required, nor is a large
amount generated internally by the movement of plane dislocations.
Experimental studies show that the temperature at the bond interface
only reaches 70-80°C, which is too low for the bond to support a thermal-
diffusion process. The softened wire is then deformed by the downward
force of the bonding wedge. Although some surface oxides may be re-
moved by this initial plastic flow of the wire over the substrate surface,
intimate contact (and the resultant bonding) do not occur. The amount of
deformation, or bond squash, is a result of this first stage.
As the bonder's vertical force and stress is distributed over the
expanding wire surface, it decreases with time and distance from the
center of the wire (Fig. 8-56). The vertical stress in the center is higher
than the horizontal stress applied by the ultrasonic energy. Therefore, no
wave formation is seen in this region.
Near the periphery of the wire, the horizontal stress applied by the
wedge is greater than the downward stress, so wave motions are observed
in this area (Fig. 8-57). This explanation coincides with the actual ultra-
sonic bond region observed as an ellipse with a central, unbonded area
(Fig. 8-58) and refutes the theory of bonding by friction or a sliding
8.4 WIRE BONDING 11-193

VERnCAL STRESS AT BOND CENTER VERnCAL STRESS DISTRIBUnON


(MACHINE LOAD CONSTANT) FOR 2 DIFFERENT TIMES

t, t, > t,

)y
t,

t, t, TIME DISTANCE FROM BOND CENTER


t- x-
a b

Figure 8-56. Vertical Stresses During Wedge Bond Formation. (a) For constant
machine load; vertical stress decreases with time because of increased contact area; (b)
magnitude of vertical stress also decreases with distance from bond center. (From Ref. 124.)

model. (Both of these cases predict preferential bonding at the center of


the wire.)
The peripheral material that is subjected to the wave motion generates
a strong cleansing action and pressure at the substrate. If aluminum wire
is bonded directly onto a hard material like silicon, the wave motion is
great enough to permanently deform the surface into ridges and troughs
that are perpendicular to the direction of the applied energy of the wedge.
The depth and distance between these features is directly related to the
amount of power applied to the wedge. Along the sides of these valleys,
two atomically clean surfaces are brought into intimate contact. Both the
wire and substrate have surface atoms that are free to interact. Therefore,
high-strength atomic bonding takes place to satisfy the atoms of both
surfaces. The result is a continuous, metal lattice structure in these areas,
which is equivalent to a single homogeneous material.

8.4.2.2 Thermosonic Welding

This method of welding is most often used with gold or copper wire.
Although the technique uses the same ultrasonic energy application, there
are two notable differences between this process and the VIS process-
the need for an external heat source, and wire material without an abrasive
oxide surface layer. Because gold and copper will not form an acceptable
weld at room temperature in a reasonable amount of time, heat is applied
to provide the activation energy levels to the materials so there is an
effective joining and intermetallic diffusion. Typical bonding temperatures
range from 150 to 200o e, with bonding times from 5 to 20 ms.
11-194 CHIP-TO-PACKAGE INTERCONNECTIONS

AT I. AT I.
ITV1: VERnCAL STRESS AT TIME I. IT V2: VERnCAL STRESS AT TIME t.
IT E: EFFEcnVE HORIZONTAL STRESS

t
IT

XII2 XC2 Xu
X- X-

INCREASED POWER

1
IT

a b

Figure 8·57. VerticaVHorizontal Stresses Causing Peripheral Bonding. Hori-


zontal stress, SE which produces wave fonn defonnation is imposed upon Sy. Bonding
occurs from XB to Xc with obvious grooving between Xc and XA• With time, bond zone
moves toward center, and previously bonded material is destroyed. Note that the SE curve
is drawn to conceptually illustrate points of model. (a) Greater machine load results in
larger undefonned central area; (b) increasing power widens wave affected region. (From
Ref. 124.)

Downward programmed forces for ball bonds are well under 100 g.
This low force means that the capillary tool's vertical motion will "stall"
and the bond squash will reach an equilibrium state before it is totally
flattened. The partially deformed ball is now ready for the welding process,
because it is physically wetted to the aluminum bond pad.
Several variables control the actual diffusion of two metals into each
other. The diffusion rate is affected by the amount of energy applied
through heat, ultrasonic energy, and time. It is also affected by the purity
(concentration) of the two materials being welded. Therefore, the bond
strength/depth is the sum of the combined energies applied for a given
time. This means that a large number of parameter combinations can
create an equally strong bond. Ultrasonic energy is most critical because
8.4 WIREBONDING 11-195

MACHINE LOAD
I

1
SUBSTRATE

a b

Figure 8-58. Peripheral Wedge Bonding Pattern. (a) Bond configuration for applied
two-force model; ultrasonic energy applied to the wedge produces horizontal (vibrational)
stress along the wire while the machine load acts vertically. (b) Bond pattern on substrate
is characterized by rough peripheral region at distance X from the bond center. (From
Ref. 124.)

it reduces the gold's yield strength [126] by increasing the mobility and
density of lattice dislocations. This plastic flow begins when ultrasonic
energy is applied to sweep away the brittle oxide layer on the aluminum
bond pad to expose a clean metal surface (Fig. 8-59).
With incomplete lattice structures at the newly exposed surface of
each metal, a migration of atoms begins from one material to the other.
These diffused atoms form bonds with their neighbors' shared outer shells.
As temperature and ultrasonic energy are applied over time, additional
interdiffusion results.

8.4.2.3 Thermocompression Welding


Although this technique was one of the original methods for the
manufacture of semiconductor packages, it is only used for a few specialty

Slip lines

##d
,
: .......... : ... ...... : ..
'. '.' . ' : .. Surface ..,'
: . . . ." • '. . 'contaminants
............. .. .... .
"..

: : ..... '.': . : Cle~ sUrface


: ·.Aluminum bOnding pad ; : "Aiumi~um' bonding
':... . .. . ... . . ...
p~d:.. " :..
;.:....~<.:: .::: :::.:.:.= ..
"...................... "" ............."

Before bonding After bonding

Figure 8-59. Gold Ball Slip Planes Form Clean Surfaces. Ultrasonic energy
applied to the gold material causes slip-plane mobility which exposes clean Au and Al
material that are in intimate contact with each other. (From Ref. 127.)
11-196 CHIP-TO-PACKAGE INTERCONNECTIONS

applications today. It utilizes only two energy sources: heat and a high
downward force (compression) to form the weld. Today the process gener-
ally requires temperatures ranging from 280° to 380°C and is much slower
than TIS bonding. The time required to make a TIC bond can approach
1 s. In most cases, the extremes of heat and time for this process make
it less attractive to manufacturing engineers.

8.4.3 Wirebonding History

Wirebonding was the earliest technique for device assembly. Bell


Laboratories published their first wirebonding results in 1957 on a tech-
nique called "thermocompression bonding" and began contacting equip-
ment manufacturers to provide production machines. After developing
several types of laboratory apparatus, Kulicke and Soffa Industries pro-
duced the first commercial bonding machines in 1963 for the infant semi-
conductor industry. With machine availability, the wirebonding technique
became generally applied. By 1970, UIS bonding and TIS bonding tech-
niques were introduced which reduced bonding temperatures.
The early machines were extremely labor-intensive, with all the
positioning and motions applied manually. The machine operator con-
trolled the entire process by moving levers that scaled down gross hand
motions through micromanipulators. This made bond placement on the
small bond pads a learned skill that varied greatly among operators. Loop
shapes were also completely dependent on the dexterity of the operator.
The TIC bonding process became the preferred method for assem-
bling commercial-grade parts. Aluminum wedge bonding provided a
"monometal" connection on the bond pads of military-grade devices. As
cam-driven, semiautomatic equipment was introduced to the marketplace,
the TIC gold-ball bonder became the favored machine for two reasons:
(1) the process was faster, as it did not require device rotation like the
wedge bonder and (2) the gold wire could be encapsulated in nonhermetic,
plastic encapsulants and not be susceptible to moisture corrosion like its
aluminum counterpart.
However, the early bonding process was not without problems. Prior
to the invention of the electronic flame-off, ball formation was accom-
plished by heat application with a hydrogen torch. This method did not
provide sufficient ball size consistency and often caused damage to the
wire just above the ball. In fact, a partial melting of the area above the
ball was considered normal. It was called the "necked down" or simply
"neck" area (a term used by many assembly engineers today, even though
the "neck" phenomenon has long since disappeared from the normal
process). Using today's highly controlled EFOs (electronic flame-off) it
is possible to form consistent balls without visibly distorting this region.
The most serious problem associated with the early bonding process
was metallurgical weld failure that led to device reliability problems. In
8.4 WIREBONDING 11-197

the early 1960s, it was common practice to wirebond at 350°C because


the high temperature provided the activation energy for interdiffusion
gold wire and aluminum bond pads. When this occurs, five intermetallic
alloys form between the pure gold and aluminum areas [126]. Some of
them are quite brittle and prone to failure during temperature cycling.
One of these brittle compounds, Au-Ah is purple in color. Its distinctive
presence at failed bond locations soon led to the term "purple plague."
If the formation of the purple Au-AI2 was not bad enough, a second
failure mechanism was found to be associated with these compounds. At
the high temperatures used for epoxy curing, ceramic lid sealing, and
stabilization baking, the Au continues to rapidly diffuse into the Au-rich
compounds like AUsAI2' leaving voids at the boundary between these two
regions, known as Kirkendall voids [126]. The same condition occurs
where the Al rapidly diffuses into the AI-rich AuAI2, leaving Kirkendall
voids at their shared boundary. In extreme cases, the voids grew/coalesced
into planes of weakness which could easily fracture, resulting in open
circuits or lifted ball bonds.
These reliability problems were so serious that minor process changes
could not be considered. To solve these problems, engineers considered
alternative connection methods such as flip-chip solder bumps, beam
leads, and TAB, and a new wirebonding process. In 1970, V/S energy
was introduced, which reduced the heat input to 200°C. This enabled
interdiffusion without a high rate of brittle phase formations. Since the
advent of T/S bonding equipment, the earlier problems are easily avoided.
In fact, most packaging engineers have never actually seen purple interme-
tallic or Kirkendall voiding outside of a historical reference. It is perhaps
ironic that the wirebonding process began with such basic problems and
has evolved to such excellent yield and reliability levels.
Today, most reliability problems are eliminated with properly con-
trolled and much improved tools and processes. Several generations of
servo controllers have long since replaced hand-manipulated mechanisms
to apply the bonding force. Free-running oscillators for the ultrasonic
energy changed to phase-locked-loop circuitry and, later, microprocessor-
controlled frequency, voltage, current, and time.
As mentioned previously, EFO circuitry now produces a completely
repeatable ball formation process. This precise control of the bonding
parameters has also been matched by the purity of materials for wirebond-
ing. Wire and plating alloys are now specified accurately. Wire mechanical
properties are now highly reproducible between lots through automated
manufacturing and tight process control of the wire chemistry and heat
treatment/annealing. The same servo control technology utilized in the
wire bonding equipment is also used in the final grinding/polishing opera-
tions for the manufacturing of bonding tools (capillaries and wedges),
which were manually produced in the past.
Today's packaging engineers can select application-specific tools
11-198 CHIP-TO-PACKAGE INTERCONNECTIONS

and wire to meet the requirements unique to their package design_ These
products can be repeatably manufactured as a result of three decades of
controlling and understanding each manufacturing process.

8.4.4 Wire-Bond Applications


It is beyond the scope of this section to list all current wire bonded
packages. An acronym dictionary, with annual updates, is required to
keep abreast of the package explosion.
Since the mid-eighties, the "end of wirebonded packages" has been
predicted to be imminent. Each new package type has presented challenges
that initially appeared too difficult for the current wirebonding capabilities.
Yet equipment advances meet these challenges and provide the required
capabilities. The semiconductor industry is strongly motivated to continue
investing time and energy into an ever-evolving wirebonding technology
because of its much improved reliability and historically successful
track record.
One such application challenge was reducing package thickness.
Portable, hand-held electronics require packages less than I mm thick.
This translates to loop heights less than 150 /lm above the chip (Fig.
8-60). Low, consistent loops (Fig. 8-61) with less than a 40-/lm, 6-sigma

Figure 8-60. Low-Loop Process. Example of very low loops without evidence of
stress damage to the wire above the ball bonds. Loop heights shown are 96 Ilm (top of
wire to the chip surface). (Courtesy of K&S .)
8.4 WIRE BONDING 11-199

Figure 8-61. Low-Loop Process Consistency. These low loops are 114 !lm above
the chip surface and illustrate 5-mm lengths. (Courtesy of K&S.)

range, are achieved using high-resolution, coordinated servo systems that


control the capillary's three-dimensional trajectory from first to second
bond. Because an error of a few microns in the actual tool path could
either buckle or break wires (Fig. 8-62), real-time microprocessor control
for each servo loop is critical.
In addition to reducing thickness, the market demanded smaller
package lengths and widths, made possible by advances in wafer process-
ing, through continuous shrinking of device features and lead-frame etch-
ing improvements. The term "fine pitch" has become an ever-decreasing,
moving target. Even with machine accuracy/repeatability made possible
by new vision systems and advanced servo systems that deliver the wire
to its intended position, two other factors needed improvement: process
control and bonding tool design. Each step of the bond formation process
that was responsible for the finished bond size needed to be improved.
These included optimiZing tail length, EFO voltage, current, time, DIS
power, and bonding force for more consistent bond squash.
However, the largest factor in achieving fine-pitch goals was the
accurate manufacturing of reduced-diameter bonding tools. Bottleneck
capillaries and side-relieved wedges were needed to ensure bonding that
did not touch the previous wire loop (Fig. 8-63). To illustrate the progress
of fine-pitch bonding from 1990 to 1995, the possible bonding pitch
(measured from center to center) moved from 150 /lm to 90 /lm for
11-200 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8-62. Prescribed Tool Trajectory for Loop Consistency. Looping incon-
sistencies may be caused by small servo lead/lag errors which are translated into irregular
wire shapes. Simultaneous coordination of three axes with real-time feedback is critical.

ball bonding and 75 !lm for wedge bonding. Several chip manufacturers
designed staggered bond pads in two rows with staggered lead-frame
fingers or two-tier ceramic packages, which produced effective pitches
of 45 and 37 !lm microns for ball and wedge bonds, respectively (Fig.
8-64). Both process development and refined tool designs continue to
provide engineering solutions for fine-pitch applications (Fig. 8-65).

Bottleneck Capillary (x100) Side Relief Wedge (x150)

Figure 8-63. Bottleneck Capillary and Side Relief Wedge. Examples of fine-pitch
tools that allow close proximity bonding to a previously bonded wire without interference.
The capillary shown is called a bottleneck. It has 360 degrees of relief because ball
bonding may be omnidirectional. Only the sides of the wedge are cut away for its
unidirectional process. (Courtesy of Micro-Swiss.)
8.4 WIREBONDING 11-201

Figure 8-64. Staggered Bonds to Reduce Effective Pitch. Shown is a test chip
with staggered ball bonds and two levels of loops to maintain a safe distance around each
wire. The effective pitch is one-half the pitch bonded on each row. (Courtesy of K&S.)

Although wirebonded packages traditionally have bond pads located


at the chip periphery, there are some applications that fall outside this
design guideline. Some loops are needed to traverse the die surface,
remaining parallel, then forming a bend and descending to the lead frame!
substrate. This bend near the die edge is particularly needed in thin
packages to avoid clearance problems between the edge of the die and

Figure 8-65. Fine-Pitch Ball and Wedge Bonding. Left: Shown is an example of
70-!lm ball bonding (1998 SIA benchmark). Note the amount of bond squash compared
to the example shown in Figure 8-53 without pitch constraints. Right: This is an example
of 60-!lm wedge bonding. High-frequency ultrasonics produces these bonds with minimal
squash out. An added benefit is the increased thickness which contributes to a larger
cross-sectional area of material at the heel of the bonds. (Courtesy of K&S.)
11-202 CHIP-TO-PACKAGE INTERCONNECTIONS

the wire. Special tool trajectories have been developed to place a calculated
angle bend into the wire and lower the preformed wire to its final position
without distortion (Fig. 8-66).
Although extra motions require additional time to form this type of
loop shape, the benefits of bonding interior bond pads and providing a
stiff wire that resists mold sweep (wire movement when viscous liquid
plastic encapsulant flows over it) far outweigh the extra time penalty of
a few milliseconds per wire.
Another application for the wire loop with a second bend at the die
edge is close proximity "down bonding" (Fig. 8-67). This connection is
used by MCM package designs as well as ICs with connections from the
die to the ground plane to which it is mounted. Once again, this application
requires proper clearance between the wire and die edge to eliminate the
possibility of an electrical short circuit.
Most ball-bonded applications place the second bond at (or below)
the first bond level for the best loop shape. Normally, second bonds
located in a plane higher than the first bond are reserved for wedge
bonding. However, there is one notable exception called lead-on-chip
(LOC). In this application, the lead frame is mounted to the top of the

Figure 8-66. Low Loops Preformed Before Attachment. Low loops shown in this
photo are 150 Ilm (from the top of the wire to the chip surface). The bends above the
ball and beyond the die edge provide proper wire clearance and the second bend adds
stiffness to resist wire movement during encapsulation. Bend angles are calculated for
each wire and preformed before attachment of the crescent bond.
8.4 WIREBONDING 11-203

Figure 8-67. Ground Bonds: Close Proximity. Using wire preform techniques simi-
lar to loops shown in Figure 8-66, these wires have second bonds placed 330 J.l.m from
the base of the chip. (Courtesy of K&S.)

die using double-sided adhesive poly imide tape to electrically isolate the
metal frame from the die. In addition to a unique looping situation, the
U/S energy can be attenuated/absorbed by the tape, forming a challenge
to achieve a reliable second bond. This application could be accomplished
by understanding the previously discussed bonding dynamics and the
variables responsible for a reliable weld. Therefore, U/S power was de-
emphasized while providing activation energy in the form of kinetic force.

8.4.5 Materials
8.4.5.1 Bonding Wires
Wirebonding materials have been reviewed by Gehman [128]. It is
customary to specify the mechanical properties of bonding wire by setting
acceptable ranges for the break strength (BS) and elongation (EL). Tensile
properties are determined from a standard stress-strain curve (Fig. 8-68).

8.4.5.2 Aluminum Wire with 1% Silicon


During the early attempts to manufacture fine aluminum wire (less
than 50 llm) it was seen that pure aluminum was difficult to draw through
11-204 CHIP-TO-PACKAGE INTERCONNECTIONS

I BREAK

/ ~r.r~7GTH
:2!:
o
Ul
,"~~~YIELD POINT - DETERMINED
z (,:,':"'SY 0.2" OFFSET,
W

I'''''"""'"""'""" ,,"
I-
',' YOUNG'S MODULUS'\.
w I ,'M. STRESS/STRAIN IN
a: '\."LINEAR REGION OF"
~ I \ " " STRESS - STRAIN :'\.
:' CURVE~~~
./ ~" ""\OUGHNESS - AREA

f=
UNDER STR [SS -
./ STRAIN
, CURVE

WIRE EXTENSION (STRAIN)

t-"- 0.2 " I


ELONGATION ---+-l
Figure 8-68. Stress-Strain Curve. Tensile test tension versus extension (stress-strain)
curve. (From Ref. 128.)

the decreasing die sets that produced the wire. It was equally difficult to
handle the wire in a bonding machine, so alloying was required to produce
a tougher wire material. Combining aluminum with silicon was a "safe"
choice because it would not introduce foreign elements into the package,
and the AISi alloy was well established in commercial practice. Today's
wire alloy (AISi 1%) became the standard composition in the early sixties
and remains today.
In retrospect, from a metallurgical standpoint, 1% Si as a solute in
Al for VIS bonding wire was an unfortunate choice. The equilibrium
solid-state solubility of Si in Al at 20° is of the order of 0.02% by weight.
Only at temperatures above 500°C is Si at I % in equilibrium solid solution.
Thus, at ordinary bonding temperatures, there is always a tendency for
Si to precipitate, forming a silicon second phase. When uncontrolled,
excessive Si segregation may degrade wirebondability or bond integrity.

8.4.5.3 Aluminum-Magnesium Wire


Aluminum-magnesium alloyed with 0.5% to 1% Mg can also be
drawn into fine wire. The finished product exhibits breaking strength and
elongation similar to that obtained in AISi 1%. This alloy wirebonds
satisfactorily and is superior to AISi 1% in resistance to fatigue failure
at first-bond heels and to break-strength degradation after exposure to
elevated temperatures. An early hypothesis that device field failures were
due to the presence of Mg in the wire alloy has been conclusively dis-
8.4 WIREBONDING 11-205

proven. In comparable production volumes, AIMg wire should be no more


expensive than AISi 1%.
The advantage of Mg over Si as a solute in Al wire alloys is that
the equilibrium solid solubility of Mg in Al is about 2% by weight. This
compares, as noted previously, with about 0.02% Si equilibrium solid
solubility in Al at 20°C. At 0.5% to 1.0% Mg concentration, there is no
tendency toward second-phase segregation, as is the case with AISi 1%.

8.4.5.4 Gold Bonding Wire


There are several parameters that must be precisely controlled to
provide acceptable gold wire for high-speed wirebonding. First, the surface
cleanliness must be highly controlled. This is a difficult task, as the entire
surface must be well lubricated during the drawing operation in order to
maintain a good surface finish and draw through the final (25-llm-diameter)
diamond die set. If the lubricant is not removed, it contaminates the
capillary tip during the EFO firing by deposition of the ionized material
and it can also clog the feed hole as it is fed through at high velocity.
Second, it reduces the area of exposed clean metal during bonding, as
this thin film prohibits the joining of the two clean materials from join-
ing/welding.
Once the wire manufacturer has produced clean wire, it must be
wound on a spool with precise tension: too little causes bunching/overlap-
ping and despooling problems, and too much results in the gold bonding
to itself, making despooling difficult.
Mechanical wire properties are determined by two parameters: wire
chemistry and heat treatment. Like aluminum, very pure gold (99.99%,
called four 9's) is very soft and difficult to draw into fine diameters. Small
concentrations of impurity atoms (dopants) make the gold workable.
Several foreign elements may be added to the pure gold to toughen
it; Be and Ca are commonly used, with the latter used exclusively by
Japan. The levels of the dopants varies from one manufacturer to another,
but typical levels of Be are 5-10 ppm by weight.
Other dopants are used in the manufacture of specialty wires for
certain applications. For example, Pd is often used in "bumping wire,"
which forms a gold ball on the chip pad (Fig. 8-69). The wire is then
intentionally broken above the ball to provide a gold bump for the bonding
of a TAB lead. The Pd dopant raises the recrystallization temperature of
the material so that the fully annealed area above the ball resulting from
the EFO heat applied to that area, called the heat-affected zone (HAZ),
is extremely short (Fig. 8-70). Conversely, other dopants, such as La,
lower the recrystallization temperature, creating a very long HAZ. This
wire can be used for specific loop-height applications by providing a
naturally bending area without wirebonder motions creating kinks in the
wire to achieve the required loop height/shape.
11-206 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8-69. Wire-Bonded Wafer Bumps. Right: Ball bonds exhibiting short tail
broken within the small HAZ created with Pd dopant. Left: Bonds that are coined, using
a secondary operation to flatten the tail feature. (Courtesy of K&S.)

CONVENTIONAL WIRE L: LONG


LOW LOOP WIRE L:SHORT

Figure 8-70. Heat-Affected Zone. Heat energy from the ball formation process is
conducted along the wire length which recrystallizes/anneals the affected region. The
recrystallization temperature of the material is controlled by the type of dopants added,
which determines the length of this "zone."
8.4 WIREBONDING 11-207

Finally, heat treatment ofthe cold-worked wire is necessary to adjust


the temper because the as-drawn properties are too stifflbrittle to withstand
the movements of an automatic wirebonder and the rigors of the molding!
encapsulation process. Annealing the material, combined with the proper
dopants, produces fine gold bonding wire that is soft enough to be shaped
into loops, yet stiff enough to hold its shape after forming.

8.4.5.5 Gold Wire Substitutes


Reducing packaging costs has often led designers to investigate
replacing expensive gold material with less expensive ones. Many materi-
als were able to be drawn into fine wire diameters and balls were success-
fully produced but only two (aluminum and copper) generated real interest
over the last two decades. While both materials required changes to the
flame retardant contents of the plastic encapsulant, aluminum was soon
discarded because of ball porosity. The amount of this porosity could not
be reliably predicted. Copper remains the major viable alternative to gold
for certain applications such as power devices, but it has been slow to
gain popularity due to reliability fears associated with oxides trapped at
second bond and corrosion issues in a non-hermetic package, (even though
its electrical/thermal conductivity, tensile strength, and price are superior
to gold).

8.4.5.6 Die Attach Materials


Shukla [129] has reviewed die-bonding materials. The die-bonding
process includes a large choice of materials, which may be broadly classi-
fied into the three categories described below.

8.4.5.6.7 Solders
These include single, binary, and ternary metallic compositions and
may be further classified into two subgroups: hard and soft solders. Hard
solders have rather high flow stresses (onset of plastic flow), offering
excellent fatigue and creep resistance. The disadvantage of using hard
solders stems primarily from their lack of plastic flow, which leads to
high stresses in the silicon chip because of the thermal-expansion mismatch
between the die and the substrate. Soft solders, on the other hand, are
low-melting binaries and ternaries which have a high degree of plastic
strain capability.

8.4.5.6.2 Organic Adhesives


Epoxies and polyimides filled with precious metals have found wide-
spread acceptance as die bond materials in low-cost packaging (plastic
packages). They offer the advantage of lowest processing temperatures
and lower cost compared to the gold-based hard solders.
11-208 CHIP-TO-PACKAGE INTERCONNECTIONS

Numerous articles have claimed that organic adhesives lower thermal


stresses in silicon. These adhesives are typically filled with a metal to
provide the thermal and electrical conductivity required in most applica-
tions (silver being the most common filler). However, the use of organic
adhesives in high-reliability packages (hermetically sealed) has been slow
in acceptance because of their poor thermal stability, especially when
filled with silver. Poor thermal stability, coupled with the outgassing of
entrapped solvents and other gaseous species, remains a challenge in the
application of these materials to high-reliability VLSI applications.
Curing epoxy requires oven baking for several hours, adding signifi-
cant processing time to the assembly cycle rate. If copper leadframes are
used, the oven must be filled with oxygen-free gas and coverage must
continue until the oven is cooled to ambient temperature, lengthing the
process even further. A considerable amount of research in epoxy chemis-
try resulted in fast-cure or snap-cure epoxies, capable of curing in a few
minutes or less. Curing may be performed in a flow-through oven con-
nected to the wirebonder, greatly reducing total product cycle time.

8.4.5.6.3 Glass Adhesives


Silver-filled specialty glass materials have emerged in the past few
years as an alternative for VLSI die bond applications. These materials
offer the possibility of a void-free, die bond interface with excellent
thermal stability. However, the currently available materials require high
processing temperatures (400°C) and oxidizing ambients for optimum
adhesion, creating some special processing concerns. In addition, the glass
die bond materials also need solvents and binders for processing purposes,
leading to the problem of complete solvent removal (similar to the case
of organic adhesives).

8.4.5.7 Bonding Tools: Capillaries and Wedges


The microscopic features of these replaceable tools are responsible
for a significant part of the bonding process. The final shape of the ball!
crescent and wedge bonds are a direct imprint of the tool itself. Consistent
loop shapes cannot be attained without correct feed hole dimensions,
leading-edge funnel shape, surface finish, and cleanliness.
A tool expert should always be consulted to select the correct combi-
nation of features for a particular application. Even then, it may be neces-
sary to create a designed experiment to select a final choice from sev-
eral candidates.
When tools were made using manual grinding techniques, there was
a significant variation between tools. A bonding process could run out of
its control limits simply by exchanging a worn tool for a fresh one. The
advent of computer-controlled, micromachining centers now produces
8.4 WIREBONDING 11-209

tools that can be assumed to have dimensions without variance from one
to another.
The most common materials for capillaries and wedges are alumina
ceramic and tungsten carbide, respectively, with both materials originating
as a sintered powder. Alternative materials such as glass, ruby, and titanium
carbide are sometimes used for unique applications. Research continues
to find better material for fine-pitch requirements.
Defining tool life is usually at the discretion of each semiconductor
manufacturer. Depending on the number of wires bonded, some high-
production facilities change tools at the end of every shift, whereas others
are able to last four to five working shifts using slower bonding equipment.
All facilities generally agree that the tool is considered a consumable and
should be replaced on a regularly scheduled basis before marginal bonds
are created, or the tool breaks.

8.4.6 Process Automation

Since the mid-seventies, process engineers and production manage-


ment for packaging have continually demanded higher levels of automation
and process control for wirebonding. It is desirable to remove human
handling of packages to improve both quality and throughput. Observa-
tions of factory production activities reveal that wirebonding defects are
not randomly distributed throughout the production shift but occur most
often in "clusters" surrounding operator intervention that interrupts the
machine's automatic operation. This machine assist is typically required
when defective/contaminated material prevents formation of a proper bond
or causes the transfer handling system to jam.
Although wirebonder operators have always been restrained from
changing bonding parameters beyond specified limits, some parameter
adjustment has always been tolerated to adapt bonding energy to slight
material differences from lot to lot. Machine designers have continually
upgraded hardware and software in the areas of bonding ultrasonics and
force application, as well as the wire feed system to reduce/eliminate the
amount of human adjustment by closing system control loops to create
self-monitoring and adjusting systems.
Host computer control for the entire wirebond manufacturing cell
has been possible for a number of years, but it was not feasible until the
mid-nineties because there was too much variation between machines.
This prohibited downloading of a common parameter set and achieving
identical bonding results. Individual machine calibration techniques and
closed-loop systems now make downloading of host parameters and
uploading of system states a reality that actually provides new levels of
bonding/looping quality.
Further process automation includes the linking of pre/postwirebond-
11-210 CHIP-TO-PACKAGE INTERCONNECTIONS

ing processes to eliminate human handling between assembly steps and to


further reduce assembly cycle time. This has been achieved by physically
linking epoxy cure ovens and die bonders (Fig. 8-71) to the wirebonder
for some chip manufacturers, whereas others have provided a variety of
robots to shuttle material on-off machine docking stations.

8.4.7 Process Choice Decisions/Guidelines


This section will provide basic guidelines for selecting a wirebonding
process for various packages. There are many devices that do not fit into
standard categories, so packaging engineers/designers should consult with
a wirebonder manufacturer to avoid designing a package that does not
allow a robust assembly process to be used.

8.4.7.1 Wire Type


The advantages of aluminum wire are that it provides the highest
reliability bond to the aluminum metallization on the chip, and it forms
a monometallic bond at room temperature. It has two disadvantages: low
resistance to corrosion in the presence of moisture and it does not form
consistent, nonporous balls when melted. This means aluminum wire is
only suitable for wedge bonding (normally in high-end hermetic packages).
It cannot be used in molded plastic packages because of its susceptibility
to corrosion, which will cause reliability problems.

Figure 8-71. Flex-Line System. Die attach (not shown), cure oven, and wirebonders
linked in-line reduce ftoorspace, work-in-process, and handling induced yield issues.
(Courtesy of K&S.)
8.4 WIREBONDING 11-211

Although gold wire is more expensive than aluminum, it remains


unharmed by moisture, making it the best choice for a plastic package.
Its two biggest disadvantages are cost and 2000 e bonding temperature.
Gold is equally suited for ball- or wedge-bonding processes.

8.4.7.2 Ball-Bonding Guidelines

Because bonding equipment, capillaries, wire, devices, die attach,


and lead frames are all changing constantly, it is impossible to set down
exact guidelines for any process. However, a few basics about ball bonding
are worth noting.
Bond pad size on the chip is dictated by the area needed for the
bond itself. The smallest bond size is a function of the unbonded ball
size, the diameter of the wire, and the bore size of the capillary. A ball
must be formed by the EFO that is large enough so that it is not extruded
up the bore as bonding pressure is applied. Therefore, these three factors
must be optimized before a minimum bond size can be defined. The
amount of pad real estate surrounding the bond allotted to accuracy/
repeatability of the bonder will continue to shrink as new generations of
machines are developed. The equipment supplier should be contacted for
the latest performance specifications.
The main factor dictating bond pad pitch is the outside dimensions
of the capillary, with the system accuracy/repeatability being statistical
error terms that must be considered when calculating the minimum space
between adjacent pads. This space is required to prevent a collision be-
tween the capillary and the wires previously bonded. Actual minimum
pitches that can be achieved change annually and generally follow the
SIA roadmap. Detailed models are available to perform the actual calcula-
tions if in doubt about a particular application.
It is somewhat ironic that silicon shrinkage and fine pitch at the first
bond has significantly increased the wire lengths within the package. The
reason is lead-frame pitch. The technology for stamping and etching lead
frames cannot match the fine geometry possible on the chip. An increasing
number of package I/Os means a larger radius around the die. Wire lengths
have increased from 1 mm to 5 mm with foreseeable need for 7 -mm
lengths. Figure 8-72 shows the wire lengths that result from various pad
and lead-frame pitches.
Wire lengths greater than 4 mm present a number of assembly
challenges. The bonder must suspend the length of wire between the die
and lead frame without vertical sagging or horizontal swaying. This is
particularly important as device signal speeds are increased and electrical
performance of the device is dependent on the shape of the loop and
distance between wires.
The lead frame should be designed so that the wire loop is routed
11-212 CHIP-TO-PACKAGE INTERCONNECTIONS

PAD PITCH = 4.5 mil


~.,---------------------------------.

400 wire length vs. pin count f---__________"7"""=---1.c=2-=rm


for various lead tip pitches
pad pitch = 4.5 mils
.!!!350
'e ~+-______________~~__~~____=-~
..r::
ac: ~+---------~~-=~--~~-------=~
J!!
~ 200 emU

.~ 150'-b-<"'------=--'''--_=;~ -/----::;::".-"F--------------j
100~=----~_r=---~-------+-----------~

50'+----,----+---,----1-------,-----,-.1.-,----.---------1
120 140 180 180 200 220 240 280 280 ~ 320

pin count

Figure 8-72. Package Wire Lengths. Example of resultant wire lengths for a radially
wired package with a 4.5-mil bond pad pitch and various second-bond pitches. (From
Ref. 130.)

over the tip of each lead finger. If the wire path is over the side of the
finger, it will have a tendency to "roll" further to the side, inducing sway
in the loop. This is more likely to happen if the angle between the wire
and the finger is greater than 20 degrees.
The top surface of the lead finger is often wider than the bottom
side. If the capillary must be targeted off the exact center of the finger,
to keep the loop within its correct path/corridor, the delicate finger can
often twist or roll. If this happens, neither the bond force or ultrasonic
power will be transmitted correctly and a substandard weld can result.
The size and shape of the crescent bond are directly related to the
diameter of the wire and the geometry of the capillary. This is a very
important consideration when reducing capillary dimensions for fine-pitch
considerations. Although it may be tempting to use a small capillary for
first-bond pitch, adequate face dimensions must be provided in order to
produce a crescent bond with minimum welded area.

8.4.7.3 Wedge-Bonding Guidelines

Advances in ultrasonic systems have made it possible to produce a


high-strength wedge bond that is only 2-3 ~m wider than the wire diame-
ter. Therefore, the minimum width of the chip bond pads is largely dictated
by the wire diameter (plus the necessary pad real estate for system inaccu-
racies). However, the pad layout is not nearly as straightforward as ball
bonding. First, the pad length must support the long dimension of the
8.4 WIREBONDING 11-213

bond (usually two to three times the wire diameter) as well as the tail
protruding from the first bond. Next, the pad's long axis should be oriented
along the intended wire path.
Generally, packages can be designed with either orthogonal wires
or radial wires. Pad pitch for orthogonal wires is mainly governed by the
wedge's outer geometry, with consideration given to the bonding system's
statistical error. Radial wiring will have the same pitch as orthogonal
wiring at the center of each side of the chip. As the pad locations approach
the comers of the chip, the bond pads must be rotated at an angle so that
it is aligned with the wire path. Furthermore, the pitch must be increased
to maintain consistent distances between wires.
Up-bonding, where the second-bond location is located above the
chip plane, is easily achieved with the wedge bonder while it is more
difficult with ball bonding. The bonding tool movement through three-
dimensional space will actually contact and "curl" the bottom side of the
wire to provide extremely consistent loop heights with this process, even
at wire lengths of 4 mm.
Second bonds are typically located on a ceramic substrate with either
embedded lead-frame fingers or printed/fired metal paste. Bonds can be
easily located in the center of these locations, but specialized vision!
lighting systems must be developed to accommodate certain low-contrast
packages. The most important application issue with second-bond location
is the position of the bonder's wire clamps to the rear of the wedge. Many
packages have a lip or wall surrounding the second-bond locations for
the attachment of the hermetic lid. Horizontal and vertical clearance must
be provided to prevent interference with the wire clamps' operation during
second-bond formation. Consult the equipment specifications for the mini-
mum distances required for various machine/wire path configurations.

8.4.8 Evaluating/Optimizing the Process


This section divides wirebonding evaluation into two groups: engi-
neering methods and production tools. Some engineering methods are
sometimes used for production purposes, depending on a need/situation
that may arise from previous experiences with a particular package or
device. Additionally, each group may be further divided into visual and
mechanical techniques for wirebond evaluation.

8.4.8.1 Engineering Evaluation Methods (Visual)


1. A toolmaker's microscope equipped with O.1-J..Lm resolution mi-
crometers is used to accurately measure two-dimensional (X and Y) bond
features while a third micrometer mounted to the focus adjustment is used
for height (Z) measurements. When this instrument is connected to a data-
collection PC, it allows the engineer to measure statistically significant
11-214 CHIP-TO-PACKAGE INTERCONNECTIONS

sample sizes quickly and is considered the standard visual tool for measur-
ing bond squash, height, concentricity, loop heights, and straightness. It
is also valuable for evaluating qualitative attributes such as cracks,
scratches, or surface finish while setting up or troubleshooting a wire-
bonder.
2. A scanning electron microscope (SEM) may also be used for
optimizing a process because it provides an infinite depth of field to the
observer so that features are seen clearly throughout the field-of-view .
Optional features include measurement at very high magnification, provid-
ing indisputable data for optimization experiments. Auger and beta-back-
scatter probes may be added to the SEM to answer questions about
contamination or elemental analysis.
3. The SEM also provides qualitative results for bond peel testing.
Tweezers are used to gently peel a crescent bond or wedge bond away
from the substrate to reveal the amount of wire material still bonded to
the surface (Fig. 8-73).
4. Bonds may also be chemically removed from the chip to gain
visual access to the underlying layers to look for cracks caused by the
bonding process. A toolmaker's microscope fitted with Nomarski filters
is ideal for finding slip-plane dislocations or cracks. Chemicals such as
sodium hydroxide and aqua-regia are used to etch away the aluminum
and gold, respectively. Both reagents can be harmful and should be used
by an experienced technician.

Figure 8-73. Engineering Peel Test: Crescent Bond. Crescent bond shown partially
peeled away from the lead-frame finger. The beta-backscatter mode on the SEM may be
used to evaluate location of Au material still bonded to a Ag substrate.
8.4 WIREBONDING 11-215

5. Perhaps the most tedious visual method is a cross-section inspec-


tion. The device is first encapsulated and placed on grinding/polishing
wheels to remove one half of the bond. A light etch following the polishing
reveals the amount of thickness of intermetallic between the bond and pad.

8.4.8.2 Engineering Evaluation Methods (Mechanical)


1. Mid-span pull testing is a traditional test for bond evaluation.
Unfortunately, the results are more dependent on the wire length and
height than on the strength of the bonding. Normally, the wire breaks
near first bond, leaving both first and second bonds intact. However,
spreadsheet calculations using the traditional formulas, with some refine-
ments can now be used to select unique pull test target strengths for
each application [131]. This makes pull testing much more valuable than
previous specifications, where one target was used for all packages. Al-
though the test is relatively easy to perform, it provides the least amount
of usable data for the process engineer.
2. Shear testing ball and wedge bonds provides a very reasonable
approximation of the weld area between the wire and substrate. Harman
describes [132] this correlation as well as analyzing failure modes associ-
ated with the shear test, which is extremely important, yet rarely used.
3. A variation of the standard pull test involves placing the hook
as close to second bond as practical, for a particular device. This second-
bond pull test is quite useful to optimize bonding parameters for second
bonds. In both ball and wedge processes, the wire above the second bond
is not weakened, so pull strengths give an indication of second-bond weld
strength because the bonds often lift before the wire breaks. (Note: In
ball bonding, the EFOlball formation recrystallizes and weakens the wire
above first bond. In wedge bonding, the heel of the first bond is weakened
by the smaller, back radius of the wedge, which is used to promote
consistent tail lengths when the wire is pulled/severed at second bond.)

8.4.8.3 Production Evaluation Methods


1. The 30-60x optical visual inspection method is the most common
technique used around the world. It is either performed on the wirebonder
itself, or after bonding at an station called "Third Op." Although an
inspector can scan an entire package quickly, looking for missing bonds/
wires, bond placement, wire straightness, chip defects, and contamination,
it becomes tedious and inefficient, like all visual inspection methods. It
is also performed too late to provide real-time feedback to the wirebonder.
2. Wirebonder machine vision systems can now replace this manual
technique, performing the inspection while the device is still located in
the bonder itself, so real process feedback is possible. Using a pattern
recognition system and high-resolution camera/optics, this technique can
11-216 CHIP-TO-PACKAGE INTERCONNECTIONS

perform sampling inspection of most visual attributes for the bonds and
wire loop, including actual variables measurements that are not part of a
typical third op visual.
3. The pull test has remained a mainstay of production testing, even
with the shortcomings listed above. Its value to the production line is to
highlight an out-of-control situation that may not be detected by the visual
inspection. Because of the destructive nature of the test, it is only performed
on a daily sampling basis for each machine.
4. Although shear testing is also destructive, it has a much higher
value for production. This test can provide meaningful trend data for each
bonder's true weld strength.
5. Open/short and electrical testing after encapsulation provide the
final test method for wirebond yield, although traceability to the wire-
bonder must be maintained to make the data completely meaningful.

8.4.9 Wirebonding 2000: The Future

The Semiconductor Industry Association (SIA) has produced a na-


tional technology roadmap, updated every 2 years. Although it avoids
specifying any preferred technology solutions or specific agendas, it does
provide an accurate description of semiconductor advancements in the
technology requirements and performance of ICs. It is the marketing and
R&D departments of the equipment and consumables suppliers that must
keep their products aligned with this roadmap.
Bond pitch will continue to be driven smaller, with the goal stated
at 50 Jlm for the years 2000-2010. This will be partially achieved through
continued capillary development, enhanced wire alloys, and "no wire
sweep" molding compounds. In addition, machine developments that elim-
inate operator "teach" error, inspection systems with high magnification
and positional feedback, and high-resolution servo systems will meet!
exceed the fine-pitch goals.
Ball-grid-array (BGA) packages will have continued growth and the
roadmap itemizes several wirebonding technology needs. The bond pad
pitch is somewhat less aggressive than other package types, but wire
interconnects will be greater than 600 per package. Faster bonding cycle
times and low-temperature gold wedge bonding (<170°C) must be devel-
oped for this package. If a strip format for BGAs is adopted for wirebond-
ing, a machine with a rotating head will be required. Otherwise, singulated
packages can still be handled by rotating the individual part and placing
into a boat!carrier.
Although many of the packaging needs listed for the future of wire-
bonding are difficult to achieve, perhaps the greatest challenge is the sheer
number of bonds that will need to be produced per square meter of
production floor space. The chip pad count (chip to package) increases
8.5 TAPE AUTOMATED BONDING 11-217

more than any other package feature in the roadmap (Table 8-1). Com-
puter-aided designlbond parameter files will need to be downloaded into
the bonder to completely eliminate the huge programming functions asso-
ciated with teaching more than 300 wires (600 bond sites). Servo computer
systems in the bonder will have to execute calculations and drive each
electromechanical axis without interrupts/delays. Low-mass bond heads
are certainly required to achieve higher-velocity, accurate movements to
create the bonding and looping motions associated with large numbers
of wires per minute. It will ultimately be the top end speed of the bonder
of the future that will determine the life of this interconnect technology.

8.5 TAPE AUTOMATED BONDING


8.5.1 Introduction
The tape automated bonding (TAB) concept originated in the 1960s
as the "miniMod" project at General Electric Research Laboratories in
New York. The name is derived from the French term "Transfert Automa-
tique sur Bande" coined in 1971 by Gerard Dehaine at Bull-General
Electric in France (later known as ClI Honeywell-Bull, and now as Bull
S.A.). In Japan, TAB is known as tape carrier packaging and a TAB
package is called a tape carrier package (TCP).
TAB was originally developed as a labor-saving, highly automatable
reel-to-reel "gang bonding" technique for packaging high-volume, low-
VO devices at a much lower cost than wirebonding [133]. It is worth
repeating that in the 1960s and early 1970s, wirebonding equipment was
slow and operator dependent. The process itself was beset by reliability
problems, mostly related to operator inconsistencies, and metallurgical
failures due to brittle gold-aluminum intermetallics and corrosion from
ionic impurities in the molding compounds. So some companies looked
for better alternatives. The result was that IBM developed the "flip-chip"
process, Bell Telephone Laboratories (now known as Lucent Technolo-
gies) developed the "beam lead" process, and General Electric developed
the "miniMod" process. All three processes allowed the electrodes (bond
pads) on the chip to be connected simultaneously in one operation to the
next interconnection level. The beam lead process featured chips with
cantilevered gold beams plated onto the bonding pads at the wafer level
[4]. After chip separation from the wafer, all the beams on the chip were
gang bonded simultaneously to gold pads on the package or substrate
using heat and pressure. The miniMod process used a three-layer, etched,
35-mm tin-plated copper tape laminated to a sprocketed polyimide film
(similar to movie film). The inner bond "fingers" were simultaneously
gang bonded to gold bumps on the semiconductor chip using a reel-to-
reel assembly machine called the "inner lead bonder" (lLB). After inner
lead bonding, the chip was tested. Good chips were then excised (cut)
11-218 CHIP-TO-PACKAGE INTERCONNECTIONS

from the tape, and the outer tips were "outer lead bonded" (aLB) to
electrodes on second-level substrates or cards.
Japan was quick to recognize the high-productivity potential of this
new technology and adopted the GE process in the 1970s for low-cost
consumer devices such as calculators and watches. Sharp Corporation
and Shindo Company Ltd. were the pioneers. Today, Sharp is the largest
user of TAB in Japan, and Shindo Company is the largest maker of TAB
tape. In Europe, Bull was the first to adopt the miniMod process for
computers in the seventies [134]. In the United States, most of the TAB
activities in the 1970s and early 1980s were focused on assembling bipolar
transistor-transistor logic (TTL) devices using single-layer or two-layer
tape by leading semiconductor makers such as Fairchild Semiconductor
(which merged with National Semiconductor in 1987), Motorola, National
Semiconductor, RCA, and Texas Instruments. 3M became the leading
supplier of two-layer tape in the seventies and eighties and still holds that
position today. Over the years, TAB activities in the United States have
been sporadic. At one time in the eighties, National Semiconductor was
the largest producer of TTL TAB devices in the world, with an annual
run rate of more than 400 million units. However, National and others
abandoned TTL TAB assembly in the late 1980s in favor of wirebonding
for cost reasons. Hewlett-Packard (HP) built calculators with TAB for a
few years in the eighties before switching to wirebonded packages. HP
also began using TAB for its inkjet printers in the eighties and is now
the largest volume user of TAB in the United States. ETA, a subsidiary
of Control Data Corporation used a unique TAB process with plated
solder bumps in its "all-TAB" supercomputer in the eighties. The TAB
parts were made by Honeywell. Around 1986, National developed a TAB
package called the "TapePak®" which used a single-layer bumped tape
and a unique molded ring with test pads and licensed it to Delco Electronics
which continues to make TapePaks today for automotive applications.
Digital Equipment Corporation (DEC) started a large TAB project for its
VAX® 9000 computer line in the eighties which led to a flurry of activity
by 3M, Olin Mesa, Rogers Corporation, and others to develop a two-
conductor tape. At about the same time, IBM licensed the 3M two-layer
tape process and set up a production line in Endicott, New York. After
DEC stopped using TAB in its VAX 9000 line a few years later, both
Olin Mesa and Rogers abandoned the TAB tape business. In the equipment
arena, Jade Corporation of the United States was an early leader of lLB
and aLB gang bonding equipment in the seventies and eighties but is
not active today. The same is true of Farco, a Swiss equipment company
which was prominent in the late seventies and eighties but left the business
in the early nineties.
As the I/O, speed, and performance requirements of lCs increased
in the eighties and nineties, TAB became more suited to applications in
8.5 TAPE AUTOMATED BONDING 11-219

liquid crystal displays (LCDs), printheads, very-high-speed integrated


circuits (VHSIC), workstations, and high-end computers (see Fig. 8-74)
[l35-l37]. The users included Bull, Casio, Cray Research, Digital Equip-
ment Corporation (DEC), Fujitsu, Hitachi, HP, Matsushita, NEC, IBM-
Japan, Seiko-Epson, Siemens, Sharp, Sun Microsystems, TI-Japan, and
Toshiba.
Figure 8-75 shows a typical TAB process using bumped chips and
planar tape. After wafer bumping, the wafer is mounted on a dicing tape
and diced. The next step is inner lead bonding (lLB) in which the metallized

TAB
500

R&D
t

400

-I
~
CD

300
g
1/1

"0
...
CD
.0
E o
J IBM Japan PC
Z
200 ~
~....
NEC SX, A2000, 930 <&
i$l0'

NEC A1500, 910,


830, 610/30
100

o 100
Lead Pitch (J£m)

Figure 8-74. TAB Technology Evolution. Tape, inner lead bonding (lLB), and outer
lead bonding (OLB).
/1-220 CHIP-TO-PACKAGE INTERCONNECTIONS

Wafer Bumping
J.
Wafer Dicing
J.
Inner Lead Bonding
J.
Testing
J.
Encapsulation
J.
Singulation
J.
Testing/Bum-in
J.
Outer Lead Bonding

Figure 8-7S. Typical TAB Fabrication Process

tape fingers are bonded to the bumps on the chip pads (see Fig. 8-76).
After ILB, the chip can be tested and/or encapsulated on one or both
sides. Figure 8-77 shows a schematic of the encapsulation step. Next, the
device can be singulated and placed in a "slide carrier" for testing and/
or burn-in prior to outer lead bonding (OLB), or it can be processed in
strips or reel. Prior to OLB, each device is excised from the tape, the
leads are "formed" if necessary, and aligned with the bonding pads on
the substrate or card. The OLB is usually a solder bond (see Fig. 8-78).
Board-mounting options include chip-up or chip-down (called "flip TAB")
configurations.
There are many variations of TAB-both as a first-level interconnect

Chip I/O Chip


Pad Pa.lvatlon

Figure 8-76. TAB Inner Lead Bonding


8.5 TAPE AUTOMATED BONDING 11-221

Figure 8·77. TAB Encapsulation. A single-orifice nozzle dispenses encapsulant on


the bonded chip.

and as a second-level package. These include: bumped tape (BTAB),


transferred bump tape (TBTAB), demountable TAB (DTAB), TapePak®,
Tape Quad Flat Pack, Mikropack, and Tape Ball-Grid Array (TBGA).
Sections 8.5.2 through 8.5.7 cover the various tape structures, bump-
ing methods, ILB and OLB processes, package options, and applications.
Electrical design, thennal design and reliability issues are covered in

Solder Polymer
Joint

Figure 8·78. TAB Outer Lead Bonding. After the TAB package is excised from the
tape and placed on the substrate or board, it is normally solder bonded.
11-222 CHIP-TO-PACKAGE INTERCONNECTIONS

Chapters 3, 4, and 5, respectively, and are not discussed separately in


this section.

8.5.2 TAB Tape


The TAB tape (also called "tape carrier" or "film carrier") comes
in a variety of shapes, sizes, base materials, and surface finishes [138-147],
as shown in Table 8-4. It has three basic formats: one layer, two layer,
and three layer (see Fig. 8-79). Table 8-5 shows the process steps for the
fabrication of the three basic tape formats.

8.5.2.1 One-Layer Tape


One-layer tape is made of metal, typically copper, 35-70 11m thick.
It can be etched or stamped and has flat (planar) lead tips or "bumped"
lead tips. Both plated and unplated single-layer tapes have been used in
the past. In addition to copper, aluminum, steel, alloy 42, and other
conductive materials can be used for the base metal.
In general, a device bonded to single-layer tape is not testable on
tape because all the leads are shorted together (except in the case of the
TapePak® design which has a special test ring). One-layer etched tape
does not need special punching tools for punching the sprocket holes and
personality window (also called "device hole") because these are etched
at the same time as the leads. This makes one-layer tape the least expensive
of the three fonnats. It is therefore widely used in cost-sensitive, low-
lead-count applications such as watches, security tags, IC cards, blood
pressure monitors, and car radios.

8.5.2.2 Two-Layer Tape


Two-layer tape typically consists of a copper layer directly bonded
to polyimide dielectric film without an organic adhesive. The fabrication

Table 8-4. TAB Tape Options

1. Fonnat One, two, or three layers


2. Conductors One, two, or more
3. Copper foil Rolled (as rolled or annealed); electrodeposited
4. Dielectric Polyimide, epoxy-glass, polyester, BT
5. Adhesive Modified epoxy, phenolic butyral, polyester, polyimide
6. Width (mm) Super 8, 8, 11, 16, 19, 24, 35, super 35, wide 35, 48, super 48, wide
48, 70, super 70, wide 70, 120, 140, 158
7. Surface structure Planar; bumped
8. Surface finish Tin, gold, nickel-gold, solder (tin-lead); selective plating
8.5 TAPE AUTOMATED BONDING 11-223

• On.. Layer Tape


Conductor

• Two-Layer Tape
Conductor

~i=&/iiii.1~'
• Three-Layer Tape
Conductor

Figure 8-79. Generic Types of TAB Tape

Table 8-5. TAB Tape Fabrication Steps

I-Layer 2-Layer 3-Layer


All Metal Additive Process Subtractive Process

I. Slit metal foil 1. Deposit metal adhesion I. Punch adhesive-coated


layer/common electrode polymer
2. Apply photoresist 2. Apply photoresist (2 2. Laminate metal foil
sides)
3. Expose 3. Expose (2 sides) 3. Cure adhesive
4. Develop 4. Develop 4. Apply photoresist
5. Etch metal pattern 5. Pattern plate 5. Expose
6. Strip photoresist 6. Etch polymer 6. Develop
7. Clean 7. Strip photoresist 7. Coat back side for lead
protection
8. Surface plate 8. Etch common electrode 8. Etch metal
9. Clean 9. Strip photoresist
10. Surface plate 10. Clean
11. Surface plate
11-224 CHIP-TO-PACKAGE INTERCONNECTIONS

process can be either subtractive (called "two-layer thin" by 3M, the


leading two-layer TAB tape supplier in the world) or additive (called
"two-layer thick" by 3M). In the subtractive process, a thin (typically 12
~m) liquid polyimide film is cast (coated) on 35-~m-thick copper, and
the circuit pattern is obtained by etching the copper and polyimide. The
process has two inherent disadvantages: The polyimide shrinks after cure,
causing dimensional and adhesion problems, and second, the polyimide
cure process anneals the copper foil. The subtractive process is not used
much today.
In the additive fabrication process for two-layer thick tape, thin layers
of chromium and copper are sequentially sputtered onto the polyimide film
which is typically 50-75 ~m thick. Chromium promotes adhesion between
the polyimide and the copper, the latter acting as the seed layer for the
final pattern plating through a photoresist.
Two-layer tape allows testing and burn-in because of the electrical
isolation between leads provided by the polyimide. Also, it offers better
handling and lead stability than one-layer tape. Two-layer tape does not
require hard tooling and has a faster turnaround time than three-layer tape.

8.5.2.3 Three-Layer Tape

Three-layer tape consists of a metal foil, usually copper, laminated


with an organic adhesive to a base film such as polyimide. The polyimide
film, typically 75-125 ~m thick, and the adhesive layer are mechanically
punched with steel dies to form sprocket holes and inner lead and outer
lead windows. The copper foil is either electrodeposited (ED) or rolled
(wrought). Each has advantages and disadvantages depending on the
application. ED foil is more popular in Japan and Europe. Rolled foil is
more common in the United States. Copper foil thickness is usually
specified in "ounces per square foot" ("ounces"). "One-ounce" copper is
35 ~m thick and weighs 1 ozlfe or 305 g/m2. A two-ounce copper foil
is 70 ~m thick, and a half-ounce foil is 18 ~m thick. Copper foil properties
such as crystal structure, grain size, hardness, tensile strength, surface
roughness, yield strength, ductility, adhesion to dielectrics, and so forth
are controlled by a variety of proprietary processes and treatments. ED
foil has a columnar grain structure which can be modified by annealing
to produce equiaxial grains. Because of the fabrication process, one side
of the ED foil is much smoother than the other side. Rolled foil has a
laminar crystal structure, with the grains aligned along the length of the
material due to the rolling process. This results in anisotropy-the material
behaves differently in the longitudinal direction (also called "machine
direction") than in the transverse direction. Like ED foil, as-rolled (AR)
foil can be annealed and treated to improve its properties for a particu-
lar application.
8.5 TAPE AUTOMATED BONDING 11-225

The most common dielectric film materials used in three-layer tape


are polyimide, polyester, glass-epoxy, and bismaleimide triazine (BT).
Polyimides such as Kapton® and Upilex®, are preferred in high-perfor-
mance applications. Lower-cost glass-epoxy and polyester base materials
are used in consumer products such as calculators and smart cards. BT
is used in applications requiring a higher glass-transition temperature than
glass-epoxy or polyester.
The adhesives used in three-layer tape are mostly proprietary formu-
lations. The formulations include modified epoxy, polyimide, phenolic
butyrals, polyesters, and acrylics. Adhesive characteristics are extremely
important for high-II0, high-performance applications where high thermal
stability, high adhesion, low moisture absorption, and low ionic impurities
are critical.
Figure 8-80 shows a three-layer TAB tape layout and cross section
listing the terms used to describe the tape parameters. Figures 8-81 and
8-82 show two tape designs: a 200-pin EIAJ 35-mm design, and a "super
slim TAB" (SST) LCD driver design, respectively. Figures 8-83 and 8-
84 show photographs of low- and high-pin-count tapes, respectively.
Figure 8-85 shows a variety of tape formats.
Because the three-layer process uses subtractive methods to etch
copper foil, the thickness of the copper foil determines the lead pitch
(width and spacing) that can be etched. Finer lead pitches therefore require
thinner copper foils [148].
Plating plays a key role in the success of inner lead bonding (ILB)
and outer lead bonding (OLB). It also affects other parameters such as
storage life, bum-in capabilities, long-term reliability, and cost. Plating
options include tin plating, gold plating, solder (tin-lead) plating, and
"selective plating" where inner leads and outer leads have different materi-
als or thicknesses.
The most common selective plating is tin for ILB and solder for
OLB. Other choices include gold for ILB and solder for OLB, tin for
ILB and gold for OLB, and thin solder for ILB and thick solder for OLB.
The plating thickness depends on the application. Typical ranges
are 0.3-0.6 J..lm for tin and solder, 0.3-4.0 J..lm for gold, and 0.3-1.0 J..lm
for a nickel underlayer for gold plating [149].
Tin plating has not been popular in the United States because of
tin's tendency to form "whiskers"-single crystals of tin which grow
under certain conditions after tin plating [150-152]. They can be as
long as a few millimeters and cause electrical shorts. Whisker growth is
associated with built-in stresses during plating. It can be minimized or
suppressed by heat treatment (annealing), alloying the tin with lead, nickel,
or other materials, reftowing the tin after plating, or using a solder mask
to cover unbonded areas. Tin also has a tendency to oxidize and has to
be stored and shipped in a nitrogen ambient.
11-226 CHIP-TO-PACKAGE INTERCONNECTIONS

a
(Index Perforation)

Test Pad

Support Ring

~--4- Inner Lead

"!::!-- - i - Pauern

Outer Lead

Tie Bar

b
PLATING 35••
COPPER
ADHESIVE

125(75),...
BASE FILM

Figure 8-80. Tape Layout (a) and Cross Section (b) of Three-Layer Tape.
(Courtesy of Shindo Co. Ltd.)

8.5.2.4 Bumped Tape

Most TAB applications use bumped chips and planar (unbumped)


tape. However, there have been some applications over the years where
bumped tape has been preferred. One of these is the TapePak which is
described later in the chapter. Figure 8-86 shows the basic structure of
(a) a bumped chip and (b) a bumped tape. A big advantage of bumped
tape is that it can be used with standard wafers (chips) just as in wirebond-
ing. Single-layer bumped tape is the easiest to manufacture and has been
used in watches, sensors, and car radios [153]. An example of a single-
.
' ::'

I
~
~
~~
~
..
,
!IOOwr;o
~
;
;

i "i
§3
." ~
Ii 1

f:
::2=
;;;t I
! t'i "i>; li
! ~~ ... 1:;r ,--.

!:
"Ij

~
Uh~ Iii '2
..J
3
.& ci
U
0
"0
c
:E
Vl
.....0
>-.
'"
Q)
t:
::l
0
~
==
.~
Q
'"
~

~
Q.
~
E-o
e
e•
I/)
~

==
Q
Q
.
=-
N
~
.:
E-o
,..;
QC)

aO
...
Q,j

0 0 =
Oil
~

227
IV

N
N
co
o SPECIFICATION
POLYIMlDE
ADHESIVE
COPPER
PLATING
D(TML OF TEST PADS
(X20)

THIS FIGURE IS SHOWN BY


DETAIL OF INt£R LEADS (XIDO) COPPER SIDE OVER POL YlMIDE
~ (o,56)
.(', COPPER
~ fif 0.02 ~(O.D5) S S" -e::pDL YIMIDE
~I '" ¢: OJ
""""IER

Figure 8-82. Super Slim TAB LCD Tape Design. (Courtesy of Shindo Co. Ltd.)
8.5 TAPE AUTOMATED BONDING 11-229

• •

b
I
• • • •

Figure 8-83. Low-I/O Tab. (a) A 28-1ead three-layer tape (courtesy of Shindo Denshi
Ltd.); (b) a 16-lead one-layer tape (courtesy of 3M Electronic Products Division); (c) a
40-lead three-layer tape (courtesy of Mesa Technology).
11-230 CHIP-TO-PACKAGE INTERCONNECTIONS

•••••

a •••••
• • • • • •

b • • • • • •

Figure 8-84. High-I/O TAB. (a) A 308-lead three-layer tape (courtesy of Shindo
Denshi, Ltd.); (b) a 328-lead two-layer tape (courtesy of 3M Electronic Products Division);
(c) a 204-lead three-layer tape (courtesy of Mesa Technology).
8.5 TAPE AUTOMATED BONDING 11-231

Figure 8-85. Examples of TAB Tape Formats. (Courtesy of 3M Electronic Products


Division and Shindo Denshi, Ltd.)

layer bumped tape is shown in Figure 8-87. Attempts to use three-layer


bumped tapes have not generally been commercially successful except
for Matsushita's "transferred bumped tape" process which is discussed
in the next section. One reason was that it was difficult to manufacture
a three-layer BTAB polyimide-up configuration, which severely limited
surface-mount options. Masuda et al. have solved this problem and devel-
11-232 CHIP-TO-PACKAGE INTERCONNECTIONS

Bumped Chip

Bumped Tape

Figure 8·86. Basic Structure of (a) Bumped Chip and (b) Bumped Tape

oped a quad tape carrier package (QTCP) for a CPU card [154]. The
various mounting options with the new BTAB tape with polyimide-up or
polyimide-down configurations are shown in Figure 8-88.
8.5.2.5 Transferred Bump TAB
A novel approach to making bumped tape was developed by Matsu-
shita Electric in the mid-eighties [155,156]. Instead of etching the bump

Figure 8·87. Bumped One-Layer TAB Tape. (Courtesy of Mesa Technology.)


8.5 TAPE AUTOMATED BONDING 11-233

a b

~_;o_S~
c d

Figure 8-88. Surface Mount Structure of TCP. (a,b) Po1yimide up; (c,d) po1yimide
down. (From Ref. 154.)

using double-sided photolithography, the bumps are first formed on a


glass substrate by selective masking and gold plating over a conductive
layer. The bumps are then transferred to the tape ILB fingers using heat
and pressure. This is called "first bonding." Next, the bumped tape is
aligned over the die pads and inner lead bonded to the die. This is called
"second bonding" (See Fig. 8-89).
This process was well suited for consumer products such as hearing
aids and LCD TVs. For high-I/O, high-performance application-specific
integrated-circuit (ASIC) devices which require a finer pad pitch and
smaller volumes, Matsushita then developed an improved process [157]
using a 127-mm square substrate with 250,000-500,000 bumps, and a
single-point bonding process for both the first bond (bump transfer to the
leads) and the second bond (ILB to the chip). Figure 8-90 shows the new
bonding technique. Toray, a Matsushita licensee, further improved the
process by developing a straight-wall bumping technique for 80-Jlm-pitch
bumps, and a method to transfer them to either the "film side" or "copper
side" of the bonding fingers [158].
Tatsumi et al. have developed a variation of the transferred bump
process by using a wirebonder to make gold balls which are then transferred
to TAB tape [159]. The technology is said to be capable of bonding to
lead pitches of 80 Jlm or less.

8.5.2.6 Two-Conductor Tape


The electrical performance of a TAB device is closely tied to the
tape design. For high-lead-count, high-performance applications, a second
metal layer can be added to be used as a ground plane or to distribute
both power and ground to the chip [160]. This type of tape is called
11-234 CHIP-TO-PACKAGE INTERCONNECTIONS

Planar
Tape

Au Bumps on
Substrate

Bump
Trans'er

Bonded
Chip

Figure 8-89. Tape Bumping Using Bump-Transfer Process. In this process, planar
tape is modified by transferring gold bumps to the tape. (From Ref. 155.)

"ground-plane TAB tape" (GTAB), "two-metal-layer tape," "two-conduc-


tor tape," "double-level metal tape," or "double-metal-layer tape"
[148,161-165]. It reduces the inductance of the power and ground lines,
provides better impedance matching, and also reduces cross-talk. Figure
8-91 compares the noise performance of a single-level metal TAB, a
double-level metal TAB, and a double-level metal area-array TAB for a
given set of constant conditions. The area-array tape had the best noise
reduction. The standard two-metal-layer tape offered at least a 50% noise
reduction compared to the single-metal tape [166].
Motorola was the first volume producer of two-metal-layer 360-lead
8.5 TAPE AUTOMATED BONDING 11-235

film carrier
/

(a) bumps transferred to lead

(b) bumps bonded to chip pads

Figure 8-90. Principle of New TBTAB Bonding Technique. (From Ref. 156.)

emitter coupled logic (ECL) TAB gate arrays for DEC's VAX 9000
computer [167]. Other companies in the United States and Japan are also
using two-conductor tape for special applications.
The starting material for two-conductor tape can be either the three-
layer tape with an additional copper foil laminated to the polyimide with
an adhesive (making it a five-layer starting material) or it can be a two-layer
adhesiveless process starting material with an additional copper plane.
For the five-layer starting material, a typical manufacturing process
would be to punch the sprocket holes, device holes, alignment holes,
debuss holes, and so forth, chemically etch the copper circuit pattern,
form the vias with a laser, interconnect the vias by an electroless prime
11-236 CHIP-TO-PACKAGE INTERCONNECTIONS

1.2 .... SINGLE LEVEL METAL TAB


-.- OCUR.£ LEVEL METAL TAB
..... DOUBLE LEVEL METAL MEA ARRAY TAB

1.0
g
~ 0.8
~
>
i O.B

!Ii
...J
0.4
t;
~
0.2

40 60 110
NlABER OF" OCO. SWITCHED

Figure 8-91. Noise Versus Simultaneous Switch for Three TAB Configura-
tions. (From Ref. 166.)

followed by electroplated copper, and, finally, cover plate with gold or


tin. Figure 8-92 shows a cross section of a two-conductor tape.
A significant benefit of two-conductor tape is that it allows "area-
array" TAB (ATAB) design (see Fig. 8-93). ATAB improves 110 density
by overcoming the limitations of peripheral-only interconnections. The
IBM tape ball-grid array, discussed in Section 8.5.6, uses this approach.

8.5.2.7 Electrostatic Discharge Protection


In reel-to-reel high-volume production of TAB devices, the tape
may be 20-100 m long. The devices are shipped to the customer in

SIGNAL-PLANE
CIRCUIT TRACES

BLIND VIA
ED-COPPER
INTERCONNECT

Figure 8-92. Cross-Section of Two Conductor Tape (After Hoffman, Ref. SK27,
© STC, 1990)
8.5 TAPE AUTOMATED BONDING 11-237

Figure 8-93. Area TAB. A ISO-lead tape fonnat is shown arranged for area bonding
to the chip. (Courtesy of 3M Electronic Products Division.)

reel fonn after inner lead bonding with a "separator" (typically made of
polyester film) which is taken up along with the tape as each device is
processed on the production line. Researchers at Sharp Corporation [168]
found this process capable of generating electrostatic discharge (ESD)
failures due to (1) the tape producing exfoliative static charges when
separated from the separator on the supply reel, (2) very large device
surface areas contributing to the generation of frictional charges, and (3)
high probability that the tape will be subjected to discharge from operators
or equipment. Typical failures included dielectric breakdown of field-
effect transistor (FET) gate-oxide layer, molten diffusion resistance in the
vicinity of 110 pins, and field-oxide layer breakdown.
In spite of taking the nonnal anti-ESD measures, Sharp researchers
found that one device mode ESD damage persisted at a low but constant
rate. This was the "charged device model" (CDM), where ESD damage
occurs when static charges built up in a device discharge through the
equipment or human body. There were two solutions: (1) increase the
material's relative dielectric constant and (2) optimize patterning. The
first approach is not desirable for high-speed devices because a higher
dielectric constant means reduced speed. So Sharp used the second ap-
proach to solve the ESD problem. Figure 8-94a shows four TAB patterns:
pattern A has no guard ring; pattern B has the devices surrounded by a
common guard ring; pattern C has each device surrounded with a separate
guard ring which is connected to the chip substrate; and pattern D is a
version of pattern B in which the guard ring is connected to the chip
substrate. Figure 8-94b shows the potential distributions of patterns A
and D. Pattern A had 23% failures; pattern B had 13.5% failures; pattern
Chad 19.5% failures; and pattern D had 3.9% failures.
11-238 CHIP-TO-PACKAGE INTERCONNECTIONS

(a)
••
••• 148Vlc.\
••
••
0
•• •
••
••
•••

•
0
••
••
ruar4 rlDr
A B c D

(b) -; +

-
'"'
.~ 1.0
~
!... 1.0
r: r:
...
GI

0
...o
QI

a.

...
()
.~

.....'"
.~~
E
....
0 t---++-----=~ 0 t---------~~
0 o time

-
~ (sec

-
()
QI o
GI
GI - ......_ _ _ _ _ _ _ _ _- ' GI - ' - -_ _ _ _ _ _ _ _ _- '

pattern A pattern 0

Figure 8-94. (a) TAB Patterns for ESD Experiment; (b) Potential Distribution
of Patterns A and D. (After Tajima et aI, Ref. SK37, © STC 1990)

8.5.3 Wafer Bumping


Wafer bumping is the process of forming "bumps" over the bonding
pads at the wafer level or at the chip level. Typical bump materials
include gold, copper, and solder (SnlPb) singly or in combination. Other
conductive metals such as aluminum, indium, nickel, and silver, and
various alloys have also been used in different applications. The bumps
serve several purposes: They provide the right metallurgy for a variety
of chip-joining processes; they provide "standoffs" which prevent shorting
of the lead fingers to the chip edge; they protect the bond pad metallization
[usually Al or AI-alloy (Sileu)] from corrosion and contamination (see
Fig. 8-95); and they act as a deformable (ductile) buffer during gang inner
lead bonding.
8.5 TAPE AUTOMATED BONDING 11-239

Exposed
AtChlp Metallurgy

Interface
Metallurgical

Figure 8-95. Improved Chip Hermeticity with TAB Bump Structure. A poten-
tia1 corrosion-resistance advantage may accompany the full coverage of the a1uminum
bonding pad.

Gold bumps are the most common. A typical gold bumping process
(see Fig. 8-96) [169] starts with sputter-cleaning of the AI-alloy bond
pads on a passivated silicon wafer in a vacuum system. A thin layer
(about 200 nm) of Ti-W alloy is sputtered over the wafer. Alternately,
Ti or Cr is used as the bottom "adhesive" layer, and W is used as a "barrier"
layer to prevent interdiffusion between Au and AI. Other diffusion barrier
metals include Cu, Mo, Ni, Pd, and Pt. Next, a thin layer of Au is sputtered
on top of the barrier layer. It protects the barrier layer from oxidation
and also serves as the plating base for the electroplated bump. Prior to
electroplating, a liquid or dry film photoresist is applied to the wafer and
serves as the plating mask for the selective gold plating of the bonding
pads. The photoresist is baked, exposed, and developed. Plasma cleaning
is used to remove any organic residues prior to electroplating to the desired
gold thickness (typically 20-25 ~m). The final step is to "anneal" the
bumps in a nitrogen atmosphere to reduce the as-plated hardness to a
level suitable for inner lead bonding.
For low-cost TAB applications, a modified wirebonder can be used to
11-240 CHIP-TO-PACKAGE INTERCONNECTIONS

Interfllce
_IIurgy

-~ b c

d e
Figure 8-96. Gold-Bump Processing. (From Ref. 169; reprinted with permission of
Solid State Technology.)

put down gold balls directly over the aluminum pads using a conventional
thermo sonic or thermocompression ball bonder [170]. A separate coining
operation with a flat tool may be used to flatten the tail above the ball
and provide a planar surface for ILB (see Fig. 8-69). NEC has implemented
an ILB method using this technology for volume production of application-
specific packages [171].
Solder bumps are becoming more important because they can be
used both for TAB and flip-chip bonding. They are formed by evaporation
of SnlPb through a metal mask over a suitable under bump metallurgy
(UBM) such as Cr-Cu-Au, or by electroplating [172,173]. Researchers
at Toshiba Corporation have developed a method of forming solder bumps
by electroless plating and ultrasonic soldering [174]. Tanaka Denshi Ko-
gyo Co. has demonstrated a method of forming solder balls over Al pads
(with Cr-Cu-Au metallization) at the chip level using a wirebonder and
a specially formulated 2 Snl98 Pb solder wire [175].

8.5.4 Inner Lead Bonding


TAB inner lead bonding (ILB) was originally conceived as a gang
bonding step in which all the TAB leads are simultaneously bonded to
the chip electrodes using heat and pressure [thermocompression (TIC)
bonding]. The objective is to provide a strong metallurgical bond with
high strength. Another ILB method used predominantly by U.S. companies
is called single-point bonding, where each connection between the chip
pad and the tape finger is made sequentially, one at a time. In addition
to thermocompression and thermo sonic bonding, other options include
the following:
8.5 TAPE AUTOMATED BONDING 11-241

• Eutectic/solderlhot gas
• Laser
• Laser sonic

8.5.4.1 Thermocompression Bonding


For TIC bonding, the main bonding parameters are temperature,
pressure, and bonding (dwell) time. They are selected on the basis of
bond reliability and manufacturing throughput trade-offs [176]. TIC gang
bonding is usually done with a constant heat thermode or a pulse heat
thermode. The thermode can be a solid tool or a bladed tool (Fig. 8-97).
Constant heat thermode shanks are typically made of low-thermal-expan-

Solid Thermode.

Heater~~~
~/J'//J~_(lnconel, BeO,
Diamond ••. )

I
I I Ii

Pul.ed Thermode.

Four-Blade Plate
Thermode Thermode

Figure 8-97. TAB Thermodes


11-242 CHIP-TO-PACKAGE INTERCONNECTIONS

sion alloys such as iron-nickel-cobalt ("Inconel"), or stainless steeV


tungsten alloy with tips of natural or synthetic diamond (single crystal or
polycrystalline), or cubic boron nitride [177]. Pulse bonding thermodes
are usually made of molybdenum or titanium blades_ Chemically vapor
deposited (CVD) diamond is finding increasing use in TAB gang bonding
because of its thermal and wear characteristics. It is a binderless polycrys-
talline material deposited on a special ceramic base [178].
The thermode has to be carefully selected for a particular application
based on cost, bonding materials, processing temperature, leadcount, die
size, and so forth. The key attributes should include excellent thermal
conductivity, stability and wear resistance, a smooth surface (0.1-0.2
Ilm), excellent adhesion resistance to metal oxides, no bending at high
temperature, and ease of cleaning by lapping with an alumina plate.

8.5.4.2 Gang Bonding Process Flow


A typical reel-to-reel gang bonding process flow is as follows [179]:

1. The TAB tape from the input reel is separated from the spacer tape
and indexed to the bonding station.
2. The sawed (diced) wafer from the wafer cassette (magazine) is
automatically fed to the wafer holder for die pickup.
3. The wafer is positioned and expanded.
4. The pattern recognition system locates the good die on the wafer.
5. Each good die is plunged up and transferred to the bonding stage.
6. The die is aligned in X, Y, and $.
7. The bonding stage with the die moves to the bonding station.
8. The tape inner leads are inspected by the lead locator and aligned
in X, Y, and $. If the site is defective, it is skipped.
9. The bonding head makes the bond with a preprogrammed bond
force, dwell time, and impact force.
10. The bonded site is taken up on the output reel with a protective
spacer tape.
11. The bonding cycle is repeated with the next good die and the next
good tape site.
12. After a certain number of bonds (typically 100), the thermode is
automatically lapped (cleaned) according to preprogrammed instruc-
tions (for x-y motion, lapping interval, and duration).

8.5.4.3 Tape and Die Presentation Options


Depending on the equipment design, there are a number of options
available for presenting the TAB tape and the chips to the inner lead
8.5 TAPE AUTOMATED BONDING 11-243

bonder. For high-yield, high-volume runs, the chips are usually presented
in wafer form on a sawed film frame as described earlier. In some cases,
the good chips are presorted from the sawed frame and put in "waffle
packs." Most bonders are designed to accept chips in waffle packs at the
input stage. Each chip is then picked up and transferred to the bonding
station. A third option is to sort the good chips and put them in an
"embossed tape" (also called "pocket tape") similar to that used for surface-
mount components such as capacitors, resistors, and molded semiconduc-
tor devices.
In addition to the reel-to-reel option described earlier, the TAB tape
can be supplied to the inner lead bonder in strips (similar to conventional
lead frames) or as individual sites in "slide carriers." The carrier, which
allows testing and bum-in, is typically made of an antistatic material and
has a high-temperature bum-in capability up to 180°C. The slide carriers
are usually handled in magazines (coin-stack tubes).

8.5.4.4 Single-Point Bonding


Single-point bonding is the process of bonding one lead at a time
just like wirebonding (wirebonding is discussed in Section 8.4). It can be
used for both inner lead bonding and outer lead bonding. Single-point
bonding offers many advantages over gang bonding [175,180-185].

1. The bond force is much lower-passivation cracking, silicon crater-


ing and other problems are reduced or eliminated.
2. There are more process choices: thermocompression, thermosonic,
ultrasonic, laser, or laser ultrasonic.
3. Because only one lead is bonded at a time, the height of each bond
is sensed and the bond force is controlled individually, eliminating
planarity problems associated with the bonding thermode, tape,
bumps, and substrates. This becomes more critical with larger die
size and higher lIOs.
4. The bonding tool is easier to install and change, and costs much
less than a gang bond thermode.
5. The process is more flexible for small lots. The same bonding tool
can be used for bonding different devices. Setup and changeover
times are much less.
6. Defective leads can be repaired one at a time.
7. The same bonder can be used for ILB, OLB and repair.
8. Thermosonic and ultrasonic single-point TAB bonding is similar to
wirebonding with a mature infrastructure and an extensive reliabil-
ity database.
11-244 CHIP-TO-PACKAGE INTERCONNECTIONS

9_ Most bonders are programmable and allow individual bond parame-


ter controL Programs can be ported from one machine to another,
assuring repeatability.
10. Single-point bonding is more effective with substrates which are not
flat, such as ceramic and printed-circuit boards.

The bonding tools (thermodes) are usually made of ceramic, tungsten


carbide, titanium carbide, and diamond-tipped ceramic. The thermode
selection is based on cost, throughput, yield requirements, and assembly
process compatibility. The thermode should have high thermal conductiv-
ity, high resistance to wear (hardness, density, and compressive strength),
and good acoustic properties for transmitting ultrasonic energy (for ther-
mosonic or ultrasonic bonding).

8.5.4.5 Laser Bonding


Laser bonding overcomes the two most serious limitations of thermo-
compression bonding, namely the need for high temperature and high
pressure, which are known to cause damage to the device and affect its
long-term reliability [186] . Because laser bonding is a noncontact method
of providing heat energy through a finely focused beam, it can be used
both on the periphery as well as in an area-array design over the surface
of the chip. Laser bonding works well with materials which have high ab-
sorpti vi ty .
Hayward [187] studied the ILB process using a Q-switched
Nd : YAG laser operating at 1.064 /lm developed by the Microelectronics
and Computer Technology Corporation (MCC) and built by Electro Scien-
tific Industries (ESI). Figure 8-98 shows the cross section of the fixture
holding the device and tape during bonding. The device and tape are held
in place by vacuum. The device had 152 gold bumps, 100/lm square and
25 /lm high on a 150-200-/lm pitch. The tape was a three-layer, tin plated
tape with 0.6 /lm of tin which had been reflowed after plating. The inner

Tape Plate

Figure 8·98. Laser Bonder ILB TapelDie Fixture


8.5 TAPE AUTOMATED BONDING 11-245

leads were 76 11m wide and 35 11m thick. The three principal parameters
studied were laser power, pulse time, and bond height (distance the die
pedestal was driven above the point of first contact between the bump
and tape). The baseline parameters were set at 37 W (power), 5 ms (pulse
time), and 25 11m (height). A comparison with TIC gang bonding was
made using an identical lot of 300 devices bonded with a Shinkawa IL-20
bonder set at an anvil temperature of 250°C, a thermode temperature of
500°C, and a force of 16 kgf. Ten units from each lot were pull tested
in the as-bonded condition. The remaining were split between temperature-
cycle testing (condition C, 1000 cycles) and high-temperature storage
(I50°C, 1000 h). Table 8-6 shows the results. The laser bonded units
have lower average bond strength but a tighter distribution than the TIC
bonded units and show a slower rate of degradation over time. Figure 8-
99 shows SEM photographs of laser bonded leads.
Spletter [188] studied Au-to-Au laser ILB with a 328-lead, 3-layer
TAB tape with 45-llm-wide, 35-llm-thick Cu leads on a lOO-llm pitch.
The leads were plated with 50 11m of gold and were bonded to a 0.63-
cm2 Si chip with 50-11m2 gold bumps, 22 11m high. A frequency-doubled
pulsed Nd : YAG laser was used to increase the absorptivity from 2-5%
at 1.064 11m to 25-50% at 0.533 11m. The ILB setup was similar to that
shown in Figure 8-98. Thirty devices were bonded and subjected to high-
temperature storage (I50°C) for 1000 h, temperature cycling (-65°C to
150°C) for 1000 cycles, and liquid-to-liquid thermal shock (-55°C to
125°C) for 1000 cycles. The average pull strength was 35 g with a
standard deviation of 2.5 g. There was no degradation of good bonds
after environmental testing.
Zakel et al. studied laser ILB of Sn, Ni-Sn, and Au tape metalliz-

Table 8-6. Comparison of Laser- and Thermocompression (T/C}-Bonding


Test Results

Mean Standard Lift


Bond Value Deviation Failure
~ 19l 19l ~

As-bonded
Laser 33.6 3.9 0.0
TIC 50.3 7.3 0.9
1000 Temperature Cycles
Laser 25.8 3.8 1.6
TIC 38.8 9.2 8.9
1000 h Temperature Storage
Laser 20.1 3.4 1.9
TIC 39.8 7.0 10.9
11-246 CHIP-TO-PACKAGE INTERCONNECTIONS

Figure 8-99. Laser Bonded Inner Leads

ations to gold and gold-tin bumps [189]. Laser soldering metallurgy was
found to be different and more critical to long-term reliability than that
of TIC gang ILB even if identical tape and bump materials are used.
Accumulation of eutectic Au-Sn (80%/20%) in the bonded interface
results in strong degradation due to Kirkendall pore formation in the
ternary Cu-Sn-Au system. The Ni barrier inhibits this effect. However,
thermal aging formed brittle intermetallics of Ni, Sn, and Au. Laser ILB
of gold plated tape to Au-Sn solder bumps showed minimal degradation
after thennal aging due to the fonnation of an intennetallic compound
with high stability.
Azdasht et al. [190] have developed a laser bonder with a glass fiber
for making a windowless flip-TAB connection on a flexible substrate.
The tip of the fiber serves to press down on the part.

8.5.4.6 Factors Affecting ILB


There are many variables which can affect ILB bond quality:

• Tape material hardness


• Tape plating thickness and uniformity
• Tape surface cleanliness, roughness, and oxidation level
• Bump hardness
• Bump flatness
• Bump design (size with respect to pad structure)
• Under bump metallurgy
• Bump adhesion and barrier layer integrity
• Thermode design
• Thermode mass
8.5 TAPE AUTOMATED BONDING 11-247

• Thermode flatness at room temperature and bond temperature


• Thermode thermal conductivity and temperature uniformity
• Thermode wear resistance and oxidation resistance
• Thermode adhesion to plating materials (especially tin)
• Tape and die alignment
• Bond interface planarity
• Interface temperature, pressure, ultrasonicllaser energy (if used), and
dwell time control
• Stage temperature
• Descent velocity and profile

Thermocompression bonding is a complex process that requires an under-


standing of a wide variety of mechanical and thermal interactions [191].
Several studies have been made to understand these interactions [151,192-
195]. Figures 8-100 and 8-101 show the dependence of ILB bond strength
on temperature and dwell time based on experimental work by Kawanobe
et al. [151] and Spenser [194], respectively. Surface contaminants also

100 r---,---,--,----r------r--,----,

80 IIIBump Bond Force


Time 1 Sec

80
.",. .... - ........
/A
,."
,,~""e
,
AU'Sn
,r ,A
,e' ,r

,
I

,,
I IAU-AU
I
I
I
/

,,
Ie

,,
20 I

Bonding Tempendure rC)

Figure 8-100. Dependency of Bond Strength on ILB Temperature. Bond strength


versus bond temperature for several metallurgies. (From Ref. 151.)
11-248 CHIP-TO-PACKAGE INTERCONNECTIONS

E"=0.2

.!
J:
8

S8SK 57&K 493K


iii B

!"
..i
~ 4

II:
2

Dwell Time (s)

10-1

-
I
.!!.

10-2

10-3L-__- L__~____~__~____~__~__~
1.2 1.& 2.0 2.4
l000/T (1("""1)

Figure 8-101. Interdependence of Dwell Time, Temperature, and Deformation


on Bond Strength. Arhenius relationship between the dwell time of the bond and the
interface temperature at several deformation levels. Data from plots of bond strength
versus dwell time, at constant strain as shown in (a), were used to study the thermally
activated process responsible for the bonding process (b). Specimens of gold plated copper
were tested in shear. (From Ref. 194.)

play a large part in Au-to-Au TIC bonding as shown in Figure 8-102


from the work of Jellison [196].
The success of the ILB process also depends on the proper choice
of interface metallurgy (IFM). A wide variety of IFM structures have
been used in the past, as shown in Table 8-7. Today, Ti-W-Au is widely
8.5 TAPE AUTOMATED BONDING 11-249

1.4,..-------------------,
1.2

Z 1.0
~

io.s
!.. 0.8
:
ii 0.4
0.2

Weld Interface Temp, 'C

Figure 8-102. Effect of Contamination on Thennocompression Bonding. Effect


of organic contamination on gold-to-gold bond strength. (From Ref. 196.)

used as the under bump metallization (UBM), along with gold-solder


bumps and tin-gold plated tape.
It is well known that gang TIC bonding of high-lead-count devices
at high temperatures (>400°C) and pressures can cause mechanical and
thermal stresses to the UBM and affect the long-term reliability of the
device [151,192,197-199].

Table 8-7. Metallurgies Used for TAB-ILB

Bum~ Structure Ta~e Structure


Adhesion Barrier Bump Plating
Com~any Metal Metal Metal Metal Lead

GE Cr Cu Au Sn Cu
CII Honeywell-Bull Au Sn Cu
Sharp Cr Cu Au Sn Cu
Honeywell Ti Pd Au Sn Cu
Fairchild Au Sn Cu
Signetics Ti Pt Au Sn Cu
Nippon Electric Ti Pt Au Au Cu
National Semiconductor Cr Cu Au-Cu Au (Thin) Cu
BTL Ti Pt Au Au Cu
Philips AI Ni Solder-Cu Au-Ni Cu
Siemens Au Sn Cu
RCA Ti Pt Au Au Cu
Motorola Au Sn Cu
TI Au Sn Cu

Source: From Ref. 151.


11-250 CHIP-TO-PACKAGE INTERCONNECTIONS

8.5.4.7 ILB Evaluation


As in wirebonding, bond evaluation methods for ILB can be classified
as destructive or nondestructive. The two most common are visual inspec-
tion (opticallSEM) and the pull test. For bumped-chip bonding, the destruc-
tive pull test is preferred for evaluating bond strength and quality. The
die is held down by vacuum; a wire hook is placed under the inner lead
(to be tested) midway between the die edge and the polyimide opening
[200]. The hook is raised at a controlled rate and the force necessary to
break the bond or the inner lead is recorded. The bump is then removed
and each bond is inspected for passivation or barrier layer cracks and
other damage such as silicon cratering [196]. Typical failure modes for
Au-to-Au T/C bonding are as follows:
• Lead break-at the bump or in mid-span
• Lead lift
• Bump lift
• Bump fracture
• Silicon cratering

In addition to opticallSEM microscopy, other nondestructive methods


commonly used to check the quality of bonded parts include scanning
acoustic microscopy, scanning beam x-ray laminography, conventional
x-ray inspection, acoustic emission detection, and laser/infrared inspec-
tion [201,202].

8.5.5 Outer Lead Bonding


Outer lead bonding (OLB) is the process of connecting the outer
leads of the tape to the package, lead frame, substrate, or card. Like ILB,
OLB can be a gang or a single-point process. There are many OLB
processes available. The choice depends on the application and includes
the following:
• Thermocompression
• Thermosonic
• Hot bar
• Laser
• Ultrasonic
• Lasersonic
• Infrared
• Hot gas
• Vapor Phase
• Anisotropic conductive film
• Mechanical (DTAB, connector, socket, elastomeric clamp, etc.)
8.5 TAPE AUTOMATED BONDING 11-251

After ILB, the chips on tape usually undergo additional processing


such as encapsulation, test, and bum-in. These steps can be performed in
reel form, strips, or individual sites mounted in slide carriers
[201,203,204]. The encapsulation process depends on the end-use applica-
tion and can include chip-size coating, potting, or transfer molding [205-
207] (see Fig. 8-103). For transfer molding, the ILB device is usually
bonded to a lead frame prior to molding (Fig. 8-1 03c). Current encapsula-
tion processes and materials are covered in Chapter 10, "Plastic Packag-
ing," Chapter 14, "Package Sealing and Encapsulation," and Chapter 17,
"Printed-Circuit Board Packaging."
Much of the previous discussion regarding ILB processes (Section
8.5.4) is also applicable to OLB processes with some modifications. Chap-
ter 16 ("Package-to-Board Interconnections") includes details of surface-
mount joining processes such as vapor phase, infrared retlow, laser retlow,
and pressure connections. This section provides a brief overview of hot
bar OLB, anisotropic conductive film OLB, demountable TAB, and some
recent experimental results on TIC-Laser OLB.

a Face-Coating

b Full Encapsulation

Figure 8-103. TAB Encapsulation. (a) face coating; (b) full encapsulation; (c) trans-
fer molding.
11-252 CHIP-TO-PACKAGE INTERCONNECTIONS

8.5.5.1 Hot Bar OLB

Pulsed hot bar soldering has been in use for over 20 years [208].
In the OLB process, the hot bar, or thennode, mechanically presses the
leads onto the bonding pads. The thennode blades are resistively heated
to provide a preprogrammed temperature profile optimized for a specific
application (see Fig. 8-104) [209]. The thennal cycle begins with the
blades in contact with the leads. The programmable parameters include
thennode idle temperature, ramp rates, flux activation time, flux activation
temperature, bonding time, bonding temperature, tool-up temperature, and
spindle-up time.
The hot bar process sequence depends on the application but can
include the following five steps [210]:

1. Excise and fonn


2. Die attach
3. Fluxing
4. Placement and alignment
5. Solder reflow

6
...........
/s
9
,,
~
,
"............
9
Temp.

4 7 10 12
t. time

User programmable parameters for the bonding cycle:


1. Idle telJ1)8rature 9. ThIrd temperature level
2. Rrst telJ1)8rature slope (optional)
3. Arst temperatureltvel 10. 'third temperature time
4. First temperature time (optional)
5. Second temperature 11. Thennode lift
slope temperature
6. Second temperature 12. Spindle lift time
Itvel
7. Second temperature time
8. Third temperature slope
(optional)

Figure 8-104. Bonding Profile and Parameters for Hot Bar OLB. (From Ref.
209.)
8.5 TAPE AUTOMATED BONDING 11-253

Figure 8·105. Sample Lead Form Configuration. (Courtesy of Intel Corporation.)

The excise and form operation includes cutting the leads from the tape
support and forming them into a gullwing or modified gull wing shape
according to specifications. Excising (cutting out) the component occurs
just before fluxing and placement. A "keeper bar" helps in maintaining
lead coplanarity and spacing during subsequent operations. (The keeper
bar is a narrow strip of polyimide which remains in place on each row
of leads after excise). Figure 8-105 shows a recommended lead form
configuration for Intel's Pentium© processor, and Table 8-8 lists the
important lead form dimensions.
Kleiner [211] has studied the effects of various parameters such as

Table 8·8. Key Lead Form Dimensions

Controlled Dimension Recommended Range

"Standoff' or "set back" of die above die pad 0.025-0.095 mm


Lead foot length minimum 0.90 ± 0.05 mm
Lead foot angle 0°
Keeper bar toe angle (minimum) 20°_35°
Keeper bar width 0.5 ± 0.1 mm
Toe radius 0.2 mm (min)
Heel radius 0.2 mm (min)
Lead shoulder length 0.3 ± 0.05 mm
Lead shoulder radius 0.2 mm (min)

Source: From Intel Corporation.


11-254 CHIP-TO-PACKAGE INTERCONNECTIONS

bend radius, material type and thickness, plating type and thickness, draft
angle, punch angle, force and speed on various lead form geometries, and
developed tooling principles and requirements for formed leads which
are parallel to the chip surface, coplanar and free of cracks at the bend radii.
If a die attach is needed, the lead form design should allow sufficient
clearance between the bottom of the silicon die (chip) and the die attach
pad. The typical die attach material is a silver-filled thermoset polymer.
The choice of flux and its method of application depend on the
interface materials. For the Pentium TCP, Intel recommends immersing
the leads in a rosin, mildly activated, halide-free, no-residue flux. The
entire surface of the foot, top and bottom, up to the top of the heel radius
should be immersed. The specific gravity of the flux, its solids content,
and the surface insulation resistance between adjacent leads should be
closely controlled to 0.80-0.81, 1-3%, and >109 n, respectively. Solder
reflow should be completed within 30 min of fluxing. A maximum ther-
mode temperature of 275°C ± 2°C with a dwell time of 5 s was shown
to yield acceptable fillets. Force should be kept at ~1.36 kg (3Ib) per blade.
The placement accuracy requirements for TAB components are much
more stringent than for conventional surface-mount devices. The factors
which affect accuracy are locational tolerances of the leads, the lands,
and fiducials on the substrates [209]. A good pick and place system with
a pattern recognition system (PRS) should be capable of ± 0.01 mm
repeatability (lead to land) in x and y and 3° of rotation [210].
The reflow process can be affected by a number of parameters such as

• Blade design
• Thermode compliancy
• Nonuniform temperature and flatness along the length of the blades
• Thermode expansion
• Substrate warpage
• Substrate support
• Bonding force
• Dwell time
• Thermode temperature

Equipment manufacturers are meeting the new challenges of sub-O.3-mm


pitch TAB by developing "self-planarizing" thermodes with spherically
compliant suspensions and better alignment and vision capabilities
[209,212-215]. Figure 8-106 shows a schematic of a self-planarizing
thermode, a thermode assembly, a titanium thermode blade, and a cleaning
station [216].
8.5 TAPE AUTOMATED BONDING 11-255

Center of Spherical Rotation

c d

Ceramic Tile

.Splash Guard

Figure 8·106. Hot Bar Process Components. (a) Self-planarizing thermode; (b)
thermode assembly; (c) titanium blade, and (d) cleaning station. (Courtesy of Universal
Instruments Corp.)

8.5.5.2 Anisotropic Conductive Film


In some applications such as LCDs, soldering is difficult due to the
temperature limitations of piece parts to be bonded. Anisotropic conductive
film (ACF) is an alternate method used in LCDs to connect the TAB
outer leads and the ITO (indium-tin-oxide) electrodes on the glass sub-
11-256 CHIP-TO-PACKAGE INTERCONNECTIONS

strate. The ACF consists of fine conductive particles uniformly scattered


in an epoxy binder. ACFs can also be made of plastic cores plated with
conductive materials. They conduct only in the vertical direction under
pressure. ACFs are also known as "Z-axis conductive adhesives." Typical
ACF conductors include nickel, gold, and carbon. Particle size and distri-
bution in the film are critical to prevent shorting between particles in a
horizontal direction. For TAB OLB of LCDs, heat and pressure are used
to bond the outer leads to the ITO electrodes.
Casio Computer Company has developed a "microconnector" (MC)
using ACF which allows OLB of 80-~m-pitch LCDs [217] . Conductive
particles with plastic cores are plated with Ni-Au, and smaller particles
coated with very thin (1000 Aor less) insulating film are applied to the
conductive particles. The latter are then combined with an epoxy binder
and TIC bonded in the same way as conventional ACF. During bonding,
the insulating film in the direction of pressure is broken to achieve conduc-
tivity. The binder can be either thermoplastic (for rework) or thermoset.
Figure 8-107 shows a schematic of the conductive particles and the micro-
connector OLB method for 80-~m pitch.

8.5.5.3 IlB and OlB Pitch and lead-Count Trends


Figures 8-108 and 8-109 show the decreasing trend in ILB and OLB
pitches for the last 5 years, and the increase in lead count since 1986,

O
Conductive particle

+ 00

o
l'

;;
0

0
. I
lnsu ;tang partlc e

___Mix""
_ _ _ __

Figure 8-107. (a) MC Conductive Particles; (b) 80-f.Lm-pitch OLB


8.5 TAPE AUTOMATED BONDING 11-257

400

Oute., L ..... d P.1.tch

300
"E 300 ---~
,,
=l
v " , 250
.c
m-----€)----~
,
0
iJ
.oj
" , 200
110 200
'"
•
'0
G
..J
Inner Lead P.1.tch

100

1990 '91 '\12 '93 "lis 'D&


Year

Figure 8-108. Decreasing Trends of TAB Lead Pitch

respectively [136,142]. The minimum ILB pitch declined from 80 J.1m in


1990-1991 to 70 J.1m in 1992-1993 and 60 J.1m in 1994-1995. It is
expected to reach 50 J.1m in 1996. The minimum OLB pitch has dropped
from 400 J.1m in 1990 to 300 J.1m in 1991-1992 and 250 J.1m in 1993-1995.
It is expected to reach 200 J.1m in 1996. The maximum lead count has
increased from about 250 leads in 1986 to over 700 leads in 1995.

8.5.5.4 Effect of High I/O and Pitch on ILB and OLB

When designing a TAB package, the outer lead pitch affects the
package size and ease of assembly. The smaller the OLB pitch, the more
difficult it is to assemble the device to the next level. The standard JEDEC
OLB pitches are 0.65, 0.5, 0.4, and 0.3 mm [166].
For tin plated tape, as the pitch becomes smaller and smaller, it is
important to reduce the tin thickness to ensure that excess tin does not
produce shorts between adjacent leads. At the same time, tin thickness
plays a major role in OLB. If there is not enough tin, the wettability of
11-258 CHIP-TO-PACKAGE INTERCONNECTIONS

",
II:
G

.,
U

j
•z••

Year

Figure 8-109. Maximum Lead-Count Trends for TCPs

the outer leads will be degraded. This will lead to weaker outer lead
bonds [218].
Saito [149] studied the relationship between the solder plating thick-
ness and bondability of outer leads using pulsed hot bar OLB. The accept-
able thickness range for 0.3-mm OLB pitch was 3-10 !lm of solder, with
5-10 !lm giving the best results. For smaller pitches (0.25 and 0.15 mm),
the acceptable range was 3-7 11m with 5-7 11m being the optimum.

8.5.6 TAB Package Applications

8.5.6.1 Tape Ball-Grid Array

The tape ball-grid array (TBGA) is a ball-grid array TAB package


developed by IBM which overcomes the OLB limitations of conventional
peripheral TAB packages [219]. Figure 8-110 shows the TBGA schematic.
It uses area-array TAB (ATAB) with a standard ground plane and has
the following performance advantages over conventional TAB and pin-
in-hole packages at the card level:

Lower lead inductance


• Lower power-supply inductance
• Lower line-to-line capacitance
• Lower average signal delay
8.5 TAPE AUTOMATED BONDING 11-259

CHIP
HEATS INK

THERMAL ADHESIVE

STIFFENER

CU I UPILEX I CU

Figure 8·110. TBGA Cross Section. (From Ref. 219.)

The TBGA is constructed using an adhesiveless two-metal-layer


tape. One side is used for signals and the other for power and ground
interconnections. Solder balls joined to the tape interconnect the completed
package to the next assembly level (PCB/card). A "stiffener plate" main-
tains planarity of both the solder bumps and the tape and serves to minimize
thermal stresses during temperature cycling because its TCE matches that
of the carrier/card.
The ILB process can be a conventional peripheral one or a modified
C4 flip-chip process known as solder attach tape technology (SATT).
SATT allows bonding over active circuitry and uses less bond pad area
than T/C ILB and wirebonding, thus reducing the die size and improving
the wafer productivity.
After ILB, an epoxy encapsulant is used to provide mechanical
protection and enhance the fatigue life performance of the inner lead bonds.
A standard infrared reflow process using 63/37 SnlPb solder paste
can be used for card assembly. The TBGA height is 1.4 mm. It weighs
less than 5 g. The TBGA is registered as a JEDEC standard S-XXGA-
NITBGA. Figure 8-111 shows a photograph of the TBGA.

8.5.6.2 TapePak®

The TapePak (Fig. 8-112) was developed by National Semiconductor


around 1986. It is a testable, plastic molded quad surface mount package
with an OLB pitch of 0.51 mm for a lead-count range of 20-124, and
0.38 mm for higher lead counts. It uses a single-layer bumped copper tape,
70 l1m thick. The bond pads are "capped" with a thin-film metallization at
the wafer level. Thermocompression bonding is used for ILB. After ILB,
the tape is cut into strips and the devices are molded with epoxy novolac.
Next, the dam bars are removed, the devices are singulated and solder
11-260 CHIP·TO·PACKAGE INTERCONNECTIONS

Figure 8-111. A 736-Lead TBGA. (Courtesy of IBM Microelectronics.)

plated. The package is designed with molded-in test pads to allow test
and bum-in. A trim-and-fonn operation is done at the pick-and-place
station prior to board assembly.
Because of its compact size, the TapePak offers significant improve-
ment in electrical characteristics. For example, a 40-lead, 7.6-mm square
TapePak has a worst-case lead length of 2.54 mm, a lead resistance of
2.4 mn, inductance of 1.2 nH, and lead-to-Iead capacitance of 0.2 pF [220].

8.5.6.3 Pentium® rep


The increasing importance of "small fonn-factor" packages for what
is called the "mobile revolution" has led to TAB's resurgence in the
United States. The world's largest semiconductor company, Intel, has
developed a TCP version of the Pentium® for mobile applications such
as laptop, notebook and palm top computers, and related products [210] .
Figure 8-113 shows the topside and bottomside of the Pentium in a slide
carrier. The chip is inner lead bonded to a JEDEC-style UO-018, 48-mm
three-layer one-ounce copper (35.56-llm-thick) tape. The inner lead and
outer lead bonding areas are gold plated over a nickel flash. The inner
lead bonding (ILB) is perfonned with a thennosonic single point bonder.
8.5 TAPE AUTOMATED BONDING 11-261

Figure 8-112. TapePak®. (Courtesy of National Semiconductor.)

After ILB, the chip and ILB area is encapsulated with a high-temperature
thermoset polymer coating. It covers the top surface of the device, the
sides, and the ILB area to the polyimide carrier ring (see Fig. 8-114).
The backside of the chip is left bare for backside bias to the printed-
circuit board (PCB).
The outer lead pitch is either 0.25 mm or 0.2 mm, depending on
the device. The OLB lead width is 0.10 mm. The test pads are 0.5 mm
square on 0.40-mm pitch outboard of the OLB window. The devices are
shipped in individual plastic slide carriers packed in coin-stack tubes.
Each component is 0.615 mm thick. After excising, lead forming, and
mounting to the PCB, the total height of the component above the PCB
is less than 0.75 mm. The package body is either 24 mm or 20 mm
square after excise, depending on the component. The 320-lead 0.25-mm
component in a 24-mm body size weighs a maximum of 0.5 g. In compari-
son, a 296-lead multilayer PQFP weighs 9.45 g.
For board mounting, Intel recommends either a hot bar, hot gas, or
laser retlow process after all other board components have been mounted
and cleaned. Figure 8-115 shows the recommended land pattern design.
The Pentium requires backside electrical and thermal contact. So it
is necessary to provide a 3.75 mm ± 0.025 mm die attach pad. Intel
recommends a silver-filled thermoset polymer with a cure profile of 6
min above 130°C.
The thermal resistance of the TCP package (0jc) is 0.8°CIW to
11-262 CHIP-TO-PACKAGE INTERCONNECTIONS

1-----------,,·....·,·-----------11
~~~--------------~~~~
L-_..J.......J.._ _ _ _ _ _ _ _ _ _ _ _ _ _ -'-I.....II_---il~· ..
272588-'
SHOWN: BO".OM OF TAPE
eo".OM OF DIE
TOP OF CARRIER

Figure 8-113. TCP Site in Carrier. (a) Bottom view of die; (b) top view of die.
(Courtesy of Intel Corporation.)

r
FRONT VIEW

(EHCU)

(DIE)

DETAIL SECTION

Figure 8-114. TCP Site (Cross-Sectional Detail). (Courtesy of Intel Corporation.)


8.5 TAPE AUTOMATED BONDING 11-263

LAND 'ATTtIIN TtRIlINAI. DIIiENSION


1---- 21.5 - - - - - I

o o

o
WIDTH
272588-8
VIAS IN DIE ATTACH PAD NOT SHOWN. SEE THERMAL VIA SECTION FOR RECOMMENDATIONS.

Figure 8-115. Land Pattern for Intel 24-mm Body Size TCP. Hot Bar Reftow
Process. (Courtesy of Intel Corporation.)

2°CfW, depending on the product. PCB enhancements such as thermal


vias with or without low-profile heat sinks brings the thermal performance
in line with mobile computing requirements which do not have forced
convection cooling. Figure 8-116 shows a schematic of heat transfer
through the PCB.

8.5.6.4 ETA Supercomputer


It is instructive to review ETA's use of TAB for its supercomputer.
ETA Systems, Inc. (which is no longer in business) was a subsidiary of

Encapsulant Tape

Thennal

Thermal and Electrical


Conductive Adhesive
Heat Sink

Figure 8-116. Heat Transfer Through the PCB. (Courtesy of Intel Corporation.)
11-264 CHIP-TO-PACKAGE INTERCONNECTIONS

Control Data Corporation in the 1980s. Its mission was to build "the
world's fastest supercomputer" [221].
ETA chose to go with TAB devices mounted on a single central
processing board. Each device was a 284-pin TAB ceramic quad flat pack,
with 20,000 CMOS gate arrays. The TAB devices were supplied by
Honeywell. The chips were solder bumped with 95% Pb-5% Sn. Each
solder bump was 100 flm in diameter. The underbump metallurgy (UBM)
was TiW -Cu-Ni. The bond pads were at a 254-flm pitch, arranged in
two rows, at an effective 127 -flm ILB pitch (see Fig. 8-117). A polyimide
"through-hole" design allowed each finger to be suspended over a hole
in the polyimide base corresponding to each individual solder bump. After
melting, the solder was constricted within the hole and reflowed to the
precise area on the lead [57].
Honeywell's solder bump reflow TAB outperformed its gold bump
thermocompression bonding process in three areas: (1) lower bonding
force (less than 1 kg for a 300-bump device versus about 11 kg for
gold); (2) inherent planarity compensation-a lOO-flm-high solder bump
typically was 50 flm after bonding, thus allowing for a 50-flm variation
in planarity versus about 10 flm for gold, and (3) area-array design capabil-
ity-the solder bumps could be put over the whole chip surface, unlike
gold bumps which were limited to the periphery.

Figure 8-117. Double-Row Solder Bump TAB. (Courtesy of Honeywell.)


8.6 PRESSURE CONNECTS 11-265

8.5.7 Future Trends


Tape carrier packages (TCPs) will continue to be the dominant
interconnect medium for the LCD industry for the immediate future. LCDs
accounted for the largest share of tape carriers in 1994 [222]. The next
largest application was smart cards, mostly in Europe, where the tape is
of the three-layer variety, and the chips are wirebonded. The largest
application in the United States today is for inkjet printers, where the
tape is of the two-layer type. The connection is by gang bonding or single-
point bonding. All three of these applications will continue to thrive as
the demand for notebook computers, smart cards, and printers grows
every year.
Intel's decision to provide Pentiums in TCPs may serve as the catalyst
for a significant broadening of the worldwide TCP infrastructure. The
next generation of microprocessors will require more 1I0s and finer ILB
and OLB pitches. Pin counts will continue to grow with microprocessor
bit size. ASICs will also act as TCP drivers with multimillion-gate sub-
OA-l1m CMOS five-level metal processes becoming common and forcing
the need for 1200+ 1I0s per chip, which can be met only by TCP and!
or flip-chip designs.
TAB's inherent advantage of being light, thin, short, and small (the
Japanese call it "Kei-haku-tan-sho"), and its flexibility will enable it to
maintain its edge in a wide variety of applications such as watches,
calculators, cameras, sensors, and hearing aids.
The proliferation oflower-cost bumping processes and materials can
expand TAB's application base into new products such as PC cards, smart
cards, and multichip modules. This could lead to a strong TAB resurgence
in the United States.

8.6 PRESSURE CONNECTS


Pressure connects are not actually bonded connections but maintain
contact only by an external, continuously applied force (such as a metal
spring or elastomer retainer). Several types have been described. Citizen
Watch [223], for example, joins Au bumped chips onto Au bumped
substrates using conductive rubber contacts embedded in a polyimide
carrier (Fig. 8-118).
Typical loading is 2 g per pad to provide adequate deformation of
the rubber for needed contact on all pads; thus, a 300-pad chip requires
600 g continuous loading. The conductive rubber is not a "good" conduc-
tor: A 200 11m x 200 11m pad has 35 mQ resistance. Thus, it has been
applied in special applications (such as liquid crystal display panels) that
can tolerate the high resistivity.
Significantly lower resistances per contact, of the order of 10 mQ,
have been achieved by using an all-metal system. However, this requires
11-266 CHIP-TO-PACKAGE INTERCONNECTIONS

! ! ! !
Au Bump

Chip
Carbon
Filled
Silicone
-===::::;
Rubber
Subtltnde

111 1
Figure 8-118. Pressure Connect (Citizen Watch). (From Ref. 223.)

a "flexible" substrate to which the chip is joined. Tektronics [224] has


joined GaAs chips to a flexible circuit, as shown in Figure 8-119. The
bumps are plated on the thin-film flexible circuit. Two hundred pads per
chip have been fabricated with 50-llm pads on lOO-llm centers. Silicon
Connection has a very similar approach called "SICONS" [223]. Applica-
tions include ROMS in ATARI computer games.
Although such applications are limited at present, there are several
reasons for the increased research in these areas. This type of contact
eliminates the chip-size restrictions of the C4 type of full bonded connec-
tion; there is no fatigue mechanism. However, this is replaced by the
concern of a sliding or point contact developing high resistance by debris
or oxidation/corrosion. The second reason for research activity is the ease
of assembly and reworking-no connections need to be "broken" or
unbonded in order to take the package part.

8.7 ADHESIVE BONDING


A number of companies are using adhesives in novel ways to provide
flip-chip options other than solder reflow and pressure connects. The
driving forces are typically cost reduction, lower process temperatures,
or fatigue elimination. Although still in their infancy, the following three
approaches are worthy of note. All require precision placement of the
chip, as no self-alignment is involved.

8.7.1 Bumps and Nonconducting Resins


Matsushita [225] has reported on a gold bump bonding process for
flip chips. Chip-to-board connection is made by using pressure to force
gold bumps on the chip into direct contact with gold pads on the circuit
8.7 ADHESIVE BONDING 11-267

Circuit Board

Figure 8-119. GaAs Chips Joined to a Flexible Circuit. (a) Ie flex circuit elastomer
pressure pad sandwich cross section; (b) 110 pressure bumps establish firm electrical
contact. (From Ref. 224.)

board through a layer of acrylic resin. Using light to cure the polymer,
an adhesive bond is made between the chip and board which keeps the
mechanical connection under compression. Experimental assemblies have
been fabricated down to a 10-Jlm pitch. A pitch of 100 Jlm has been used
on thermal print heads, LED array sensor, and a memory card.
A modification of this approach has been described by TOSHIBA
[226] for chip-on-glass applications in a video camera. The modification
involves having tin-lead lands on the substrate, gold bumps on the chip,
11-268 CHIP-TO-PACKAGE INTERCONNECTIONS

and supplying thermal energy below the eutectic temperature of 183°C


to allow diffusion between the two metal contacts. Again, acrylic resin is
used in the gap between the chip and substrate to strengthen the assembly.

8.7.2 Anisotropic Conductive Adhesives


Anisotropic conductive materials display unidirectional conductivity
[227]. For flip-chip bonding, conductive particles are suspended in a
dielectric medium so that they do not come into physical contact. Conduc-
tivity, in the z direction perpendicular to the mating surfaces, only occurs
when protruding conductor pads on the chip or substrate (or both) are
squeezed against the z-axis film. The electrically isolated particles simply
make contact with the opposing conductor pads to provide electrical
conduits. Seiko Epson [228,229] has used NiAu-coated resin balls, 7.S /.lm
in diameter, embedded in adhesive. The process temperature used to cure
the adhesive is ISO-200°C with a load of approximately S kg/cm2. The
application was chip-on-glass for high-resolution liquid crystal displays.
One unsuccessful implementation [230] demonstrated that clusters
of small particles gave more reproducible electrical contact than did larger
discrete particles. Many different particle materials have been explored
including solid metals (Ni, solder), carbon, and metal-coated plastic parti-
cles. Thermoplastic, thermosetting, and light setting polymers have been
tried as adhesives with varying degrees of success. Coupling very small
particles has proven to be difficult, with intermittents showing up in
thermal-cycle or humidity-cycling tests. Improvements in reliability are
expected as materials are optimized.

8.8 ELECTRICAL PARAMETERS OF INTERCONNECTIONS


The electrical performance of the packaged chip may well be deter-
mined by the mode of chip interconnection. The electrical parasitics are
very different for the three major interconnection techniques, with the
flip chip offering the best properties due to its small size-nearly a
point contact (see Table 8-8). Package geometry, however, also plays a
significant role. In fact, it can reduce the intrinsic differences between
the three interconnection approaches. For example, a wirebonded silicon
device is normally associated with a lead frame, which, for fabrication
reasons or board-level interconnection geometry requirements, may re-
quire rather long leads, well separated from a ground or reference plane.
(Alternative wirebonded structures, such as pin-grid arrays or direct-chip
attach to a substrate, may provide a much different electrical signal envi-
ronment). High-end computer packaging designers are aware of the electri-
cal performance issues associated with packaging, usually for their own
customized applications (see Chapter 1). For this reason, as well as because
low-performance systems are not so often package limited, focus on
8.9 DENSITY OF CONNECTIONS 11-269

mid-range structures are frequently gated by package configuration and


performance. To be considered are chips having 84 and 180 IIOs packaged
using each of the three primary chip interconnection approaches addressed
in this chapter: C4 solder-joined, wirebonded and TAB-bonded packages.
The packages to be compared are (1) a C4-joined, metallized ceramic
single-chip carrier having swaged pins, (2) a C4-joined, multilayer ce-
ramic, single-chip carrier having brazed pins, (3) a wirebonded pin-grid
array, (4) a wirebonded plastic-leaded chip carrier (PLCC) (84 IIOs only),
and (S) a TAB-packaged device. The pin grid for packages 1-3 is fixed
at 2.S4 mm; the PLCC is fixed at 1.27-mm lead pitch; the TAB package
is fixed at O.S-mm lead pitch. (Note, package 2 is a somewhat artificial
configuration, given that usually more than one C4-joined chip is joined
to an MLC module.) The objective here is to give a rough basis for
comparing package performance, using the lead geometry of the packages.
The basis for comparison is the calculated range in the value of lead
inductance. This provides a starting point for estimating the effective
package inductance, Leff. This quantity is used to predict the inductance-
based voltage swing (or M noise), d V =Leff n(dlldt), where n is the number
of simultaneously switching drivers, having a given current surge dlldt
(as discussed in more detail in Chapter 3). Shown in Table 8-9 is the lead
inductance calculated for various packages using a three-dimensional
inductance program [231]. Lead inductance may be used to estimate
the effective package inductance, given a detailed knowledge of chip
configuration and the number and location of power and ground leads
(all of which should be optimized for the package and chip of interest
[232]). The relation between lead inductance and Leff is quite geometry-
specific and will not be expanded upon further. Signal coupled noise,
which is equally dependent on a detailed package description, will also
not be considered here. Referring to the table of package inductances, it
should be noted that for most of the package geometries chosen, a broad
range of lead inductance is observed, indicating the need for personaliza-
tion of signal and power/ground assignments.

8.9 DENSITY OF CONNECTIONS


A dramatic difference in attainable connection density between area
arrays and peripheral pads was shown previously in Table 8-3 and Figure
8-1S. With the advent of area TAB and multiple wirebond layers, improve-
ments in density are being made (Fig. 8-120). High performance and II
o technologies in both leading edge CMOS and bipolar chips will favor
area-array technologies over peripheral. The present area of applicability
of wirebond, TAB, and C4 is shown in Figure 7-16 in Chapter 7, "Micro-
electronics Packaging-An Overview."
A typical process comparison of C4, wirebond, and TAB for a
11-270 CHIP-TO-PACKAGE INTERCONNECTIONS

Table 8-9. Calculated Lead Inductance for Single-Chip Modules

C4 Bonded, C4 Bonded, Wirebonded Wirebonded


MC Pin MLC Pin Pin Plastic Leaded TAB
Grid Array Grid Array Grid Array ChiQ Carrier Package

84 110 Module

Substrate size (mm) 24 24 24 30 18

Lead length (mm)


(min.) 2 2 12.5 6.5
(max.) 17 13.5 17 17.7 9.0

Lead inductance (nH)


(min.) 3.5 2.5 7.0 16.0 5.0
(max.) 13.3 6.3 19.0 23.0 7.2

180 110 Module

Substrate size (mm) 36 36 36 30

Lead length (mm)


(min.) 2 2 11.3
(max.) 23 20 23 16.3
Lead inductance (nH)
(min.) 3.5 2.5 7.0 8.5
(max.) 17 8.2 24 12.5

Source: From Ref. 232.

multichip module application is shown in Table 8-10. Although process


costs and density are major influencing factors, reliability, component and
tooling costs, thermal performance, hermeticity requirements, need for
reworking, turnaround time, and electrical parameters have all been shown
to play major roles in the selection process. Although reworking is easy
for C4 interconnections, inspectability of the joints is not. Although wire-
bonding and TAB do not restrict chip sizes and thermal-expansion coeffi-
cients of the substrates, these restrictions are inconsequential for C4 when
matched expansion substrates are used. High thermal capability options
have been shown for all of the technologies.

8.10 SUMMARY
It has been shown that wirebond, TAB, and C4 have all evolved
and improved over the past 25-40 years. All have found niches and
8.10 SUMMARY 11-271

a
TAB Wire Bonding
r----
A....

I.
Active
e• •c r°l..===
"Ol-
:"r A~
.c: E .or
t&
I.e
-~
c.
I!c
·ou
&A.

Radueeto
93%
b Wiring Area TAB I Multiple Wire Bonding
T:'~" i ..~-
I ~ccccc~!
: ~r-=B~EJ:--\lg 0 0 ~

.~~
i I ~aara7
L . , .~
n
Reduce to
84%

c 0000000
0000000
Flip 0000000
Chip 000000
C4 0000000
0000000
0000000

Figure 8-120. Relationships Among Various Bonding Methods and Bonding


Space. (From Ref. 233.)
11-272 CHIP-TO-PACKAGE INTERCONNECTIONS

Table 8-10. Chip Interconnection Assembly Comparison for Multichip


Module

Wire Bond TAB

1. Flux site 1. Solder or epoxy preform 1. Reel tape into bond


place position
2. Align and place chip(s) 2. Align and place chip 2. Align and place chip
3. Reflow to bond all pads 3. Die bond 3. Inner lead bond (one chip
(on all chips) at a time)
4. Clean flux 4. Wirebond each wire on 4. Encapsulate
chip and SIS (one wire at
a time)
5. Test 5. Test 5. Testlbum-in (optional)
6. Encapsulate/finish module 6. Encapsulate/finish module 6. Excise chip
assembly assembly
7. Align and place assembly
on SIS
8. Outer lead bond
9. Test

applications which make them valuable technologies; but, as of today,


wirebonding remains the incumbent king of the broad market and is the
chip interconnect of choice for about 95% of the packages. The reasons
are many. First, the process has transformed itself from where it was in
1960 to a highly reliable, inexpensive, very manufacturable process with
greatly appreciated flexibility and versatility which does not constrain
or commit the manufacturer. Au ball-bond wires coupled with greatly
improved plastic molding processes and materials have made plastic dips
and quad flat packs the low-cost, highly reliable package for the majority.
Virtually all memory and all commodity electronics utilize these packages.
The wirebonding infrastructure is so huge worldwide that there would
have to be extremely strong and compelling reasons not to use today' s wire-
bonding.
TAB was invented to replace wirebonding in the early 1970s. Its
fine-pitch, peripheral nature was intended to extend wirebonding to sig-
nificantly higher liDs on chips with greatly improved reliability and manu-
facturability. Some referred to it as "poor man's beam lead," alluding to
Bell's own replacement technology for the TIC wirebond. The technology
is good. It did what it was designed to do, but the rate of improvement
and cost of wirebonding have made it extremely difficult for TAB to
displace WB. The Japanese manufacturers have had much more success
with TAB than American or European manufacturers. The biggest use of
TAB there is in the niche of LCD peripheral connections. Sharp has also
8.11 REFERENCES 11-273

been very successful with commodity items because of the inherently


manufacturable nature of the process, combined with the lightness and
low profile.
The overriding weakness of WB and TAB has been the peripheral
nature of the interconnect, as opposed to C4, which has its greatest strength
in area-array connectability. The rapid rate of integration of all devices
in the ULSI era has driven the need for I/O, particularly in logic ASICs
and microprocessors. The ease with which C4's satisfy the 1000-5000
I/O needs of current and future semiconductors has made its choice easy.
The coupling of multilayer chip wiring and multilayered substrate wiring
has also brought the best performance for high-speed computers and
telecommunications devices. Also of value is the fact that the same chip
functionality can be realized in a much smaller chip with the resulting
economies of more chips per wafer.
The development of chip underfills has made possible direct chip
attach to FR-4 cards in a reliable way for the first time, making the chip
the ultimate "smallest package." Miniature pagers and telephones are
beginning to appear, and low-cost bumping consortia have the support
of government and industrial groups to make low-cost chip-on-board a
growing interest. For the first time, flip chip is a serious consideration
for low-cost commodity electronics as well as the up-scale applications.
It has been gratifying to watch these three interconnect technologies
grow and develop, particularly the sharing of ideas. Features of one
system are used freely in another. For example, single-point thermosonic
wirebonders are excellent for TAB inner lead bonds, Au ball-bond bump-
ing on flip chips provides bumping without major reinvestment. Flip-on-
flex has a lot of commonality with TAB.
All three interconnect technologies will continue to coexist for a
long time. Wirebonding will continue to dominate memory and commodity
packaging requiring relatively small I/O counts. Flip chip will dominate
high-end packages with greater than 500 I/O, and perhaps some commodity
applications; TAB will find the special niches featuring thinness and
lightness.

8.11 REFERENCES
I. C. T. Goddard. "The Role of Hybrids in LSI Systems," IEEE TR, TR-2: 367, 1979.
2. The National Technology Roadmap for Semiconductors. Semiconductor Industry
Assn., 1994.
3. O. L. Anderson. Bell Lab. Rec., November 1957.
4. M. P. Lepselter. "Beam Lead Technology," Bell Sys. Tech. J., 45: pp. 233-253,1966.
5. E. M. Davis, W. E. Harding, R. S. Schwartz, and J. J. Corning. "Solid Logic Technol-
ogy: Versatile High Performance. Microelectronics," IBM J. Res. Devel., 8:
p. 102, 1964.
6. J. A. Perri, H. S. Lehman, W. A. Pliskin, and J. Riseman. "Surface Protection of
Silicon Devices with Glass Films," Electrochemical Society Meeting, p. 102, 1961.
11-274 CHIP-TO-PACKAGE INTERCONNECTIONS

7. P. A. Tottaand R. P. Sopher. "SLT Device Metallurgy and Its Monolithic Extension,"


IBM J. Res. Devel., 5: pp. 226-238, 1969.
8. P. A. Totta. "Hip-Chip Solder Terminals," 21st Electronics Components Conference,
p. 275, 1971.
9. L. J. Fried, J. Havas, J. S. Lechaton, 1. S. Logan, G. Paal, and P. Totta. "A VLSI
Bipolar Metallization Design with Three-Level Wiring and Area Array Solder Con-
nections," IBM J. Res. Devel., 26: pp. 362-371, 1982.
10. K. H. Brown, D. A. Grose, R. C. Lange, T. H. Ning, and P. A. Totta. "Advancing
the State of the Art in High-Performance Logic and Array Technology," IBM J. Res.
Devel., 36: p. 821, 1992.
11. L. F. Miller. "Controlled Collapse Reftow Chip Joining," IBM J. Res. Devel., 13:
pp. 239-250, 1969.
12. R. Darveaux, I. Turlik, L. Hwang, and A. Reisman. "Thermal Analysis of a Multichip
Package Design," IEEE Trans. Components Hybrids Manuf Technol. CHMT-12(4):
pp. 663-672, 1989; J. Electron. Mater., 18(2): pp. 267-274, 1989.
13. D. J. Bendz, R. W. Gedney, and J. Rasile. "CostlPerformance Single Chip Module,"
IBM J. Res. Devel., 26: pp. 278-285, 1982.
14. A. J. Blodgett. "A Multilayer Ceramic Multichip Module," IEEE Components Hy-
brids Manuf. Technol. CHMT-3: p. 634, 1980.
15. L. S. Goldmann and P. A. Totta. "Area Array Solder Interconnections for VLSI,"
Solid State Technol., 1983.
16. S. E. Greer. "Low Expansivity Organic Substrate for Hip-Chip Bonding," 28th
Electronic Components Conference Proceedings, pp. 166-171, 1978.
17. T. Inoue, H. Matsuyama, E. Matsuzaki, Y. Narazuka, M. Ishino, M. Tanaka, and T.
Takenaka. "Micro Carrier for LSI Chip Used in the HITAC M-880 Processor Group,"
Proc. ECTC, pp. 704-711, 1991.
18. B. T. Clark. "Design of the IBM Thermal-Conduction Module," IEEE Components
Hybrids Manu! Technol. CHMT-4: 1981.
19. A. J. Blodgett and D. R. Barbour. ''Thermal Conduction Module: A High-Perfor-
mance Multilayer Ceramic Package," IBM J. Res. Devel., 26: pp. 30-36, 1982.
20. D. R. Barbour, S. Oktay, and R. A. Rinne. "Thermal-Conduction Module Cradles
and Coops up to 133. LSI Chips," Electronics, June 1982.
21. S. Ahmed, R. Tummala, and H. Potts. "Packaging Technology for IBMs Latest
Mainframe Computers (S13901ES 9000)," Proc. ECTC, pp. 682-688, 1991.
22. J. Oberschmidt and J. Humenik. "A Low Inductance Capacitor Technology," Proc.
ECTC, pp. 284-288, 1990.
23. N. G. Koopman and P. A. Totta. "Controlled Collapse Chip Connections (C4)," ASM
World Material Congress, 1988.
24. 1. P. Dietz. "East Fishkill Advanced Technology in New 9370 Series," IBM Press
Release, October 1986.
25. T. Kamei and M. Nakamura. "Hybrid IC Structures Using Solder Reftow Technol-
ogy," 28th Electronic Components Conference Proceedings, pp. 172-182, 1978.
26. P. Schmid and H. Melchior. "Coplanar Hip-Chip Mounting Techniques for Picosec-
ond Devices," Rev. Sci Instrum., 55(11): p. 1854, 1984.
27. W. Weston. "High Density 128 x 128 Area Arrays of Vertical Electrical Interconnec-
tions," 4th Annual Microelectronic Interconnect Conference, 1985.
28. S. K. Lahiri, H. R. Bickford, P. Geldermans, K. R. Grebe, and P. A. Moskowitz.
"Packaging Technology for Josephson Integrated Circuits," IEEE Components Hy-
brids Manuf. Technol., CHMT-5(2): p. 271, 1982.
29. C. Y. Ting, K. Grebe, and D. Waldman. "Controlled Collapse Reftow for Josephson
Chip Bonding," J. Electrochem. Soc., 129(4): 859-864, 1982.
8.11 REFERENCES 11-275

30. K. R. Grebe. "Orthogonal Solder Interconnects for Josephson Packaging," J. Electro-


chem. Soc., 127: 1980.
31. I. Lang. "Surface Mounting of Leadless Chip Carriers on Various Printed Circuit
Board Type Substrates," Electron. Sci. Techno!., 9: p. 296, 1982.
32. 1. D. Leibowitz and W. E. Winters. "Shirline A Coefficient of Thermal Expansion
(CTE) Controllable Family of Materials," IEPC, 1982.
33. 1. Fisher. "Cast Leads for Surface Attachment," ECC, p. 219, 1984.
34. Y. Tsukada. "Surface Lamilar Circuit and Flip Chip Attach Packaging," ECTC,
1992.
35. R. K. Wassink. Soldering in Electronics, Electrochemical Publications Limited, Ayr,
Scotland, 1984.
36. 1. Gow III and A. R. Leach. "Memory Packaging at IBM," Circuits Manu!, p. 72,
May 1983.
37. M. Ohshima, A. Kenmotsu, and I. Ishi. "Optimization of Micro Solder Reflow
Bonding for the LSI Flip Chip," The Second International Electronics Packaging
Conference, 1982.
38. T. Takenaka, F. Kobayashi, and T. Netsu. "Reliability of Flip-Chip Interconnections,"
1984 ISHM Conference Proceedings, p. 419, 1984.
39. F. Kobayashi, T. Takenaka, M. Ishida, and T. Netsu. "Fabrication and Reliability
Evaluation of the Multi-Chip Ram Module," Proc. 1982 IMC, p. 425, 1982.
40. K. C. Norris, and A. H. Landzberg. "Reliability of Controlled Collapse Interconnec-
tions," IBM J. Res. Deve!., 13(3): pp. 266-271, 1969.
41. V. K. Nagesh. "Reliability of Flip Chip Solder Bump Joints," Reliability Physics
Symposium, 1982.
42. D. L. Brownell and G. C. Waite. "Solder Bump Flip-Chip Fabrication Using Standard
Chip and Wire Integrated Circuit Layout," 1974 ISHM Conference Proceedings,
p. 77, September 1974.
43. W. Roush. "Effect of Gold Additions on High Strain-Low Cycle Fatigue Behavior
of 95/5 Pb/Sn Solder Connections," Proc. 2nd International Conference on Behavior
of Materials, p. 780, 1976.
44. O. Clatterbaugh and H. Charles. "Design Optimation and Reliability Testing of
Surface Mounted Solder Joints," 1SHM, p. 31, 1985.
45. T. Yamada, K. Otsutani, K. Otsukani, K. Sahara, and K. Otsuka. "Low Stress Design
of Flip Chip Technology for "Si on Si Multichip Modules," IEPS, pp. 551-557,1985.
46. H. Curtis. "Integrated Circuit Design Production and Packaging for System 38,"
IBM System 38 Technical Development, December 1978.
47. L. S. Goldmann. "Self-Alignment Capability of Controlled-Collapse Chip Joining,"
Proceedings 22nd Electronic Components Conference, p. 332, 1972.
48. T. Kawanobe, K. Miyamoto, and Y. Inaba. "Solder Bump Fabrication by Electro-
chemical Method for Flip-Chip Interconnections," 31st Electronic Components Con-
ference Proceedings, p. 149, 1981.
49. J. Marshall and F. Rode. "Solder Bump Interconnected Multiple Chip Thick Film
Hybrid for a 40-Character Alphanumeric LCD Application," Solid State Techno!.,
p. 87, 1979.
50. E. Rasmanis. "Fabrication of Semiconductor Devices with Solder Terminals," 1973
ISHM Conference Proceedings, pp. 3B-3-1, 1973.
51. A. VanDer Drift, W. G. Gelling, and A. Rademakers. "Integrated Circuits with Leads
on Flexible Tape," Philips Tech. Rev., 34(4): p. 85, 1974.
52. T. R. Myers. "Flip-Chip Microcircuit Bonding System," Proc. ECC, p. 131, 1969.
53. N. Tsujimoto. "Flip Chip and Micropackaged Chips," Electron. Parts, 19(5):
p. 51, 1980.
11-276 CHIP-TO-PACKAGE INTERCONNECTIONS

54. N. Nuzaki, K. Nakamura, and I. Taubokawa. "A Thick Film Active Filter for PCM
Communication Systems," 28th Electronic Components Conference Proceedings,
p. 183, 1977.
55. G. Adema, C. Berry, N. Koopman, G. Rinne, E. Yung, and I. Turlik. "Flip Chip
Technology: A Method for Providing Known Good Die With High Density Intercon-
nections", 3rd International Conference and Exhibition on Multichip Modules, 1994.
56. P. Magill, P. Deane, D. Mis, and G. Rinne. "Flip Chip Overview," IEEE Multichip
Module Conference, 1996.
57. C. J. Speerschneider and J. M. Lee. "Solder Bump Reflow Tape Automated Bonding,"
Proceedings 2nd ASM International Electronic Materials and Processing Congress,
pp. 7-12, 1989.
58. D. J. Genin and M. M. Wurster. "Probing Considerations in C-4 Testing of Wafers,"
Microcircuits Electron. Packaging, 15-4: p. 229, 1992.
59. N. Koopman. "Flip Chip Interconnections," in Concise Encyclopedia of Semiconduct-
ing Materials and Related Technologies, pp. 184-187, Pergamon Press, London,
1992.
60. V. Iinuma, T. Hirohara, K. Inove, "Liquid Crystal Color Television," ISHM Proc
1987, pp. 635-640.
61. Y. Kondoh, M. Saito, "A New CCD Module Using the Chip-On-Glass (COG)
Technique, ISHM Proc, 1990, pp. 487-494.
62. R. Bache, P. Burdett, K. Pickering, A. Parsons, D. Pedder. "Bond Design and
Alignment in Flip-Chip Solder Bonding, Proc. IEPC, pp. 830-841, 1988.
63. M. Goodwin, C. Kirkby, A. Parsons, I. Bennion, and W. Stewart. "8 x 8 Element
Hybridized PLZT/Silicon Spacial Light Modulator Array," Electron. Lett., 25(18):
pp. 1260-1262, 1989.
64. P. Burdett, K. Lodge, and D. Pedder. "Techniques for the Inspection of Flip Chip
Solder Bonded Devices," Proc IEPS, pp. xxx-xxx, 1988, reprinted in Hybrid Circuits,
19: pp. 44-48, 1989.
65. S. Patra and Y. Lee. "Modelling of Self Alignment in Flip Chip Soldering-Part 2:
Multichip Solder Joints," Proc. ECTC, pp. 783-788, 1991.
66. M. Wale. "Self-Aligned, Flip Chip Assembly of Photonic Devices with Electrical
and Optical Connections," Proc. ECTC, pp. 470-476, 1990.
67. A. Burkhart. "Recent Developments in Flip Chip Technology," Surface Mount Tech-
nol., pp. 41-45, 1991.
68. R. T. Howard. "Optimization of Indium-Lead Alloys for Controlled Collapse Chip
Connection Application," IBM J. Res. Devel., 26(3): pp. 372-389, 1982.
69. K. J. PuttIitz. "Flip-Chip Replacement Within the Constraints Imposed by Multilayer
Ceramic (MLC) Modules," J. Electron. Mater., 13(1): pp. 29-46, 1984.
70. S. Teed and V. Marcotte. "A DSC Technique to Measure the Amount of Indium or
Tin Leached from Lead Base Solders by Rosin Fluxes," NATAS, p. 77, 1981.
71. J. Temmyo, K. Aoki, H. Yoshikiyo, S. Tsurumi, and Y. Takeuchi. "Solder Bump
Height Dependence of Josephson Chip to Card Interconnection Inductance Using
Flip-Chip Bonding Technique," J. Appl. Phys., 54(9): p. 5282, 1983.
72. N. Koopman and S. Nangalia. "Fluxless Flip Chip Solder Joining," NEPCON
WEST, 1995.
73. N. Koopman, G. Adema, S. Nangalia, M. Schneider, and V. Saba. "Flip Chip Process
Development Technique Using a Modified Laboratory Aligner Bonder," IEMT Sym-
posium, 1995.
74. H. Markstein, "Multichip Modules Pursue Water Scale Interaction,", EP&P, Octo-
ber, 1991.
75. K. Lodge, D. Pedder. "The Impact of Packaging on the Reliability of Flip Chip
Solder Bonded Devices," Proc. ECTC, pp. 470-476, 1990.
8.11 REFERENCES 11-277

76. A. Munns. "Flip-Chip Solder Bonding in Microelectronic Applications," Metals


Mater., pp. 22-25, 1989.
77. K. Puttlitz. "Preparation, Structure, and Fracture Modes of Pb-In Terminated Flip
Chips Attached to Gold-Capped Microsockets," Proc. ECTC, pp. 360-367, 1990.
78. B. Le Pape, "Use of a Tinned Copper Slug for Module Reworking." IBM Tech.
Disclosure Bull., 24(7A): pp. 3481,1981.
79. A. Ingraham, J. McCreary, and J. Varcoe. "Flip-Chip Soldering to Bare Copper
Circuits," Proc. ECTC, pp. 333-337, 1990.
80. T. Adams. "Acoustic Micro-Imaging of Multichip Modules," Suiface Mount Tech-
no/., June 1991.
81. B. Gilbert, W. McNeff, P. Zabinski, W. Komrumpf, S. Tead, R. Maki, K. Carlson,
and J. Gerber. "Development of Deposited Multichip Modules with Unique Features
for Application in GaAs Signal Processors Operating Above I GHz Clock Rates,"
Proc.IEPS, pp. 526-541, 1992.
82. N. Koopman, G. Rinne, P. Magill, S. Nangalia, C. Berry, D. Mis, V. Rogers,
G. Adema, and M. Berry. "Solder Flip Chip Developments at MCNC," 1996 Interna-
tional Flip Chip, Ball Grid Array, TAB and Advanced Packaging Symposium-
ITAP, 1996.
83. D. A. Jeannotte. "Solder as a Structural Member for Chip Joining," Electronic Compo-
nents Conference Proceedings, p. 334, 1969.
84. L. F. Coffin. "Low Cycle Fatigue: A Review," App/. Mech. Res., 1(3): pp. 129-
141,1962.
85. P. A. Tobias, N. A. Sinclair, and A. S. Van. "The Reliability of Controlled-Collapse
Solder LSI Interconnections," ISHM Proc., p. 360, 1976.
86. N. A. Sinclair. "Thermal Cycle Fatigue Life of LSI Solder Interconnections," Interna-
tional Electrical Proceedings, p. 56, 1982.
87. H. J. Shah and 1. H. Kelly. "Effect of Dwell Time on Thermal Cycling of the Flip-
Chip Joint," 1970 ISHM Proc., paper 3.4, 1970.
88. E. Levine and 1. Ordonez. "Analysis of Thermal Cycle Fatigue Damage in Micro-
socket Solder Joints," 31st Electronic Components Conference Proceedings, pp.
515-519, 1981.
89. L. S. Goldmann. "Geometric Optimization of Controlled Collapse Interconnections,"
IBM 1. Res. Deve/., 3: pp. 251-265, 1969.
90. E. Anderson. "Flip-Chip Bonding: The Effect of Increased Solder Bump Spacing,"
Microelectronic Interconnect Conference, 1984.
91. R. T. Howard, S. W. Sobeck, and C. Sanetra. "A New Package-Related Failure
Mechanism for Leadless Ceramic Chip Carriers (LC-3s) Solder-Attached to Alumina
Substrates," Solid State Techno!', 26(2): pp. 115-122, 1983.
92. P. M. Hall. "Solder Post Attachment of Ceramic Chip Carriers to Ceramic Film
Integrated Circuits," 31st Electronic Components Conference Proceedings, pp. 172-
180, 1981.
93. R. Satoh, M. Ohshima, H. Komura, I. Ishi, and K. Serizawa. "Development of a
New Micro-Solder Bonding Method for VLSI," IEPS, p. 455, 1983.
94. N. Matsui, S. Sasaki, and T. Ohsaki. "VLSI Chip Interconnection Technology Using
Stacked Solder Bumps," Proceedings IEEE 37th Electronic Components Conference,
pp. 573-578, 1987.
95. E. Horikoshi, K. Hoshimoto, T. Sato, and H. Nakajima. "Published unexamined
patent application," Japanese Patent No. 62-117346, 1987.
96. G. Cherian. "Use of Discrete Solder Columns to Mount LCe's on Glass/Epoxy
Printed Circuit Boards," IEPS, p. 710, 1984.
97. L. S. Goldmann, R. J. Herdizk, N. G. Koopman, and V. C. Marcotte. "Lead Indium
for Controlled Collapse Chip Joining," 27th Electronic Components Conference
Proceedings, 27: p. 25, 1977.
11-278 CHIP-TO-PACKAGE INTERCONNECTIONS

98. R. T. Howard. "Packaging Reliability and How to Define and Measure It," 32nd
Electronic Components Conference Proceedings, pp. 367-384, 1982.
99. W. Roush and J. Jaspal. "Thermomigration in Lead-Indium Solder," 32nd Electronic
Components Conference Proceedings, p. 342, 1982.
100. C. J. Bartlett et al. "Multichip Packaging Design for VLSI-Based Systems," Proc.
ECTC, p. 518, 1987.
101. J. H. Lau and D. W. Rice. "Solder Joint Fatigue in Surface Mount Technology State
of the Art," Solid State Technol., p. 91, 1985.
102. W. Engelmaier. "Effects of Power Cycling on Leadless Chip Carrier Mounting
Reliability," IEPS, p. 15, 1982.
103. F. Nakano, T. Soga, and S. Amagi. "Resin Insertion Effect on Thermal Cycle
Resistivity of Flip Chip Mounted LSI Devices," Proceedings 1987 ISHM Conference,
pp. 536-541, 1987.
104. B. Han and Y. Guo, "Thermal Deformation Analysis of Various Electronic Packaging
Products by Moire and Microscope Moire Interferometry," J. Electron. Packaging,
117: p. 185, 1995.
105. V. Marcotte, M. Ricker, and N. Koopman. "Metallography of Soldered Joints," IMSI
ASMIAWS Symposium, 1985.
106. V. C. Marcotte and N. G. Koopman. "Palladium Depletion of 80Ag20Pd Thick Film
Electrodes," ECC, pp. 157-164, 1981.
107. D. Bouldin. "The Measurement of Alpha Particle Emissions from Semiconductor
Memory Materials," J. Electron. Mater., 10(4): p. 747, 1981.
108. R. Howard. "Characterization of Low-Alpha-Particle Emitting Ceramics," ECC,
p.453, 1984.
109. Cominco/Johnson Matthry. Private Communications, 1987.
110. K. Scheibner. "Laser Isotope Purification of Lead for Use in Semiconductor Chip
Interconnect", ECTC, 1996.
111. J. L. Chu, H. R. Torabi and F. J. Towler, "A 128 Kb CMOS Static Random-Access
Memory," IBM J. R&D pp. 321-329, 1991.
112. S. Oktay. "Parametric Study of Temperature Profiles in Chips Joined by Controlled
Collapse Technique," IBM J. Res. Devel., 13: pp. 272-285, 1969.
113. V. Kurokawa, K. Utsumi, H. Takamizawa, T. Kamata, and S. Noguchi. "AlN Sub-
strates with High Thermal Conductivity," IEEE TR., Tr-2: p. 247, 1985.
114. S. Oktay and H. C. Kammerer. "A Conduction-Cooled Module for High-Performance
LSI Devices," IBM J. Res. Devel., 26(1): pp. 55-66, 1982.
115. R. C. Chu, U. P. Hwang, and R. E. Simons. "Conduction Cooling for an LSI Package:
A One-Dimensional Approach," IBM J. Res. Devel., 26: pp. 45-54, 1982.
116. F. Kobayashi, Y. Watanabe, M. Yamamoro, A. Anzai, A. Takahashi, T. Daikoku,
and T. Fujita. "Hardware Technology for Hitachi M880 Processor Group," Proc.
ECTC, pp. 693-704, 1991.
117. R. E. Simmons and R. C. Chu. "Direct Immersion Cooling Techniques for High
Density Elect. Packaging and System," ISHM, p. 314, 1985.
118. S. Aoki and Y. Imanaka. "Multilayer Ceramic Substrate for HEMT Packaging (Liquid
Nitrogen Packaging) for GaAs Devices," Proc. IEPS, pp. 329-341, 1991.
119. H. Tong, L. Mok, K. Grebe, and H. Yeh. "Paralyne Encapsulation of Ceramic
Packages for Liquid Nitrogen Applications," Proc. ECTC, pp. 345-350, 1990.
120. K. Casson, B. Gibson, and K. Habeck. "Flip-on-Flex: Solder Bumped ICs Bond to
a New High Temperature, Adhesiveness, Flex Material," Surface Mount International
Proceedings, pp. 99-104, 1991.
121. K. Gilleo. "Direct Chip Interconnect Using Polymer Bonding," Proc. 39th ECC,
pp.37-44, 1989.
8.11 REFERENCES 11-279

122. The VLSI Manufacturing Outlook, vol. I, sects. 52 and 54, VLSI Research Inc., 1992.
123. Pashby et al.International Business Machines Corporation. "Lead on Chip Semicon-
ductor Package," U. S. Patent 4,862,245.
124. Winchell and Berg. "Enhancing Ultrasonic Bond Development," IEEE Trans. Com-
ponents Hybrids Manu! Technol., CHMT-I (3): 1978.
125. Harman and Leedy. "An Experimental Model of the Microelectronic Ultrasonic Wire
Bonding Mechanism," Proceedings 10th Annual Reliability Physics Symposium,
pp.49-56,1972.
126. E. Philofsky. "Design Limits When Using Gold-Aluminum Bonds," Proceedings
9th Annual IEEE Reliability Physics Symposium, pp. 177-185, 1970.
127. Mitsubishi Semiconductor IC Packages, Databook, pp. 2-14, 1988.
128. B. L. Gehman, "Bonding Wire Microelectronics Interconnections," IEEE Trans. on
Comp., CHMT-3(3): p. 375, 1980.
129. R. Shukla and N. Mencinger, "A Critical Review of VLSI Die-Attachment in High
Reliability Applications," Solid State Tech., p. 67, July 1985.
130. Presented at: IEElCHMT International Electronics Manufacturing Technology Sym-
posium, San Francisco, Ca., September 16-18, 1991.
131. Cindy Enman and Gil Perlberg, "Wirebond Pull Testing: Understanding the Geo-
metric Resolution Forces," Advanced Packaging, pp. 21-22, 1994.
132. G. Harman, "Reliability and Yield Problems of Wire Bonding in Microelectronics,"
ISHM, pp. 19-20, 1991.
133. A. D. Aird. "Method of Manufacturing a Semiconductor Device Utilizing a Flexible
Carrier," U. S. Patient 3,689,991, 1972.
134. G. Dehaine and M. Leclercq. "Tape Automated Bonding, a New Multichip Module
Assembly Technique," Electronic Components Conference Proceedings, pp. 69-
73, 1973.
135. T. Watari and H. Murano. "Packaging Technology for the NEC SX Supercomputer,"
Proc. 35th ECC, p. 192-198, 1985.
136. K. Saito. "TAB Applications in Japan," Proceedings Fifth International TAB and
Advanced Packaging Symposium, pp. 9-19, 1993.
137. 1. Lyman. "How DOD's VHSIC Is Spreading," Electronics, pp. 33-37, 1985.
138. R. L. Cain. "Beam Tape Carriers-a Design Guide," Solid State Technol., pp. 53-58,
March 1978.
139. TAB Design Guide, Shindo Company Ltd., 1993.
140. K. Fukuta, T. Tsuda, and T. Maeda. "Optimization of Tape Carrier Materials,"
Proceedings Fourth International TAB Symposium, pp. 283-312, 1992.
141. J. M. Smith and S. M. Stuhlbarg. "Hybrid Microcircuit Tape Chip Carrier Materials!
Processing Tradeoffs," Proc. 27th ECC, pp. 34-47, 1977.
142. K. Saito, K. Nomoto, and Y. Nishimoto. "The Properties of Adhesive in Three-Layer
TAB Tape," Proceedings Fourth International TAB Symposium, pp. 210-235, 1992.
143. K. Doss and S. Holzinger. "Materials Choices for Tape Automated Bonding," in
Handbook of Tape Automated Bonding, ed. by J. Lau, pp. 99-134, Van Nostrand
Reinhold, New York, 1992.
144. T. Mugishima, O. Seki, and K. Ohtani. "Improvement of Adhesion Strength in Two-
Layer Tape," Proceedings Fourth International TAB Symposium, pp. 198-209, 1992.
145. R. S. Dodsworth and R. T. Smith. ''TAB Tape Design and Manufacturing," in Hand-
book of Tape Automated Bonding, ed. by J. Lau, pp. 135-175, Van Nostrand
Reinhold, New York, 1992.
146. P. Chen, K. Blackwell, and A. Knoll. "Curl and Residual Stress Control of Metallized
Polymides," Proceedings Fifth International TAB/Advanced Packing Symposium,
pp. 85-91, 1993.
/1-280 CHIP-TO-PACKAGE INTERCONNECTIONS

147. H. Nakayama, Y. Satomi and T. Natsume. "Characteristics of New TAB Products


using Special Copper Alloys," Proceedings Third International TAB Symposium,
pp. 180-l9l, 1991.
148. K. Saito. "Fine Lead Pitch Technology in Tape Carriers and Double Metal Layer Tape
Carrier," Proceedings Second International TAB Symposium, pp. 199-213, 1990.
149. K. Saito. "Selective Plating," Proceedings Third International TAB Symposium,
pp. 146-156, 1991.
150. S. Britton. "Spontaneous Growth of Whiskers on Tin Coatings: 20 Years ofObserva-
tion," Trans. Inst. Metal Finishing, pp. 95-102, 1974.
151. T. Kawanobe, K. Miyamoto and M. Hirano. "Tape Automated Bonding Process for
High Lead Count LSI," Proc. 33rd ECC, pp. 221-226, 1983.
152. P. Hoffman, T. Kleiner, and H. Phan. "Characteristics of Tin Plated TAB Tape,"
Proc. IEPS, pp. 1307-1326, 1989.
153. A. Juan. "TAB Applications for Miniaturization," Proceedings Fourth International
TAB Symposium, pp. 26-41, 1992.
154. M. Masuda, K. Sakurai, K. Haley, Y. Hiraki, and S. Inoue. "New Configuration Three-
Layer BTAB Technology Application for CPU," Proceedings Fifth International TAB
Symposium, pp. 112-117, 1993.
155. K. Hatada, J. Okamoto, M. Hirai, K. Matsunaga, and I. Kitahiro. "New Film Carrier
Assembly Technology: Transferred Bumped TAB," Nat. Tech. Rept. Japan, 31:
pp.412-420, 1985.
156. K. Hatada. "Application to the Electronic Instrument by Transferred Bump-TAB
Technology, Proceedings International Symposium on Microelectronics (ISHM),
pp.649-653,1987.
157. K. Hatada. "New Bonding Technology Using Transferred Bumped TAB and Single
Point Bonding," Proceedings Third International TAB Symposium, pp. 124-130,
1991.
158. K. Taguchi. "Small Bump Development on using TB-TAB for Fine Pitch Intercon-
nect," Proceedings Fourth International TAB Symposium, pp. 236-241, 1993.
159. K. Tatsumi, T. Ando, Y. Ohno, M. Konda, Y. Kawakami, N. Ohikata, and
T. Maruyama. "Transferred Ball Bump Technology for Tape Carrier Packages,"
Proceedings International Symposium on Microelectronics, pp. 54-59, 1994.
160. P. Hoffman. "Design, Manufacturing and Reliability Aspects of Two Metal Layer
TAB Tape," Proceedings Second International TAB Symposium, pp. 124-146, 1990.
161. K. Saito. "Fine Lead Pitch Technology in Tape Carriers and Double Metal Layer Tape
Carrier," Proceedings Second International TAB Symposium, pp. 199-213, 1990.
162. Y. Mashiko and S. Nishiyama. "Comparison of Characteristics and Reliability Be-
tween Two Types of Adhesiveless 2-Metal TAB Tape," Proceedings Fifth Interna-
tional TAB/Advanced Packaging Symposium, pp. 62-69, 1993.
163. D. Aeschliman. "Multiconductor TAB Tape Materials," Proceedings Second Interna-
tional TAB Symposium, pp. 250-257, 1990.
164. B. Lynch. "Two Metal Layer COT Package Enhancement," Proceedings Fifth Inter-
national TAB/Advanced Packaging Symposium, pp. 74-84, 1993.
165. G. Cox, J. Dorfman, C. Fedetz, T. Lantzer, R. Lawton, and J. Lott. "DuPont Process
for High Performance Two Conductor TAB," Proceedings Second International TAB
Symposium, pp. 222-229, 1990.
166. J. M. Milewski and C. G. Angulas. ''TAB Design for Manufacturability and Perfor-
mance," Proceedings Second International TAB Symposium, pp. 121-132, 1992.
167. S. C. Lockard, J. M. Hansen, and G. H. Nelson. "High Leadcount Multimetal Layer
TAB Circuits," Proceedings Second International TAB Symposium, pp. 214-221,
1990.
8.11 REFERENCES 11-281

168. N. Tajima, Y. Chikawa, T. Tsuda, and T. Maeda. "TAB Design for ESD Protection,"
Proceedings Second International TAB Symposium. pp. 77-87, 1990.
169. R. G. Oswald and W. R. de Miranda. "Application of Tape Automated Bonding
Technology to Hybrid Microcircuits," Solid State Technol., pp. 33-38, 1977.
170. P. Elenius. "Au Bumped Known Good Die," Proceedings First International Sympo-
sium on Flip Chip Technology & Sixth International TAB/Advanced Packaging
Symposium. pp. 94-97, 1994.
171. M. Bonkohara, E. Hajimoto, and K. Takekawa. "Utilization of Inner Lead Bonding
Using Ball Bump Technology," Proceedings Fourth International TAB Symposium.
pp. 86-96, 1992.
172. J. Wolf, G. Chmiel, J. Simon, and H. Reichl. "Solderbumping-A Comparison of
Different Technologies," Proceedings First International Symposium on Flip Chip
Technology & Sixth International TAB/Advanced Packaging Symposium. pp. 105-
110, 1994.
173. T. Yokoyama, H. Ikeda, K. Oshige, M. Kimura, and K. Utsumi. "Bare Chip Bump
Technology Using Advanced Electroplating Method," Proceedings 8th International
Microelectronics Conference. pp. 402-407. 1994.
174. M. Inaba, K. Yamakawa, and N. Iwase. "Solder Bump Formation Using Electroless
Plating and Ultrasonic Soldering," Proceedings 5th IEEElCHMT IEMT Symposium.
pp. 13-17. 1988.
175. S. Khadpe. Introduction to Tape Automated Bonding. pp. 28-29, Semiconductor
Technology Center, Inc., 1992.
176. K. Atsumi, N. Kashima, and Y. Maehara. "Inner Lead Bonding Techniques for
500 Lead Dies Having a 90 ~m Lead Pitch," Proceedings 39th IEEE Electronic
Components Conference. pp. 171-176, 1989.
177. R. C. Kershner and N. T. Panousis. "Diamond-Tipped and Other New Thermodes
for Device Bonding," IEEE Trans. Components Hybrids Manu! Technol.. CHMT-2,
pp. 283-288, 1979.
178. K. Tanaka, "The Performance of a CVD Diamond Bonding Tool," First VLSI
Packaging Workshop of Japan Abstracts. pp. 105-107, 1992.
179. Shinkawa Inner Lead Bonder Brochure, Shinkawa Ltd., Japan, 1992.
180. L. Levine and M. Sheaffer. "Optimizing the Single Point TAB Inner Lead Bonding
Process." Proceedings Second International TAB Symposium. pp. 16-24, 1990.
181. S. Patil. "Chip-on-Tape Single Point Bonding and Interconnect Reliability," presented
at the Second International TAB Symposium. 1990.
182. G. Silverberg, "Single Point TAB (SPT): A Versatile Tool for TAB Bonding,"
Proceedings ISHM Symposium. pp. 449-456. 1987.
183. G. Dehaine, "Ceramic Package Leadframe Replaced by TAB Carrier,"Proceedings
Second International TAB Symposium. pp. 54-61, 1990.
184. G. Kelly. "Automated Single Point TAB-High Quality Bonding for High-Density
Chips," Microelectronic Manufacturing and Testing. 1989.
185. G. Dehaine. "Single Point ILB at Narrow Pitch," Proceedings Fifth International
TAB/Advanced Packaging Symposium. pp. 153-157, 1993.
186. J. D. Hayward. "Preliminary Evaluation of Laser Bonding for TAB ILB," presented
at the Third International TAB Symposium. 1991.
187. J. D. Hayward. "Optimization and Reliability Evaluation of a Laser Inner Lead
Bonding Process," Proceedings Fifth International TAB Symposium. pp. 52-61, 1993.
188. P. Spletter. "Gold to Gold TAB Inner Lead Bonding with a Laser," Proceedings
Fourth International TAB Symposium. pp. 58-71, 1992.
189. E. Zakel, G. Azdasht, and H. Reichl. "Investigations of Laser Soldered TAB Inner
Lead Contacts," Proceedings 41st ECTC. pp. 497-506, 1991.
11-282 CHIP-TO-PACKAGE INTERCONNECTIONS

190. G. Azdasht, E. Zakel, and H. Reichl. "A New Chip Packaging Method Using
Windowless Flip-TAB Laser Connection on Flex Substrate," Proceedings Interna-
tional Flip Chip, BGA, TAB and Advanced Packaging Symposium, pp. 237-244, 1995.
191. N. G. Koopman, T. C. Rieley, and P. A. Totta. "Chip-to-Package Interconnections,"
in Microelectronics Packaging Handbook, pp. 409-435, 1989.
192. T. A. Scharr. ''TAB Bonding a 200 Lead Die," Proceedings ISHM Symposium,
pp.561-565, 1983.
193. M. Hayakawa, T. Maeda, M. Kumura, R. H. Holly, and T. H. Gielow. "Film Carrier
Assembly Process," Solid State Technol., pp. 52-55, March 1979.
194. T. H. Spenser. ''Thermocompression Bond Kinetics: The Four Principal Variables,"
Int. J. Hybrid Microelectron., pp. 404-410, 1982.
195. N. Ahmed and J. J. Svitak. "Characterization of Gold-Gold Thermocompression
Bonding," Proceedings ECC, pp. 52-63, 1975.
196. J. L. Jellison. "Effect of Contamination on the Thermocompression Bondability of
Gold," Proceedings ECC, pp. 271-277, 1975.
197. P. Hedemalm, L.-G. Liljestrand, and H. Bemhoff. "Quality and Reliability of TAB
Inner Lead Bond," Proceedings Second International TAB Symposium, pp. 88-
102, 1990.
198. E. Zakel, R. Leutenbauer, and H. Reichl. "Investigations of the Cu-Sn and Cu-Au
Tape Metallurgy and of the Bondability of TAB Inner Lead Contacts after Thermal
Aging," Proceedings Third International TAB Symposium, pp. 78-93, 1991.
199. D. B. Walshak, Jr. ''TAB Inner Lead Bonding," in Handbook of Tape Automated
Bonding, ed. by J. Lau, pp. 202-242, Van Nostrand Reinhold, New York, 1992.
200. B. Lynch. "Optimization of the Inner Lead Bonding Process Using Taguchi Methods,"
Proceedings Fourth International TAB Symposium, pp. 72-85, 1992.
201. T. C. Chung, D. A. Gibson, and P. B. Wesling. ''TAB Testing and Bum-in," in
Handbook of Tape Automated Bonding, ed. by I. Lau, pp. 243-304, Van Nostrand
Reinhold, New York, 1992.
202. G. G. Harman. "Acoustic-Emission Monitored Tests for TAB Inner Lead Bond
Quality," Trans. IEEE Components Hybrids Manu! Technol., CHMT-5: pp. 445-
453, 1982.
203. R. W. Shreeve. "Automated On-Chip Bum-In System for TAB Reels," Proceedings
39th IEEE Electronic Components Conference, pp. 187-189, 1989.
204. J. L. Kowalski. "Individual Carriers for TAB Integrated Circuit Assembly," Proceed-
ings 29th ECC, pp. 315-318, 1979.
205. K. Fujita, T. Onishi, S. Wakamoto, T. Maeda, and M. Hayakawa. "Chip-Size Plastic
Encapsulation on Tape Carrier Package," Int. J. Hybrid Microelectron .• 8(2),
pp.9-15, 1985.
206. K. Mori, M. Ohsono, and T. Maeda. "High-Reliability TAB Encapsulation Technol-
ogy," Proceedings Third International TAB Symposium, pp. 43-53, 1991.
207. B. Bonitz, M. DiPietro, and 1. Poole. "Encapsulation of TAB Devices," Proceedings
Fourth International TAB Symposium, pp. 133-139, 1992.
208. G. Zimmer. "Using Advanced Pulsed Hotbar Solder Technology for Reliable Posi-
tioning and Mounting of High Lead Count Flat Packs and TAB Devices," Proceedings
Second International TAB Symposium, pp. 230-249, 1990.
209. G. Westby, R. Heitmann, and S. Haruyama. "Hot Bar Soldering for TAB Attachment,"
Proceedings Fourth International TAB Symposium, pp. 332-341, 1992.
210. Advanced Packaging Applications: Tape Carrier Package, Addendum to Intel's
1995 Packaging Databook, pp. 17-18, Intel Corporation, 1995.
211. T. Kleiner. "Excise & Form Tooling Principles," Proceedings Second International
TAB Symposium, pp. 62-76, 1990.
8.11 REFERENCES 11-283

212. E. Zakel, G. Azdasht, P. Kruppa, and H. Reichl. "Reliability Investigations of


Different Tape Metallizations for TAB-Outer Lead Bonding," Proceedings Fourth
International TAB Symposium, pp. 97-120, 1992.
213. M. C. Miller and 1. Glazer. "Equipment Selection for TAB Outer Lead Bonding,"
in Handbook of Tape Automated Bonding, ed. by J. Lau, pp. 338-357, Van Nostrand
Reinhold, New York, 1992.
214. M. Kasahara, T. Oyanagi, and Y. Honma. "Hot Bar Technology for Gang Bond,"
Proceedings 8th International Microelectronic Conference, pp. 399-401, 1994.
215. A. Eldred, R. Heitmann, and M. Yingling. "TAB Outer Lead Bonding at 96 Micron
Pitch," presented at Fifth International TAB/Advanced Packaging Symposium, 1993.
216. Fine Pitch Bonder Brochure, Universal Instruments Corporation, 1991.
217. I. Kurashima. "The New Technology of Anisotropic Conductive Adhesive for Fine
Pitch TAB," Proceedings Third International TAB Symposium, pp. 212-220, 1991.
218. E. Hosomi, C. Takubo, Y. Hiruta, T. Sudo, F. Takahashi, K. Saito, N. Nakajima,
and M. Nakazono. "The Wettability and Bondability Degradation Mechanisms of
Tin-Plated Copper Leads of a Fine-Pitch TCP during Thermal and Humid Aging,"
Proc. IMC 1994, pp. 128-133, 1994.
219. F. Andros and R. Hammer. "IBM's Area Array TAB Packaging Technology,"
Semicond. Packaging Update, 8(3): pp. 1-8, 1993.
220. J. Walker. "TapePak: The Semiconductor Package for the Future," Natl. Semicond.
Technol. Rev., 1: pp. 69-72, 1988.
221. D. Card. "How ETA Chose to Make a Megaboard for its Supercomputer," Electron.
Bus., pp. 50-52, July 1, 1988.
222. S. Kbadpe. "Worldwide Activities in Flip Chip, BGA and TAB Technologies and
Markets," Proceedings 1995 International Flip Chip, Ball Grid Array, TAB and
Advanced Packaging Symposium," pp. 290-293, 1995.
223. "New Mounting Methods to Directly Connect LSI Chips with a Large Number of
Pins," Nikkei Electronics Microdevices, June 11, 1984.
224. K. Smith. "An Inexpensive High Frequency High Power VLSI Chip Carrier,"
IEPS, 1985.
225. K. Hatada, H. Fujumoto, and T. Kawakita. "Insulation Resin Bonding: Chip on
Substrate Assembly Technology," ICICE, 1987.
226. K. Chung, R. Fleishman, D. Bendorovich, M. Yan, and N. Mescia. "Z-Axis Conduc-
tive Adhesives for Fine Pitch Interconnections," ISHM Proceedings, pp. 678-689,
1992.
227. Epoxy Technology, Inc. Product Literature, 1992.
228. N. Masuda, K. Sakuma, E. Satoh, Y. Yamasaki, H. Miyasaka, and J. Takeuchi.
IEEElCHMT Japan IEMT Symposium, pp. 55-58,1989.
229. H. Kristiansen and A. Bjomeklett. "Fine Pitch Connection of Flexible Circuits to
Rigid Substrates Using Non-Conductive Epoxy Adhesive," IEPS, pp. 759-773, 1991.
230. K. Sakuma, K. Nozawa, E. Sato, Y. Yamasaki, K. Hanyuda, H. Miyasaka, and J.
Takeuchi. "Chip on Glass Technology with Standard Aluminized IC Chip," ISHM,
pp. 250-256, 1990.
231. M. F. Bregman. Private communication.
232. G. A. Katopis. "Delta-I Noise Specification for a High Performance Computing.
"Machine," Proc. IEEE, 73: pp. 1405-1415, 1985.
233. K. Ohtsuka. "Special Article 2 Diversifying LSI Packages 1," Nikkei Electronics
Microdevices 2, June 11, 1984.
9
CERAMIC PACKAGING
RAO R. TUMMALA-Georgia Tech
PHIL GARROU-Dow Chemical
TAPAN GUPTA-Westinghouse
N. KURAMOTO-Tokuyama Soda
KOICHI NIWA-Fu;itsu
YUZO SHIMADA-NEC
MASAMI TERASAWA-Kyocera

9.1 INTRODUCTION
Ceramics and glasses, defined as inorganic and nonmetallic materials,
have been an integral part of the information-processing industry. Glasses
and ceramics are used for data processing by the use of semiconductor
devices interconnected onto packages. They are also used for storing
information by the use of such magnetic materials as iron and chromium
oxides and ferrites; for displaying information by the use of low-tempera-
ture solder glasses sealed to transparent faceplate glasses; for printing
information by the use of ceramic-metal print heads and corrosion-resis-
tant glasses; and for transferring information by the use of very-low-loss
glass fibers. The materials currently used for these applications are shown
in Table 9-1.
Integrated-circuit (IC) packaging is one of the most important appli-
9.1 INTRODUCTION 11-285

Table 9-1. Application of Ceramics in Microelectronics


Information
Processing Ceramic
Function Application Materials

Information processing
Package Substrate AhO], BeO, AlN
Device Dielectric Sh N4, Si02
Mask Borosilicate glass
Information storage
Disk Iron oxide, Ferrite
Tape Chrome oxide
Head Ferrites, glass
AhO] + TiC substrate
Information display
Dielectric Lead borosilicates
Seals Lead zinc borosilicates
Faceplate Soda-lime glass
Information printing
Ink jet Zr02-containing glass
Electro-erosion Ceramic-metal composites
Information transfer
Optical fiber

cations of ceramics in microelectronics. Ceramic materials in the form


of dual-in-line packages, chip carriers, and pin-grid arrays are used
throughout the industry, from consumer electronics to mainframe comput-
ers. Although plastic packages outnumber ceramic packages, ceramic
packages constitute roughly two-thirds of the total packaging market,
which is currently considered a multibillion dollar industry worldwide.
The Japanese market illustrated in Figure 9-1 is considered roughly two-
thirds of the total market, as it does not include the ceramic substrates
manufactured by United States and European computer and ceramic manu-
facturers. The ceramic package market in the United States along with a
list of ceramic package suppliers to this market and their market share
are shown in Tables 9-2 and 9-3, respectively [1]. It should be noted,
again, that these numbers do not include those markets for internal con-
sumption of individual companies, such as IBM, which manufactures
ceramics for its own high-performance applications.

9.1.1 Unique Attributes of Ceramic Packages


9.1.1.1 Unique Properties
Ceramics possess a combination of electrical, thermal, mechanical,
and dimensional stability properties unmatched by any group of materials.
11-286 CERAMIC PACKAGING

1000 l#tWl@fl Others


US$ 1 = 200 Yen
900 ~Narumi

mwmNTK Total
800 .2 Bill.
~Kyocera
700

800
0
Ul
c
~ 500
~
400

300

200

100

0
1994
Year

Figure 9-1. Japanese Ceramic Companies' Shipments

Table 9-2. Package Suppliers to U.S. Market. Market share based on sales volume.

Kyocera 55-70
NTK 15-25
Sumitomo 5-10
ALCOA <4
COORS <4
Secondary
American Hoescht (Metceram) <2
Cabot <2
RaychemlInteramics <2
General Ceramics <2
Brush Wellman (EPI) <2
LDC <2
Diacon <2
Bourns <2
Shinko <2
TOTAL 100

Source: Data from Ref. 1.


9.1 INTRODUCTION 11-287

Table 9-3. Worldwide Ceramic and Other Packaging Materials Market


Worldwide Materials Consumption ($ millions)

1982 1988 1993 1996 2000

Ceramic Packages, Metallized 230 1,073 1,230 1,480 1,300


CERDIP 200 363 IIO 80 50
Multilayer Capacitors 350 610 1,800 2,080 2,470
Leadfrarnes 600 1,146 1,830 2,460 3,010
Encapsulation Resins 180 394 565 680 780
Bonding Wire II7 253 363 479 710
Seal Lids with Preforms 190 222 IIO 135 120
Die Attach Materials 55 75 100 ll8 140
Headers and Cans IIO 105 90 85 60
Hybrid Materials 340 402 314 350 290
Other 90 175 200 230 390
$2,462 $4,208 $4,912 $6,097 $9,320

For example, ceramics have dielectric constants from 4 to 20,000, thermal-


expansion coefficients matching silicon (3 x lO-{i/°C) to matching copper
(17 x lO-{i/°C), and thermal conductivities from that of insulated brick
(0.1 W1m K) to better than aluminum metal (220 W1m K), which is one
of the best metallic thermal conductors known. Dimensional stability as
measured by either change of dimensions as a function of temperature and
time or shrinkage control during substrate processing has been achieved at
better than ±O.05% nominal shrinkage, allowing as many as 100 layers
of ceramics and metal to be cofired into one substrate [2]. Such a substrate
is illustrated in Figure 9-2 and is discussed in the glass-ceramic section
of this chapter. A number of manufacturers, such as IBM, are routinely
manufacturing these type of substrates but with a lower number of layers
with nominal zero shrinkage and tolerance better than ±0.1 %. Recent
technology breakthroughs have led to the use of ceramics as superconduc-
tors, offering no electrical resistance to the flow of current at low tempera-
tures. Figure 9-3 illustrates the development in this area over the last eight
decades. As a result, a number of commercial applications in medical,
electronic, and transportation industries are being developed. If the temper-
ature could be pushed closer to room temperature, a direct and profound
application for power distribution of electronic packages could be de-
veloped.

9.1.1.2 Highest Wiring Density


Ceramic substrates provide the highest wiring density of all substrate
technologies. Even though the line and via dimensions as well as their
11-288 CERAMIC PACKAGING

Figure 9·2. State-or-the Art 166mm Ceramic Substrate with 107 Layers.
(Courtesy of IBM.)

pitch are coarser and further apart compared to thin-film technologies, as


illustrated in Figure 9-4, and hence the wiring density is lower per layer,
the total wiring density that has been achieved is much higher, because
of cofiring of as many as 63 layers in full production and lOO layers in
development. This advantage over thin-film (MCM-D) and printed wiring
board (MCM-L) is reflected in Figure 9-5 for a 20-layer cofired ceramic
and a 5x projection for a lOO-layer substrate.

9.1.1.3 Best Reliability and Hermeticity

The only packages that achieved the highest reliability have been
ceramic packages and, as a consequence, the packages that are in use
today for the very stringent reliability requirements are ceramic packages.
These uses include most of the defense and aerospace applications and
the leading-edge microprocessors (Pentium and Power PC) for commercial
applications. This reliability superiority over plastic or printed wiring
board is due to three fundamental reasons. First, by their very nature,
ceramics are hermetic, as discussed in Chapter 14, "Package Sealing and
Encapsulation." They do not absorb and retain moisture nor do they allow
permeation of gases. Second, their dimensional stability during and after
high-temperature processing is exceptional. Some of the advantages of
9.1 INTRODUCTION 11-289

v.s..cu.()
100 O·························~ No Mojor Change.
: Feb. 18, 1..7

80

Uquid N,
80

70

1910 1930 1950 1970 1990 2010

Figure 9-3. Progress and Role of Ceramics in Superconductors

this stability come from the intrinsic low thermal expansion, similar to
that of silicon IC devices. Third, the chemical intemess of most of the
ceramics to water, acids, solvents, and other chemicals is outstanding.
The IBM's better than six-sigma achievement for its ceramic multichip
module (MCM) reliability, as illustrated in Figure 9-6 and as discussed

Thin Film MLC PWB

• • o o

Via/land: Via/land: Via/land:


0.030/0.04mm DIA. 0.20/0.40mm DIA. 0.30/0.7Smm DIA.
Une/space: Une/space: Une/space:
0.02S/0.050mm O.l00/O.l00mm 0.127/0.127mm

30 Iinesl1 OOMIL 9Iines/100MIL Slines/100MIL

Figure 9-4. Multilayer Ceramic and Printed Wiring Board Compared. (Cour-
tesy of Kyocera Corp.)
11-290 CERAMIC PACKAGING

100
umber or Llyns

Figure 9-5. Wiring Density Comparison

in Chapter 6, "Manufacturing Considerations for the Packaging Technolo-


gist" is a good example of ceramic reliability which has not been achieved
by any organic package.

9.1.2 Drawbacks of Ceramics


Ceramics are brittle, making them prone to catastrophic failure and
sensitive to stress corrosion unless designed carefully for each application

1,000,000 IRS Tax Advice

10,000 66.8K Restaurant Billing


6.21 K Payroll Processing
233 Airline Baggage Handling
.5 100 3.4

IBM'sTCM:
0.01
5 Billion Component 0.018
Hours Without Fail

0.0001
2 4 6 8
Sigma

Figure 9-6. Multilayer Ceramic Reliability. (Courtesy of IBM.)


9.2 EARLY CERAMIC PACKAGING 11-291

for which they are used. In very-high-performance applications requiring


very low dielectric constants and very high packaging densities, ceramics
cannot compete with thin-film organic packages. However, ceramics are
expected to be used as stable building blocks on which high-performance
thin-film wiring can be developed with the use of polymeric materials
for high-performance applications.
Ceramic packages offered in a variety of forms, such as dual-in-
line packages (DIPs), chip carriers, flat packs, and pin-grid arrays are
capable of supporting from 1 chip to as many as 100 or more chips per
package, as discussed later in this chapter.
In this chapter, the history and evolution of ceramic packaging are
discussed, from simple dual-in-line and solid logic packages to the state-of-
the-art ceramic processing, as in fabricating multilayer packages capable of
interconnecting in excess of 30 ceramic layers with 30 layers of metal to
support devices in excess of 100 chips on a single substrate with all the
required power distribution and thermal dissipation to and from each of
the devices. Recent developments in the use of ceramics with thermal-
expansion coefficients near that of silicon, such as AIN, mullite, glass ±
ceramics, are also included. Future technology trends and the scientific
areas required to develop the understanding necessary to support the new
product developments are indicated.

9.2 EARLY CERAMIC PACKAGING


Early transistors were either solder bonded to glass-epoxy cards or
packaged in simple plastic packages (as shown in Fig. 9-7). The need for
hermetic sealing led to the development of the outline metal packages,

.~~
.
,:.
-. ...
~~,~
/ '~'A.
. / '/ , /' ~~),.
;-'
.'
/
/ . /. # i ) ...... ,
/
" . ".' "

~ ;:~;
. ........
'
.,

Figure 9-7. TO Transistor Bonded to Glass-Epoxy


11-292 CERAMIC PACKAG ING

which were sealed by welding nickel lids to gold-plated transistor


metal bases.
The initial transistor outline (TO) package was manufactured with
as many as 14 leads. In the 1960s the need for many more leads led to
the development of rectangular ceramic packages with leads on two oppo-
site sides, known as dual-in-line packages or planar packages. These
packages were introduced by Fairchild.

9.2.1 Ceramic Solid Logic Technology

At about this time (1963), IBM introduced for its own use a ceramic
package called SLT (Solid Logic Technology) [3], 12 x 12 mm in size,
made of 96% dry-pressed alumina on which conductor and resistor pastes
were screened and fired at about 800°C. Sixteen connector pins of copper
were swaged into the holes in the substrate, and the entire assembly was
immersed in a solder bath to coat the conductors and pins, forming
electrical contacts between pins and conductor lines. Semiconductor chips
were soldered in place and the cap sealed to the module with epoxy (as
shown in Fig. 9-8).

9.2.2 Ceramic Dual-in-Line Packages

The availability of dual-in-line packages (DIPs) and the ease with


which they can be transported and automatically inserted into the holes of
printed-circuit boards and soldered in place made them a industry standard.

Figure 9-8. SLT Package


9.2 EARLY CERAMIC PACKAGING 11-293

Ceramic DIPs were fabricated using either ceramic greensheets or


dry-pressing processes. Figure 9-9 illustrates the tape processes used
in fabricating DIPs, and Figure 9-10 illustrates the same with a dry-
pressing process.
With either process, the maximum number of leads provided with
DIP packages has been 64. This limit is due to fabrication difficulties,
the question of the reliability of solder joints to the board, and excessive
lead inductance, among other reasons. The tape process involves forming
a slurry of well-dispersed alumina and glass in a suitable organic vehicle,
and doctor blading to form tape, which is metallized with tungsten metal
paste. The desired number of layers is laminated and sintered in a con-
trolled atmosphere between 1500°C and 1600°C. Exposed tungsten metal-
lizations are nickel and gold plated, and then brazed with a copper-
silver alloy.
Pressed ceramic (also known as "Cerdip"), as the name implies,
involves molding a mixture of alumina lubricants and binders to form
the desired shape, which is then sintered to form a monolithic body. Metal
frames are bonded to this body by using low-temperature glasses, as
shown in Figure 9-10.
Hermetic sealing posed problems early in the development of DIPs.
These problems were traced to the incomplete removal of water prior to
sealing. Glasses in the system PbO-B203-Si02-AI203-ZnO, capable of

Vehicle
"Green" Ceramic Tape Mfg.

t
Screen Metallized (Tungsten) Petterns

t
Laminate Green Ceremic Layers

•
Fire Ceramic 1600'C N2/H2 ATM

Nickel Plating of Exposed Metallization


r¥,
Pressure

Braze Metal Components


to Ceramic Body ICu-Ag Breze)

Final Gold Plate of Exposed


Metal Surfaces

Figure 9-9. DIP-Laminated Ceramic Package. Also called side-brazed package.


11-294 CERAMIC PACKAGING

Lld---

- - - O l e (Wire
Bond After
Ole Attach)

Sidebraze
Package -

Figure 9-10. Laminated Ceramic Package. Also called a side-brazed ceramic DIP.

meeting all the requirements shown in Table 9-4, have been successfully
developed [4] and used to solve the hermeticity problems.

9.2.3 Advanced SLT

Advanced SLT (ASLT) [5] was developed as a follow-up to SLT


during the mid-1960s. The basic fabrication technology followed the SLT
process closely, and the additional circuit wiring is added by screening
conductors on both sides of the substrate. Two such substrates are stacked
vertically, as shown in Figure 9-11. Electrical connections between the

Table 9-4. Glass Requirements for Cerdip Sealing

Low-temperature sealing «450°C)


Good adhesion to both Ab03 ceramic and alloy 42
Thermal expansion matched to ceramic and alloy 42
Electrically insulating
Good chemical resistance
No outgassing of moisture of other contaminants
Low alpha-particle emission for soft error protection
Good fracture toughness
High thermal-shock resistance
9.2 EARLY CERAMIC PACKAGING 11-295

Figure 9-11. Stacked Solid Logic Technology

substrates are provided by soldering the pins from the bottom of the top
substrate to the top-surface metallization of the bottom substrate.

9.2.4 Monolithic Systems Technology

The monolithic systems technology (MST) [6], consistent with im-


provements in semiconductor integration at about 25 circuits per 2-mm-
size chip, was developed in the late 1960s. This technology is very similar
to the SLT and ASLT, but the number of chip connections increased from
3 and 12 with the previous technology to 16 and 18, respectively.

9.2.5 Thick-Film (Crossover) Technology

Parallel to dry-press ceramic technology, thick-film technology be-


gan to evolve as a way of depositing thick-film dielectrics and conductors
by screen-printing processes. Thick-film technology involving the physics
and chemistry of metal and ceramic powders, the organic vehicle systems
capable of providing stable thick-film systems, the measurement systems
of viscosity, surface tension, and dispersion are discussed in a European
handbook [7] . The conductor materials typically used are those that can
be fired below lOOO°C with good electrical conductivity. These include
silver, gold, copper, gold-platinum, silver-palladium, silver-platinum,
and palladium-gold. The choice of conductor is usually dependent on
one or more of the following considerations for the pastes shown in Table
9-5:
11-296 CERAMIC PACKAGING

Table 9-S. Primary Components of Some Common Early Resistor Systems

Manufacturer Series Components

E.1. du Pont 7800 Palladium oxide/silver/palladium


8()()()
Electroscience Laboratories 6900 Palladium oxide/silver/palladium
7()()()
E.1. du Pont Birox Bismuth ruthenate
1100
1400
EMCA Firon Ruthenium oxide/iridium oxide/gold
Plessy Co. (formerly Alloys Unlimited) A Ruthenium oxide/silver
B
BB
Chicago Telephone Supply Ruthenium oxide
Airco Spear TGA Thallium oxide
TFG
Electroscience Laboratories 3800 Iridium oxide

1. Resistivity
2. Line definition
3. Compatibility with resistor, dielectric, or other conductor materials
4. Solderability and solder leach resistance
5. Adhesion strength
6. Suitability for ultrasonic wirebonding
7. Suitability for chip bonding
8. Long-term stability (e.g., resistance to migration effects under the
influence of voltage gradients)

Glass (frit) additions, generally in the lead-borosilicate family, were


included in the paste to provide adhesion to the ceramic on which they
were fired. The resistivity that can be achieved with these thick films is
in the range of 5-25 J.1Q cm.
The dielectrics generally fall into two categories: those involving
low-dielectric-constant materials and others involving very-high-
dielectric-constant capacitor materials. Each of these have been further
classified into those involving crystallizing glasses and those involving
mixtures of glasses and ceramics. In addition, thick-film resistors (Table
9-5) are processed in a similar fashion.
9.2 EARLY CERAMIC PACKAGING 11-297

9.2.6 Ceramic Packages with Thin Films


With the development of more highly integrated semiconductors, the
package-wiring-density requirement became so great that either mUltiple
layers of thick film or a higher density of surface wiring, using thin films,
was required. The metallized ceramic shown in Figure 9-12 was fabricated
with thin-film wiring on the surface of a dry-pressed alumina substrate
[8]. A very thin layer of chromium, followed by a thicker film of copper,
and finally a thin layer of chromium were either sputtered or evaporated
onto the surface of the ceramic. The first chromium layer was expected
to form chromium oxide during substrate processing to form an excellent
bond to alumina. The metal conductor patterns are defined by subtractively
etching fine lines by conventional lithographic processes. The pinning of
the substrate is performed in a way similar to the swaged-pin processes
described earlier. The chip connections are performed by the flip-chip
solder connections, described in detail in Chapter 8, "Chip-to-Package
Interconnections. "

9.2.7 Earlier Multilayer Ceramic Technology


The origin of multilayer ceramic (MLC) technology can be traced
to RCA in the late 1950s [9], starting with the doctor-blade process for
casting greensheets, laminating a number of metallized layers [10], and
finally leading to the concept of "via" interconnection from layer to layer
[11]. At about this time, the U.S. Army Signal Corps and the Radio

Figure 9-12. Thin Film on Ceramic Package


11-298 CERAMIC PACKAGING

Corporation of America, in Somerville, New Jersey, started the Micromod-


ule Project [12]. The goal was to find a practical means to increase the
value of capacitance available to module designers by a factor of 10. Both
fabrication technology and dielectric composition improvements were
studied. This project was designed to use 7.6 x 7.6-rnrn substrates, with
three notches on each of the four edges and thicknesses greater than 0.25
mm. The first goal was to develop a laminated capacitor in this form
factor. To accomplish this goal, it was necessary to meet the capacitance
requirements of up to 3000 pF for a precision capacitor and up to 300,000
pF for a general-purpose capacitor. The equation used to make these
calculations was

C = 9.6 KAN [9-1]


t

where
A = the area of the electrode
C = the capacitance (in pF)
K = the dielectric constant
N = the number of layers
t = the thickness of each layer

For a precision capacitor, the dielectric constant was about 30, and for a
general-purpose capacitor it was about 3000, which meant that at least
11 layers of 25 J.lm thickness would have to be achieved in order to meet
these specifications. The area of each electrode was about 26 J.lm2. Many
technologies were tried concurrently until doctor-blading started to show
the most promise. Fortuitously, the paint industry was then producing
vinyl-based products, and after several months of trials, a plasticized
copolymer of polyvinyl chloride acerate binder was discovered. These
capacitors had 11 layers, each 25 J.lm thick, which when sintered met the
specifications described. The process steps used are shown in Figure 9-13.
In order to produce a stable slurry with high viscosity and specific
gravity that could be used for casting films, the interactions of the ceramic
powders, binders, plasticizers, deftocculants, and solvents had to be opti-
mized [13,14]. The early studies correctly pointed out, as we know now,
that steric hindrance is the primary mechanism for dispersing ceramic
powders in organic systems. Because of the increased interest in tape
casting of ceramics for much better dimensional requirements, new re-
search programs have recently been initiated [15] which should improve
the understanding of the fundamentals controlling the dispersion behavior
of these systems. The casting process was first done with a hand doctor
blade, but it was quickly replaced with a continuous caster using a plastic
9.2 EARLY CERAMIC PACKAGING 11-299

1. Ball Mill Slip


• Ceramic Dielectric (Titanates) Powder
• Thermo-Plastic Binder and Plasticizer
• Deftocculant
• Solvents
2. Doctor Blade a Tape on a Plastic Carrier
3. Dry, Strip, and Cut into Sheets
4. Screen Noble-Metal Paste Electrodes and Dry
5. Laminate Sheets to Design
6. Cut Units to Size
7. Connect Alternate Layers at Edges (Notches)
8. Sinter in Air
9. Tin Terminals
10. Test
Figure 9-13. Early Multilayer Ceramic Flow Sheet for Capacitors

carrier material. The green sheets could be easily stripped from it and in-
spected.
These sheets were metallized with several of the existing noble metal
pastes then commercially available. Palladium was preferred for capacitor
applications, because it could be fired in air and withstand the temperatures
needed to sinter the titanates. Screening processes provided adequate
resolutions for the electrodes.
Assembling the titanate sheets was the most difficult problem to
solve. The first attempt was to sequentially cast the ceramic sheets, dry,
screen the electrodes, dry and so forth. This method had so many problems
with drying, thickness control, yield, and layer-to-Iayer connections that
it led to the development of lamination technology. The criteria for proper
greensheet structure and laminating conditions have been described more
quantitatively elsewhere [16]. The bond strength between the green ce-
ramic layers is determined by the following relationship:

where
Kh K2 = constants for the system used
J> = the pressure
T = the temperature
t = the time of hold at the laminating temperature.
11-300 CERAMIC PACKAGING

All the process steps in fabricating these capacitors are listed in


Table 9-6.
About this time in the micromodule program, a new need arose for
an alumina hermetic package for a quartz crystal oscillator application
that led to the development of alumina multilayer ceramic. This was
believed to be the first laminated alumina multilayer ceramic package
application. All of the processes described were used, except for the
binder. In this application a polyvinyl butyral resin was used for the thicker
green sheets. This package used a Mo-Mn paste for its metallization, in
order to achieve hermeticity. The iron-nickel-cobalt cap was ultrasoni-
cally soldered to a nickel and gold-plated sealing ring on the surface of
the three-layer substrate. Because of this breakthrough with the quartz
crystal oscillator package, other packages were made for diodes and
transistors. Next, the concept of a "via," as illustrated in Figure 9-14, was
introduced [11]. This is a hole punched in a green ceramic sheet and
subsequently filled with a metal paste in order to make the interconnections
from layer to layer within the micromodule substrate. These historical
developments are discussed further elsewhere [8]. These basic develop-
ments provided the background for the development of the state-of-the-
art multilayer ceramic described later in the chapter.

9.2.8 Ceramic Chip Carriers, Flat Packs, and Hybrid Packages


The ceramic dual-in-line package with up to 64 pins on a 2.54-mm
pitch became too large for high-performance applications. The ceramic
chip carriers, which can be considered as square DIPs with terminals on
all four sides at half the pitch of the DIP, have filled an important gap
between DIPs and pin-grid arrays. Chip carriers can also be used exten-

Table 9-6. Thick-Film Processes Considered in 1959

Estimated Minimum
Thickness (1lII1)

Pressing 125 Too thick


Extruding 75 Control
Spraying 75 Shorts
Doctor-blading Unknown Lamination
Screening 15 Shorts
Electrophoresis 2 Shorts
Sputtering or evaporation 0.1 Voltage rating

Source: Data from Ref. 12.


9.2 EARLY CERAMIC PACKAGING 11-301

3,1....,.

rU04 AprU :n. 1_

lOS'
110'
112'

Figure 9-14. Patent Drawings Showing First Ceramic Via. Figure numbers shown
in art-work are from patent. (From Ref. 10)

sively in low pin counts to replace DIPs as, for example, in high-density
memory packaging. Chip carriers (CC) can be either leaded (LCC) or
leadless (LLCC), depending on how it is applied to the second-level card
or board technology. An example of a ceramic chip carrier is illustrated
in Figure 9-15. These packages are made by using the tape process for
multilayer structures and the dry-press process for single-layer DIPs and
other plug-in packages (see Fig. 9-16). The materials and processes for
11-302 CERAMIC PACKAGING

...-BASE

ASSEMBLED

Figure 9-15. Ceramic Chip Carrier. (Courtesy of Intel Corp.)

multilayer substrates are similar to the state-of-the-art ceramic packaging


covered in the next section, but the dry-press technology has also been
applied to the formation of metallized substrates by first firing in air,
followed by screening and sintering the desired metallization. Cavity
metallization is accomplished similarly to end up with Cerdip bases. A
number of designs (leadless designs A to D and others) of ceramic packages
[17], both in pinned and nonpinned versions, have been developed by
manufacturers exclusively for their customers' needs. One of these is the
ISOPAK pin-grid array developed by General Dynamics [18], where the
pins of Kovar are glass-sealed into a chemically milled Kovar pan, which
is flush filled with glass and planarized to provide the chip-bonding circuits.
The chip itself is mounted on a Kovar platform at the center of the package.
So-called multiple-in-line packages (MIP) are rectangular pin-grid arrays,
in which the pins are in four rows of 2.5 mm spacing, and the spacing
between the first and second, and third and fourth is 7.5 mm. PINBELL,
used by AT&T Bell [19], is of this type. In addition to other chip carriers,
one variation of leadless chip carrier type, known as single-layer metallized
(SLAM), is used primarily by the military market for direct solder attach-
ment to a ceramic substrate [20]. This package has no recess and is sealed
by glass as discussed in Chapter 14 "Package Sealing and Encapsulation."
9.2 EARLY CERAMIC PACKAGING 11-303

Multilayer Ceramic Press Ceramic

Sintering
(In Reduced
Atmosphere)

i l i 1
r-l--+
Finalinspedi~ Rnalinspectl~ I Final Inspection I
L
--T--
and O.A.

Chip Carriers
...J L
--l---J
and O.A.

Plug-In Packages
I
L andO.A. ~

Cerdip Bases Cerdip Caps


Ceramic Boards Single Layer DIPs Cerpac Bases Cerpac Caps
Plug-In Packages Hybrid Packages Frit Lids
Flat Packages
Multilayer DIPs
Lead Frame
Hybrid Packages
Embedded Bases
Opto-Electronic Packages
Hysubpak®s

Figure 9-16_ Process of Making Alumina Packages (Chip Carriers, Flat Packs,
DIPs). (Courtesy of Kyocera Corp.)
11-304 CERAMIC PACKAGING

Hole-grid array, as opposed to pin-grid array, with holes in ceramic is


another variation. The available ceramic chip carriers and DIPs, and their
characteristics, are given in Table 9-7, which illustrates the availability
of ceramic carriers with I/O spacing down to 0.50 mm (20 mils) and the
number of I/Os in excess of 200. The I/O configurations available from
Kyocera for these chip carriers are illustrated in Figures 9-17 and 9-18.
Figures 9-19 and 9-20 compare the three most common ceramic packages
[21] for the area they occupy on second-level organic packages and
the signal delay involved in interconnecting 12 silicon chips. For high-
performance logic applications, it can be concluded from these figures
that the pin-grid array is the best solution, as it uses the least area on the
board while providing the fastest signal transmission.
The best package solution may not be entirely technical. In addition
to the technical aspects, package decisions are made based on existing
assembly methods and the costs of developing and implementing a total
package system with minimum risk. Thus, each company must consider
its own requirements before selecting a particular package solution.
The use of leadless ceramic chip carriers has fulfilled an important

Table 9-7. Available (1987) Ceramic Chip Carriers, DIPs, PGAs, and
Their Leads

110 Spacing
Package ~ 110 Range (mm)

BOA Area array balls >300 1.00


PGA Cavity up 64-312 2.54
1.27 Staggered
2.54
Cavity down 64-240 1.27 Staggered
Leadless JEDEC A 28-156 1.27
JEDEC B 28-156 1.27
JEDEC C 16-96 1.00
Custom 1 200 0.50
Custom 2 284 0.28
Leaded Flat package 10-28 1.27
Quad lead 24-124 1.27
Quad lead 36-64 1.00
Quad lead 172-224 0.63
Quad lead 35-224 0.50
DIP Standard 8-64 2.54
Skinny 24-32 2.54
Shrink 24-64 1.78
Narrow 24-44 1.27
9.3 ALUMINA CERAMIC PACKAGING 11-305

Leaded Package

PIN/LEAD CENTERLINE SPACING


CONFIGURATION
STANDARD CUSTOM HDCM

1) SIDE BRAZE fBbs0~ -2.54


.100
1.27
-.050
1.27
-.050

2) TOP BRAZE ~
1.27
.OSO
-0.508
.020
0.S08
.020

1.27
3) BOnOM BRAZE
(Flat leads) ~~~ -0.508
.020
-0.5OB
.020
.050

~'~~ .11L J2L


4) eonOM BRAZE 2.54
(L shape leads) .100 .050 .050

-1.778 1.778

~
5) PIN GRID ARRAY 2.54 .070 .070
(Nail head pin) .100 (Diagonal) (Diagonal)

~~~
1.905 1.905
6) SIP
(Nail head pin)
-2.54 -.075 -.075
.100

J""LcCl 1.27 1.27


7) SIP
(Flat leads) P~~OTHER80AR~1 -2.54
.100 .050 .050

Unit: : :

Figure 9-17. Leaded Package 110 Configurations. (Courtesy of Kyocera Corp.)

need in military applications but posed limitations in commercial applica-


tions [22]. Although leadless chip carriers can be fabricated to as low as
0.25 mm (10 mil) lead spacing, surface mounting poses a limit to the
size or number of pins a chip carrier can have. This limit has been observed
to be about 30-50 mm in size for a standard 96% alumina substrate
surface soldered to FR-4 epoxy-glass board [23]. Surface mounting of
larger ceramic chip carriers, therefore, requires the development of second-
level packages that are closer in thermal expansion than conventional
epoxy-glass boards [24], as discussed in Chapter 17 , "Printed-Wiring
Board Packaging." The surface mount limitations of chip carriers are
discussed in Chapter 16, "Package-to-Board Interconnections."

9.3 ALUMINA CERAMIC PACKAGING


The state-of-the-art multilayer ceramic described in this section is
excerpted from an earlier publication [25]. The multilayer ceramic packag-
11-306 CERAMIC PACKAGING

L.adl ... Package

PIN/LEAD CENTERLINE SPACING


CONFIGURATION
STANDARD CUSTOM I HDCM
i.27 I
.050 i
1) CHIP CARRIER
(Castellatiom) ~ &
1.016
0.762
.030
I
.Qru
.030
.040
-.
1.27
2 CHIP CARRIER .050

~
0.762 0.635
&
) (Edge Metallization)
1.&1§ .030 :025.
.040

1.27

~.
.050
3) CHIP CARRIER 0.635 0.508
(Via Holes) &
.025 .020
1.016
.040

4) PAD GRID ARRA't .~~--1 .Ln.


.050
0.635
.025
0.608
.020

~
Unit:~

Figure 9-18. Leadless Package 110 Configurations. (Courtesy of Kyocera Corp.)

ing capable of providing interconnections for as many as 100 chips cur-


rently in a single monolithic alumina-molybdenum substrate is considered
the state of the art in ceramic packaging [26]. It should be noted, however,
that alumina-tungsten multilayer ceramic, the technology of which is
very similar to the technology of alumina-molybdenum technology dis-
cussed in this section, is the technology most commonly used in the
majority of commercial applications. This technology, which has been in

40
N'
E
..2-
"E 30
as
0
aI
·s
e
u 20
"C
CD
E
it
'0 10
as
e
-<
0
24 64 100 144
Number of Pins/Package

Figure 9-19. Comparison of DIP, Chip Carrier, and PGA for Second-Level
Area. (Courtesy of Standard Elektrik.)
9.3 ALUMINA CERAMIC PACKAGING 11-307

64 100
Number of Pins/Package

Figure 9-20. Comparison of DIP, Chip Carrier, and PGA for Signal Delay.
(Courtesy of Standard Elektrik.)

existence since the early 1980s [24], has been the basis of ffiM's main-
frame technology, as it provided a revolutionary approach in interconnect-
ing a very large number of chips, providing power to each chip and
removal of heat from each chip in a way that was not practiced earlier.
A number of Japanese and American manufacturers have developed their
own single and multichip, multilayer, ceramic substrate technologies that
are variations of the materials, processes, and tools discussed here. Table
9-8, for example, illustrates the characteristics of multilayer ceramic sub-
strates as manufactured by Kyocera. The impact of this technology has
been to provide very high performance by minimizing the interchip wiring.
It also provides very high reliability due to minimization of the number
of interconnections and elimination of one level of packaging. In addition,
this ceramic package provides hermeticity for the silicon devices. It is
expected that this technology will continue as the basis on which advances
will be made in interconnecting more chips with more circuits on each
chip, each dissipating more power.
The single-line reference planes have been designed to provide a
characteristic impedance of 55 Q for the selected line geometry by the
use of a combination of nominally 0.15-mm- and 0.2-mm-thick ceramic
layers with a nominal dielectric constant of 9.4. The bulk resistivity of
the sintered molybdenum metallurgy is approximately 10 llQ-cm.
11-308 CERAMIC PACKAGING

Table 9·8. Multilayer Substrate Dimensions

Standard Custom High Density

Maximum substrate size (mm) 150 x 150 300 x 300 150 x 150
Tolerances-as fixed
x-y Dimensions +1% ±O.5% ±O.5%
Thickness ±1O% ±5% ±5%
Flatness (mmlmm) 0.102125.4 0.076125.4 0.051125.4
Number of layers
Tape layer 7 14 30+
Screened dielectric 4 8
Vias diameter (mm) 0.203 0.203 0.102
Pitch 0.762 0.635 0.254
Lines
Width (mm) 0.254 0.102 0.102
Pitch (mm) 0.508 0.203 0.203

Source: Courtesy of Kyocera Corp.

9.4 STATE-OF·THE·ART ALUMINA PACKAGING


AND APPLICATIONS

Table 9-9 illustrates the state-of-the-art evolution in alumina-Mo


ceramic packaging during the last 15 years, starting with the original
thermal conduction module (TCM) in 1980_ The substrate size has grown
from 90 mm to 127.5 mm, the number oflayers from 33 to 63, the number
of vias in the substrate from 350 k to 2 million and the wiring length
from 130 to 400 meters. The state-of-the-art is the last one in this table
and is described briefly here, but the details can be found elsewhere. [27]
ffiM's application of the latest ceramic package for an air-cooled
mainframe system is illustrated in Figure 9-21. Assembled, the module
measures 166 mm wide by 146 mm deep by 169 mm high; it contains
up to 121 logic and array chips with 144 decoupling capacitors, all mounted
using controlled collapse chip connection (C4) technology. The alumina
thermal-conduction module (TCM) utilizes a new MLC substrate, top-
surface thin-film redistribution wiring, and a new air-cooling technology
which allows the package to dissipate 600 W and uses 2772 pins to
connect with the second-level package. In this module, CMOS and bipolar
chip technologies are packaged together on a single TCM for the first time.
The process of fabricating the multilayer ceramic package is outlined
in Figure 9-22, and the resulting substrate cross section is illustrated in
Figure 9-23.
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-309

Table 9·9. Evolution of Alumina Packaging at IBM

System 3080 System 3090 Systeml390


TCM TCM air-cooled TCM
(1980) (1985) (1990)

Size (mm) 90 x 90 1l0.5 x 117.5 127.5 x 127.5


Layers 33 36-45 63
Via count 350K 470K 2000K
Wiring (m) 130 180 400
Available C4 connections 16K 24K 80.7K
Chip sites 100-133 132 121
110 pin connections 1800 1800 2772
Terminal metallurgy Plating Plating Plating/thin film
Cooling capacity (W) 300 (water) 520 (water) 600 (air)

Source: From Ref. 27.

9.4.1 Ceramic
Pure alumina, although capable of much higher mechanical strength
when combined with glass, presented two basic problems. The thermal-
expansion coefficient of pure alumina is 70 x 1O-7fOC compared to

Heat sink

Figure 9·21. Assembly/Cutaway View of the S/390 Alumina TCM


11-310 CERAMIC PACKAGING

Slurry Plparatlon Nickel/Gold Plating

1
:r
Caltlng/Blanklng Substrata Electrical Test

1
Via HoleiUnChlng
1
r-\_ _ _ _ p_ln'_FI.. Braze

Metallization
Jr
"'r'.
nlng

1
Chip Chip

SlaCklng'rmlnatlon
Ror ·...
Sintering Module rsemblY

Helium Fill

Figure 9-22. Process of Fabricating Multilayer Ceramic

58 X 1O-7jOe for molydenum in the temperature range 20-800°e. This


difference resulted in a radial cracking of alumina around each molybde-
num via. Additionally, pure alumina required sintering temperatures
around 1900o e, which is much too high for the economic operation of
manufacturing furnaces. Both these problems were solved by the addition
of glass. The properties of alumina used are described in Table 9-10. The

Redistribution
layers

~
c::::::::-....· o - ~ Signal
-:=:~~o~-;:? ~ distribution
==="'~~ layers
'" ~ <:>

Power
distribution
layers

Figure 9-23. Multilayer MuItichip Module


9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-311

Table 9-10. Properties of Alumina Powder for Multilayer Ceramic

Average particle size 3.4-4.0 11m


Surface area 0.75-0.95 m2/g
Leachable soda Nominal ± 20 ppm
Bulk density 1.0 g1cml

ceramic consists of approximately 90% alumina and 10% glass. A number


of glass additives added to alumina powder and the resulting mechanical,
thermal, electrical, and dimensional control properties of ceramic led to
the choice of ca1cia-magnesia-alumina silicate glass. The influence of
this composition of glass on the nominal shrinkage of the substrate is
illustrated in Figure 9-24, further defining the chemical composition of
the glass used [28].

9.4.2 Greensheet
The basic building block used in the multilayer ceramic process is
the "ceramic greensheet," nominally 0.2 mm or 0.28 mm thick (unfired),
which is a mixture of ceramic and glass powder suspended in an organic
binder as described below. A key factor in achieving acceptable yields
is the formulation of a greensheet that exhibits the necessary strength for
handling and processing. In addition, the green sheet must be dimensionally
stable to ensure accurate plane-to-plane registration when the layers are

+0.2%

•
Nominal
iCD

...51'

I
C
1:
.c;
Ul
-0.2%
.........

-0.4%
16 20 22 24
Total MgO + CaO (Wt.%)

Figure 9·24. Shrinkage of Multilayer Ceramic as Influenced by Glass Compo-


sition. (Courtesy of ffiM Corp.)
11-312 CERAMIC PACKAGING

stacked and laminated. Strength and stability have been achieved by the
proper selection of binder constituents, a controlled casting process, and
the use of a molybdenum paste vehicle that does not interact with the
green sheet binder.
A typical green sheet vehicle system [29] consists of binder, solvent,
and plasticizer. The purpose of the binder is to "bind" the ceramic particles
temporarily in forming greensheets and allowing sheets to be screened
with an appropriate paste of the metal powders.
Power is distributed from the substrate pins to the power planes and,
in turn, through parallel paths to chip power pads on the surface of the
substrate. The maximum voltage drop in the substrate is 16.5 mY, and
the power distribution design allows a minimum of 450 off-chip drivers
to be switched simultaneously. Solvents playa number of key roles,
ranging from deagglomeration of ceramic particles during the ball-milling
operation, due to the low viscosity of the grinding liquid the solvent
provides, to formation of microporosity as the solvent evaporates from
the sheet. The formation of microporosity in the sheet is considered one
of the most important features, because it allows the sheet to be compressed
around metal lines during lamination. Plasticizers allow the sheet to be
"plastic" or flexible due to the lowering of the glass-transition temperature
of the binder by the plasticization process.

9.4.3 Binder and Slurry

The most effective binder for providing all the required thermoplastic
properties and bond strength between layers is polyvinyl butyral (PVB).
Other binders considered and used in certain applications are polyvinyl
chloride acetate, polymethyl methacrylate (PMMA), polyisobutylene
(PIB), polyalphamethyl styrene (PAMS), nitrocellulose, cellulose acetate,
and cellulose acetate butyral. The chemical formula of PVB is shown in
Figure 9-25 and the different types of PBVs available, and the solvents
and plasticizers used with PVB are listed in Tables 9-11, 9-12, and 9-
13 respectively.
The green sheet organic raw materials are combined with the alumina
and glass in a ball mill and deagglomerated to achieve a uniform slurry
dispersion. A dispersion model, illustrated in Figure 9-26, shows how the
greensheet density relates to the breakdown of agglomerates during initial
milling. Once the agglomerates are broken, the greensheet density and
hence the shrinkage of ceramic remains insensitive to the milling time.
The slurry is then fed to a continuous caster and deposited on a constantly
moving plastic carrier to form 200-mm-wide ceramic tapes. The caster
is fitted with a constant-level slurry reservoir and a doctor-blade assembly
to control greensheet thickness. The ceramic webs are passed through a
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-313

H H
HI
R-C + R -OH
I
~ I
R-C-OR + R'-OH ~ I
R-CI-OR\ + H2O
II I
0 OH

Alcohol Alcohol Acetal

Aldehyde Hemia.cetal

H H
I I
CH -C CHz-C
2 I I
OH o
I
C=O
I
CH 3

A B c
PVAcetal PV Alcohol PV Acetate

Figure 9-25. Chemical Formula of Polyvinyl Butyral

series of drying ovens with controlled temperature and humidity and are
subsequently separated from the plastic carrier and spooled.
Each spool of ceramic tape is fully inspected by a scanning laser and
cut into square blocks in preparation for personalization. A representative
sample of the blanks is evaluated for density, compressibility, bond and
yield strength, and shrinkage.

9.4.4 Via Punching

Computer-controlled step and repeat equipment is used to punch via


holes in each greensheet layer. A pallet, used for mounting the greensheets,
is an integral part of a precision x-y table that moves the greensheet relative
to a stationary die set containing 100 punches at a rate of approximately 10
steps per second. A location hole in each comer of the greensheet is used
to position the greensheet on the punch pallet. Up to 36,000 via holes
are punched in a single green sheet layer. The punch tool also includes an
inspection station that contains a light source and an array of photodiodes to
determine the exact location of the hole patterns. If any holes are clogged,
the sheet is either repaired or discarded.
An alternative to the via punching process is the photoforming
process developed by DuPont and others, as illustrated schematically in
Figure 9-27. The photoformed ceramic module (PCM) is formed by the
use of photosensitive dielectric and conductive films. These films are
subsequently exposed to ultraviolet (UV) light. Linewidths less than
50 ~m on 1OO-~m spacings are being developed with these processes.
Table 9-11. Properties of Polyvinyl Butyral White Free-Flowing Powder Form

Betvar® Butvar® Butvar® Butvar® Butvar® Butvar® Butvar®


~ B-72A B-74 B-83 B-76 B-79 B90 B98

Volatiles (%)-max. 3.0 3.0 3.0 5.0 5.0 5.0 5.0


Molecular weight 180,000- 100,000- 50,000- 45,000- 34,000- 38,000- 30,000-
(weight average) 270,000 150,000 80,000 55,000 38,000 45,000 34,000
Hydroxyl content 9.0-17.5 9.0-17.5 17.5-18.0 18.0
expressed as weight % 21.0 20.0 21.0 13.0 13.0 20.0 20.0
polyvinyl alcohol
Acetate content 0-2.5 0-2.5 0-2.5 0-2.5 0-2.5 0-1.0 0-2.5
~
.... expressed as weight %
polyvinyl acetate
Butyral content 80 80 80 88 88 80 80
expressed as weight %
polyvinyl butyral
Specific gravity 1.100 1.100 1.100 1.083 1.083 1.100 1.100
(23°CJ23°C)
Viscosity (cps)' -1570 -700 -400 -175 -55 -195 -75
Viscosity (CpS)b 8,000-18,000 4,000-8,000 1,000-4,000 500-1,000 100-200 600-1,200 200-450

'Viscosity of a 10% solution in 95% ethanol at 25°C using an Ostwald Viscometer.


bViscosity of a 15% solution in 60: 40 toluene/ethanol at 25°C using a Brookfield Viscometer.
Note: Butvar® is a trademark of the Monsanto Co.
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-315

Table 9-12. Solvents for Polyvinyl Butyral Resins

Butvar® Butvar® Butvar®


Solvent B-72A, B-73, B-74 B-76, B-79 B-90, B-98

Acetic acid (glacial) S S S


Acetone I S SW
n-Butyl alcohol S S S
Butyl acetate PS
Carbon tetrachloride I PS I
Cyclohexanone S S S
Diacetone alcohol PS S S
Dioxane (1, 4) S S S
Ethyl alcohol, 95% S S S
Ethyl acetate, 85% I S PS
Ethyl cellosolve S S S
Ethylene chloride SW S PS
Isophoronet PS S S
Isopropyl alcohol, 95% S S S
Isopropyl acetate I S I
Methyl alcohol S SW S
Methyl acetate S I
Methyl cellosolve S S S
Methyl ethyl ketone SW S PS
Methyl isobutyl ketone S
Pentoxol PS S PS
Pentoxone PS S PS
Propylene dichloride SW S PS
Toluene I PS SW
Toluene ethyl alcohol 95% S S S
(60 : 40 by weight)

Key: S = soluble; PS = partially soluble; I = insoluble; SW = swells.


Note: Butvar® is a trade mark of the Monsanto Co.

Table 9-13. Plasticizers for Polyvinyl Butyral Resin

Phthalate Polyester
Phosphate Ricinoleate
Polyethylene glycol ether Rosin derivates
Glyceryl mono oleate Sabacate
Petroleum Citrate
11-316 CERAMIC PACKAGING

t
""-/
No Agglomerates /

-----~£~------------

Minimum
- - Agglomerated

Dispersion Milling Time

Figure 9-26. Ceramic Dispersion Model

9.4.5 Screening
The screening pastes are tailored for the various patterns shown in
Figure 9-28 and consist of molybdenum powder uniformly dispersed in
a resin and solvent mixture. Non-IBM multilayer ceramic substrates use
tungsten instead of molybdenum, but the paste technology is similar. As
with alumina milling for greensheets, the milling process is critical to
achieving a proper dispersion of the metal powder in the organic materials
used in the paste binder system. Three-roll mills are used for this purpose.
The rolls rotate at different speeds to produce a shearing action that
breaks up the metal agglomerates and distributes the molybdenum particles
uniformly. Metallization of the green sheet is accomplished by extruding
the molybdenum paste through a nozzle as the nozzle traverses a metal
mask in contact with the green sheet. The vias are filled and the pattern
is defined on the surface of the greensheet simultaneously.
Following screening, the metallized layers are dried in a forced-air
circulation oven. The drying cycle has been optimized and is carefully
controlled to avoid dimensional change and damage to the greensheet,
particularly on sheets with dense via arrays.
One of the key advantages of the multilayer ceramic process is the
ability to inspect and repair individual metallized layers prior to stacking
and lamination. Inspection is accomplished automatically with a system
that detects deviations in the screened pattern relative to an optimum
configuration.
In order to achieve the required thermal, mechanical, electrical,
and dimension control properties of the substrates, three aspects of raw
materials technology need to be understood: (1) the physics and chemistry
of ceramic and glass powders as well as what can be expected between
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-317

Laminated Photosensitive
PCM Conductor Film;
Expose and Develop
Circuit Unes; Fire

Laminated Photosensitive
PCM Dielectric Film;
Expose and Develop
ViAs; Fire

Screen Print to Fill


ViAs; Dry; Fire

Repeat Process to
Form Multilayer

Figure 9-27. Photoformed Ceramic Modules (PCM). (Courtesy of DuPont Corp.)

these two at sintering temperatures; (2) the intrinsic properties of organics


such as molecular weight and viscosity as well as the reaction between
organics and inorganics to form stable steric forces; (3) properties of
slurry and greensheet as affected by the particular slurry, and greensheet
tool design. These interactions between organic and inorganic materials
as they are processed through the tools are shown in Figure 9-29. A
thorough understanding of each of the three, as well as of simultaneous
interactions between all three, is necessary to provide the dimensional
control required in the green and fired states to align 10,000 vias each of
0.125 mm diameter from layer to layer through 33 layers to form a network
of 350,000 interconnections.
The chronological development of multilayer ceramic illustrating
the improvements in substrate parameters is shown in Table 9-14 [30].
MLC packaging, as the current dominant multichip technology, owes its
11-318 CERAMIC PACKAGING

Figure 9-28. Greensheet Screening Patterns

RAW MATERIAL
I '1f'1I111
Control

h~
Colt..
Control
r- GREEN SHEET

...... ~ ....... I L-
Composition I o..ulomeretion
Shrlnk-ee
I

--
ALUMINA

t
I
Surf"e",.. ~ I

_... -_.
Partlc" SU. I
N. Cont.nt, p ~ L_':
-
r_peeI Dlnetty
A",ome"-"
~-:
r--.
Vi.coslly, O,.. nShHt
r-
Strength ~-- p
-~--~ Don""
FAIT
I
I
I
- t:: A
I
I ,.-~
I I I r"
I I I I
I I I I I
~:
OI..,Tr.nl. T

L.t Lll-!_
IoftemngPt. T
Exp.".lOn Coef Specific
0<....
~
Compo.Hlon
~
Stripping

r+-f-
".rtlcfeSlze CherKt.,lsUct

BUTVAA
Function"
Group Chem
R..ln
(PVOH)
DtItrlbution
8lne" Sol.
VllICOflty

Lr
Moleeultr Wt.

8c,""~ng
Chl,.ct.n.tlct

Figure 9-29. Greensheet Materials, Processes, and Tool Interactions. (Butvar®


is a trade mark of Monsanto Company. Courtesy of IBM Corp.)
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-319

Table 9-14. Laminated Multilayer Ceramic Development

Min. Via Min. Line


Diam.lMin. WidthlMin.
Year Spacing Spacing Shrinkage Substrate Area
Reported Metallurgy lli!!!l lli!!!l Tolerance Layers (mm2)

1961 Mo, Mo-Mn 125/250 125/250 3-24 200


1967 Mo 200 6 161
1968 200/125 375 1,290
1969 Pd 125/250 100/200 15 323
1971 Noble 2501500 125/250 ±O.5% 6 1290
Mo 2501500 200 5-10 323
WlMolPdIPt 125/250 1001200 ±O.5% 10-20 1,935
1972 Mo-Mn 125/250 2001400 ±O.6% 11 323
1973 W 200/400 125/250 10 3226
1975 WlMolMo-Mn 1251500 1001275 ±O.5% 11 11,613
1977 WlMo 1751500 1001200 ±O.5% 15-20 6,653
1978 WlMolPdlPt 250 ±O.1%
1980 Mo 125/250 23 2,500
Mo 100/200 100/250 ±O.1% 33 8,100
1983 W ±O.3% 10,000
1992 Mo 75/200 90/225 ±O.05 63 16,256

Source: From Ref. 30.

importance to the processes described in this section, leading to the shrink-


age control of ±O.1 % over as large an area as 127.5 x 127.5 mm with as
many as 63 layers. If not for the basic understanding of the materials,
slurry, paste, and greensheet processes and the special tools developed
to go with each of the processes, multichip ceramic technology would
not be the high-performance package it is today. Accordingly, high-perfor-
mance organic technologies would have been chip-packaging technolog-
ies. The figures of merit discussed in Chapter 1, "Microelectronics Packag-
ing Overview," for these two package options brings further evidence to
bear on this discussion.

9.4.6 Sintering
The sintering process is the most complex of all the processes used
in forming multilayer ceramic substrates. The organic removal of the
solvent and plasticizer followed by the pyrolysis of the binder and the
oxidation of residual carbonaceous materials in wet H2 are all complex
chemical processes that prevail in excess of lOOO°C. The sintering of the
11-320 CERAMIC PACKAGING

substrate begins with the densification of the glass, which leads to the
glass-alumina reactions that cause crystallization. The crystals, however,
melt at about l450°C into a fluid glass, the viscosity of which keeps
dropping until the maximum temperature is reached. The metal paste
similarly goes through organic removal and densification processes. The
glass enters the porous tungsten or molybdenum metal vias, thus providing
mechanical anchoring as well as any chemical reactions and bonding
depending on the oxygen partial pressure of the furnace system. The
allowable oxygen partial pressures could be decided from Figure 9-30
for several of the metals and oxides typically used in cofired structures.
The properties of most common metals and ceramics that are typi-
cally considered for multilayer cofired ceramic substrates are given in
Tables 9-15 and 9-16.
The pyrolytic mechanisms for binder burnout during early stages of
firing are shown in Figures 9-31 and 9-32. The thermal decomposition
of 23-layer ceramic greensheets, shown in Figure 9-31, were studied
by Knudsen Effusion-Mass Spectrometry. This technique is a powerful
method for examining decomposition reactions, as it allows one to analyze
both equilibrium and activated reactions. The order and activation energy
of kinetic processes and heats and entropies of equilibrium reactions can
be derived directly from an analysis of the time, temperature, and intensity
data provided by the modulated beam mass spectrometric technique.
Figure 9-31 shows the total gas desorption rate as a function of
temperature when heated at 4°C/min from 25°C to 800°C. Two organic
materials, the binder and the plasticizer, are the principal components that
generate gas with this thermal treatment. The plasticizer desorbs without
decomposition in a simple first-order process with an activation energy
of 90 kJ/mole. The desorption rate peaks at 140°C for a single layer. The
relative amounts of gas released for the major components are shown in
Figure 9-32. For comparison, the total gas generated as a function of
temperature is plotted in the bottom curve of the graph. In the middle
two curves, the spectra for water and butyraldehyde (represented by CHi
CHO+) show similar behavior, as they are released during the initial
breakdown of the side chains of the polymer. The curve for the ion species,
mlz =44, is predominantly a fragment from the butyraldehyde molecule.
However, this curve also shows a small amount of carbon dioxide being
released around 550°C. The sensitivity of the technique is amply demon-
strated in the top curve, where the formation of high-molecular-weight
aromatic species occurs in the pyrolysis. In this instance, toluene (C6Hs-
CH3), which is representative of the typical aromatic species desorbing,
peaks around 440°C, where the polymer backbone is undergoing decompo-
sition. After this stage, the laminate structure is left with nonvolatile
residues of carbon and tars that will require oxidation to carbon dioxide
and water for their complete removal.
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-321

Hu'><,O RATIO

Xo8.\'01 X~ Xo4\
~
:rOS
Cc:vt:Oj~ATIJ
\ \ I
0 y'08 Y,O' XOb XOS XOol )(03
0

~ -20

-'0

I
H'

!
c'

,
q, _1("1411~

~' j~100b

4, 10"'8 'cf

:ol~ I~
-ll~O~~~~~~~~~~~~lroo~~I'~OO~lbOO~~I~~~~~~~-~
-ABSOLUTE 7.ERO TEM'ERATUR£
CO,cot 'ATIO '9.4

"';"1° '.ATIO I I

Figure 9-30. Oxygen Partial Pressures for Representative Metals and Oxides.
These diagrams are used to detennine fonnation and stability of inorganic oxides as a
function of partial pressure of oxygen.

This delay in the gas release would lead to a pressure buildup at


the sheet interface and thus be a contributing cause of delamination. This
stress at the ground plane is most serious because of the very large amount
of metallurgy on this surface, resulting in a weak bond between the layers.

9.4.7 Plating
Nickel is plated on the surface features and is diffusion bonded
to the molybdenum base metal to enhance adhesion. Following nickel
11-322 CERAMIC PACKAGING

Table 9-15. Typical Cofired Metals

Melting Electrical Thermal Thermal


Point Resistivity Expansion Conductivity
.L9 (I~n m) (I0-7f'C) (W/m K)

Ag 960 1.6 197 41S


Au 1063 2.2 142 297
Cu IOS3 1.7 170 397
Pd 1552 10.S 110 71
Pt 1774 10.6 90 71
Mo 2625 5.2 50 146
W 3415 5.5 45 201

diffusion, a layer of gold is applied to prevent formation of nickel oxide


and to enhance wettability during subsequent soldering and brazing pro-
cesses. The final plating step is the application of heavy gold on the wiring
pads surrounding each chip site to accommodate ultrasonic joining of
discrete engineering-change wires on the module surface. After the sub-
strate electric test, the plated pins and flange are simultaneously brazed
to the substrate in nitrogen gas atmosphere using a gold-tin alloy. Chips are
joined to the substrate using the flip-chip solder-reflow process described in
Chapter 8, "Chip-to-Package Interconnections."

Table 9-16(A) Dielectric Constant and Thermal Coefficient of Expansion of


Candidate Oxides

Thermal
Dielectric Expansion Coefficient
Constant (I0-7/0C)

AI20) 9.6 6.S


AIN
MuIIite (3AI20). 2Si02) 6.6 4.0
BeO 6.S 250
Forsterite (2MgO. Si02) 6.2 9.S
Cordierite (2MgO. 2Ah 0). 5Si02) 4.5-5.3 2.5
Spodumene (Li20. AhO). 4Si02) 6.0 2.0
Steatite (MgO-Si02) 5.7 4.2
Quartz, Si02 3.9 1.3
Borosilicate Glass 4.0 3.3
Pb-Borosilicate Glass 7.0 7.0
Mg-Aluminosilicate 5.0 3.S
9.4 STATE-OF-THE-ART ALUMINA PACKAGING AND APPLICATIONS 11-323

Table 9-16(8). Laminated Ceramics and Their Properties

Thennal Thennal Bonding


Alumina Glass Dielectric Conductivity Expansion Strength
.r&l .r&l Constant (W/m K) (lO-7/°q (MPa)

96 4 10.2 20.9 71 276


92 8 9.5 16.7 69 315
55 45 7.5 4.2 42 296
Mullite 40 6.2 10 4.5 165
(3AIz0 3·2Si02)
BeO 0 6.8 250 54 231

9.4.8 Chip Joining


The chip joining by solder, tape automated bonding, and wirebonding
is covered in great detail in Chapter 8, "Chip-to-Package Interconnec-
tions." The chip joining in this section addresses only the joining of the
chip to the multilayer multichip substrate. The chip sites, fluxed with a
water-white rosin, and the LSI devices, which are terminated in 120
lead-tin solder pads, are positioned semiautomatically on the correspond-
ing substrate. Up to 118 chips are positioned on the substrate in this

1.0

~
c::
0.8
::J

.e
, ,,
I
I
50.6 I ,
.!l I
C
D!: ,
c::
,g 0.4 binder
::J

] ),
8 0.2 \

0.0
..... - '- --
o 100 200 300 400 500 600 700 800
Temperature (OC)

Figure 9-31. Binder Burnout-Processes. The desorption of the plasticizer (Benzo-


flex) and the decomposition of the binder (Butvar) are plotted for two multilayer ceramic
samples heated at 4°/min: single sheet (-) and 23 layers with metallization (-). (From
Ref. 29.)
11-324 CERAMIC PACKAGING

0.50

0.00
~---+---+---+---+---+---+---+---+~
12
m/z =44

o ~---+---+---+---+---+---+---+---+~
7

o ~---+---+---+---+---+---+---+---+~
100
Integrated

o
o 100 200 300 400 500 600 700 800
Temperature (OC)

Figure 9-32. Gases Released from Ceramic Greensheets. (From Ref. 29.)

manner and reflowed in a nitrogen environment. Thousands of chip-to-


module connections are made simultaneously in this manner. Following
chip joining, the assembly is submerged in a nonconducting coolant and
tested by contacting the I/O pins and probing the surface pads with power
applied to the substrate.

9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING


Alumina has been the only workhorse for more than three decades,
from the 1960s until the beginning of the 1990s. It has become increasingly
clear during the last 5-10 years that ceramic materials with improved
9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING 11-325

properties and lower in cost than alumina are required. These materials
fall into two categories: low-temperature ceramics, sometimes referred to
as glass-ceramic or glass + ceramic, and aluminum nitride. In limited
applications, mullite ceramic has been developed as well. The properties
of these substrate materials are compared in Tables 9-17 and 9-18. These
materials are being developed very actively so that packages will have im-
proved:

• Signal propagation
• Thermal-expansion matching to silicon, allowing larger chips with
more circuits directly solder bonded.
• Electrical conductors for added power distribution
• Thermal conductors for improved power dissipation
• Input/output connection to support more circuits

The recent progress in ceramic substrate materials is discussed in


the following sections.
There is a clear trend as indicated below for both ceramic and organic
packaging. For most part, the ceramic packages have been pin-grid arrays
(PGA) both for pin-through-hole insertion and surface-mount assembly,
including butt joining:

PGA ~ SMT ~ BGA ~ DCA with SLIM


1970s ~ 1980s ~ 1990s ~ 2000

Table 9-17. Recent Ceramic-Package Materials Developments

Package Material Package Property

Glass ± ceramics Dielectric constant as low as 5


Thermal expansion coefficient near that of silicon
Cofired metallization of high-conductive metal such as
gold and copper
Low thermal conductivity
Mullite Thermal-expansion coefficient near silicon
Cofired W metallurgy
Electrical insulation
Low dielectric constant «7)
Aluminum nitride Thermal-conductivity approaching aluminum
Thermal expansion near silicon
Cofired multilayer metallurgy (W)
Dielectric constant (-9)
11-326 CERAMIC PACKAGING

Table 9-18. Properties of AIN, Glass ± Ceramics as Compared with 90%


Alumina

Substrate Glass ±
Properties AIN BeD Ceramics 90% AIPl Mullite

Thermal conductivity (W/m K) 230 260 5 20 5


Coefficient of thermal expansion 43 75 30-42 67 42
(RT - 400°C)(xlO-7/oC)
Dielectric constant at 1 MHz 8.9 6.7 3.9-7.8 9.4 5.9
Tan ii (I x 10-4) 2 3 2 2 2
Flexural strength (Kg/cm') 3500-4000 2500 1500 3000 2200
Thin-film metals TilPd/Au Cr/Cu, Au Cr/Cu Cr/Cu
NiCr/Pd/Au
Thick-film metals Ag-Pd Au, Cu, Ag-Pd Au, Cu
Cu Cu, Au Ag-Pd
Cofired metals W W Au, Cu, Ag-Pd W,Mo W,Mo
Cofired sheet resistance 10 10 3 10 10
(mQ/[J)
Cooling capability
Air (OCIW) 6 5 60 30 60
Water (OCIW)' <1 <1 <1 <1 <1

'External Cooling

It appears that because of thin, light, and portable needs of current


and future electronic systems, ball-grid arrays (BGA) and direct chip
attach (DCA) of chip to ceramic and organic boards referred to as SLIM,
Single Level Integrated Module under development at Georgia Tech, will
be the leading-edge technologies.

9.5.1 Aluminum Nitride


Aluminum nitride, a synthetic III-V compound with wurtzite struc-
ture, was discovered in the 1860s. Although single-crystal AIN has a
theoretical thermal conductivity of 320 W/m K [31], single-crystal or
polycrystalline AIN shows a much lower thermal conductivity due to the
incorporation of oxygen and metallic impurities into the lattice [32]. This
inability to achieve high thermal conductivity in fired parts has historically
hindered the use of AIN in commercial applications.
The recent development of effective sintering aids such as CaO and
Y203 has resulted in high-purity and thus higher terminal-conductivity
AIN [33]. This, in tum, has led to the commercialization of AIN products
for packaging and interconnect applications.
9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING 11-327

9.5.1.1 AIN Powder and Sintering


Commercial AIN powders are manufactured by two processes: (A)
carbothermal reduction of alumina and (B) direct nitridation of alumi-
num metal:

(A) Al20 3 + 3C + N2 ~ 2AIN + 3CO


(B) Al + 0.5N2 ~ AIN

In comparison to the carbothermal reaction, the direct-nitrided reaction


produces a broader particle size distribution and a larger average particle
size which results in a less sinterable material. Worldwide production of
all AIN powders was estimated to be 150 tons in 1991.
Alkali-earth and rare-earth oxides are effective sintering aides for
AIN pressureless sintering. Yttrium oxide or calcium oxide are generally
used in production. Sintering is done under nitrogen at approximately
1800°C. The sintering aids get oxygen from the surface and/or the lattice
ofthe AIN and form a liquid grain-boundary phase (Y-AI-O or Ca-AI-O)
which concentrate at grain-boundary triple points or migrate to the surface
of the sintered AIN body. This promotes densification and grain growth
and, at the same time, prevents oxygen atoms from diffusing into AIN
grains so that high thermal conductivities up to 200 W1m K can be attained
[34]. Sintering in a carbon atmosphere is one means of achieving thermal
conductivity of> 200W/m K.
The thermal conductivity of AIN is very sensitive to the presence
of oxygen and other metallic and nonmetallic impurities such as Mg, Fe,
Ca, and Si [32,35]. Oxygen diffusion into the AIN lattice forms nitrogen
vacancies, which impede phonon heat transfer. The relationship between
thermal conductivity and oxygen content in a fully dense sintered AIN
is shown in Figure 9-33.
High-quality AIN ceramics are being produced by conventional ce-
ramic processes such as dry-pressing, tape casting, and injection molding.

9.5.1.2 Low-Temperature Firing


The AIN sintering temperature of > 1800°C prevents the use of
traditional alumina furnaces. Recent reports indicate that AIN can be
sintered at 1600°C by the addition of fluoride sintering additives [36].
The low-temperature sintering of AIN, if it can be accomplished in alumina
furnaces, will lead to cheaper AIN substrates that will expedite its commer-
cialization.

9.5.1.3 AIN Properties


Table 9-18 compared AIN properties with other commercial ceramic
substrates and related ancillary materials. The electrical properties of AIN
11-328 CERAMIC PACKAGING

260
240 0. by Kuramoto
::ii' 220 • by Slack
E x bySokal
~ 200
~ 180
~::I 160
"1:1 140
c
0
u 120
1i 100
~ 80
~" 60
40
20
0
0.03 0.1 1.0 10
Oxygen content in A1N ceramic (wt%)

Figure 9-33. Relationship Between Thermal Conductivity and Oxygen in AIN

are comparable to that of alumina and beryllia. AIN shows a better thermal-
expansion match to silicon than either alumina or beryllia, as shown in
Table 9-18. This results in less stress generation due to thermal mismatch
between the silicon chip and the substrate as structures are exposed to
temperature excursions. This becomes very important as chips get larger
and when silicon-based ICs (2-3.5 cm on a side) are packaged.
The flexural strength of AIN is greater than alumina, whereas the
Vickers hardness is half that of alumina. This lower hardness and superior
strength allows AIN to be machined into complex shapes. The lower
density of AIN results in a 20% weight reduction when compared to
equivalent alumina parts.
AIN is resistant to most process chemicals with the following nota-
ble exceptions:

• The oxidation characteristics of fired AIN have been examined in


detail [37]. The oxidation starts at temperatures in excess of 800°C.
Thus, although oxygen must be excluded during AIN firing, air
oxidation is not an issue during subsequent fabrication or use.
• Early concerns about AIN centered around its stability in moist
environments. The inherent hydrolytic instability of AIN under accel-
erated aging test conditions is corrected by oxidation of the surface
by a short exposure to air at 100°C [38-40] A thin layer of alpha
alumina forms, which protects the surface from further oxidation
and or hydrolysis. No hydrolysis is observed after 1000 h at 85°CI
85% rh and/or pressure cooker testing (PCT). Loss in thermal con-
ductivity by this process is reported to be 10% or less.
9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING 11-329

• Strong alkaline solutions attack AIN ceramics [41]; thus basic elec-
troless plating solutions must be avoided. The thin oxide coatings
described above also serve to protect the AIN surface from alkaline
solutions [40].

9.5.1.4 Metallization of AIN


A summary of thick- and thin-film metallizations is presented in
Table 9-19. Thick-film metallizations containing conventional glass frit
powders developed for 96% alumina cannot be used on AIN surfaces.
Studies have concluded that, during firing, PbO, CuO, and Bi20 3 glasses
cause oxidation of the AIN surface. This oxidation results in blistering
and poor adhesion of the metallized surface [42]. When proper thick-film
pastes (PdlAg,Au), or thin-film metallizations (Cr/Au, TiIPtIAu, TiIW,Al)
were examined, all were found to give acceptable properties[43,44]. Braz-
ing and resistor paste technologies have also been developed [45-47].

9.5.1.5 AIN Applications


Table 9-20 shows AIN's practical application areas. In the field of
power devices, bare substrates and thick-film metallized substrates are used
in various thyristor and transistor modules. Direct-bond copper (DBC) AIN
substrates has overcome early thermal coefficient of expansion (TCE)
mismatch problems [48] and are being used in various power applications
including IGBT (insulated gate bipolar transistor) modules.
Submount substrates for LD and LED packaging, which are patterned
by thin-film TilPtIAu, require mirror surface finish (Ra: 0.5 /lm) and high-
thermal-conductivity (>200 W/m K) of AIN. Aluminum nitride use in
computer packaging has been led by Fujitsu which has used AIN in its

Table 9-19. Thin- and Thick-Film Metallizations for AIN

Process Metallurgy Process Temp.

Thick-Film Metallization
Postfired Ag-Pd. fritted 920°C in air
RuO,. fritted 850°C in air
Copper 850°C in N,
Gold 850°C in air
Molten Metallized Cu-Ag-Ti-Sn 930°C in H2
Cofired Tungsten 1900°C in H,
Thin-Film Metallization
Sputtering NiCr-Pd-Au lOO-200°C
Evaporation Ti-Pd-Au lOO-200°C
11-330 CERAMIC PACKAGING

Table 9-20. AlN Applications in Electronics

Application Metallization Example

Power device Thyristor stack for trains, relay Not metallized


module
TO-type power module Mo-TiN, Mo-Mn
RF package Mo-Mn
Heat sink for car ignition Cofiring W, Mo-TiN
High-power switching module Direct bond copper, Ag-Cu-Ti
brazed copper
Laser diode Submount for LD and LED TiIPtIAulSn-Pb or Au-Sn
Package Heat sink for computer packaging TilNilAu
MCM PGA package Cofiring W, copper polyimide
QFP Glass sealed

supercomputer, the VP 2000, as a heat spreader and surface-mount PGA


package (17 mm2 with 462 pins) as shown in Figure 9-34 [49]. Fijitsu
has also developed an AIN surface-mount quad flat pack (QFD) for mount-
ing large high-power application-specific integrated circuits (ASICs) (Fig.
9-35). Thermal resistivity measured on a representative QFP is shown in
Figure 9-36. This face-down package with a pin-fin can be applied to a
chip of about 4 W with air cooling at 1 m1s or 2.5 W without forced air.
These packages will be used for the MUSE decoder of high-definition
TVs [50].

AIN heat sink solder

LSI
):=d:!~::::::::;::;::::::::;::;::::::::-, thin films
l'JmmL..c~::==:;~=::::;===::}(polyimide-copper)
solder
via

AINPGA
(17mmsq.)

Figure 9-34. Schematic of Fujitsu VP 2000 Module Using Surface-Mount AlN


PGA and Heat Sink
9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING 11-331

AlN base Metal film

cap

Figure 9-35. Glass-Sealed AlN Quad Flat Package

The Hitachi M-880 processor series adopted a water-cooled thermal


solution. The heat generated from the LSI chips goes through the hermetic
AIN cap enclosing the chip, the comb-shaped AIN microfins and the large
AIN "board" which interfaces the water-cooling jacket as shown in Figure
9-37. The heat density generated on this module (36-41 chips on 106 mm2)
is about 7.5 W/cm2). In this application, the excellent machinability of
AIN is a key factor as well as AIN's thermal conductivity and TCE
properties [51].
Multilayer AIN packages with tungsten cofire metallurgy are close
to practical application for high-pin-count PGA packages [52,53] The
NEC PGA cross section shown in Figure 9-38 is for a 208 liD AIN
package with thermal conductivity of 200 W1m K and tungsten sheet

60
Chip = 9mm sq.
~ 50
~

~ 40
...
~
1.. 30
i
..c 20

J 10

o0 1 2 3 4 5 6 7 8
Forced air speed (m/sec.)

Figure 9-36. Heat Resistivity of AIN QFP 160 (Face Down)


11-332 CERAMIC PACKAGING

Water cooling jacbt

AIM board

He gas
t -____"i:f:=""""__"'7'Iu_solder

thin films (6 layers)


mullite (4 layers)
Pz.::z::z::z=....!:ll:D:lilD:D:I::J:[J~u:Do:::rlr::n:::a::le:::::::.....lOlder bump

L -_ _ _ _ _ _-:Mullite (44-layers./06mm sq.)

Figure 9-37. Schematic of Hitachi M-SSO Processor Module Using AIN


Heat Spreaders

resistivity of 7.2 mWD. Compared to an alumina package, this AIN


package can handle a 3.7 times higher-powered Si chip (up to 40 W).
Large hermetic enclosures for silicon-on-silicon MCMs [7,25] and
MCM cofired substrates (with or without copper-polymer thin-film wiring
layers) are also in active development [54,55].
The NOSC (Navan Ocean Systems Center) package (Fig. 9-39),
developed by Hughes/Coors/Grace, is a 10-cm2 hot-pressed package built
to hold a large silicon-on-silicon MCM. AIN was chosen because of the
need to match TCE with this large silicon area [56].

A1N Layer Substrate


Seal Ring
~'--_
cap
.________
_ _ __
AgCu Brazln
Hi

Metallizatian
AgCu Brazing
1/0 Pin

Figure 9-38. AIN PGA Package by NEe


9.5 RECENT DEVELOPMENTS IN CERAMIC PACKAGING 11-333

Koyor
Leod Frame
II-
......,..--:---'It-Ji--L..~'*==; Refl'CK1Oly Metal
I/O Pads

Refnldory
Metal Leod.

Figure 9-39. Silicon-on-Silicon MCM with AIN

9.5.2 Beryllium Oxide


For years, beryllium oxide has been considered the only ceramic
with very-high-thermal conductivity and good dielectric properties for
packaging applications. Beryllium oxide packaging, even though not a
recent development, is covered here because of its very high thermal
conductivity, similar to AIN. The processes described previously in this
chapter for alumina using dry-pressing and tape casting processes are
applicable to beryllium oxide. Cofired, postfired, and conventional thin-
film metallizations have been developed for forming conductors to form
ground planes, mounting pads, microstrip circuitry, and other features.
Beryllia is used as a heat sink for standard transistor cases, TO-3, TO-8,
TO-9, TO-II, and TO-33. It is also used as an electrically insulating heat
spreader for high-power transistors covering the frequency range from 60
Hz to 10 GHz. Power hybrids, bipolar chips, LSI devices, and microwave
integrated circuits are often housed in beryllia packages. Beryllia is used
as a substrate for microwave integrated circuits. The microstrip circuitry
is usually applied by means of sputtered or evaporated thin films. Direct
bonding of copper using copper-copper oxide eutectic has been reported
[57]. Details of these and additional information on other aspects of
beryllium oxide can be found elsewhere [58]. The pertinent properties of
beryllium oxide are included in Table 9-18. Although the outstanding
technical properties of beryllium oxide justify its use, its toxicity-requir-
ing a label, "warning, this product contains BeO" -has limited its use.

9.5.3 Mullite
Conventional mullite has been known for over 30 years. The mullite
that is getting the recognition in electronic packaging, however, is mullite
with additions to lower the dielectric constant and its sintering temperature.
11-334 CERAMIC PACKAGING

Table 9-21 lists the hardware technology using mullite that Hitachi devel-
oped for its M-880 mainframe computers using small-scale chip carriers
called microchip carriers (MCC) with solder ball-grid array connections
to the multilayer mullite-glass-ceramic board [59].
To mount such high-speed high-density LSI chips, Hitachi has devel-
oped multilayer ceramic boards made of mullite ceramics and tungsten
conductors [59]. In contrast to previous mullites, Hitachi mullite has very
low dielectric constant, presumably due to Si02 addition to conventional
mullite (3A120 3 x 2Si02).
This mullite ceramic board features the following:

1. Excellent electrical characteristics; specific dielectric constant: 5.9;


conductor resistance; 0.6 Wcm
2. Thermal expansion coefficient: 3.5 x 1O-7/oC, Very close to that of
silicon (3.0 x 1O-7/0C)
3. Flexural strength: 20 kglmm2 or more; strong enough to bear MCC
soldering, module sealing, and brazing of 110 pins

Table 9-21. M-880 Mainframe Hardware Technology Based on Mullite


Ceramic (From Ref. 59)

M-880 M-680H

High-density module Size 106 x 106 x 7 mm 420 x 280 x 3 mm


(M-880) or Dielectric materials Mullite ceramic Polyimide
package (M-680H) Linewidth loo/.lm 100 /.lID
No. of conductor layers 44 20
No. of 110 pins 2521 1776
LSI package MCC Flat package
No. of LSIs 36-41 72
Cooling Water Forced air
Power density 7.5 W/cm 2 0.4 W/cm 2
Thermal resistance S.2°C/W 5°C/W
Junction temperature 50-60°C 6O-80°C

Processor Board Size 730 x 534 x 7 mm 570 x 420 x 3 mm


Dielectric materials MS resin Poly-imide
Linewidthlthickness 70/.lrnl65 /.lID 100 /.lrnl40 /.lm
No. of conductor layers 46 22
No. of modules 20 21

Source: From Ref. 59.


9.6 LOW-TEMPERATURE CERAMIC OR GLASS + CERAMIC PACKAGING 11-335

4. Excellent dimensional stability because of similar thermal-expansion


coefficients of the ceramics and the conductor (tungsten)

9.6 LOW-TEMPERATURE CERAMIC OR GLASS + CERAMIC


PACKAGING
The limitations of alumina and the advantage of low-temperature
ceramic, as indicated in Table 9.22, fall into three categories. The dielectric
constant (9.4) of alumina is too high compared to low-temperature ceram-
ics which can be fabricated around 5. The thermal-expansion coefficient
of alumina (7 x 1O-7fOC) similarly is too high in those applications in
which the silicon chip of TCE, 3 x 10-7/ oC, is bonded directly to the
ceramic. This is clearly the case with the flip chip, particularly with next
generation of large chips, as indicated in Figure 9-40 for solder-joint
fatigue life between the substrate and silicon chip. The resistance of the
package conductors has as much, if not more, influence than the dielectric
constant on the interconnect delay, particularly in long critical-length
applications. Figure 9-36 plots the interconnection delay as a function of
linewidth for various line lengths in the critical path of mainframe or
workstation applications.
Glass-ceramics, with their dielectric constant around 5, together
with their cosinterability with copper or gold, make them potentially one
of the best high-performance multilayer ceramic substrates, meeting most
of the requirements of an ideal substrate (see Table 9-22). Glass additions
to alumina [60-62] to form substrates with a dielectric constant down to
5 and are sinterable with copper or gold have been reported by a number
of Japanese and American companies. In this book, "glass + ceramics"
refers to those ceramics formed by the addition of glass to the ceramic
in forming slurry or greensheet. In contrast, "glass-ceramics" refers to

Table 9-22. Ideal Properties of Ceramic Package

Ideal Ceramic Alumina Ceramic

Dielectric constant <5 9.4


Coefficient of thermal expansion 30 x 1O-7/ oC 66 x 1O-7/ oC
Package metallization resistivity 211n cm 5.6!lO cm
Metallization layers
Thick Film 95 11m on 500 11m Same
up to 100 layers
Thin Film :0;25 11m on 75 Ilffi, Same
up to 10 layers
Number of contacts to chip Up to 5000 per chip site Same
Number of package output pins >20/cm2 Same
11-336 CERAMIC PACKAGING

r
Z
II)
10,000
QI
U
>-
<1 D. Epoxy-Kevlar
oS! 1000
::J

..
QI Silicon
::s
C)
IBM's Glass-Ceramlc
ftI 100
U. 1rf Coffin-Manson Equation ~

0 40 80 120 160 200


Coefficient of Thermal Expansion (10.7 fOC)

Figure 9-40. Fatigue Life of Flip-Chip Solder Joint as a Function of TCE


of Substrate

ceramics formed by controlled crystallization of glass in the slurry or


greensheet resulting in ceramics of high crystallinity during firing. "Glass
± ceramics" refers, therefore, to a generic family of both these materials.
Table 9-22, which lists ideal properties of substrate materials, shows
that glass-ceramics approach these ideal requirements more than any
other group of ceramics. These substrates have low-dielectric-constant,
thermal-expansion matching silicon and contain the gold or copper metalli-
zation required for high-power distribution. In addition, thin-film metalli-
zation and dielectrics could be deposited onto the surface of these sub-
strates, thus making them extendable for high performance. The major
disadvantage seems to be their low thermal conductivity. With solder
connection (C4) technology, which allows the heat to be removed from
the back of the chip, this limitation is largely overcome. This is particularly
true for those package applications where the chip is bonded face (active
area) down toward the substrate, leaving the back side for air or water
cooling. In this case, minimum heat removal takes place through the
substrate. Thus, high thermal conductivity of the substrate is desirable
but not absolutely required. The impact of poor thermal conductivity in
wirebonding applications of these ceramics can be minimized significantly
by die bonding to metal or ceramic heat sinks which are placed in the
single-chip package cavity and then brazed or glass sealed to the
glass ± ceramic package. This subject is further discussed in Section 9.8,
"Chip Attachment and Thermal Dissipation of Ceramic Substrates." In
this case, minimum heat removal takes place through the substrate, hence,
high thermal conductivity of the substrate is desirable but not absolutely
required. In addition, for high-performance applications involving bipolar
silicon or high performance CMOS, where the integration is expected to
9.6 LOW-TEMPERATURE CERAMIC OR GLASS + CERAMIC PACKAGING 11-337

be very high, cooling capabilities may be required in excess of 40


W/chip. At these power levels, even the very-high-thermal-conductivity
ceramics such as AIN will require thermal enhancements both in face-
down and face-up chip-to-substrate interconnections. Low-dielectric-con-
stant glass, glass-ceramic, and various glasses added to various low-
dielectric-constant ceramics are listed in Table 9-23.
The concept of glass-ceramic cofired, multilayer substrate contain-
ing copper, gold, or silver conductors had its origin in 1976 in IBM. The
first multilayer glass-ceramic substrate was reported in 1978 based on
vitreous densification of cordierite and spodumene glass powders, fol-
lowed by crystallization below 1000°C into cordierite and spodumene
glass-ceramics [63]. This IBM technology is discussed in great detail
later in this chapter. The subject of glass-ceramics is covered in a textbook
by McMillan [64]. This process allowed fabrication of high-temperature
ceramics using low-temperature vitreous systems, thereby providing op-
portunities for cofiring in a controlled atmosphere with copper metalliza-
tion to form multilayer ceramic substrates as reported first in 1980 [65].
Subsequent to this invention, a number of glass ± ceramics, most of them

Table 9-23. Low-Dielectric-Constant Glasses, Ceramics, Glass + Ceramics

Dielectric Constant
Material @ I MHz

Borate glass 3.2


Silica glass 3.8
Borosilicate glass 4.0
Lead borosilicate glass 7.0
Magnesium aluminosilicate glass 5.0
Calcium aluminosilicate glass 6.0
Lithium silicate glass 6.5
Cordierite 4.5
Mullite 6.7
Alumina 9.8
Borosilicate + silica 3.9
Borosilicate + alumina 5.6
Borosilicate + mullite 4.5
Borosilicate + cordierite 4.2
Lead borosilicate + alumina 7.8
Calcium aluminosilicate + alumina 7.9
Lithium silicate + silica 5.0
Magnesium aluminosilicate + silica 5.0
Magnesium aluminosilicate + alumina 6.5
Magnesium aluminosilicate + mullite 6.0
11-338 CERAMIC PACKAGING

Table 9-24. Properties of Glasses, Glass + Ceramics, and Glass-Ceramics

Coefficient of
Thermal
Dielectric Expansion
Materials Constant (l0-7FC) Conductor

Glasses
B203-Si02-AI203-Na20 4.1 32 Cu
Glass + Ceramics
PbO + B20 3 + Si02 + (AI20 3) 7.5 42 Au, Ag, Ag + Pd
MgO + Ah03 + Si02 + B20 3 + (AI 20 3) 4.5 30 Au, Ag, Ag + Pd
B20 3 + Si0 2 + (AI20 3) 5.6 45 Au, Ag, Ag + Pd
2Mg02A120 3·5Si02 + (Ah0 3) 5.5 30 Au
CaO + Ah03 + Si02 + B2 + 0 3 + (Ah03) 7.7 55 Ag, Ag-Pd
Li 20 + Si0 2 + MgO + Al20 3 + Si0 2 + (AI20 3) 7.3 59 Au, Ni, Ag-Pd
Li20 + Al20 3 + SiOz + (Ah03) 7.8 30 Au
Glass-Ceramics
MgO-AI203-SiOz-Bz03-PzOs 5.0 30 Cu
Li20-AI203-Si02-B203 6.5 25 Cu

as glass additions to alumina, at roughly 50/50 wt. % level were reported.


Table 9-24 lists a number of these compositions, the metallization devel-
oped to cofire with each, and the properties achieved for each multilayer
ceramic substrate. It should be noted, however, that glass-ceramic material
technology originated from Corning Glass Works in the late 1950s, and,
subsequently, glass-ceramic pastes became available from DuPont and
others for sequential, crossover, multilayer applications onto fired alu-
mina substrate.
In those applications that require wirebonding or tape automated
bonding (TAB) of chips to these low-dielectric-constant, low-thermal-
conductivity ceramics, heat dissipation becomes a problem. A number of
compositions containing high-thermal-conductivity additions of beryllia,
silicon nitride, and artificial diamonds, shown in Table 9-25, are being
explored [66]. In addition to those previously discussed [63], glass-
ceramic compositions based on cordierite formation, with a dielectric
constant around 5 and thermal expansion near that of silicon [66], are
given in Table 9-26.
Copper or gold metallization with glass-ceramic to form a high-
conductive (electrical) multilayer ceramic is the most desirable substrate
and perhaps the hardest to achieve technically. Copper has the best combi-
nation of electrical conductivity, sintering temperature, and wettability to
braze and solder. The material is available in abundance, the powder
processing is generally known with desired physical and chemical proper-
9.6 LOW·TEMPERATURE CERAMIC OR GLASS + CERAMIC PACKAGING 11-339

Table 9-25. High-Thermal-Conductivity Glass + Ceramics

Various ComI!ositions
ComQonent I 2 3 4 5 6

Al20 3 50wt.% 33.3 wt.% 33.3 wt.% 33.3 wt.% 50wt.% 96wt.%
B20 3-Si02 50wt.% 33.3 wt.% 33.3 wt.% 33.3 wt.% 40wt.%
BeO 33.3 wt.%
Si3N. 33.3 wt.%
Artificial 33.3 wt.% lOwt.%
Diamond
Property
Thennal 4 48 32 48 20 20
conductivity
(WIM K)
Dielectric 5.6 5.2 5.0 5.5 unknown 9.4
constant

Source: From Ref. 46.

ties, and the paste technology is readily available. Cofired gold metalliza-
tion is not as desirable, primarily due to the high cost of gold raw material.
Thick-film copper metallization into glass-ceramic multilayer ceramic
poses chemical and mechanical incompatibility problems, however. The
chemical incompatibility arises as follows.
Basically, any conventional atmosphere in which organics could be
removed from greensheets during multilayer ceramic fabrication results
in oxidation of copper. Conversely, in any reduced atmosphere in which
copper could be maintained as pure metallic copper, the green sheet organ-
ics remain or "char" the ceramic, thus spoiling its electrical properties.
Although it would seem that the standard nitrogen firing of copper thick
films should be applicable for greensheet multilayer ceramic laminate,
this is not the case for two reasons. Paste organics are chemically different
from greensheet organics, and the paste thickness is much thinner than
the green laminate thickness (20 ~m versus 5000 ~m).
Silver, gold, and silver-palladium metallizations requiring no con-
trolled atmospheres for bum-off of binders have been reported [67]. The
substrate sintering processes are simpler than those involving copper, but
the materials are considered expensive for volume manufacturing. One
example of these processes is illustrated in Figure 9-39, wherein Ag-Pd
is used in the external layers of substrates which contain silver in the
internal layers. Silver-palladium has been reported to provide "hermetic-
ity" to the package.
Both glass + ceramics, by control of volume fractions or the selection
11-340 CERAMIC PACKAGING

Table 9-26. Selected Glass-Ceramic Compositions Based on Cordierite


Formation

Coefficient of
Dielectric Thermal
Material (wt.%) Ceramic Constant Expansion
~ AI,OJ MgO B,O) Others Addition I MHz 1O-7O C

ZnO P,Os
56.0 23.5 15.0 1.0 3.5 1.0 5.5 30
ZnO P,Os
57.5 25.5 12.0 2.0 2.5 0.5 5.2 25
TiO, P,Os + AI,OJ
52.0 20.0 17.0 2.0 7.0 2.0 20 vol. % 5.5 37
TiO, P,05 +BN
52.0 20.0 17.0 2.0 7.0 2.0 20 vol. % 5.0 32
CaO zrO,
45.0 32.0 12.0 4.5 4.5 2.0 5.5 22
CaO zrO,
50.0 30.0 13.0 3.0 3.0 1.0 5.6 16
Y,OJ zrO,
58.0 22.0 12.0 1.0 6.0 1.0 5.7 35
Y,03 zrO,
55.0 17.0 21.0 3.0 3 .0 1.0 5.5 30

Source: From Ref. 66.

of ceramics and glass, and glass-ceramics, by selection of glass composi-


tions and heat treatment profiles, could readily be developed with thermal-
expansion coefficients near GaAs, the thermal-expansion behavior of
which is shown in Figure 9-41. Alumina substrate is considered a good
match. However, because it can only be metallized with such metals as
molybdenum or tungsten, which have low electrical conductivity, high-
performance packaging of GaAs with alumina MLC is considered a major
limitation. Low- and medium-performance applications can readily be
handled by alumina with C4 bonding or by TAB and wirebonding of
chips which pose no thermal mismatch problems.

9.7 STATE OF THE ART IN LOW-TEMPERATURE


CERAMIC PACKAGING
9.7.1 IBM's Glass-Ceramic Substrate
IBM's glass-ceramic is the first and true glass-ceramic that starts
out with glass and crystallizes to glass-ceramic. IBM developed and
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-341

-
,52
0
...
I
80

70
..-
~

c
0 60
'iii
c
cd
a. 50
x
W
Iii 40
E
"-
(I)

r- 30
~

'0
"E 20
CD
'u
!E 10
CD
0
()
0
200 400 600 800 1000 1200 1400
Temperature (I<)

Figure 9-41. Thermal Expansion of Gallium Arsenide

perfected the manufacturing of this technology for all of its top-of-


the-line mainframe computers based on a 63-layer substrate. The key
technologies in developing the 63-layer glass-ceramic/copper substrate
are discussed in the following subsections, excerpted from a recent IBM
publication [68]; glass-ceramic material, greensheet, copper powder and
paste, personalization, stacking, and lamination, glass-ceramic/copper
sintering, and substrate machining and finishing.

9.7.1.1 Glass-Ceramic Material

Table 9-27 lists various potential systems considered for forming


glass-ceramic substrates in terms of their thermal and electrical properties
as well as sintering temperature required to cofire with highly conductive
metal such as copper. None of the systems shown in Table 9-27 met
the requirements of a thermal-expansion coefficient near that of silicon
(30 x 1O-7/0C), a dielectric constant of 5.0, and sintering compatibility
with copper. In addition to these glass-ceramic systems, glass and ceramic
systems involving, for example, the addition of a greater amount of glass
to alumina than in typical alumina substrates were also considered, such
as a low-viscosity lead glass added to alumina ceramic. For the glass-
plus-ceramic compositions, binder removal was anticipated to be more
difficult because of the need for a low-viscosity glass required to sinter
the ceramic high density at about 950°C [69].
The glass developed to form the glass-ceramic (USP: 4,301,324
11-342 CERAMIC PACKAGING

Table 9-27. Glass-Ceramic Systems Considered

Thermal Dielectric Sintering


Crystal type Expansion Constant Temperature

Cordierite Too low Good Too high


2MgO . 2Al 20 3 . 5Si02
Beta-spodumene Too low Too high Too high
Li 20 . Al 20 3 . 4Si02
Celsian Good Too high Too high
BaO . Al20 3 . 2Si02
Anorthite Too high Too high Too high
CaO . Ab03 . 2Si02
Glass + alumina High High 950°C

filed: Feb. 1978) has the following composition-Si02: 50-55 wt.%;


A1 20 3: 18-23 wt.%; MgO: 18-25 wt.%; it contains P205 and B20 in 0-3
wt.% [70]. Table 9-28 indicates the role of each ofthe oxides in determin-
ing final properties of ceramic.
Crystallizable cordierite glass offers process advantages for multi-
layer ceramic-Cu substrates. These advantages can be examined with the
use of differential thermal analysis (DTA), as shown in Figure 9-42. To
form the glass-ceramic substrate with copper, it is important that no
sintering occurs below 780°C and that complete crystallization occurs
below the melting point of copper (1083°C). In Figure 9-43, both of these
conditions are met, allowing adequate binder removal at temperatures
below those at which sintering occurs, complete densification prior to
crystallization, and crystallization well below 1083°C, the melting point

Table 9-28. Glass Composition (wt. %) to Form IBM's Glass-Ceramic

Component wt. % Range


Si0 2 50-55 Control properties
Al20 3 18-23 Control properties
MgO 18-25 Control properties
P20S 0-3 Nucleating agent, sintering aid
8 20 3 0-3 Delays crystallization, sintering aid

Glass transition temperature: 780°C


Complete densification temperature: 800°C
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-343

IBM cordlerite (10 OCImln)

1
1: 780 "C glass transltlan temp.
2: B50-925 "C slntering 3
3: looo"C crystallizatian
4: 132O·C melting
1
2
J/ ,...,...
I: 1 1!
) ._ _ _--..:...1

600 800 1000 1200 1400


Temperature ( 'C)

Figure 9-42. DitTerential Thermal Analysis of Cordierite

of copper. The crystallization peak occurs at approximately 1000°C in


the dynamic DTA experiment and at lower temperatures around 950°C
for actual substrate formation.
When a stoichiometric cordierite glass is heated in a DTA experi-
ment, the crystallization exotherm initiates at a lower temperature and
proceeds at a higher rate. With the shift from stoichiometry and with
small amounts of B 20 3 and P 20 5, the onset of crystallization is delayed to
a higher temperature and the crystallization is reduced. This crystallization
temperature increase and crystallization rate decrease help enable the glass
to sinter to 100% of theoretical density [71]. Further studies on glass

14

12 Cordierite

Log 10
Poise
8 No Densification

400 600 800 1000


Temperature DC

Figure 9-43. Viscosity-Temperature Relation for Borosilicate and Cordier-


ite Glasses
11-344 CERAMIC PACKAGING

viscosity [72,73], sintering [74], and crystallization [75,76] show the


effects of compositional modifications on final density.
The preferred powder particle size is about 3 Ilm in diameter, and
because of surface nucleation, the final grain size is also typically less
than 3 Ilm. The density of this material is more than 99% of theoretical
value, and the pores generally measure less than 5 Ilm. The microstructure
and x-ray diffraction data indicate the glass-ceramic to be approximately
95% alpha-cordierite, and the remainder to be enstatite. The properties
of this material are shown in Table 9-29.
To meet the stringent requirements of the glass-ceramic substrate,
a modification of the binder system and the continuous casting process
used previously for the alumina system was developed. It is capable of
producing greensheets with the following characteristics:

• Stable dimensional properties after punching and handling


• Microporosity levels which enable sufficient deformability to enclose
screened lines and permeability to allow escape of gases during the
sintering process
• Wetting by copper paste solvents without dimensional change
• Complete removal of organics in the sintering cycle

Glass powder is formed by grinding the glass in a high-alumina ball mill


to an average particle size of about 3.0 Ilm. The particle size, particle-
size distribution, and surface area of the glass prior to incorporation into
greensheets are critical control parameters in the fabrication of reproduc-
ible greensheets. [77]
The green sheet obtains its stability, porosity, and chemical inertness
from a binder system consisting of polyvinylbutyral (PVB), dipropylene-
glycol dibenzoate (BPGDB) plasticizer, and a binary solvent of methanol
and methylisobutyl ketone [78]. This formulation provides a stable suspen-
sion of the glass particles in the slurry with a viscosity suitable for
continuous casting of high-quality greensheets having the required me-

Table 9-29. Properties of Cordierite Glass-Ceramic

Dielectric constant 5.0 at I MHz


Thermal-expansion coefficient 30 x 1O-7/ O C (25-200°C)
Mechanical strength 210 MPa (30 kpsi)
Maximum process temperature 950°C
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-345

chanical strength and dimensional stability. The resulting greensheet has


about 90% glass powder and 10% binder (PVBIDPGDB) by weight.
The polyvinyl butyral binder and solvent system provides excellent
slurry properties. It has a viscosity of about 1500 CP and has a behavior
which is pseudosplastic at lower shear rates and Newtonian at higher
shear rates typically experienced in continuous casting. These properties
permit precise control of the slurry in the caster, resulting in a precise
greensheet thickness necessary for the dimensional control of the sint-
ered substrate.

9.7.1.2 Copper Powder and Paste Technology


IBM chose copper as the most desirable metal for cofiring with
glass-ceramic because of its high electrical conductivity and low cost as
indicated in Table 9-30, compared to other metals. This metal, however,
presents major difficulties in forming suitable powders and pastes to
fabricate 63-layer glass-ceramic/copper substrates, as discussed pre-
viously [69].
A powder suitable for screening 90- and 100-J.lm vias on 225-m
centers as well as fine lines (75 m) must have a small particle size. This
requires that the powder be deagglomerated so that it can be screened
through a metal mask without clogging the holes in the mask. Unlike
many other metal powders, copper is difficult to de agglomerate during
paste manufacture. Copper particles tend to deform rather than deagglom-
erate. When copper thick-film technology originated, no known powder
or paste met the requirements of the glass-ceramic/copper substrate de-
scribed earlier.
As expected, most fine, commercially available copper powders
sinter between 600°C and 780°C, whereas the glass begins to densify at

Table 9·30. Package Metallurgy Options

PropertylMetal Mo Cu ,ig ~ Au

Bulk resistivity (~Q cm) 5.2 1.72 1.59 7.0 2.35


Thick-film Resistivity 10-15 4-6 3.7 2.0 3.75
(~Q cm)
Sintering temp. (OC) 1500 950 950 850 950
Atmosphere f-Reducing-t ( Air )

TCE X ppml°C 60 180 170 130 140


Concerns Sintering Electromigration Cost
atmosphere
control
11-346 CERAMIC PACKAGING

about 800°C and completes by about 860°C. This mismatch, presented


in Figure 9-44, presents major problems in the fabrication of multilayer
glass-ceramic substrates, as discussed elsewhere [69].
As stated elsewhere [79], in fabricating the 63-layer glass-ceramic
substrate, the copper powder utilized has been suitably modified [80,81]
and cosintered with glass.
The interface between the copper and the glass-ceramic was studied
by means of a number of material and process options. The bonding
methods investigated were mechanical, chemical, and a combination of
the two. Compatible glasses and glass-ceramics were used for mechanical
interfacial bonding, whereas chemical bonding was achieved by incorpora-
tion of suitable oxides in the paste. With both bonding methods, a con-
trolled atmosphere was utilized during the sintering process to enhance
the interfacial bonding. The details of these technologies are consid-
ered proprietary.

9.7.1.2.1 Copper Thick Film


The IBM molybdenum thick-film (paste) technology used to person-
alize the alumina TCM as well as thick film used by others are well
established as discussed in the alumina section, 9.5. The personalization
of the new glass-ceramic/copper package has a higher density with de-
creased grid spacing as indicated in Table 9-31, requiring significant
improvement in the paste technology. For example, via diameters are 90
and 100 ~m on 225-~m spacing for glass-ceramic, compared to via
diameters of 140 and 150 ~m on approximately 300-~m spacing for

25
11/
CI 20
~
c
c
.;:
.c
15
...c
V)

10
III
...U
III
Q" 5

0
500 600 700 800 900 1000
Temperature (0C)
Activation Energy for Sintering of Copper -50 KCaVMole

Figure 9·44. Shrinkage Mismatch Between Cu and Glass-Ceramic


9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-347

Table 9-31. Comparison of Feature Parameters for Alumina and


Glass-Ceramic Substrates

Technology AluminaMLC Glass-Ceramic


MLC

Linewidth (~m) 116 75


Line spacing (~m) 300 225
Wiring length (~m) 130 400
Smallest via diameter (~m) 140 90
Largest via diameter (~m) 150 100
Via spacing (~m) 300 225
Vias per layer 33,000 78,500
Vias per substrate 350,000 2,000,000

original alumina. In addition, more than 78,500 vias are screened simulta-
neously with lines on each green sheet. These vias extend through the
green sheet and are filled simultaneously as the lines are screened.
To form these densely wired structures, paste interaction with the
greensheet must be considered. The interaction between the solvent in
the paste and the binder of the sheet must be small but not zero. Control
of this interaction is essential for dimensional stability of the sheets after
screening, thus ensuring proper alignment of the vias in as many as 63
discrete layers during lamination. Solid content, particle-size distribution,
rheology stability, and cross-sectional area are essential parameters for
reproducible pastes.
A significant challenge in screening was to achieve good leveling
of the screened features without spreading of the screened geometry.
This is especially important for the glass-ceramic substrate, where the
requirement for high electrical conductivity makes it necessary for the
paste to have high solid content. The choice of paste solvent (or vehicle
system) was critical, as its interaction with greensheets places a limit on
the amount of leveling that can occur. A minimal amount of surfactant,
combined with an appropriate amount of thixotrope, resulted in a paste that
sets up adequately at the edges of the screened features while maintaining
sufficient fluidity in the body of the paste to provide excellent leveling,
thus producing copper features with a desirable cross-sectional area.
Extensive experimentation was carried out in order to understand
the trade-off among the various properties required. Vehicle systems
containing various polar and nonpolar solvent such as acetates, oils, esters,
and glycols were considered together with polymer systems containing
compatible functional groups. Screening process optimization was carried
11-348 CERAMIC PACKAGING

out by means of statistical studies to simultaneously adjust three major


parameters as a function of paste composition: line-height variation, line-
width variation, and greensheet dimensional control.
In-line controls such as grind gauge for the dispersion of particles,
viscosity, and theological measurements were instituted.
The resultant thick film is a balanced blend of pseudoplasticity,
thixotropy, and yield strength. A typical thick-film composition includes
80-90 wt. % copper and 20-10 wt. % paste organics and solvents. This
thick film is suitable for a wide variety of screened layers (e.g., signal,
redistribution, and vias only), which in the past required many differ-
ent formulations.

9.7.1.3 Personalization, Stacking, and Lamination Technology

After the greensheet is cast, dried, stripped from the carrier and
blanked into squares, it is "personalized" in one screening operation. A
specific wiring is deposited and a through-via pattern obtained by punching
vias through the sheets and filling them. Subsequently, the appropriate
individual layers are aligned one on top of another and then laminated
together. For the alumina-based thermal-conduction module, the vias were
formed by means of mechanical punches, and the screening was done
with various formulations of molybdenum metallurgical paste. Although,
in principle, the same procedure was used for the glass-ceramic program,
several factors presented major challenges with respect to the personaliza-
tion of the latter system. These included the chemistry of new materials,
and, most importantly, the dramatically increased density offeatures (both
vias and lines). Table 9-31 compares some feature parameters for the
two systems.

9.7.1.3.1 Punching
The change in materials from alumina to glass-ceramic presented
interesting challenges with respect to the personalization of the green-
sheets. It is well known that the stresses induced in a green sheet during
punching cause macroscopic deformations in the sheet [82,83]. All other
factors (organic binder, sheet thickness, and punching parameters) remain-
ing constant; this deformation differed significantly in glass-ceramic and
alumina greensheets.
The magnitude and causes of distortion influenced by a number
of materials and process parameters were investigated. In particular, a
comprehensive study of both mechanical and chemical properties of the
system was performed.
One mechanical study involved the effect on via location error of the
thickness of the greensheet being punched and the difference in diameter
between the punch and its associated die bushing, as illustrated in Figure
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-349

Greensheet

Die bushing

Figure 9-45. Schematic of Greensheet Punching

9-45 . Figure 9-46 shows a figure of merit for via location error as a
function of one of the controllable parameters, the punch-to-die clearance.
It can be seen that if the sheet thickness is kept low enough, the sheet is
relatively insensitive to other factors. Beyond a certain critical sheet
thickness, relatively small changes in punch-to-die clearance have a large
effect on the eventual distortion of the punched sheet. On the other hand,

... 25 x 102mm
...0... t:,. 152
QI
C
20 0203
+229
0 0254
',0: • 279
C
u
..2
15 *406 *
.....
QI 0...-
....
-
:::J 10
C
QI
QI
> 5
-.0:
c t::,..
x it ~
Qj
a:: 0
0 20 40 60 80 100
Punch-to-die clearance (J.Im)
Figure 9-46. Relative Greensheet Feature Location Error as Function of
Punch-to-Die Clearance
11-350 CERAMIC PACKAGING

for any given set of materials and sheet thicknesses, an increase in the
clearance between the punch and its associated die bushing results in a
decrease in the distortion of the punched sheet.
A fundamental reason for the difference between the alumina and
glass-ceramic sheets during punching was found to be the interfacial
chemistry between the inorganic components of the greensheets (glass
versus alumina) and the organic binder, as discussed in Refs. 84 and 85.
This effect was studied utilizing materials with a known degree of surface
aciditylbasicity (measured by their isoelectric point). There was a direct
correlation between feature location accuracy and the known degree of
aciditylbasicity.
9.7.1.3.2 Screening
A new screening process was developed for the copper paste because
it was drastically different in composition and rheology from that used
on alumina sheets. In the latter system, the pastes were customized with
respect to their solvent and binder materials for individual patterns. For
the glass-ceramic system, a single materials set for solvent and binder
was used. This required the use of a system which was tailored for limited
reaction with the green sheet. This provided sufficient adhesion of the
screening paste to the greensheet without causing significant dimen-
sional instability.
The properties of the conductive metal within the paste were radically
different. Molybdenum particles have high hardness and a platelet shape,
whereas copper particles are malleable and more nearly spherical. The
change in density and surface area required extensive cooperation between
the screening and the paste formulation development teams in order to
fine-tune the rheology of different paste formulations and the screening
parameters (pressure, wipe speed, etc.) for the individual layers. Major
developments in the fabrication of screening masks were another key
ingredient in the final success of the actual screening operation in the
manufacture of glass-ceramic substrates.
9.7.1.3.3 Inspection
Automated inspection of screened greensheets ensures that via fill
and line dimensions meet specifications. Advances in this technology
were required because of the reduction in feature size for glass-ceramic
with copper thick film. This operation is critical to achieving high electrical
test yields after sintering, because a single defective layer, if not rejected,
can cause the loss of the 62 other defect-free layers which combined
make up a substrate.
9.7.1.3.4 Stacking and Lamination
During stacking and lamination, the individual sheets must be stacked
one on top of another; they must be aligned so that any additional distortion
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-351

is kept to a minimum during this process. The registration of all the sheets
in the stack must be so precise that good electrical contact is made from
vias in one layer to vias in the next layer for every one of the nearly two
million vias in the laminate. Even after this is achieved, the stack must
survive the lamination process, in which the stack is heated (under pres-
sure) to a temperature just above the glass transition temperature of the
binder in the sheet. Aside from the tooling challenge of building a press
in which the platens have very tight tolerance placed on their flatness,
parallelism, and temperature uniformity, a delicate balance had to be
struck between conditions of temperature and pressure which would permit
enough flow to allow for the consolidation of the individual sheets into
a single unit, but were not so severe as to introduce additional distortion
due to lateral flow.

9.7.1.4 Glass-Ceramic/Copper Sintering Technology


Sintering a multilayer ceramic substrate requires cofiring a composite
matrix of organics, metal powders, and ceramic and glass powders into
a complex network of electrically conductive wiring embedded in a dense,
insulating ceramic body. The glass-ceramic/copper substrate technology
presented the following challenges to co-sintering:

• Low-temperature organics removal «800°C), without copper oxida-


tion and low-temperature densification «1000°C)
• Crystallization of glass to form a high-strength ceramic
• Densification of copper powder to high-conductivity copper
• Copper/ceramic interface integrity
• Improved dimensional control to satisfy requirements for via align-
ment and for thin-film wiring

9.7.1.4.1 Binder Removal and Carbon Oxidation


Various approaches were evaluated to reduce the residual carbon
content to the desired low levels. These investigations included polymers
that pyrolyze primarily into gaseous species ("unzip"), atmospheric firing
cycles for oxidation of carbon and copper, and subsequent reduction of
copper oxide to copper and catalysts to aid the binder bum-off and varia-
tions in compositions of glass as well as glass added to ceramics. As
illustrated in Figure 9-43, the binder bum-off requirement based on viscos-
ity of glass have led to the composition selected. Differences in thermal
degradation were noted for the free binder and the binder absorbed on a
glassy surface. When the binder was mixed with powdered glass to simu-
late tape casting, all candidates left excessive residual carbon in the glass.
This occurred because the surface of the glass was chemically active and
11-352 CERAMIC PACKAGING

Dielectric Constant
6

5.8

3000 5.6

5.4

5.2 --..,../
2000 5~ ___________________

200 250 300 350

Residual Carbon, PPM


1000

0L-___~==========~~____
o 200 400 600 800
Residual Carbon, PPM

Figure 9-47. Dielectric Constant as a Function of Residual Carbon in Glass-


Ceramic. (Courtesy of lliM.)

bonded with the polymer. The result of this residual carbon as affecting
the dielectric constant is illustrated in Figure 9-47.
The challenges in cofiring copper paste with glass greensheet to form
a 63-layer structure of glass-ceramic/copper substrate can be overcome by
the use of steam and hydrogen atmospheres [69]. Figure 9-48 indicates

o
-2

2eu + H20 = Cu,o + H2

-14
• 0~--~20~0----400~----600~--~8~00~--~1000

Temperature (DC)

Figure 9-48. Thermodynamics of Cu-O-C System


9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-353

the PH/PH20 ratio necessary at 800D C to oxidize carbon and reduce copper
oxide according to the following chemic reactions:

C + H20 -7 CO2 + H2
CU20 + H2 -7 2Cu + H20

Whereas the oxygen partical pressure required to drive the above


reactions can be achieved by a number of ambients [86], the only kinetic-
ally acceptable process for forming the 63-layer structure in a few hours
is the one based on the use of steam.

9.7.1.4.2 Densification
Densification of the glass-ceramic/copper substrate depends on the
composition of the glass-ceramic and the particle-size distributions of
the glass and metal powers. Greensheet laminate properties such as the
ratio of organics to inorganics and laminate density play a role in the
control of shrinkage.
The objective of the sintering process is to define a temperature
profile, atmospheric conditions, and fixtures to achieve a dense, high-
strength ceramic and dense, highly conductive copper wiring. Copper/
ceramic interface integrity also depends on materials properties and sinter-
ing atmosphere control. The glass-ceramic/copper sintering process devel-
opment grew out of experience gained in developing the alumina-
molybdenum sintering technology and required improvements in tools
and processes. Steam furnaces were developed and designed to sinter
glass-ceramic/copper substrates, which are believed to be an industry
first. The furnaces required significant advances in mass flow control,
atmosphere injection systems, and reaction chamber materials. Unique
fixtures helped achieve the required mass transport for carbon oxidation
and removal which affects materials densification and substrate character-
istics. The results of this development effort were illustrated previously
in Figure 9-2. This substrate sets new standards for multilayer ceramic
packaging, as evidenced by the final properties presented in Table 9-32
[69]. The overall process developed for the glass-ceramic substrate is
indicated in Figure 9-49.

9.7.1.4.3 Dimensional Control


Dimensional control during the densification segment of the glass-
ceramic/copper sintering is a function of processing before and during
sintering [87,88]. Variations in greensheet properties, paste properties,
paste loading, and lamination conditions have a significant impact on
dimensional control and must be carefully controlled. Improved furnace
materials, fixtures, and methods for temperature and atmosphere control
11-354 CERAMIC PACKAGING

Table 9-32. Sintered Glass-Ceramic/Copper Substrate Properties

Fired ceramic density (glcml) (99.5%) 2.66


Shrinkage tolerance (%) ±O.lO
Camber or sintered flatness (lim/cm) 2.3
Ceramic flexural strength (MPa) 210
Copper resistivity (lin cm) 3.5
Dielectric constant 5.0
Thermal expansion coefficient (lO-10C) 30
Number of layers 63
Wiring density (cm/cm2) 800

Source: Courtesy of IBM.

have been developed for sintering glass-ceramic/copper substrates. These


advances in sintering technology have resulted in significant improvements
beyond state-of-the-art alumina-molybdenum sintering with respect to
dimensional control.
With respect to requirements for dimensional control in a sintered
substrate, the goal is to have all surface features located exactly within
specifications relative to one another and the edges of the substrate. In a
multilayer ceramic substrate, deviations from ideal locations are allowed
within specified limits. For example, a perfect substrate, where metal
features have no deviations from ideal locations, is said to have 0% nominal
movement. An MLC substrate tends to deviate from ideal locations by
expansion or contraction around the desired value. On the other hand, if

1000

u 800
0

.
II

.a 600
E
..
c
II 0
a.. +I
E 400
'tJ
~ ';(

200 ...CII
0
'tJ
C
0 iii
Time
Figure 9-49. Overall Glass-CeramiC/Copper Cosintering Cycle
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-355

all the surface features expand or contract linearly, there is zero radial
feature displacement or distortion. Radial feature displacement is measured
in microns using an optical measurement system.
Figure 9-50 compares the dimensional control in two populations
of sintered substrates. The larger area represents alumina-molybdenum
substrates, and the smaller area represents glass-ceramic/copper sub-
strates. Perfect dimensional control with ideal linearity (zero distortion)
is represented by the two dark lines forming a "V." If, for example, a
sintered substrate had a global expansion of +0.10% with ideal linearity,
the radial feature displacement would be on top of the right-hand dark
line just above the +0.10% mark. Therefore, the goal in dimensional
control is twofold: (a) It should be close to zero nominal expansion, and
(b) when deviations occur, the expansion should be linear.
As the data in Figure 9-50 illustrate, dimensional control in the
glass-ceramic/copper population is significantly better than in the
alumina-molybdenum population. Dimensional control in the glass-
ceramic/copper population deviates less from the ideal dimensions and
is closer to perfect linearity.

9.7.1.5 Substrate Machining and Finishing Technology

As described in the previous sections, the glass-ceramic/copper


substrate is a culmination of extensive materials and process development.
However, the sintered substrate, advanced as it is at this stage, is not
yet complete. The thin-film redistribution wiring, top-surface terminal
metallurgy, and back-side liD pins must still be applied to the top and

....c::
cu

Q.~1
VI
:0 §
~t
........
:I 0
c VI

-
~:o
c 0 ...
:0
c
~

-0.30 -0.20 -0.10 0 0.30


Nominal expansion (%)

Figure 9-50. Dimensional Control of Alumina TCM Compared to Glass-


Ceramic TCM
11-356 CERAMIC PACKAGING

bottom surfaces and preparation made for mounting and finishing pro-
cesses prepare the substrate for these final operations.
The application of thin films onto the glass-ceramic substrate gener-
ates new requirements not encountered in the fabrication of earlier sub-
strates. Specific requirements pertain to the substrate dimensions, flatness,
surface roughness, and the condition of the metal vias exiting the surface.
Additionally, a flange must be ground into the periphery of the substrate
for module encapsulation. These requirements are met using a series of
operations summarized in the process flow given in Table 9.33.

9.7.1.5.1 Postsinter Sizing


Precision diamond-saw sizing is the first machining operation for
the sintered substrate. This operation precisely sizes the substrate to
127.5 mm. In this operation, the substrate is cut square, the comers are
truncated, and a 45° bevel is machined on all edges. The truncating of
the comers and beveling of the edges help to minimize chipping and
damage to edges during subsequent operations.
The sizing process uses resin-bonded diamond blades on a double
arbor wet saw. A specially formulated coolant is used to minimize the
buildup of heat and prevent mechanical damage to the substrate. The saw
sizing operation is completely automatic, first cutting the sides of the
substrate and then truncating the comers. Precise dimensional control as
well as pattern centrality are important for subsequent processing. This
is achieved by the use of an integrated optical alignment technique that
targets pattern alignment fiducials on the substrate and transfers these
alignment data to the saw. After the cutting operations are complete, the
newly formed edges are beveled on a deburring machine.

Table 9-33. Details of Substrate Machining and Finishing Processes

Process Process elements

Sizing Diamond sawing Size = 127.5 mm


Planarization Abrasive slurry on lapping table Flatness = 2.3 mlm
Vias above substrate surface
Flange grind Abrasive wheel Surface roughness = 4000 A
Thickness/parallelism = 50 mm
Ceramic polish Submicron abrasive slurry with Surface roughness = 450 A
polishing pad Vias coplanar with surface
Metal polish Submicron abrasive slurry with Reflective vias
polishing pad
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-357

9.7.1.5.2 Planarization
The accurate application of thin films to the substrate requires that
it be planarized to produce flat surfaces. Further, because the substrate is
electrically tested after planarization, the process must leave the copper
vias in the surface in order to ensure probe contact during electrical test.
A free-abrasive lapping process is used to planarize the substrate.
The process utilizes an abrasive slurry and a lapping table to planarize
both the top and bottom surfaces of the substrate. Slurry consistency and
its distribution to the part, as well as lapping table speed and the pressure
applied to the part, are important in achieving the desired flatness. The
machining mechanics of a free-abrasive lapping operation result in the
enhanced removal of the ceramic over the more ductile copper [89]. Thus,
after planarization, the copper vias protrude above the ceramic surface,
as required for electrical testing.
Measurement of the planarized surface is critical to the assessment
of process integrity. An automated laser scanning interferometer not only
measures the overall flatness of the part but also gives a contour map
of the surface, showing the location and the degree of deviation from
ideal flatness.

9.7.1.5.3 Seal Flange Grind


Once a substrate has been planarized and electrically tested, those
parts passing an electrical test have a seal flange ground into their periph-
ery. The flange is used to mount and seal the top surface of the substrate
into the module hardware.
Key elements of the flange are its surface finish and its parallelism.
The top surface of the flange must be smooth enough to permit a hermetic
seal to the module. The flange surfaces must be parallel, so that an even
distribution of pressure to the module seal is obtained without bending
moments on the ceramic.
A ground seal flange is not new to ceramic technology; alumina
TCM substrates have had this feature for several years [90]. However,
machining such a flange into the glass-ceramic/copper substrate required
the development of a new process, largely because of the differences in the
physical properties between alumina and the crystallized glass-ceramic.
The seal flange is machined into the substrate by means of a cam
grinder using a specially shaped, abrasive wheel that grinds both the
top and bottom flange surfaces simultaneously. A combination coolant!
lubricant is used in conjunction with the process. This coolant, along with
the speed of the grinding wheel and the feed rate of the substrate, is critical
in preventing damage to the ceramic during processing. Considerable effort
has also gone into designing the fixturing used to hold the substrate during
grinding. The fixtures must be able to resist the bending and vibrational
modes placed on the substrate during grinding while maintaining the
11-358 CERAMIC PACKAGING

dimensional accuracy of the resultant flange. After the flange has been
machined, the same abrasive wheel is used to bevel the flange edges to
prevent chipping of the ceramic. By means of this process, the flange
thickness and parallelism are controlled to within 50 J.lm, and the surface
roughness is maintained at 4500 A.

9.7.1.5.4 Polishing
The complete surface finishing of a glass-ceramic/copper substrate
requires two polishing operations. The first operation, the ceramic polish,
produces a smooth surface suitable for the application of thin-film metalli-
zation. The second operation, the metal polish, brightens the surface of
the copper vias to enable them to be seen by the automatic mapping tools
used in thin-film processing. These polishing operations are used on both
the top and bottom surfaces of the substrate.
The ceramic polishing operation was developed to provide a smooth
surface, maximize surface flatness, and achieve copper vias which are
coplanar with the ceramic surface. The ceramic polish uses a submicron
abrasive slurry and polishing pad to achieve these goals. Process parame-
ters such as applied pressure, table speed, substrate rotation speed, slurry
concentration, and slurry delivery rate are also critical to the quality of
the polish. Their relative importance in the process was determined using
a robust experimental design methodology [91]. Considerable effort has
also gone into developing fixturing for the substrate that ensures a nearly
uniform polishing rate across the entire substrate surface. The ceramic
polish routinely yields substrates with a surface roughness of 450 A,
minimal change to the surface flatness achieved during the lapping process,
and copper vias coplanar with the ceramic surface.
The metal polish operation also uses not only a pad but also a still
finer submicron abrasive slurry. The improved reflectivity of the copper
achieved by this polish allows for precise optical mapping of via location
for subsequent thin-film processing.

9.7.2 Fujitsu's Glass-Ceramic Substrate


Fujitsu developed, manufactured and applied to its mainframe appli-
cation a very large (245 mm) glass + ceramic substrate with 61 layers of
copper wiring [92]. It explored two glass + ceramic systems, one in
cordierite + glass and the other in alumina + glass. Alumina-glass, for
the reasons discussed below, became a reality.

9.7.2.1 Glass-Cordierite System


Cordierite was chosen as the original ceramic material because of
its very low dielectric constant and its low thermal expansion near that
of silicon.
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-359

Glass powder and ceramic powder were milled with binder and
solvent for 20 h. The glass used in this system and in the alumina system
is a borosilicate glass. The slurry was cast into 500-llm-thick greensheets
using the doctor-blade method. These greensheets were then laminated,
giving a thickness of 1 mm, then fired in a N2 atmosphere at 1000°C for
5 h. The heating rate was 200°CIh.
The glass-cordierite substrate so fabricated exhibited high TCE. The
calculated TCE of the system is (2-3) x lO-6fOC, which is slightly lower
than that of silicon. The TCE actually measured for that system, however,
is 17 x lO-6/oC, or about seven times higher than the calculated value, as
shown in Figure 9-51. The formation of cristobalite explains this high TCE.

9.7.2.2 Glass-Alumina System

The same tests as described above were performed on the glass-


alumina system. No large difference was observed between the TCE
values calculated and those measured for the glass-alumina system. The
x-ray diffraction pattern of glass-alumina revealed no formation of cristo-
balite. The densification mechanism of the alumina-glass system can be
explained as liquid-phase sintering. The alumina is uniformly dispersed
in the glass matrix. Wettability between the glass and the ceramic powder
is essential to densification. A proper amount of glass yields high density
with high strength. An optimum firing point for the maximum densification
also exists for any given glass-to-ceramic ratio. A combination suitable
for co-firing with Cu or Au has an alumina content between 40 and 50

0.6 , . . - - - - - - - - - - - - - - - - - - - - - .

g Glass-30% cordierite
c
.21/1 0.4
c

f
Formation

"6 0.2
~
~
0~0-=~~~1~OO~-----=20bO~----~3~OO
Temperature COc)

Figure 9-51. Thermal-Expansion Curve of Glass-Cordierite System and Alu-


mina-Added Glass-Cordierite System
11-360 CERAMIC PACKAGING

1000 6.0

..
a
t...
2!
l 900 5.5 ~
E! 1:
II
A.
E
;c
! 8
~ U
'1:
Q:
E
800 5.0 i
~
E ~
a
0 Glass: 820 3- Si02- A120 3

700 4.5
20 40 60
Alumina content (wt%)

Figure 9-52. Changes in Dielectric Constant and Sintering Temperature of


Glass-Alumina System

wt.% (see Fig. 9-52). The powder used for glass-alumina-composite


materials is listed in Table 9-34.
To avoid carbonization of the binder, Fujitsu used a newly developed
binder system that easily decomposes without oxygen when heated. The
binder structure is shown in Figure 9-53. Most binders generally used in
ceramic boards show side-chain reaction or random rupture, but the new
binder shows depolymerization as illustrated in Figure 9-53 [6]. A 61-
layer lamination of greensheets, however, makes the burning out of all
binder very difficult, so Fujitsu also developed a new firing process which
can decrease the carbon residue in a fired body.
Carbon residue exceeding 100 ppm affected the densification of the
fired body as well as the flexural strength. A fired body containing less
than 30 ppm of carbon residue approached 100% of theoretical density,
but a specimen containing 1000 ppm shows 93% of theoretical density.

Table 9-34. Powder Properties for Glass-Alumina-Composite Material

Specific gravity Particle size Specific Surface Area


(gm/cm3) (mm) (m2/g)

Alumina 2.5 4.4 2.8


Borosilicate glass 2.2 4.5 4.2
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-361

Carbon concentration (ppm)

. : Carbon atom
0: Side chain

Figure 9-53. Novel Binder for Easier Binder Burn-ott. (Courtesy of Fujitsu.)

Carbon also affects the breakdown voltage; a specimen with more than
100 ppm of carbon shows 0 kV/cm voltage breakdown, as illustrated in
Figure 9-54. The proposed mechanism involves trapped carbon that is
intermixed with the glass matrix, creating short circuits that cause break-
down between the patterns. The multilayer substrates processed, however,
in the special fashion do not include more carbon than bodies fired in an
air atmosphere.

9.7.2.3 Mainframe Application of Glass-Ceramic


Fujitsu applied the multilayer circuit board for the VP 2000 series
supercomputer with total number of 61 layers, including 36 signal layers.
About 40,000 signal patterns are formed on 1 layer, and the total wiring
length of a signal pattern is about 1 km. The size of the substrate is 24.5
x 24.5 cm and 13 mm thick. The characteristic impedance of the signal
pattern is controlled to 65 Q, and the resistance of the copper pattern is
only 100 mQ/cm. Two signal layers are sandwiched between ground
and voltage layers to decrease the cross-talk noise and deviations in
characteristic impedance.
The properties of the multilayer glass-ceramic (MLG) substrate are
listed in Table 9-35. The MLG made it possible to develop a high-density
package that compares favorably with conventional organic printed-circuit
boards previously used for large-scale, high-speed computer systems. The
11-362 CERAMIC PACKAGING

-
E
c
60

>:
-
~

cu
CI
a
n
0
0
0

1
~ 30 r
]
~
ca

O~--------------~I~==~~------~
101 1()2 103
Carbon concentration (ppm)

Figure 9-54. The Relation Between Carbon Concentration and Breakdown


Voltage

gate density of the MLG used in the VP 2000 series is 10 times greater
than that of the Fujitsu M-780 organic printed-circuit board.

9.7.3 NEC's Lead Borosilicate-Alumina Glass + Ceramic


NEe developed its glass-ceramic similar to Fujitsu's except for the
nature of glass used [93]. The dielectric material used for the multilayer
glass-ceramic (MGC) package is a mixture of alumina and lead borosili-
cate glass. The composition of the mixture is 55 wt. % alumina and 45

Table 9-35. Properties of 61-Layer VP 2000 Circuit Board

MLG Characteristic Specifications

Conductor: copper (mn/cm) 100 (95 11m width)


Wiring length (m) 1000
Signal layers 36 (40,000 wiring patterns)
Characteristic impedance (n) 65
Dielectric constant (at I MHz) 5.7
Grid (mm) 0.45

Source: Courtesy of Fujitsu.


9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-363

wt. % lead borosilicate glass. This material can be sintered at as low as


about 900°C. The dielectric constant for this material is 7.8, less than
that for alumina [93]. The thermal-expansion coefficient achieved is 42
x 10-7/ o C, close to that of silicon chip. When sintered at about 900°C,
crystallization takes place in the sintered ceramic body as a result of
reaction between glass and ceramic. This reaction leads to high mechanical
strength for the multilayer glass-ceramic (MGC) substrate.
Crystallization takes place during sintering around 800°C. If coarse
glass particles are used, crystallization within the glass particles takes
place, forming crystobalite. However, if fine particles of glass are used,
glass reacts with alumina to form anorthite. The resulting substrate has
high mechanical strength, a low thermal-expansion coefficient, and a good
dielectric constant, as shown in Table 9-36.
This substrate has high flexural strength-more than 300 MPa. Cam-
ber is extremely small, even though the substrate size is 225 mm. The
dielectric constant is 7.8, and the dissipation factor is less than 0.3% at
1 MHz.

9.7.4 Panasonic's Glass-Ceramic


In contrast to IBM's, Fujitsu's and NEC's glass-ceramics described
above, Panasonic developed its own low-temperature cofired technology
(LTCC) and began to apply for such consumer products as VCRs at its
plant at Saijo. The required low-cost for these type of products necessitated
low-cost metallization, which was achieved by simple air firing using
Ag-Pd conductors. Figure 9-55 shows the structure, including the embed-
ded capacitors and surface resistors that Panasonic was able to achieve.
Panasonic also designed and developed a low-cost process technol-

Table 9-36. Typical MGC Substrate Properties

Material Characteristic

Flexural strength (MPa) 300


Dielectric constant at 1 MHz 7.8
Dissipation factor at 1 MHz <0.3%
Thermal-expansion coefficient eC-1) 42 X 10-7
Thermal conductivity (W/mK) 3.6
Linear shrinkage (%) 13.0
Shrinkage tolerance (%) ±<O.2%

Source: Courtesy of NEC.


11-364 CERAMIC PACKAGING

Conductor
(Au)
o 5 _ 1 , _ , ""..... Ie)
-D_ ....._ 7.4
.n.-al_1on _dont 6.1x lct'fC
-F1oxu...I_ngth 2500 kglcm'
o ConcIuc:tw
·Top 15mQ.1]
5mQ.1]
-I..., 3mQ.1]
1+--f-lnh,maICapaclto,
15mOJ[]
o R..lltar
·Top RuO, 110-1 mQ.1])
-I...,
o Cc!pdw
.Top
-Inn...
Internal conductor Capacitance 3OnF/cm'
lAg) Resistor (RuO,) Dlnlpotion Factor 1.21'
T...,.Cha,. YST

Figure 9-55. Consumer Electronic Ceramic Substrate. (Courtesy of Panasonic.)

ogy for cofiring with copper that is illustrated in Figure 9-56 [94]. This
process involves forming thick films with CuD and cofiring in air to
initially remove organics from greensheets and pastes and then reducing
the oxide in forming gas and finally forming a good bond between the
glass that flows from ceramic walls and slightly oxidized copper in the
via that is formed in the N2 atmosphere used in sintering. Panasonic
expects to integrate capacitors and resistors into this substrate as well.
Figure 9-57 illustrates the glass-ceramic/copper with surface resistors
currently practiced by Panasonic.

1000 Densification

u
•
!
i 500
!.
~
o~===+~~==~~==~~
15hr 1hr 1hr
Binder Cgi'bOn--=:""""
Conductor CuO Z
Cu
Dielectric _ _ _ _ N_on_-_si_nte_red
________Z'--S_i_nte_red_
layer

Figure 9-56. Low-Cost Ceramic Cofiring Process with Cu. (Courtesy of Pana-
sonic.)
9.7 STATE OF THE ART IN LOW-TEMPERATURE CERAMIC PACKAGING 11-365

LSI
c:...IuoIo<
f"'iIl!p.....!limD-Eil_--liIimD.--IIIPL, ~AeINl

• Subotrate (gIG.. ceramic)


• DWecIrIc: canstant 7.4
• Thermal epanaion coefficient 6.1 x I O~
• ........ 1INngIh 2500~

.,-
• Conductor
• Top Cu 2mOID
• NVAII plating
Cu(CuO) 3mQ/[J

..........
•
.......
• Top T~ (lOONIOKCIDI

Figure 9-57. Consumer Electronic Ceramic Substrate with Cu. (Courtesy of Pan-
asonic.)

9.7.5 David Sarnoff's Glass-Ceramic on Metal (LTCC-M)


A variation of glass-ceramic that is receiving great attention at
David Sarnoff Research Center is that illustrated in Figure 9-58 which is
accomplished by lamination of the desired number of layers on to a metal
substrate and cofiring the entire substrate [95]. The substrate, obviously,
must meet two major requirements. Its thermal-expansion coefficient

LTCC

IC in Cavity
Cu/Stainless Steel/Cu Metal Core

Few processing steps equal fast


turn around, high yield, and low cost

Figure 9-58. Low-Temperature Cofired Ceramic on Metal


11-366 CERAMIC PACKAGING

should be equal to or slightly higher than the glass-ceramic so as to put


the ceramic in compression. It should also form an excellent bond to
the metal or alloy. A copper-stainless steel-copper-composite structure
is proposed.
The process of fabricating the substrate is illustrated in Figure 9-59
and the advantages of fabricating this way as compared with conventional
cofiring of ceramic are listed in Table 9-37. Among the most notable
features is the zero shrinkage that is intrinsic to the process because the
glass-ceramic is bonded to the metal core all through the process. The
thermal conductivity of the substrate is enhanced greatly, and in those
applications that require even higher dissipation, the chip can be back-
bonded to the metal core and wirebonded to the substrate metal. Silver
conductors that provide the very high conductivity and yet provide a fast
firing cycle in air are proposed.

9.7.6 Other Ceramic Packages


Silicon nitride, boron nitride, diamond, and mixtures of carbides,
nitrides, and oxides with appropriate sintering additives are expected to
be explored for ceramic packaging. The addition of such low-dielectric-
constant glasses as borosilicates or fused silica (with dielectric constant
around 4.0) could be added to alumina, cordierite [93], and other ceramics
of low dielectric constant to produce ceramic substrates of dielectric

Slurry with ~ITape cast I


ceramic powder
IBlank tape I
Layer 1 Layer 2 Layer X
r-----------~------~~r-------~-,

Figure 9-59. LTCC·M Process


Table 9-37. Technology Comparison

PWB MCM-L MCM-D LTCC LTCC-M

Wiring density (cmlcm2) 65 200 800 800 300-600


Thermal conductivity om 0.01 10 0.1 30
(% of Cu)
...en.... Buried passives Resistors only Resistors only Resistors only Resistors & capacitors Resistors & capacitors
Limited range Limited range Limited range Poor yields Wide range
Expensive Expensive Expensive Good yields
Ruggedness High High Medium Medium High
Large-format processing Yes Yes Beginning No Yes
Component assembly Fair Fair Poor Fair Good
flexibility
Flip-chip capability Good Good Good Good Good
TCE Match to SilGaAs Poor Poor Fair Good Good
Cost ($/in.2) 0.95 3-6 10-50 3-5 1-5
11-368 CERAMIC PACKAGING

Polylmide Thin-Film
Signal Layers with
Ground Planes

~_-...;=~..:=~~==:- Signal
- - Ground

}
V222222Z2ZJ ; ; Po_.A.m""'

Ceramic Base Substrate


with Power & Ground
1/0 Pin Planes

Figure 9-60_ Hybrid Thin- and Thick-Film Package (42-Chip Module). (Cour-
tesy of NEC Corp.)

constant around 4.0. Any further improvements can only come from the
incorporation of porosity to achieve a dielectric constant around 1 or the
use of polymers with a dielectric constant between 2.0 and 4.0. The net
result would be composite ceramic substrates fabricated with dielectric
constants of less than 4.0 [96]. Because the lowest dielectric constant that
can be achieved is always with polymer materials and because most
polymer materials have thermal-expansion coefficients much in excess of
current and future device materials, hybrid packages of thin films of
polymer layers on top of ceramic substrates are expected for very-high-
performance single-chip and multichip applications. An example of such
a hybrid package, capable of supporting 42 chips [97], by NEC Corporation
is shown in Figure 9-60. It is expected that these kinds of packages are
the wave of the future, as they use ceramics as building blocks to control
thermal and mechanical properties as well as for power distribution, and

Conductor Layer (Copper FoiQ

Figure 9-61. Package Structure of Epoxy-Filled Cordierite Ceramic. (Courtesy


of Ibiden.)
9.B CHIP ATTACHMENT AND THERMAL DISSIPATION OF CERAMIC SUBSTRATES 11-369

Forming
Drilling
~
Firing
1
Through Hole Plating

!
Subtractive Process

1
Coupling

~
Glass Cloth laminate
Solder Resist Printing
~
Epoxy Resin Impregnating

~
!
Metal Finishing

!
Copper Foil laminate
~
Curing Epoxy Resin Cutting

!
Finishing

~
1
Inspection
Inspection

Figure 9-62. Process of Fabricating Epoxy-Filled Cordierite Ceramic Board.


(Courtesy of Ihiden.)

polymers for their low dielectric constants as thin-film interchip wiring.


In addition, polymer-filled ceramics will be developed to achieve the
lowest dielectric constant. An example of this technology involving the
fabrication of a porous cordierite-based substrate that is subsequently
filled with epoxy, laminated with copper foil, and mechanically drilled
to form plated-through via holes has already been reported [98], as shown
in Figure 9-61. Such a substrate, fabricated with the processes shown in
Figure 9-62 and the properties in Table 9-38, allows ceramics to be
fabricated into large-area, low-cost, and high-performance substrates using
board technologies. Such substrates could be developed with thermal-
expansion coefficients matching silicon or GaAs devices.

9.8 CHIP ATTACHMENT AND THERMAL DISSIPATION OF CERAMIC


SUBSTRATES
As discussed in Chapter 8, "Chip-to-Package Interconnections," all
three chip-bonding technologies are used with ceramic packages. With
solder connection (C4) technology, thermal conductivity of the ceramic
is not as important. The thermal limitations of the package, therefore, are
the same as for silicon, because the heat is removed from the back of the
silicon chip. However, wirebonding is still the predominant technology
11-370 CERAMIC PACKAGING

Table 9-38. Properties of Epoxy-Filled Ceramic in Comparison with the


Other Materials

Cordierite SiC 96% Epoxy-


Property (30 vol.%) (30 vol.%) Alumina Glass

Reinforce layer Nothing 0.1 mm 0.2mm 0.1 mm


glass cloth thickness x2 x2 x2
Thickness (mm) 1.0 1.0 1.2 1.0 1.0 1.0
Bending strength (kg/mm') 12.0 18.0 19.0 35.2 35.0 45.0
Toughness (kg cm'/cm) 3.5 11.4 29.2 4.6 54.6
Coefficient of thermal
expansion (lO-'/°C) 38 38 55 47 70 ISO
Thermal conductivity (W/m K) 1.3 0.9 0.6 2.8 20.0 0.2
Relative dielectric constant 4.0 9.3 4.8
Density (g/cml) 1.9 1.8 1.8 2.5 3.9 1.8
Machinability 2,000 2,000 100 impossible 12,000

Source: From Ref. 98.

in the industry, and it requires that the chip be backbonded to the substrate
and wirebonded. In high-power applications, this poses two problems:
(1) bonding the die to the ceramic with high-thermal conductive materials
and (2) heat removal requirement from the package, sometimes in excess
of the capabilities of the alumina or other ceramic used. Bonding of die
to the substrate is accomplished by either eutectic alloys (gold-silicon),
epoxies, or polyimides filled with precious metals or silver-filled glasses.
Eutectic die attachment utilizes gold or silver eutectic composition that
forms as a result of a reaction between the gold-silicon preform and the
silicon die, as shown in Figure 9-63. The adhesive attachments using
organic or inorganic materials filled with such thermally conductive mate-
rials as silver are usually cheaper and lower in stress than the eutectics,
allowing larger dies to be bonded at lower temperatures. These are further
discussed in Chapter 8, "Chip-to-Package Interconnections."
A number of ingenious designs have been reported [99] to handle
heat dissipation in applications involving die bonding and wirebonding
chips to low thermal-conductive ceramics. These designs, which can be
extended to plastic packages and printed-circuit boards, involve bonding
of thermal-expansion-matched alloys (Cu-W) or ceramics (AIN, SiC)
of very high thermal conductivity to alumina or other ceramics, using
copper-silver, gold-tin, or other brazes. This is schematically illustrated
in Figure 9-45a for alumina with copper-tungsten alloys and (b) for
glass-ceramics, with AINheat sinks. As shown in Table 7-15 of Chapter 7,
9.9 CERAMIC PACKAGE RELIABILITY 11-371

Ole Attllch ay.tem

Die Si

Au·SI Preform (1-3 o/w Si) +


+
Substrate

J So"", - "'"C, N, '""'"

Au·Si Eutectic Structure

Figure 9-63. Die Attach System

these heat sinks have thermal conductivities in excess of 200 W/m K,


providing several watts of capability from each chip. Removal of heat
from the chip that is C4-bonded is discussed in Chapter 8, "Chip-to-
Package Interconnections." Removal of heat with TAB-bond chips can
be accomplished in either of two ways. In those applications involving
die bonding, heat removal can be aided with brazed or adhesive-bonded
heat sinks, as discussed earlier with wirebonding. In addition, heat sinks
can be bonded to chip encapsulation, as discussed in the TAB section of
Chapter 8.

9.9 CERAMIC PACKAGE RELIABILITY


Ceramic packages are generally considered "hermetic." However,
there have been a number of designs that depended on organic sealants or
encapsulants to provide reliability. Reliability of these designs, therefore, is
meant to meet the requirements in the absence of hermeticity.
Because ceramic packages consist of ceramics and metals, they
are prone to two general types of degradation: those involving metallic
corrosion, resulting in open circuit lines, and those involving degradation
or cracking of ceramics in the presence of mechanical stresses aided by
water vapor. The latter results in stress-corrosion cracking of ceramic.
The metal degradation mechanisms are covered in great detail in Chapter
11-372 CERAMIC PACKAGING

5, "Package Reliability." Stress corrosion of ceramics and glasses is briefly


described here.
It is well established that although freshly drawn fibers of glasses
and ceramics have tensile strengths around 7xlO3 MPa (one million pounds
per square inch) and that their short-term tensile strengths are in the range
of 70-350 MPa (10,000 psi for glasses to about 50,000 psi for ceramics),
the strengths to which they can be designed are roughly one-third to one-
half of these numbers. The degradation mechanism involves growth of
Griffith cracks under the influence of residual mismatch stresses and other
applied stresses during processing, and are aided by such stress-corrosion
atmospheres as water, plating solutions, and so forth. It has been estimated
that the residual mismatch stress as a result of cooling from l600°C
alumina-molybdenum multilayer ceramic substrate can be as much as
150 MPa (25,000 psi) at the interface of molybdenum via. Cracks often
seen at this interface are directly attributable to the lowering of the strength
of the interface as a result of mismatch stress in the presence of water to
a level where stress exceeded the strength. More quantitative data on this
subject can be found in the literature [100]. Glasses suffer from this type
of degradation as well, and the design strengths can often be no more
than 35 MPa (5,000 psi) in tension.

9.10 FUTURE CERAMIC PACKAGING


The overall trend in ceramic packaging over the last three decades
is summarized in Figure 9-64, indicating how ceramics have evolved from
dual-in-line packages to single-chip pin-grid arrays in alumina systems
and to low-temperature glass ± ceramic systems in recent years. This
figure also shows the trend in mullite and aluminum nitride development.
It appears that the trend in future packages, including ceramics, is very
clear and is summarized in Table 9-39. The technology trend that can be
expected is discussed below for each of the major solutions.

Integrated Low-
Cost Ceramics
Based on
Large Area
and Fine Une/Via
Technaloges,
Including:
Capacitors, Resistors,
and Inductors

1965 1970 1975 1980 1985 1990 1995


Product Year

Figure 9-64. Ceramic Packaging Evolution


9.10 FUTURE CERAMIC PACKAGING 11-373

Table 9-39. Future Requirements of Ceramic Packages


Package Trend Potential Ceramic Substrate Solution

Thin and light Fewer layers, thinner layers lower-dielectric-constant materials


Ultralow cost Low-cost materials and processes
Large-area substrate fabrication
Compact size Chip scale packaging
Ball-grid array, flip-chip connections
Direct chip attach to ceramic board
Multiple functions Integrated ceramic substrate with capacitors, resistors, inductors, and
other passives.

9.10.1 Future Low-Dielectric-Constant Ceramic Materials


The search for low-dielectric-constant ceramic materials for ad-
vanced electronic packages has been reviewed in this section from the
viewpoint of technology and manufacturing. Recognizing that speed, size,
reliability, and cost are the principal driving forces for the advanced
packages, this section then presents a discussion and problems that the
presently available low-dielectric constant ceramics offer to meet the
above goals. It follows that a material with an ever-decreasing dielectric
constant is not enough to increase the packaging speed without a concurrent
reduction in circuit dimensions and metallization with the low-resistance
metals. In this regard, the lack of a fine-line and via printing technology
presents a barrier to the full implementation of the desired circuit feature
size that the packaging design rule would allow for low dielectric systems.
Similarly, benefits of low-dielectric-constant systems have not been incor-
porated into the cost equation, and cost remains a complex but highly
desirable challenge. This discussion is then followed by a broad overview
ofthe commercial packaging activities, with emphasis given to a currently
successful cordierite system with a dielectric constant of about 5. The
next breakthrough involves the development of a new, phase-transforma-
tion-free, nearly pure Si02 dielectric with a dielectric constant of 3.9-4.2,
the lowest limit achieved for a single-phase inorganic compound, offering
a formidable competition to the high end of a polyimide dielectric with
a dielectric constant of 3.5-4.0. Given that this new dielectric has also a
coefficient of thermal expansion matched to that of Si, the SiOrbased
ceramic multichip module (MCM-C) can be made as an alternative path
to the polyimide-based thin-film multichip module (MCM-D).

9.10.1.1 Borosilicate Glass + Silica


Of particular interest in this category is the dielectric made by react-
ing a mixture of borosilicate glass + nearly pure Si02 filler [101,102] into
a single-phase material, with a dielectric constant of 3.9-4.2 and a thermal
11-374 CERAMIC PACKAGING

coefficient of expansion matching that of Si (Table 9-40). Figure 9-65


compares the dielectric constant versus frequency of Si02 dielectric as
described earlier with the several commercially available glass-ceramics.
Dense, multilayered, cofired test vehicles (to characterize both physical
and electrical properties as partially indicated in Table 9-41. The unique
feature of this dielectric system is that the destructive martensitic phase
transformation that is so common to crystalline Si02 has been totally
prevented by incorporating a crystal growth inhibitor (similar to grain
growth inhibitor) into the reactive materials systems. As a result, a large
quantity of nearly pure Si02 can be accommodated in the dielectric compo-
sition. The borosilicate glass, to the extent of 30-40 vol. %, has been used
to provide a reactive liquid-phase sintering [103] of the components at
about 950°C in air. The coefficient of thermal expansion was matched to
that of Si by alloying with dopants having TCE substantially lower than
that of crystalline Si02• In other words, a careful design of materials and
process has resulted in the prevention of the phase transformation of Si02
to cristobalite. The net result is a substrate with a TCE match to Si while
allowing sintering to occur below l000°C by using a glass that reacts
with other components to form a single-phase, nearly pure Si02 as deter-
mined by X-ray diffraction (XRD). By extending the range of doping,
TCE can be matched closely to that of GaAs with some sacrifice in
dielectric constant from about 4 to about 5.8. Figure 9-66 shows the
relation between dielectric constant and TCE versus amount of dopant.
The above progress represents the end of a journey toward achieving
a cofireable lowest dielectric constant, single-phase inorganic material
with a signal propagation gain of about 50%, and places this dielectric
in direct competition with the high end of the polyimide with a dielectric
constant of 3.5. In fact, given that polyimides [104], and to a lesser extent
BCB [105], are characterized by such unattractive features (Table 9-42)
as large TCE, compared to that of Si, combined with poor strength, poor
thermal conductivity, lower adhension between metal and polymers, large
dissipation factor, significant moisture absorption, and limited number of
signal layers (due to bowing of the substrate and sequential processing),
makes Si02, an attractive alternative for large-scale MCM production,
using a well-developed, well-understood multilayer, cofired technology.
Investment in plant and equipment is also less for this technology. From
the comparison of two systems in Table 9-4 wherein data for the polyimidel
Cu system were obtained from AVP technology [106,107] and those for
Si0 2/Au were developed at Alcoa, it can hardly be denied that a ceramic
multichip module (MCM-C) made from a Si02-borosilicate system pres-
ents a formidable challenge to deposited multichip module (MCM-D)
made from polyimide dielectric, the only limitation being the lack of a
fine-line printing technology equivalent to that of photolithography, but
this is also being developed as discussed in large-area fine-line processing.
Table 9-40. L TCC Packaging System as Reported by Different Companies

Dielectric Dissipation TCE


Company Glass Matrix Ceramic Filler Metallization Constant Factor (ppmf'C)

Alcoa Borosilicate glass Si02, proprietary dopants Au, AglPd (internal) 3.9-4.2 <0.003 2.5-3.5
Asahi Glass BA-Ah03-Si02-B203 Ah0 3 Au 6.3 0.001 3.8-6.8
Coming Crystallizable glass Crystalline cordierite Au 5.2 3.4
DuPont Alumino borosilicate Ah0 3 Ag, Au 7.8 0.002 7.9
DuPont Crystallizable glass Crystalline cordierite Au 4.8 <0.003 4.5
Ferro Crystallizable glass Crystalline materials Ag, Au, AglPd 6.0 0.002 7.0
unknown
Fujitsu Borate glass Ah03, Si02 Cu 4.9 4.0
Hitachi Pb-alumino-borosilicate Ah03, caZr03 AglPd 9-12 0.001-0.003
w
IBM Crystallizable glass ~-spodumene Cu 5.0-6.5 0.002 2.0-8.3
.....
U1
IBM Crystallizable glass Crystalline cordierite Cu 5.2-5.7 0.001 2.5-5.5
Kyocera Pb-borosilicate glass Ah03' Si02 Au 7.9 7.9
Kyocera ZnO-borosilicate Ah03, Si02 Cu 5.0 4.4
Matsushita MgO-CaO-aiumino-borosilicate Ah0 3 Cu 7.1 0.0025
Matsushita Na20-CaO-aiumino-borosilicate Ah0 3 Cu 7.4 0.002 6.1
Murata BaO-B20rAI203-CaO-Si02 Cu 6.1 0.0007 8.0
Narumi CaO-aiumino-borosilicate Ah0 3 Ag, Au 7.7 0.0003 5.5
NEC Borosilicate glass Si02, cordierite, 13-49% Au 2.9-4.2 0.002 15-3.2
pores
NEC Pb-borosilicate Ah03, Si02 AglPd 7.8 0.003 7.9
NGK ZnO-cordierite glass Crystalline cordierite Ag,Au 5.2-5.5 0.001 1.5-1.3
Nihon Ah0 3 Ag,Au 7.8 7.0
Shoel BaZr(B03)2 Cu 7.0 0.001 7.7
Taiyo Yuden AI203-CaO-Si02-Zr02-MgO-B203 Cu 6.7 0.001 4.8
Taiyo Yuden AI20 3-Si02-Zr02-MgO AglPd, Ni 7.3 0.002 5.1
Tektronix MgO-CaO-silicate Ah0 3 Ag, Au 5.8 0.0016 4.6
Toshiba BaO-Sn02-Ti02-B 20 3 Au 7-13 0.0005-0.0008
Westinghouse CaO-B203 x Ah03 Si02 Au 4.6 0.001 9.6
11-376 CERAMIC PACKAGING

6 Glass-ceramic (Ferro)

..
..
c
.E
c 5 Glass-ceramic (IBM)
0
u
.
U
u
.;:

II
...
Glass-ceramic (DuPont)
... ... ... ~
1i
C
4
Borosilicate + Si02 (Alcoa)

102 1()3 104 lOS 106 107


Frequency

Figure 9-65. Dielectric Constant Versus Frequency for Various Low-Temper-


ature Ceramics

9.10.1.2 Porous Ceramics and Glasses


Moving away from the real world of packaging (i.e., the ability to
manufacture repeatedly with high yield), a considerable amount of research
work is ongoing to reduce the dielectric constant to values even lower
than 4, that of Si02• As Si02 represents the lowest limit of dielectric
constant for a single-phase inorganic material due primarily to its open
crystal structure, it is only appropriate that the next phase of advancement
in the reduction of dielectric constant must come from the creation of
void space in the microstructure. (Note that the ultimate dielectric is a
vacuum with dielectric constant of 1.) This is equivalent to introducing
a second phase in the microstructure with a dielectric constant of 1. Of
significant practical value in this regard is the dielectric mixing rules
[108] that can be faithfully applied to design the dielectric constant of
the composite. In a variety of studies [109-112], the second phase has
been chosen to be the pores, deliberately created in the microstructure or
hollow microspheres, again deliberately incorporated into the microstruc-
ture. It must be acknowledged, however, that the introduction of any open
space in the microstructure to reduce the dielectric constant will also
degrade both mechanical strength and thermal conductivity and make the
material prone to moisture absorption if the pores are open. Compounding
these problems are the enormous difficulty of repeatedly manufacturing
a body of porous ceramics into a controlled multilayer structure with
Table 9-41. Effects of Matched TCE, Low-k and High Conductivity of Au on Package Design and Cost; 4.5 in. (11.43 cm)
square module

Design Advantage Cost Advantage of


Parameters Al.ol Module SiO, Module of SiO, Module SiO, Module

Chip size 0.5 cm 1.0 cm Assumed Assumed


Chip area 0.25 cm' 1.0 cm' 4 times area in larger chip 4 times area in larger chip
Circuits per chip 104 4 x 104 4 times # of circuit in larger chip 4 times # of circuit in larger chip
# of chips per module 20x20=400 10 x 10 = 100 1I4th chips on the SiO, module 1I4th cost for chip mounting
Circuit per module 400 x 10' = 4 x 106 100 x 4 X 104 = 4 x Same for both modules Same for both modules
106
110 per chip; Rent's rule: 50 100 110 per chip twice in larger chip 110 per chip twice in larger chip
w
.....
..... I/O = 0.5 (# circuit)O.5
Total signal 110 per module 50 x 400 = 20,000 100 x 100 = 10,000 50% less 110 in SiO, module 50% less interconnect cost
# of signal planes per More Fewer Less routing of signal planes in Lower routing cost in SiO,
module SiO, module module
Off-chip delay (microstrip) 86 pslcm 55 pslcm 56% gain in speed in SiO, Higher selling price for SiO,
module module
Wiring density for same Zo I (relative) 1.5 1.5 times closer spacing in SiO, 50% denser lines, fewer planes,
module lower cost
Wiring length I (relative) 0.1-0.5 At least 50% less wire length At least 50% less gold
Wiring Rs for 25-llm-thick w= 8 mOlD Au = 1.6 mOlD 5 times more line conduction 115 less Au for equivalent
line with Au resistance
Dielectric thickness 1 (relative) 112 Thinner SiO, module Less materials consumption;
lower cost
Density 3.9 glcm3 2.2 glcm3 44% less weight for SiO, module Lower shipping cost for SiO,
module
11-378 CERAMIC PACKAGING

8 6
--+

-~
6
4 a .
1:
VI
C
0

-
Q. U
Q. 4 .~
w
~
j
2 CII

2 is

o~ __ __ ____ ____ __
~ ~ ~ ~ ~o

o 10 40 50
Vol % Dopant
Figure 9-66. TCE and Dielectric Constant as a Function of Dopant in SiOr
Borosilicate System

reasonable yield, reliability, and throughput. Given that a solution exists


for exercising such a high degree of control at the level of manufacturing,
convincing the customers to accept the product will be a daunting task.
The most advanced among these approaches that utilize porosity
as a second phase is the investigation conducted by NEe on several
compositions, as reported [111] in Table 9-40. The amount of porosity
introduced in the microstructure varied from 13% to 49%, corresponding
to a dielectric constant of 4.2 to 2.9. Predetermined amounts of polystyrene
microspheres were added to the glass-ceramic material systems, composed
of quartz glass, cordierite, and borosilicate glass, to control the porosity
which was generated by the decomposition of the spheres during heat
treatment. The merit of the concept was demonstrated by carrying the
fabrication process through the entire range of multilayer processing with
Au metallization. Different techniques were used by others to introduce
porosity as a second phase, although controlling the amount of porosity
and keeping the pores isolated were found to be difficult in many instances.
Porous Si02 films of about 40% porosity and ranging in thickness from
1 to 25 ~m with k -2.2 and tan B of <0.005 were demonstrated by a sol-
gel process [113,114]. Reactively sputtered Si02 films [113,115] were
prepared with a thickness of 5-10 ~m that exhibited k -3.4 and tan B of
-0.005. The porous Vycor glass [113] with a pore size of 40A exhibited
dielectric constant values of 2.6-3.3. High-strength Si02 microballoons
[116] with an average diameter of 65 ~m and wall thickness of 1-2 ~m
9.10 FUTURE CERAMIC PACKAGING 11-379

Table 9-42. Comparison of Thin-Film Versus Low-k Ceramic Packaging


Parameters

Parameters Thin Film Polymide/Cu Low k SiO,/ Au

Intrinsic
Dielectric constant 3.0-3.6 kHz 3.9-4.2 kHz
Dissipation factor 1-2% 0.1%
Volume resistivity 10 15 _10 16 cm >10 16 cm
Expansion coefficient (Si =3 pprn/0C) 20-70 pprn/°C 3 pprn/°C
Thermal conductivity 0.17 W/m °C 2.1 W/m °C
Moisture absorption 1-4% <0.1%
Density 1.42 g/cml 2.2 g/cml
Strength 5 ksi 15 ksi
Decomposition/melting temperature 4OO-450°C 17l0°C
Conductor resistivity 1.7 Q cm 2.2 Q cm
Extrinsic
Linewidth 10 11m 4-6 mil'
Line thickness 2-3 11m 10 11m
Printing technology Photolithography Screen printing
Line resistance 8Q/cm 1.6 Q/cm
Sheet resistance 8mQ/O 1.6 mQID
Speed I (relative) 90% of polyimide
Wiring density I (relative) 85% of polyimideb
Interconnect delay Low Low
Process steps >200 <50
Package support Requires a substrate Stand-alone package
Substrate Si, AI 20 J, glass-ceramic Optional: Si, AbO) metal
TCE match to Si substrate Poor Excellent
Aip-chip die attach Will produce stress No stress
Equipment/process/compliance cost High Low to moderate
Similarity to high-temperature cofire None High

'Process limited at present, design optimum 25 11m.


bSubject to enhancement of printing capability.

were incorporated into calcium aluminate cement to reduce the dielectric


constant from a value of about 7 to a value of about 4.7. Further studies
have reduced the dielectric constant to 2.7. In a slight variation of this
approach [109], quartz bubbles were incorporated into a composite of
alumina in glass, resulting in a dielectric constant of about 3.7. In a
separate study [117], hollow silica glass rnicrospheres were bonded together
with a lead glass frit that also acted as a flux. Samples with 55-68% porosity
were fabricated at a temperature of about 550°C. The dielectric constants
of these composites ranged from 3.9 to 3.3 and were nearly independent of
temperature. Hollow microspheres of mullite [118] with a diameter of 20-
11-380 CERAMIC PACKAGING

80 mm have recently been used as a second phase in a cordierite glass-


ceramic matrix, resulting in a dielectric constant of -2.5. Contrary to the
case of silica micro spheres, it is claimed that upon cooling from the sintering
temperature, such composites will exhibit higher strength due to mullite
microspheres being in tension and matrix glass-ceramic in compression.
In search of a still lower dielectric constant, polymer was introduced as yet
an additional phase to the pore phase. A Si02-polyimide composite with
controlled porosity was formed from a spray-dried mixture of colloidal silica
and silica fibers that were partially sintered and infiltrated with polyimide
[119]. A dielectric constant of -1.6 and a loss of -0.005 were reported for
this composite. Colloidal Si02 film [105] processed by a sol-gel technique
also gave a dielectric constant of 1.6. Judging from all these activities,
it appears that the days of a perfect material with a dielectric constant of
1 are not too far out in the future! What remains as a challenge in the low-
dielectric-constant system is to develop a perfect processing technology to
match the perfect material, leading to low-cost. An interesting consequence
of such a perfect accomplishment, however, is to put a severe strain on the
design rules. Figure 9-67 summarizes the ceramic materials development
discussed in this book as a function of dielectric constant.

9.10.2 Low-Cost Based on Large-Area Processing


Referring to Figure 9-68, in which cost of various substrates is
plotted against wiring density, it is clear that printed wiring board (PWB),

Ceramic Materials

11
E
.......
C
III 10
>:
ICI
Ii 9
C Mullite Glass/Cerami
c
0
;
8 Alumina + Glass
ICI Current Polymers Systems
ICII
Future Polymers
ICI 7 Glass-Ceramic
e
Q,

1
A.
6 Silica + Borosilicate
Silica + Polymer
5
2 3 4 5 6 7 8 9 10
Dielectric Constant
Figure 9-67. Interconnect Density versus Cost for Various MCMs
9.10 FUTURE CERAMIC PACKAGING 11-381

25.00

20.00

.~ 15.00 MCM-D
t:r (projected)
1/1
~
8 10.00
1/1

5.00 Next Generation


"MCM-D/L

200 400 600 800 1000


Interconnect Density (in./sq. in.)

Figure 9-68. Cost of ceramic, Thin-Film, and PWB Technologies as a Function


of Wiring Density

even though it is very inexpensive in a few layers and low wiring density,
becomes very expensive with high number of layers. This is borne out
by at least two examples in which Fujitsu and Hitachi have used a 40-
layer board, supposedly at a cost in excess of $100,000 per board [120].
At these densities, cofired alumina ceramic has been established to be
very much cheaper, perhaps more like $50/in.2 for a 30-layer substrate
or one-third of this at $15/in. 2 for a 1O-layer substrate. This realization
that wiring density in ceramic is cheaper than PWB has led to the develop-
ment of very extensive developments by the three mainframe companies:
IBM, Fujitsu, and Hitachi. NEC, Semens-Nixdorf, and Digital, however,
pursued thin-film multichip technology on either ceramic, metal, or printed
wiring board. In Figure 9-68, it is apparent that thin film is a very expensive
way of packaging these needs. Nevertheless, because the price of millions
of instructions per second (MIPS), is of the order of $25,000 or more,
the high cost of packaging has not been a problem in the 1970s and 1980s.
The challenge now is quite different. Almost all the electronic needs in
computer, telecommunications, consumer, and automotive are now based
on CMOS-both for hand-held consumer and desktop high-performance
systems-and the cost is number 1 driver of technology. Large-area pro-
cessing, similar to large-area processing of thin-film packaging discussed
in Chapter 12, Thin-Film Packaging," is expected to lower cost as indicated
in Figure 9-68.
There are clear indications that a number of larger substrates are
being developed throughout the community. IBM has demonstrated a
166-mm substrate including green sheet fabrication, screening, lamination,
and sintering. Such a substrate previously shown in Figure 9-2 had 107
layers of glass-ceramic and copper. This substrate was fabricated to zero
11-382 CERAMIC PACKAGING

shrinkage. NEC used a 22S-mm alumina substrate with power and ground
on which it deposited five to seven layers of thin film wiring using
polyimide and gold. This substrate was used in all its mainframe and
supercomputers. Similarly, Fujitsu developed and used a 24S-mm glass-
ceramic/copper substrate for its recent mainframes. It is perhaps conceiv-
able that a 300-mm size substrate from which tens of single-chip and
multichip substrates can be fabricated can be thought of to further lower
the cost of ceramic packages. It is abundantly clear that large ceramic
substrates developments in the future are based on need to lower the cost
per unit area of single-chip and multichip substrates as opposed to the
use of very large substrates for packaging ofleading-edge central processor
involving 100 or more ICs on one substrate.

9.10.3 Fine-Line and Via Technologies


Figure 9-69 illustrates the trend in ceramic packaging for linewidth,
line pitch, and via dimensions and compares these with thin-film technol-
ogy based on the polymer-metal structure as discussed in Chapter 12,
"Thin-Film Packaging." IBM routinely manufactures its ceramic sub-
strates with 7S-Jlm lines and 90-Jlm vias by major enhancements in mask
fabrication, thick-film paste, and screen-printing technologies.
An alternative to the via punching process is the photoforming
process developed by DuPont and others, as illustrated schematically in
Figure 9-27. The photoformed ceramic module (PCM) is formed by the
use of photosensitive dielectric and conductive films. These films are
subsequently exposed to ultraviolet light. Linewidths less than SO Jlm on
100-Jlm spacings are being developed with these processes.

-
400 40
E
3300
III
line pitch 30 f
c
o ~
·~200 20 ..9
C\I
E
~100
c
10
·c
Via hole
~ 0
~------~--------~--------~
1980 1990 2000
Year
Figure 9-69. Ceramic Packaging Trends
9.10 FUTURE CERAMIC PACKAGING 11-383

9.10.4 Ceramic Ball-Grid Array and Chip-Scale Packages


Consistent with the need for ultrasmall, thin, light, and low-cost
packages, ultimately no bigger than the chip itself, ceramic packaging is
being enhanced. Ceramic ball grid arrays are beginning developed at a
cost of approximately half of pin grid arrays and size roughly one-fourth.
Figure 9-70 illustrates the cross sections of two ceramic ball-grid
arrays, one from IBM [121], called miniball-grid array, and the other
from Hitachi, called microchip carrier (MCC) [59], that uses polymer-
copper thin film and a TaSi 2 resistor on top of the ceramic ball-grid array.
IBM's ceramic BGA is a 21-mm-size alumina-Mo substrate with 0.25-
mm balls placed on a 0.5-mm pitch, resulting in 1521 1I0s (39 x 39).
This BGA also shows, as does the Hitachi, the incorporation of decoupling
capacitors. The Hitachi BGA, on the other hand, is a 12 mm with 604
1I0s placed on centers. This subject matter is discussed in greater detail
in package-to-board interconnections, along with other nonceramic BGAs
and micro-BGAs.

(a) Thermal Poste

iiil?~ll~~~!~~r~~rDecoUPlIng
COp<lcitor
C4 Solder
Balls

Ceramic
Substrate

Mlnl·Boll Grid Array (8GA)


(0.25 MM Balls on 0.5 MM Pitch)

(b) Soldered

Solder bump
(minimum)
250 mm pitch)
seal
Thin tilm layers
(5 wiring layers,
1 r•• istor layer)

Sold.r bump
(450 jim pitch)

Figure 9-70. Two Types of Ball Grid Arrays: (a) IBM; (b) Hitachi
11-384 CERAMIC PACKAGING

9.10.5 Integrated Ceramic Packaging with Passives


Discrete ceramic components, as pointed out in Section 9.1, is a
large electronic ceramic market estimated at roughly 700 billion units a
year. Passive components and their miniaturization are important contribu-
tions to the miniaturization of most electronic products. Passives include
capacitors, resistors, inductors, transformers, and others. Japanese compa-
nies that include Murata, TDK, Kyocera (AVX in the United States), and
others lead this market. Some of the ceramic materials and their properties
in use today as discrete passives are indicated in Table 9-43. A recent JTEC
study [94] revealed the Japanese packaging strategy to be based on the min-
iaturization of voltage-controlled oscillators (VCO) and capacitors as indi-
cated in Table 9-44, leading to components packaging density that are about
20 currently to almost 50 by year 2000 (Fig. 9-71). It is clear that further
enhancements in package miniaturization come from the integration of pas-
sives into the ceramic substrate as part of the substrate process.
Potential candidates for integration into ceramic substrates include
the following:

• Decoupling capacitors
• Resistors
• Inductors

Table 9-43. High-Dielectric-Constant Ceramics

SM210 SZ310 SH800 SHIIO


ltemlMaterials MgTiO, caZrD, Ba Nd,TiO H Ba Nd,TiOI~

Dielectric constant E, 21 ± I 31 ± I 89 ± I 1I0±3


Q value (= lItan 0) >8000 >6000 >1800 >1200
(6.0 GHz) (4.0 GHz) (2.6 GHz) (2.3 GHz)
QxF >48000 >24000 >4680 >2760
T, (temperature coefficient of -25 ± 30 -30 ± 30 -35 ± 30 -70 ± 30
dielectric constant, ppmfOC)
Coefficient of linear thermal 9.2 8.5 10.1 10.1
expansion (40-400°C),
ppmfOC
Thermal conductivity 0.018 0.011 0.005 0.005
(callcmls°C)
Bulk density (g/cm') 3.7 4.5 5.7 5.7
Flexural strength (kg/cm') 2000 2000 1800 1800
Water absorption (%) 0 0 0 0
Note Cu thick film
is available
9.10 FUTURE CERAMIC PACKAGING 11-385

Table 9-44. Mass Production Japanese Strategy for Low-Cost Electronic


Products

Plastic
Technololn: Packages Discrete Coml!onents PWB SMT

Today QFP 1.0 x 0.5 mm size lOO-l1m lines O.4-mm pitch (QFP)
250-1Lffi vias
,j, ,j, ,j, 6-8 layers ,j,
,j,
Tomorrow TQFP 0.8 x 0.4 mm 50-11m lines 0.15-mm pitch (QFP)
(Year 2000) TAB built-in capacitors, 50-11m vias
resistors, inductors lOO-l1m pitch 0.5-mm (BOA)
8 layers

Source: From Ref. 94.

• Optoisolators
• Thin-film batteries
• Driver transistors
• Power transistors
• Power conditioners

9.10.5.1 Requirements for Wireless Communications [122]


Ceramic technologies have become the essence of modem portable
radio communicators. The use oflow-Ioss, high-dielectric-constant ceram-
50 .......................................................................... Camcorder
(2000)
40

Camcarder (1993)
Cellular phone (1993)
Palm top (1993)
10 Notebook (1993)
1980 1990 1995 2000
Year

Figure 9-71. Japanese Consumer Product Component Density Trend. (Cour-


tesy of Sony.)
11-386 CERAMIC PACKAGING

ics for radio frequency (RF) and infrared frequency (IF) filtering, voltage-
controlled oscillators, and capacitors has enabled the portable radio to be
miniaturized to a practical size while maintaining high performance. In
response to the continuous quest for a smaller and thinner profile, ceramic
devices are transforming from discrete components into a monolithic
lumped circuit component and finally into a totally integrated substrate.
Resonators, capacitors, inductors, and resistors are being integrated into
a high-density multilayer structure. Compared to the organic circuit board,
the multilayer ceramic substrate offers advantages of ease of processing
of these integrated passives for lower RF insertion loss, better interference
isolation, higher component density, and better temperature properties
(Fig. 9-72).
Generally speaking, every passive component behaves like a distrib-
uted circuit at RF frequency. The parasitic effect, in many cases, over-
whelms the component's original intent. As the frequency increases, the
parasitic resistance (skin effect) and the parasitic reactance become so
appreciable that an impedance peak can be produced at or near the applica-
tion frequency, thereby making the electrical properties uncontrollable.
This problem is more pronounced at frequencies greater than 500 MHz.
Recent ceramic technologies enable the component to be fabricated in
chip format and greatly reduce the unwanted lead wire's inductance,
distributed parasitic capacitance, and the temperature variation effect.
Using a high-k dielectric, one can reduce the coaxial resonator values
a thousand times. The coaxial resonator performs better than the equivalent
lumped circuit with discrete components. Although higher-dielectric-
constant ceramics can logically be used to reduce the size further, the
insertion loss and manufacturing problems appear insurmountable.
Higher-dielectric-constant materials have two disadvantages at high
frequency. They tend to have higher dielectric loss and temperature sensi-
tivity. To maintain proper impedance, the center conductor becomes too
narrow to be formed, due to processing limitations. Narrow center elec-
trodes also lead to Q degradation.

Ie Resistors

Figure 9-72. Concept of Integrated Ceramic Packaging


9.10 FUTURE CERAMIC PACKAGING 11-387

9.10.5.2 Integrated Ceramic Size Requirements


It is becoming obvious that only the Personal Computer Memory
Card Induced Association (PCMCIA) format or the so-called credit-card-
size format provides true portability. The thickness, therefore, needs to
be reduced to less than 2 mm. In order to reduce the thickness, the
surface-mount approach faces very tough challenges. Instead of mounting
components on a printed wiring board, the ceramic components themselves
should also function as the board and IC chip carrier (see Fig. 9-72).
All components are embedded into the substrate; recess cavities are
provided for mounting active IC chips using the dropped-in, flip-chip
technology. Because the resistance of the metal interconnects contribute
to RF loss, refractory metals cannot be used. Therefore, the ceramics need
to be cofireable with either copper or silver. The sintering temperature
should be less than 950°C while maintaining high dielectric and magnetic
Q factors. The structure will be like the one in Figure 9-72. All the circuits
and components are integrated into three-dimensional structure with mini-
mized thickness. For RF resonators and transmission lines, a virtual space
is created using new design technology to provide the required geometry
ratio for impedance matching.

9.10.5.3 Some Examples of Passive Integration


Murata [123] has developed LC filters, RF diode switches, hybrid
couplers, and delay lines using low-temperature cofireable ceramics with
copper electrodes. The low-temperature ceramic is used in the barium-
alumina-silica system and the buried capacitor developed to cofire with
this ceramic is a calcium zirconate plus glass. The properties of these
materials are shown in Table 9-45.

Table 9-45. Ceramic Integrated Capacitors by Murata


Material Name
CZG

Property
Basic fonnulation BAO, Ah03, Si02 caZr03, Glass
Dielectric constant (1 MHz) 6.1 25
Q
(l MHz) 1400 1700
(5 MHz) 300 700
Temperature coefficient of 80 ± 20 o ± 10
dielectric constant (ppmJ°C)
Resistivity (0 cm) 10- 12 10-14
Flexural strength (kglcm2) 1600 2150
Thennal Expansion 11.6 7.0
coefficient (x lQ-6{°C)
11-388 CERAMIC PACKAGING

9.11 REFERENCES
I. J.R.H. Black. "Technology and Market-Trends in Multilayer Ceramic Devices," in
Advances in Ceramics, ed. by J. Blum and W. R Cannon, pp. 3-11, American
Ceramics Society, Columbus, OH, 1986.
2. R R Tummala. "Multichip Packaging in IBM, Past, Present, and Future," Proceed-
ings of International Conference on Multichip Modules, pp. I-II, 1995.
3. E. M. Davis, W. E. Harding, R S. Schwartz, and J. J. Coming. "Solid Logic Technol-
ogy: Versatile High-Performance Microelectronics," IBM J. Res. Devel., 8(2):
pp. 102-114, 1964.
4. R G. Frieser. "A Review of Solder Glasses," Electrocomponent Sci Technol. 2(6):
pp. 163-199, 1975.
5. RH.F. Lloyd. "ASLT: An Extension of Hybrid-Miniaturization Techniques," IBM
J. Res. Develop., 11(4): pp. 86-92, 1967.
6. P. E. Fox and W. J. Nestork. "Design of Logic-Circuit Technology for IBM System
370 Models 145 and 155," IBM J. Res. Devel., 15(2): pp. 384-390, 1971.
7. P. J. Holmes and R G. Loasby. Handbook of Thick Film Technology, Electrochemical
Publications, Ayr, Scotland, 1976.
8. R W. Gedney. "Trends in Packaging Technology," 16th Annual Proceedings of
Reliability Physics, pp. 127-129, 1978.
9. H. Stetson. "Multilayer Ceramic Technology," Ceram. Civilization, 3: pp. 307-
322, 1987.
10. W. J. Gyuvk. "Methods of Manufacturing Multilayered Monolithic Ceramic Bodies,"
U.S. Patent No. 3,192,086, 1965.
II. H. Stetson. "Methods of Making Multilayer Circuits," U.S. Patent No. 3,189,978,
1965.
12. B. Schwartz and H. Stetson. "Ceramics and the Micromodule," RCA Eng., 5(4):
p. 56-58. 1960.
13. J. C. Williams. "Doctor Blade Process," in Treatise on Materials, Science and Tech-
nology, pp. 173-198, Academic Press, New York, 1976.
14. R E. Mistler, D. J. Shanefield, and R. B. Runk. Tape Casting of Ceramics in Ceramic
Processing Before Firing, ed. by G. Y. Onoda and L. L. Hench, pp. 411-448, John
Wiley and Sons, New York, 1978.
15. M. Sacks. "Milling of Agglomenated Powders," International Conference on Ultra-
structure Processing of Ceramics, Glasses and Composites, pp. 13-17, 1983.
16. R. A. Gardner and R. W. Nufer. "Properties of Multilayer-Ceramic Green Sheets,"
Solid State Technol. 17(5): pp. 38-43, 1974.
17. J. W. Blade. VLSI and the Substrate Connection, The Technological Trade-Offs of
the Package-Board Interface, D. Brown Associates, Warrington, PA, 1986.
18. R B. Lomeson. "High Technology Microcircuit Packaging," International Electronic
Packaging Society Proceedings, pp. 498-503, 1982.
19. C. Deisch, C. Gopal, and J. Stafford. "Designs of a High Performance DIP-Like Pin
Array for Logic Devices," 33rd Electronic Components Conference, 1983.
20. R Vernon. "Technology vs. Inflation: The USAFIUSNffi Low Cost Carrier," Semi-
conductor/Microelectronics Symposium, 1979.
21. H. Reiner. "VLSI Packaging," Elect. Commun., 58: pp. 440-446, 1984.
22. J. W. Balde. "Status and Prospects of Surface Mount Technology," Solid State
Technol., 29(6): pp. 99-103, 1986.
23. W. Engelmaier. "Effects of Power Cycling on Leadless Chip Carrier Mounting
Reliability and Technology," Proceedings of International Electronic Packaging
Society Conference, pp. 15-23, 1982.
9.11 REFERENCES 11-389

24. W. Engelmaier. Surface-Mount Technology, pp. 87-114, ISHM Press, Silver Spring,
MD,1984.
25. A. J. Blodgett and D. R. Barbour. "Thermal Conduction Module: A High-Perfor-
mance Multilayer Ceramic Package," IBM J. Res. Deve!., 26(1):pp. 30-36, 1982.
26. A. J. Blodgett. "Microelectronic Packaging," Scientific American, 249(1): pp. 86-
96, 1983.
27. 1. U. Knickerbocker, G. B. Leung, W. R. Miller, S. P. Young, S. A. Sands, and R. F.
Indyk. "IBM System Air-Cooled Alumina Thermal Conduction Module," IBM J.
Res. Deve!., 35(5): pp. 330-341, 1991.
28. R. R. Tummala and R. R. Shaw. "Glasses in Microelectronics," Adv. High-Techno!.
Ceram., 18: pp. 87-102,1987.
29. L. Anderson, R. W. Nufer, and F. G. Pugliese. "Ceramic Dielectrics," U.S. Patent
No. 4104345, 1978.
30. W. S. Young. "Multilayer Ceramic Technology," in Ceramic Materials for Electron-
ics, ed. by R. C. Buchanan, pp. 403-424, Marcel Dekker, Inc., New York, 1986.
31. G. A. Slack, R. A. Tanzilli, R. O. Pohl, andJ. W. Vanderande. "Thelntrinsic Thermal
Conductivity of AlN," J. Phys. Chem. Solids, 48: p. 641, 1987.
32. N. Kuramoto, H. Taniguchi, and I. Aso. "Sintering and Process of Translucent AIN
and Effect ofImpurities on Thermal Conductivity of AIN Ceramics," Yogyo-Kyokai-
Shi, 93: p. 41, 1985.
33. K. Komeya, H. Inoue, and A. Tsuge. "Effects of Various Addatives on the Sintering
of AlN," Yogo Kyokai-Shi, 89: p. 330, 1981.
34. N. Kuramoto, H. Taniguchi, and I. Aso. "Sintering and Properties of High Purity
AlN ceramics," Adv. Ceram., 26: p. 107, 1989.
35. A. K. Knudsen, T. A. Guiton, N. R. Nicholas, K. L. Mills, P. J. Bourns,
J. E.Volmering, 1. L. Board, D. R. Beaman, D. Susnitzky, P. S. deBranda, and E. Ruh.
"The Influence of Imperities on the Optical, Thermal and Electrical Properties of
Sintered AlN," Proceedings International Microelectronics Conference, p. 270, 1992.
36. N. !chinose, K. Katakura, and I. Hazeyama. "Electrical and Thermal Properties of AlN
Ceramic with Addatives," Proceedings International Microelectronics Conference,
p. 265, 1992.
37. V. A. Lavrenko and A. F. Alexeev. "Oxidation of Sintered AIN," Ceram. Int., 9:
p.80, 1983.
38. R. Chanchani. "Processability of Thin Film Fine Line Patterns on AIN Substrates for
HICs," IEEE Trans. Components Hybrids Manu! Techno!., CHMT-l1: p. 633,1990.
39. D. Suryanarayana, L. Matienzo, and D. Spencer. "Behavior of AIN Ceramic Surfaces
Under Hydrothermal Oxidation Conditions," Proceedings Electronic Components
Conference, p. 29, 1989.
40. S. Smith and B. Hazen, "Thin Film Protective Coating for Processing AlN Substrates,"
Proc. ISHM, p. 48, 1991.
41. N. Miyashiro, N. Iwase, A. Tsuje, F. Ueno, M. Nakahashi, and T. Takahashi.
"High Thermal Conductivity AlN Ceramic Substrates and Packages," IEEE Trans.
Components Hybrids Manu! Techno!., CHMT-13: p. 313, 1990.
42. T. Yamaguchi and M. Kageyama. "Oxidation Behavior of AIN in the Presence of
Oxide and Glass for Thick Film Applications," IEEE Trans. Components Hybrids
Manu! Technol., CHMT-12: p. 402, 1989.
43. A. Mohamed, A. Abdo, G. Scarlett, and F. Sherrima. "Effect of Lot Variations
on the Manufacturability of Thick and Thin Film AIN Substrates," Proc. ISHM,
p.7, 1990.
44. Y. Kurokawa, H. Hamaguchi, Y. Shimada, K. Utsumi, H. Takamizawa, T. Kamata,
and S. Noguchi. Proceedings Electronics Components Conference, p. 412, 1986.
11-390 CERAMIC PACKAGING

45. N.Iwase, T. Yasumoto,K.Iyogi, T. Kawakami, Y. Yokono,K. Itoh,andT. Takahashi.


"Thin Film and Pin Brazing Technologies for High Performance AIN Packages,"
Proc. IEPS, p. 191, 1990.
46. Y. H. Chiao, A. K. Knudsen, and I. F. Hu. "Interfacial Bonding in Brazed and Cofired
Aluminum Nitride," Proc. ISHM, p. 460, 1991.
47. Y. Kurihara, S. Takahashi, and K. Yamada. "Thin Film Resistors for AIN Ceramics,"
IEEE Trans. Components Hybrid Manu! Technol., CHMT-14: p. 199, 1991.
48. J. B. Blum, "Aluminum Nitride Substrates for Hybrid Microelectronics Applica-
tions," Hybrid Circuit Techno!., 1989.
49. A. Kaneko, K. Kuwabara, S. Kikuchi, and T. Kano. Fujitsu Scient. Tech. J. vol. 27:
p. 171-178, 1991.
50. S. Hamano, A. Wanatabe, and H. Takahashi. "High Pin Count AIN Cer-Quad
Package," Electron. Packaging Tech., 8: p. 69, 1992.
51. F. Kobayashi, W. Wanatabe, N. Yamamoto, A. Anzai, A. Takahashi, T. Daikoku, and
T. Fujita. "Hardware Technology for Hitachi M-880 Processor Group," Proceedings
Electronic Components Conference, p. 693, 1991.
52. N. Iwase, H. Sawaya, and T. Takahashi. "AIN Multilayer Package Technologies for
High Speed ECL and High Pin Count Bi-CMOS Devices," Proc ISHM, p. 269, 1992.
53. A. Shibuya, Y. Kurokawa, and Y. Shimada. "High Thermal Conductivity AIN PGA
Package," Proceedings International Microelectronics Conference, p. 285, 1992.
54. P. E. Garrou, and l. Turlik, "Materials of Construction: Substrate, Dielectric, Metalli-
zation," in Thin Film Multichip Modules, ed. by G. Messner, l. Turlik, J. Balde, and
P. Garrou, ISHM Press, Silver Spring, MD, 1992.
55. Y. Iseki, F. Shimizu, and T. Sudo. "Multichip Module Technology using AIN
Substrate for 2-Gbitls High Speed Switching Module."
56. L. E. Dolhert, J. W. Lau, J. H. Enloe, E. Y. Luh, A. L. Kovacs, and J. Stephan.
"Performance and Reliability of Metallized AIN for MCM Applications," Int. J.
Hybrid Microelectron., 14: p. 113, 1991.
57. D. E. Neil. Power Modules Utilizing Beryllium Oxide Direct Bonded to Copper,
Proceedings of International Microelectronic Symposium, pp. 491-495, 1975.
58. A. J. Rothman. "Beryllium Oxide," in Ceramics for Advanced Technologies, ed. by
J. E. Hove and W. C. Reiley, John Wiley and Sons, New York, 1965.
59. F. Kobayashi, Y. Watanbe, M. Yamamoto, A. Anzai, A. Takahashi, T. Daikoku,
and T. Fujita. "Hardware Technology for Hitachi M-880 Processor Group," Proceed-
ings of 42nd IEEE-ECTC, pp. 693-703, 1992.
60. Y. Shimade, K. Utsumi, M. Suzuki, H. Takamizowa, M. Nitta, and T. Watari. "Low
Firing Temperature Multilayer Glass-Ceramic Substrate," IEEE Trans. Components,
Hybrids Manu! Technol., CHMT-6(4): pp. 382-388, 1983.
61. N. Kamehana, K. Niwa, and K. Murakawa. "Packaging Material for High Speed Com-
puters," 33rd Electronic Components Conference Proceedings, pp. 388-392, 1983.
62. T. Noro and H. Tozaki. "Constituent Materials for via Conductor in Multilayer
Glass-Ceramic Board," Japanese Patent No. 6070799A, 1986.
63. R. R. Tummala, A. H. Kumar, and P. W. McMillan. "Glass Ceramic Structures and
Sintered Multilayer Substrates Thereof with Circuit Patterns of Gold, Silver, or
Copper," U.S. Patent No. 4,301,324, 1981.
64. P. W. McMillan. Glass-Ceramics, Academic Press, New York, 1974.
65. R. R. Tummala, L. W. Herron, and R. Master. "Method of Making Multilayered
Glass-Ceramic Structures Having an Internal Distribution of Copper-Based Conduc-
tors," U.S. Patent No. 4,234,367, 1980.
66. See for examples Japanese Patients 61-142759 and 60-245154.
67. S. Nichigaki, S. Yano, H. Kawake, J. Fukuta, T. Nonomura, and S. Hebishima.
"A New Low-Temperature Multilayer Ceramic Substrate with Gold (Top)-Silver
(Internal)-TilMo/Cu (Bottome) Conductor System," ISHM Proceedings, 1987.
9.11 REFERENCES 11-391

68. R. R. Tummala, J. U. Knickerbocker et al. "High Performance Glass Ceramic/Copper


Multilayer Substrate with Thin Film Redistribution," IBM J. Res. Devel., 36:
pp. 889-903, 1992.
69. R. R. Tummala. "Ceramic and Glass Ceramic Packaging in the 1990's," J. Am.
Ceram Soc., 74: p. 895, 1991.
70. A. H. Kumar, P. W. McMillan, and R. R. Tummala. "Glass-Ceramic Structures and
Sintered Multilayer Substrates Thereof with Circuit Patterns of Gold, Silver, or
Copper," U.S. Patent No. 4,301,324, 1981.
7l. S. H. Knickerbocker, A. Kumar, and L. W. Herron. "Cordierite Glass-Ceramics for
Multi-Layer Ceramic Packaging Applications," Ceram. Bull., in press.
72. E. A. Giess and S. H. Knickerbocker. "Viscosity of MgO-AI203-Si02-B203-P20j
Cordierite-Type Glasses," J. Mater. Sci. Lett., 4: p. 835, 1985.
73. E. A. Giess, J. P. Fletcher, and L. W. Herron. "Isothermal Sintering of Cordierite-
Type Glass Powders," J. Am. Ceram. Soc., 67: p. 549, 1984.
74. H. E. Exner and E. A. Giess. "Anisotropic Shrinkage of Cordierite-Type Glass Pow-
der Cylindrical Compacts," J. Mater. Res., 3: p. 122, 1988.
75. K. Watanabe, W. A. Giess, and M. W. Shafer. "The Crystallization Mechanism of
High Cordierite Glass," J. Mater. Sci, 79: p. 508, 1985.
76. K. Watanabe and E. A. Giess. "Coalescence and Crystallization in Powdered High-
Cordierite (2MgO-2AI20 3-5Si0 2) Glass," J. Am. Ceram. Sci., 68: p. C102, 1985.
77. T. C. Patton. "Film Applicators," in Paint and Pigment Dispersion, p. 581, John
Wiley and Sons, New York, 1979.
78. R. W. Nuffer, L. E. Anderson, and F. G. Pugliese. "Ceramic Dielectrics," U.S. Patent
No. 4,387,131, 1983.
79. R. N. Master, L. W. Herron, and R. R. Tummala. "Cosintering Process for Glass-
Ceramic/Copper Multilayer Ceramic Substrate," IEEE Trans. Components Hybrids
Manu! Technol. CHMT-14: p. 780, 1991.
80. L. W. Herron, R. N. Master, and R. W. Nufer. "Methods of Controlling the Sintering
of Metal Particles," U.S. Patent No. 4,671,928, 1987.
81. L. W. Herron, R. N. Master, and R. W. Nufer. "Methods of Controlling the Sintering
of Metal Particles," U.S. Patent No. 4,776,978, 1987.
82. J. R. Piazza and T. G. Steele. "Positional Deviation of Preformed Holes in Substrates,"
Bull. Am. Ceram. Soc., 51: p. 516, 1972.
83. R. Iwamura, H. Murakami, K. Ichimoto, and M. Takasaki. "Punching of Holes at
High Density in Alumina Greensheet," Proceedings of the American Ceramic Society
Meeting, 1986.
84. F. M. Fowkes and M. A. Mostafa. "Acid-Base Interactions in Polyimide Adsorption,"
I&EC Product Res. Develop., 17: p. 3, 1978.
85. M. Marmo, H. Jinnai, M. A. Mostafa, F. M. Fowkes, and J. A. Manson. "Acid-Base
Interactions in Filler-Matrix Systems," I&EC Product Res. Devel., 15: p. 206, 1978.
86. T. B. Reed. Free Energy of Formation of Binary Compounds: An Atlas of Charts
for High-Temperature Chemical Calculations, MIT Press, Cambridge, MA, 1971.
87. W. Herron, R. Master, and R. R. Tummala. "Method of Making Multilayered Glass-
Ceramic Structures Having an Internal Distribution of Copper-Based Conductors,"
U.S. Patent No. 4,234,367, 1980.
88. Y. Shimada, Y. Kobayashi, K. Kata, M. Kurano, and H. Takamizawa. "Large Scale
Multilayer Glass-Ceramic Substrate for Supercomputer," IEEE Trans. Components
Hybrids Manu! Technol., CHMT 13: p. 751, 1990.
89. J. H. Indge. "Flat Precision Machining Ceramic Materials," Advanced Ceramics
'88, 1988.
90. M. Bennett. "Grinding Ceramics at IBM," Inter-Society Symposiumfor the Machining
of Ceramic Materials (A. Cer. S., ASME, A. Abrasive Soc.), 1988.
11-392 CERAMIC PACKAGING

91. P. L. Flaitz and M. Neisser. "Optimizing a Polish Operation," Proceeding of 91st


Annual Meeting of the American Ceramics Society, p. 92, 1989.
93. Y. Shimada, Y. Shiozawa, M. Suzuki, H. Takamizawa, and Y. Yamashita. "Low
Dielectric Constant Multilayer Glass-Ceramic Substrate Ag-Pd Wiring for VSLI
Package," 36th Electronic Components Conference Proceedings, pp. 395-405, 1984.
94. R. R. Tummala and M. Pecht. "Japan's Electronic Packaging Technologies," Japa-
nese Technology Evaluation Committee (JTEC) Panel Report, pp. 59-95, 1995.
95. A. H. Kumar. "LTCC Multichip Modules," Proceedings of International Multichip
Module Conference, 1995.
96. L. E. Gross and T. R. Gurnaraj. "Ultra-Low Dielectric Permittivity Ceramics and
Composites for Packaging Applications," Materials Research Society Symp. Proc.,
72: pp. 53-65, 1986.
97. T. Watari and H. Murano. "Packaging Technology for the NEC SX Supercomputer,"
37th Electronic Components Conference Proceedings, pp. 192-205, 1985.
98. Y. Iwata, S. Saito, Y. Satoh, and F. Okamura. "Development of Ceramic Composite
Porous Ceramic and Resin Composite with Copper Foil," International Microelec-
tronic Conference in Tokyo, pp. 65-70, 1986.
99. M. Terasawa, S. Minami, and J. Rubin. "A Comparison of Thin Film, Thick Film,
and Co-Fired High Density Ceramic Multilayer," Int. Hybrid Microelectron., 6:
pp. I-II, 1983.
100. J. N. Humenik and J. E. Ritter. "Stress Corrosion of Alumina Substrate," J. Mater.
Sci., 14(5): pp. 626-632, 1979.
101. J. H. Jean and T. K. Gupta. "Design of Low Dielectric Materials for Microelectronic
Packaging Applications," Am. Ceram. Soc. Bull., 7(9): p. 1326, 1992.
102. J. H. Jean and T. K. Gupta. U.S. Patent Nos. 5,071,793 and 5,079,194, 1992.
103. J. H. Jean and T. K. Gupta. "Isothermal and Nonisothermal Sintering Kinetics of
Glass-Filled Ceramics," J. Mat. Res., 7(12): pp. 3342-3347, 1992.
104. G. M. Adema, M. J. Berry, and I. Turlik. "Dielectric Materials for Use in Thin Film
Multichip Module," Electron. Packaging Prodution, vol. 11, pp. 72-76, 1992.
105. E. W. Rutter, Jr., E. S. Moyer, R. F. Harris, D. C. Frye, V. L. St. Jeor, and F. L.
Oaks. "A Photodefinable Benzocyclobutene Resin for Thin Film Microelectronic
Applications," ICMCM Proceedings, pp. 394-400, 1992.
106. C. J. Bartlett. "Advanced Packaging for VLSI," Solid-State Technol., pp. 119-123,
June 1986.
107. C. J. Bartlett, J. M. Segelken, and N. A. Teneketges. "Multichip Packaging Design
for VLSI-based systems," IEEE Trans. Components Hybrids Manu! Technol.,
CHMT-12: p. 518, 1987.
108. W. D. Kingery, H. K. Bowen, and D. R. Uhlmann. in Introduction to Ceramics,
chaps. 12, 15, and 18, John Wiley and Sons, New York, 1976.
109. D. W. Kellerman. "The Development and Characterization of a Low Dielectric
Constant Thick Film Material," Proceedings of 37th Electronic Components Confer-
ence, pp. 316-327, 1987.
110. S. J. Stein et al. "Controlled Porosity Dielectrics and Etchable Conductors for High
Density Packages," ISHM 1990 Proceedings, pp. 725-732, 1990.
Ill. K. Kata, A. Sasaki, Y. Shimada, and K. Utsumi. "New Fabrication Technology of
Low Dielectric Permittivity Multilayer Ceramic Substrate," ISHM 1990 Proceedings,
pp. 308-315, 1990.
112. J. K. Yamamoto, et al. "Dielectric Properties of Microporous Glass in the Microwave
Region," J. Am Ceram. Soc. 72(6): pp. 916-21, 1989.
113. L. E. Cross and T. R. Gururaja. "Ultralow Dielectric Permittivity Ceramics and
Composites for Packaging Applications," Mater. Res. Soc. Symp. Proc., 72:
pp. 53-65, 1986.
9.11 REFERENCES 11-393

114. W. A. Yarbrough, T. R. Gururaja, and L. E. Cross. "Materials for IC Packaging Very


Low Permittivity via Colloidal Sol-Gel Processing," Am. Ceram. Soc. Bull., 66(4):
pp. 692-698, 1987.
115. A. Das. "Sputter Deposited Silica Films as Substrate for Microelectronic Packaging
Applications," Ph.D. thesis, Pennsylvania State University, 1988.
116. P. Silva, L. E. Cross, T. R. Gururaja, and B. E. Scheetz. "Relative Dielectric Permittiv-
ity of Calcium Aluminate Cement-Glass Microsphere Composites," Mater. Lett.
4(11-12): pp. 475-480,1986.
117. M. 1. Leap. "The Processing and Electrical Properties of Hollow Microsphere Com-
posites," M. S. thesis, Pennsylvania State University, 1989.
118. D. 1. Vemetti and D. L. Wilcox, Sr., "Synthesis of Hollow Mullite Microspheres: A
New Approach for Low Dielectric Constant Inorganic Packaging Materials," ICMCM
Proceedings, pp. 300-307, 1992.
119. M. 1. Duggan. "Fabrication and Properties of Low Permittivity Silica Fiber Reinforced
Silica Substrate," M. S. thesis, Pennsylvania State University, 1991.
120. G. Messner. Private communication.
121. R. N. Master, M. Cole, and G. Martin. "Ceramic Column Grid Array for Flip Chip
Applications," Proceedings of IEEE-ECTC, pp. 925-929, 1995.
122. Wei-Yean Howng. "Ceramic Technologies for Portable Radio Applications." Pri-
vate communication.
10
PLASTIC PACKAGING

MICHAEL G. PECHT-University of Maryland


LUU T. NGUYEN-National Semiconductor

10.1 INTRODUCTION

This chapter presents the information needed to design, manufacture


and test microelectronic devices which are encapsulated in "plastic" mold-
ing compound. The chapter begins with an historical overview, then
discusses advantages and disadvantages of plastic versus ceramic packag-
ing. The next three sections present the materials, manufacturing processes
for molding and the handling methods of the finished product. This is
followed by a section on test and reliability issues. The chapter concludes
with a view of the future of plastic encapsulated microcircuits. The inter-
ested reader is referred to the books titled "Plastic Encapsulated Microcir-
cuits" [1], and "Integrated Circuit, Hybrid and Multichip Module Design
Guidelines" [2] for more detailed information.
10.1 INTRODUCTION 11-395

Electronic devices have been packaged in a variety of ways. One


of the first methods was a Kovar (high nickel alloy) preformed package.
The device was bonded to the bottom, and the top was later secured.
Ceramic packages, similar in construction to the Kovar casing, later ap-
peared as a cheaper alternative. The first evidence of plastic encapsulation
was seen in the early 1950s, using compression molding of phenolics.
The phenolics material was compressed around electrical devices, leaving
the electrical connectors protruding. The problem with this method was
that the delicate connections were often severed due to the high pressure.
By the early 1960s, plastic encapsulation emerged as an inexpensive,
simple alternative to ceramics and metal encasing, and during the 1970s,
virtually all high-volume Integrated Circuits (ICs) were encapsulated in
plastic. For example, one cost comparison of a 14-pin package in 1979
for quantities of 500,000 packages and a gold price of $200 per ounce
sets the cost per plastic package at $0.063 and the cost of a ceramic dual-
in-line package (DIP) at $0.82.
Although most early devices were compression molded, potting soon
emerged as a suitable alternative. It involved positioning of the electrical
circuit in a container and pouring the molten encapsulant into the cavity.
Figure 10-1 shows a typical transistor encapsulated using the "can and
header" method. Transfer molding then gained worldwide acceptance as
a economical method best suited for mass production. In transfer molding,
the chip is loaded into a mold cavity, constrained, and the encapsulant is
transferred from a reservoir into the cavity under pressure. The encapsu-
lant, typically a thermosetting polymer, cross-links and cures in the cavity
to form the final assembly.
Figure 10-2 shows one of the first molds used for transfer molding
of a full-wave rectifier. In this method, the thin-wirebonds used to form
connections from the chip to the outside world were frequently damaged
due to the high velocities of the encapsulant. A novel approach to this

Figure 10-1. "Can and Header" Transistor. (LanzI, Patent # 3,235,937.)


11-396 PLASTIC PACKAGING

Second.". Nnner
Die

Figure 10-2. Transfer Molding of a Full-wave Rectifier. (Doyle, Patent


# 3,367,025.)

problem was to use a bottom-sided gated process in which the molding


compound entered the cavity from the opposite side and remote from the
bond wires (Fig. 10-3), thereby reducing the chances of breaking the
delicate wires.
Although epoxy novolac was the first material used for plastic encap-
sulation, phenolics and silicones were the dominant plastics of the 1960s.
At that time, plastic encapsulated microelectronics (PEMs) were plagued
by numerous reliability problems owing largely to the poor quality of the
encapsulation system. Over the decade, numerous formulations of epoxies

Device Runner

Figure 10-3. Transfer Molding with Bottom-side Gating. (Birchler et al., Pa-
tent # 4,043,027)
10.1 INTRODUCTION 11-397

were developed which had lower curing shrinkages and contamination


levels and thus took over as the main encapsulating material in the early
1970s. Although silicones are still sometimes used, a typical encapsulant
used today is a complex mixture of cross-linkers, accelerators, flame
retardants, fillers, coupling agents, mold release agents, and flexibilizers
in an epoxy resin matrix.
Today, a typical plastic package or PEM (Fig. 10-4) consists of a
silicon chip, a metal support or leadframe, wires that electrically attach
the chip circuits to a leadframe which incorporates external electrical
connections, and a plastic epoxy-encapsulating material to protect the
chip and the wire interconnects. The leadframe is made of a copper alloy,
alloy 42 (42 Nil58 Fe) or alloy 50 (50 Nil50 Fe), and is plated with gold
and silver or palladium, either completely or in selected areas over nickel
or nickel/cobalt. The silicon chip is usually mounted to the leadframe
with an organic conductive formulation of epoxy. Wires, generally of
gold but also of aluminum or copper, are bonded to the aluminum bonding
pads on the chips and to the fingers of the leadframe. The assembly is
then typically transfer molded in epoxy. Following the molding operations,
the external leads are plated with a lead-tin alloy, cut away from the
strip, and formed into desired shapes.
Plastic packages can be premolded or postmolded. In the former, a
plastic base is molded; the chip is then placed on it and connected to an
110 fan-out pattern with wire. The die and wirebonds are usually protected
by an epoxy-attached lead which forms a cavity. Premolded packages are

Die Passivation

Lead:fra.me Die attach


Figure 10-4. Typical Plastic Package Construction (From Ref. 1.)
11-398 PLASTIC PACKAGING

most often used for high-pin-count devices or pin-grid arrays that are not
amenable to flat leadframes and simple fan-out patterns.
In the postmolded type, the die is attached to a leadframe, which is
then loaded into a multicavity molding tool and encapsulated in a thermoset
molding compound via the transfer-molding process. Postmolded pack-
ages are less expensive than premolded ones because there are fewer parts
and assembly steps. In the 1990s, about 90% of plastic packages were
made using postmolding techniques.
Plastic-encapsulated microelectronics are made in either surface-
mount or through-hole configurations. The common families of surface-
mounted PEMs are the small-outline package (SOP), the plastic leaded
chip carrier (PLeC), and the plastic quad flat pack (PQFP). The common
families of through-hole-mounted PEMs are the plastic dual-in-line
(PDIP), the single-in-line (SIP), and the plastic pin-grid array (PPGA).
These are discussed in overview chapter Sections 1-5 and illustrated in
Fig. 1-28.

10.1.1 Through-Hole-Mounted Devices


The common families of through-hole-mounted PEMs are the plastic
dual-in-line (PDIP), the single-in-line (SIP), and the plastic pin-grid array
(PPGA). The plastic dual-in-line package, the most commonly used PEM
in the 1980s, has a rectangular plastic body with two rows of leads, often
on 2.54-mm (IOO-mil) centers on the long sides. The dual-in-line package
is conducive to high-volume manufacturing at low cost. Their in-line
structure of the leads allows the packages to be shipped in plastic tubes
end to end, without contacting the leads. The leads are bent up at a small
angle from the package body for through-hole insertion mounting. The
package layout allows automated board mounting and offers highly reli-
able joints.
Single-in-line packages are rectangular in shape with leads on one
of the long sides. The leads are typically on 1.27-mm (50-mil) centers at
the plastic body interface, formed into two staggered rows spaced 2.54
mm (100 mils) and 1.225 mm (50 mils) apart. These packages offer a
high profile but small footprint on the board, maintaining a 2.54-mm
(100-mil) hole-mounting standard. They offer all the advantages of dual-
in-line packages in ease and low manufacturing cost. Plastic pin-grid
arrays are packages with leads located in a grid array under a plastic
body. The pins are often located on 2.54-mm (lOO-mil) centers, although
1.715-mm (70-mil) centers are also available. The pin-grid array package
takes advantage of mature through-hole board mounting technology for
high-pin-count devices. Plastic pin-grid arrays are thus the highest density
through-hole packages and offer the highest available pin counts for PEMs.
The grid-array layout of the interconnects allows impedance matching of
the package to the chip. Using multilayer printed-board technology, ground
10.1 INTRODUCTION 11-399

and power planes can be incorporated into the package for high-perfor-
mance devices. Heat sinks can also be incorporated, if needed. Figure
10-5 shows some typical through-hole plastic configurations.

10.1.2 Surface-Mounted Devices

Common families of surface-mounted PEMs are the small-outline


packages (SOP), the plastic leaded chip carrier (PLCC), and the plastic
quad fiat pack (PQFP). These packages are designed for low-profile mount-
ing on printed wiring boards.
Small-outline packages are similar to dual-in-line packages, with
leads on two sides of the package body, but the leads are typically formed
in a gull-wing configuration.
A variation on the small-outline package is the small-outline J-Ieaded
(SOJ) package, in which the lead is formed in a J-bend configuration and
folded under the body. The advantage is an even smaller footprint than
that of a gull wing, although solder-joint inspection becomes more difficult.
Plastic leaded chip carrier packages are molded PEMs with leads
on all four sides of the plastic body. Typically, from 18 to 124 leads are
present on 1.27-mm (50-mil) centers and formed in the J-bend configura-
tion. Because the leads are on all four sides of the plastic body, this
package style offers the advantages of dense mounting on the board.
Electrically, the interconnect leads of plastic leaded chip carriers are
shorter on average and more consistent in length than those of equivalent
dual-in-line packages, resulting in a better match of the impedances of
the package leads.
Plastic quad fiat pack packages (PQFPs) are square or rectangular
plastic packages, typically with 40-304 leads distributed on all four sides.
Figure 10.5 shows typical surface-mount plastic package configurations.
A new package called Very Small Peripheral Array Package can be
considered a type of plastic package because the pins in three rows of
periphery are inserted into the preformed plastic as illustrated in Chapter
one Fig. 1-54. This simple and elegant package offers the "best of both
worlds" in that it is superior to QFP in size and to BGA in pin-to-board
reliability. The cost per pin is about the same as QFP. Another major
improvement worth noting is its outstanding heat transfer capability that
is achieved as a result of the copper heat sink.

10.1.3 Plastic Versus Ceramic

Plastic-encapsulated microelectronics offer many advantages over


hermetic packages in the areas of size, weight, performance, cost, reliabil-
ity, and availability. It is not surprising that plastics account for more
than 97% of the worldwide commercial chip-encapsulation market [3].
Mold" Plastic Packll"

~_ . . . f"..
..... ~ .... ry.. ~
P, • ....
Co .. ", ~ekag. Ty.,. Pitch D_le•• U••d '0' I_~"_ol ~=
Count
... ......
__ IItide..
G.,• ...".., _
1N00"_',....IIMO .. \ .... ,. \ \
~ ,. 1.' ....25.1( DRAM

IOT·n.sen
Squ.r. Oued "'IM JllIII:
~ "Lee
0 '''''''I

JIIoo."f.. ".,.~ ... D.rt... WI_'


"""CorttoolLOOIe

~JllM;k ••• 2<>28 1.' One' Fo"r Me.abft DRAM


I'I..,;:tllgular0.....,.., JllIII:
Outllna
~laad
,..
-,~
(SO.1)
9-.,
0 .... u"..... LOIII' ... DftAIII" ,IiIAIII,.
0 III'c..,pI'Oc ....... IIKl"'•• HlOIII,
OIM"""'" 111-30 10&5 1 0,,",""
.~ 20. ... ... Lo.lc. PROMS, Mlc,oll'o<:",o,,_
Gata Array a, Standard Call V.,.,!IIn"~.~k.
~ (YSOP)

Pla.tlc La.d" Chip Ca"'er


D""" . . . . ."""... (PleC,
0 ~
IJlU ... JII.c..... I ]2·100 I OS 0. ............,
I
".z~
Il'"

>
0.84 Cat. Art., a Standard Cell LoQlc v.,., Srn-'l Quid
U ....... Loe'e•• tItA ... o ... o::...,
~.- ....~,.....,.. .....
~ {Y"'"
""'' 111:

,......... (110",
Pla.tle Ouad I'lat Pac
.~ fPOFPJ

Figure 10-5. Plastic Package Configurations


10.1 INTRODUCTION 11-401

10.1.3.1 Size and Weight


Commercial PEMs generally weigh about half as much as ceramic
packages. For example, a 14-lead plastic dual-in-line package (DIP)
weighs about 1 g, versus 2 g for a 14-lead ceramic DIP. Although there
is little difference in size between plastic and ceramic DIPs, smaller
configurations, such as small-outline packages (SOPs), and thinner con-
figurations, such as thin small-outline packages (TSOPs), are available
only in plastic. The use of SOPs and TSOPs also enables better-performing
circuit boards due to higher packing density and consequent reduced
component propagation delays. A smaller form factor naturally implies
higher board density and more functionality packed into the same precious
board real estate. Similarly, a lighter package results in a smaller overall
payload for the same board functionality, a concern of critical importance
for the avionics industry. The relative size, weight and mounting height
of various plastic packages were discussed and illustrated in Chapter 7
("Microelectronics Packaging-An Overview") Figs. 7-23, 7-24, 7-25
and 7-26.

10.1.3.2 Performance

Plastics have better dielectric properties than ceramics. Although


plastic packages, such as DIPs, are not the most effective in propagating
high-frequency signals [4], plastic quad flat pack (PQFP), pin-grid arrays,
and ball-grid arrays are favored for minimizing propagation delays. For
the typical commercial applications, in which frequencies do not exceed
2-3 GHz, plastic packages perform better than their ceramic counterparts
in the same form factor. Two main features account for this characteristic:
the lower dielectric constant of the epoxy compared with that of the
standard cofired ceramic, and the smaller lead inductance of the copper
leadframes next to the Kovar leads.
When the application calls for much higher frequencies (up to 20
GHz) however, better and more predictable performance is obtained with
ceramic packages. The dielectric constant of typical ceramics stays the
same over a wider frequency range than that of plastic molding compounds.
This is because the dielectric constant is influenced by the moisture content
in the package.

10.1.3.3 Cost
The cost of a complete plastic package is driven by several factors,
such as die, encapsulation, production volume, size, cost of assembly and
yield, screening, pre-burn-in and its yield, bum-in final testing and its
yield, and the mandatory qualification tests. Table 10-1 presents the relative
cost for various microcircuits packaging options. Because more than 97%
11-402 PLASTIC PACKAGING

Table 10-1. Relative Cost of Packaging Options (From Ref. 1.)

Package Type Relative Cost

Small-outline integrated circuit (SOIC) 0.9


Plastic dual-in-line package (PDIP) 1.0
Plastic leadless chip carrier (PLCC) 1.2
Ceramic dual-in-line package (CERDIP) 4.0
Plastic pin-grid array (PPGA) 10.0
Plastic quad fiat pack (PQFP) 50.0
Pin-grid array (PGA) 130.0
Leaded chip carrier (LCC) 150.0

of the integrated-circuit market is plastic packaged, the cost has been


lowered by high demand, competition, and high-quality automated volume
manufacturing.
Hermetic packages usually have a higher material cost and are fabri-
cated with more labor-intensive manual processes. For example, Thomson-
CSF reported a 45% purchase cost reduction for each of 12 printed wiring
boards in a manpack transceiver application implemented with plastic
rather than ceramic components [5].
Another factor that drives up the cost of hermetically packaged inte-
grated circuits is the rigorous testing and screening required for the low-
volume hermetic parts [6]. When both types were screened to customer
requirements, ELDEC [7] estimated that purchased components for plastic
encapSUlation of integrated circuits cost 12% less than their hermetic coun-
terparts, primarily due to the economics of high-volume production.
The cost benefits of PEMs decrease with higher integration levels
and pin counts, because of the high price of the die in relation to the total
cost of the packaged device. Although these cost benefits may not be
realized for complex monolithic very-large-scale integrated circuits, cost
advantages may accrue for complex package styles, such as multichip
modules, because of the ease of assembly. Indeed, the trend toward future
multichip modules in laminates (MCM-L) packaged in form factors such
as PQFPs or ball-grid arrays (BGAs) will make plastic packages even
more popular. In the MCM-L, several dies and passive components can
be combined on a printed-circuit board substrate and integrated into a
leadframe to enhance part functionality. This approach shortens the time
to market (by using readily available dies to eliminate the need for die
integration), increases the process yield (there are no large dies and no
mixture of such different technologies as Complimentary Metal Oxide
Semiconductor and bipolar), and lowers package cost (compared with a
10.1 INTRODUCTION 11-403

ceramic equivalent). MCM-L products have already been produced for


some time and the cost savings achieved will impact the proliferation of
plastic packages [8].

10.1.3.4 Reliability
The reliability of plastic-encapsulated microelectronics, which was
traditionally a major concern, has increased tremendously since the 1970s,
due largely to improved encapsulating materials, die passivation, and
manufacturing processes. In particular, modem encapsulating materials
now have lower ionic impurities, better adhesion to other encapsulation
materials, a higher glass-transition temperature, higher thermal conductiv-
ity, and better matching of the coefficients of thermal expansion to the lead-
frame.
Advances in passivation include better die adhesion, fewer pinholes
or cracks, low ionic impurity, low moisture absorption, and thermal proper-
ties well matched to the substrate. The failure rate of plastic packages
has decreased from about 100 per million device-hours in 1978 to about
0.05 per million device-hours in 1990 [9]. Figure 10-6 summarizes pub-
lished improvements in reliability of different PEM technologies since
1976 [7].
Figure 10-7 presents comparative failure-rate data for PEMs and
hermetically packaged devices from first-year warranty information on
commercial equipment operating primarily in ground-based applications

1000

... 100
o=
..::
n, ~lin-:~tiC (Prairie 1984J
Linear
10 plastic Motorola.
ITT 1988) ~tic CMOS LFergua 1990)
-..
~ 1.0
..a
;f ",,"___ lDlel,
0.1 _ VSU plastic [Crook 1990J

0,01 L-......I-~-.......-'---I.-....--'
76 78 80 82 84 86 88 90
Year
Figure 10-6. Microcircuit Reliability Improvement Trends. (From Ref. 7.)
11-404 PLASTIC PACKAGING

0.6
Devices tested - 589.148.859
0.5 Failure devices - 57.274
Tow hours - 445.376.007,880
on 0.4
.a
-
~

...'"
.2
Col
0.3 - - Hermetic
- - Non-hermetic
;E 0.2

0.1

0.0 !---'--~~-""-"--""-~--.---r-T"-';:::'"
78 79 80 81 82 83 84 85 86 87 88 89 90
Year
Figure 10-7. Comparative Failure Rate as a Function of Year. (From Ref. 10.)

(office, laboratory, and transportable equipment) from 1978 to 1990; the


rates are for the same part (or part function) over time [10]. As shown
in Figure 10-8, during this period, both types of packaged devices im-
proved by more than one order of magnitude in early-life failure rate. For
PEMs, the early 1990s failure rate was between 0.3 and 3.0 failures per
106 device-hours, with less variability between encapsulant materials and
their vendors. This very closely correlates with the figures for hermetic
parts [11].
A study by Rockwell International, Collins Group [12] compared
plastic dual-in-line packages and plastic surface-mounted devices to ce-
ramic dual-in-line packages (CERDIPs) during extended temperature cy-
cling of -40°C to +85°C for 3532h in (4-h cycles). The parts were mounted
on circuit boards, biased with 5 V during testing, and subjected to 98%
relative humidity. These conditions were designed to stimulate a worst-
case avionics environment. The failure rate for the plastic parts was as
low as 0.0016% per 1000 h, whereas the failure rate for the hermetic parts
was 0.0061 % per 1000 h, both at 60% confidence. The large difference in
failure rates, shown in Table 10-2, is due to the loss of hermeticity of
glass seals. The failure mechanism for both plastic and hermetic parts
was metallization damage due to moisture. The sample size of the various
lots was 2920 plastic parts and 1200 ceramic parts from 5 different vendors
(Texas Instruments, Motorola, National, Signetics, and Fairchild); there
were 174.6 million equivalent field device-hours.
Lidback [13], of Motorola's Semiconductor Product Sector, studied
10.1 INTRODUCTION 11-405

Table 10-2. Thermomechanical Reliahility (From Ref. 1.)

Number Failed!
Study [Ref.] Package Number Tested Failure Rate

Griggs Plastic hermetic 112,920 0.0342%/IK h


(Rockwell Collins) [12] 211,200 O.l667%/IK h
Villalobos Plastic hermetic 23/9,177 0.2S06%/IK cycles
(Motorola TED) 7/1,844 0.3796%/IK cycles
Lidback Plastic hermetic 111133,747 0.083%
(Motorola SPS) 46/46,473 0.099%

133,747 PEMs and 46,473 hermetic packages (obtained from Motorola's


Semiconductor Product Sector) that underwent 1000 temperature cycles
from -65°C to + 150°C. The failure rate was 0.083% for the PEMs and
0.099% for the hermetic packages.
Villalobos [14] ran 1000 cycles at -65°C to +150°C using a smaller
sample size of 9240 plastic packages and 1848 ceramic packages (also
obtained from Motorola). The ceramic failure rate was 0.38%, and the
plastic failure rate was 0.44%. Both of these studies indicated similar
failure mechanisms for plastic and hermetic packages under worst-case
conditions.
The automotive industry, which uses some of the commercial indus-
tries' most stringent qualification standards, conducts temperature cycling
for 1000 cycles, thermal shock (liquid-to-liquid) for 500 cycles, 85°C and
85% rh testing for 1000 h, life testing for 1000 h, high-temperature reverse
bias for 1000 h, intermittent operational life testing for 20,000 cycles,
and autoclave (live steam) testing for 96 h. The number of rejects allowed
for all these tests is zero. Most vendors pass these tests without problems,
indicating a broad, industrywide ability to meet or exceed harsh automotive
standards [15].
In general, PEMs lend themselves well to automatic assembly tech-
niques, which eliminate manual handling and operator error, resulting in
high yields and low assembly costs. On the other hand, even for high-
volume production of hermetic packages, automated assembly pick-and-
place machines have been known to crack hermetic seals or chip ce-
ramic packages.

10.1.3.5 Availability
Plastic devices are assembled and packaged on continuous produc-
tion lines, as opposed to the on-demand production of hermetic parts,
thus making PEMs much more readily available than hermetic devices.
11-406 PLASTIC PACKAGING

This is mainly because market forces (cost and volume) encourage most
designs to be developed first as plastic encapsulated. Lead times for plastic
packages are significantly shorter. The problems associated with restarting
a hermetic line are absent in continuous plastic production lines.
Hermetic packages are developed only when there are perceived
high-performance requirements and sufficient market interest. Thus, some
parts are simply not available from major manufacturers in hermetic form.
Furthermore, the u.s. military and government, the major purchasers of
hermetic parts, have become a relatively small portion of the total electron-
ics market (- 1% in 1995), although they accounted for nearly 80% of
the total market in the 1960s. With package technology moving to surface
mount, development of ceramic packages has lagged further in the micro-
electronic market, making adaptation of plastic-packaged integrated cir-
cuits to government and military applications more critical. With global
competition, industrial research in materials and manufacturing processes
will continue to focus on PEMs. In 1993, over 97% of all integrated
circuits were packaged in plastic. It is estimated by suppliers that, at any
given time, 30% more part functions are available in plastic than in
ceramic [7].

10.1.4 Summary
Plastic encapsulation has come a long way. It offers attractive benefits
over ceramic technology in terms of form factors, weight, performance,
cost, and availability. The reliability of plastic packages is no longer a
stumbling block to their widespread application.
Plastic-encapsulated microelectronics have generally encountered
formidable challenges in gaining acceptance for use in government and
military applications. In fact, it was only in the early 1990s that the
industry dispelled the notion that hermetic packages were superior in
reliability to plastic packages, in spite of the fact that hermetic packages
typically had low production and procurement volumes, along with out-
dated government and defense department standards and handbooks asso-
ciated with their manufacture and use.
Prior to the 1980s, moisture-induced failure mechanisms, such as
corrosion, cracking, and interfacial delamination, were significant [16].
However, improvements in encapsulants, die passivation, metallization
technology, and assembly automation have made plastics the encapSUlation
technology of the future. Currently, the best endorsement for PEMs is
from automotive manufacturers, which consume plastic integrated circuits
at a rate of over 2.7 million a day [9].
Plastic-encapsulated microcircuits will continue to account for the
vast share of the integrated-circuits market in coming years, but hermetic
packages, with their special characteristics, will continue to have a unique
market in the electronics industry.
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-407

10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS


A PEM encapsulant is generally an electrically insulating plastic
material formulation that protects an electronic device and die-Ieadframe
assembly from the adverse effects of handling, storage, and operation.
The encapsulant must possess adequate mechanical strength, adhe-
sion to package components, manufacturing and environmental chemical
resistance, electrical resistance, matched coefficient of thermal expansion
to the materials it interfaces with, and high thermal and moisture resistance
in the application temperature range. Thermosetting encapsulation com-
pounds (i.e., molding compounds) based on epoxy resins or, in some
niche applications, organosilicone polymers, are the most widely used to
encase electronic devices. Polyurethanes, polyimides, and polyesters are
used to encase modules and hybrids intended for use under low-tempera-
ture and high-humidity conditions. Thermoplastics are rarely used for
PEMs, because they require high-temperature and high-pressure process-
ing conditions, are low-purity materials and can result in moisture-in-
duced stresses.
The molding compound is a multicomponent mixture of an encapsu-
lating resin with various types of additives. The principal active and
passive (inert) components in a molding compound include curing agents
or hardeners, accelerators, inert fillers, coupling agents, flame retardants,
stress-relief additives, coloring agents, and mold-release agents. Table
10-3 lists these components and their functions in epoxy molding com-
pounds [17]. Each of these components will be discussed in this section.

10.2.1 Resins
The earliest resin materials used for plastic packaging of microelec-
tronic devices were silicones, phenolics, and bisphenol A epoxies because
of their excellent molding characteristics. Silicones were used as molding
compounds because of their high-temperature performance and purity,
but poor adhesion of silicones with the device and metallization resulted
in device failure in salt-spray tests and flux penetrants during soldering.
Integrated circuits (ICs) molded with phenolics experienced early corro-
sion failures caused by the generation of ammonia during postmold curing,
the presence of high levels of sodium and chloride ions, and high moisture
absorption. Early epoxy materials such as bisphenol A (BPA) offered
much better adhesion than silicones but have a low glass-transition temper-
ature in the range of 100-120°C and high concentrations of corrosion
agents such as chlorine. As a result, novolac epoxies displaced BPA mate-
rials.
Table 10-4 summarizes PEM properties of commonly used polymeric
materials. Except for epoxy resins, all others presented have only niche
applications in PEM manufacture.
11-408 PLASTIC PACKAGING

Table 10-3. Components of Epoxy Molding Compounds Used in Electronic


Packaging

Concentration
ComQ2nent % of Resin)
(WI. Major Function TYl~ical Agents

Epoxy resin Matrix Binder Cresol-novolac


Curing agents Up to 60 Linear/cross-polymerization amines, phenols, and acid
(hardeners) anhydrides
Accelerators Very low «I) Enhance rate of polymerization Amines, imidazoles,
organophosphines, ureas,
Lewis acids, and their
organic salts
Inert fillers 68-80 Lower coefficient of thermal ground fused silica
expansion, higher TC' (widely used), alumina
(w/AI,03)' higher E," reduce
resin bleed, reduce shrinkage,
reduce residual stress
Flame retardants -10 Retard flammability Brominated epoxies,
antimony trioxide
Mold-release Trace Aid in release of package from Silicones, hydrocarbon
agents mold waxes, fluorocarbons,
inorganic salts of organic
acids
Adhesion Trace Enhance adhesion with IC Silanes, titanates
parameter components
Coloring agents -0.5 Reduce photonic activity; Carbon black
reduce device visibility
Stress-relief Up to 25 Inhibit crack propagation, Silicones,
additives reduce crack initiation, lower acrylonitrile-butadiene
coefficient of thermal rubbers, polybutyl
expansion acrylate

'Thermal conductivity.
bE modulus.
Source: From Ref. 17.

All epoxy resins contain compounds from the epoxide, ethoxylene,


or oxirane group, in which an oxygen atom is bonded to two adjacent
(end) bonded carbon atoms. Because the resins are cross-linkable (thermo-
setting), each resin molecule must have two or more epoxy groups. Most
commonly, the epoxide group is attached linearly to a chain of -CH 2-
and is called a glycidyl group.
This group is attached to the rest of the resin molecule by an oxygen,
nitrogen, or carboxyl linkage forming glycidyl ether, amine, or ester. The
resin is cured by the reaction of the epoxide group with compounds
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-409

Table 10-4. Common Polymeric Encapsulants for Microelectronic Packages

Polymeric Advantages and


Material Properties Disadvantages

Epoxies Good chemical and mechanical protection High stress


Low moisture absorption Moisture sensitivity
Suitable for all thermosetting processing Short shelf life (can be
methods extended under low-
Excellent wetting characteristics temperature storage
Ability to cure at atmospheric pressure conditions)
Excellent adhesion to a wide variety of
substrates under many environmental
conditions
Thermal stability up to 200°C

Silicones Low stresses Low tensile tear strength


Excellent electrical properties High cost
Good chemical resistance Attacked by halogenated
Low water absorption solvents
Thermal stability up to 315°C Poor adhesion
Good UV resistance Long cure time

Polyimides Good mechanical properties High cure temperature


Solvent and chemical resistant Dark color (some)
Excellent barriers Attacked by alkalis
Thermal stability from -190°C to 600°C High cost
Require surface priming and/or
coupling agents to improve
adhesion properties
Low moisture resistance
Low dielectric constant
Lower thermal stability
Provides improved stress relief

Phenolics High strength High shrinkage


Good moldability and dimensional Poor electrical properties
stability High cure temperature
Good adhesion Dark color
High resistivity High ionic concentration
Thermal stability up to 260°C
Low cost

Polyurethanes Good mechanical properties (toughness, Poor thermal stability


flexibility, resistance to abrasion) Poor weatherability
Low viscosity Flammable
Low moisture absorption Dark colors
Ambient curing possible High ionic concentration
Thermal stability up to about 135°C Inhomogeneities due to poor
Low cost mixing of two-part system
11-410 PLASTIC PACKAGING

containing a plurality of reactive hydrogen atoms (curing agents) like


primary, secondary, or tertiary amines, carboxylic acid, mercaptan, and
phenol; as such, they are coreactants. Catalytic cures are affected by Lewis
acids, Lewis bases, and metal salts, resulting in the linear homopolymeriza-
tion of the epoxide group forming polyether linkages. The viscosity of
the base resin for the encapsulant-grade materials is kept low by limiting
the average degree of polymerization of the novolac epoxy to values
around 5. The different commercial formulations differ primarily in the
distribution of the degree of polymerization, with an average molecular
weight of about 900. Longer and shorter chain lengths provide looser or
tighter cross-link structures, with varying ductility and glass-transition
temperatures.
In electrical and electronic applications, three types of epoxy resins
are commonly used: the diglycidyl ethers of bisphenol A (DGEBA) or
bisphenol F (DGEBF), the phenolic and cresol novolacs, and the cycloali-
phatic epoxides. Liquid DGEBAs synthesized from petrochemical deriva-
tives are most common. They are readily adaptable for electrical and
electronic device encapSUlation. DGEBF is less viscous than DGEBA.
The epoxy novolacs, essentially synthesized in the same way as DGEBA,
are primarily solids. Because of their relatively superior elevated-tempera-
ture performance, they are widely used as molding compounds. The
cycloaliphatic epoxides or peracid epoxides, usually cured with dicarbox-
ylic acid anhydrides, offer excellent electrical properties and resistance
to environment exposure.
A new class of popcorn-resistant (during retlow soldering in printed
wiring assembly) ultralow-stress epoxy encapsulating resins has been
synthesized. Ultralow-stress epoxy molding compounds exhibit almost
no shrinkage while curing. The absence of stress is achieved by including
chemicals that inhibit cross-links between polymer chains in the plastic.
Without cross-links, the chains are less likely to pull toward one another
as the plastic cures. These new compounds result in PEMs that routinely
survive solder immersion shock tests without failure.

10.2.1.1 Fillers and Coupling Agents


Fillers are employed in epoxy resins to modify the properties and
characteristics of epoxies. Inert inorganic fillers are added to the molding
compound to lower the coefficient of thermal expansion, increase thermal
conductivity, raise the elastic modulus, prevent resin bleed at the molding
tool parting line, and reduce encapsulant shrinkage during cure (and,
thus, reduce residual thermomechanical stress). Microstructurally, particle
shape, size, and distribution in the chosen filler dictate the rheology of
the molten epoxy molding compound. The advantages and disadvantages
of the use of the epoxy are given in Table 10-5. Common fillers and the
properties they affect are given in Table 10-6.
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-411

Table 10-5. Advantages and Disadvantages of Using Fillers

Advantages Disadvantages

Reduced shrinkage Increased weight


Improved toughness Increased viscosity
Improved abrasion resistance Increased dielectric constant
Reduced water absorption
Increased heat-deflection temperature
Increased thermal conductivity
Reduced thermal-expansion coefficient

Historically, the filler with the optimum combination of required


properties has been crystalline silica or alpha quartz. A typical crystalline-
silica-filled molding compound, loaded to 73% by weight, offers a coeffi-
cient of thermal expansion of about 32 ppmrC and a thermal conductivity
of around 15 kW/m ac. Crystalline silica was replaced by ground fused
silica, which provided lower density and viscosity. A formulation with
similar moldability, as one with 73% crystalline silica, requires 68% fused
silica and produces a coefficient of thermal expansion of around 24
ppmfOC and a thermal conductivity about 16 kW/m ac.
The addition of particulate fillers generally reduces strength charac-
teristics such as tensile strength and flexural strength. Fillers do not usually
provide any significant enhancement of glass-transition temperature or
other measures of heat-distortion temperatures. Fillers also allow the
possibility of modifying various thermal characteristics, including the
thermal conductivity and thermal expansion coefficient. Thermal conduc-

Table 10-6. Common Fillers and Their Property Modification

Alumina Abrasion resistance, electrical resistivity, dimensional stability, toughness,


thermal conductivity
Aluminum trioxide Flame retardation
Beryllium oxide Thermal conductivity
Calcium Silicate Tensile strength, flexural strength
Copper Electrical conductivity, thermal conductivity, tensile strength
Silica Abrasion resistance, electrical properties, dimensional stability, thermal
conductivity, moisture resistance
Silver Electrical conductivity, thermal conductivity
11-412 PLASTIC PACKAGING

tivity can be increased by a factor of about 5 by the addition of fillers


such as alumina and copper. Generally, an increase in filler concentration
increases thermal conductivity. Figure 10-8 shows data on the effect of
filler content on thermal conductivity. Thermal conductivity can also be
enhanced by the addition of other filler materials such as aluminum nitride,
silicon carbide, magnesium oxide, and silicon nitride [18]. However,
Proctor and Sole [19] have shown the futility of using filler materials for
thermal conductivity more than 100 times of that of the base resin and
estimated that a practical limit of thermal conductivity improvement is
about 12 times that of the thermoset resin. The influence of varying the
thermal conductivity of the epoxy molding compound on the thermal
performance of the IC package has also been evaluated [20]. In this case,
by varying the filler types (e.g., fused and crystalline silica, aluminum
oxide, aluminum and boron nitride, silicon carbide, and diamond), the
particle sizes, and the filler distribution, the thermal conductivity of the
composites was estimated. The resulting effectiveness in dissipating heat
away from the package was computed for three different surface-mount
packages, namely SOIC 8-lead, 16-lead wide, and 24-lead wide, The
results indicated that an enhanced mold compound can cause a decrease
in thermal resistance of the package equivalent to what can be achieved
with a molded-in heat spreader. However, when thermally enhanced lead-
frames are used, the modified compounds are less effective.
In low-stress epoxy resins, spherical silica particles are usually
blended with crushed silica to further lower the coefficient of thermal
expansion, by increasing filler loading with a nonlinear increase of viscos-

.E
£:I
(J

c
~ 100
:~

lo
(J

~ 10

20 40 60 80 100
Filler content (vol %)

Figure 10-8. Relationship of Filler Content and Thermal Conductivity of the


Melting Compound [18]
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-413

ity. However, an increase in the melt viscosity of the molding compound


can increase void density and increase the difficulty of achieving a uniform
encapsulant flow over large areas. Variables such as particle size, particle-
size distribution, particle surface chemistry, and particulate volume frac-
tion have been found to be the most important variables necessary for
optimum property enhancement. The effect of the filler figure ratio (angu-
lar : spherical) as measured by the molding compound spiral flow length
for two volume-percentages filler loadings is shown in Figure 10-9. How-
ever as Figure 10-10 shows, moisture ingress susceptibility of the molding
compound as measured by the "popcorn" effect during 215°e, 90-sec,
vapor-phase soldering increases with higher percentages of spherical sil-
ica filler.
RosIer [18] evaluated various other fillers for optimizing the coeffi-
cient of thermal expansion and the thermal conductivity for specific appli-
cations. Figure 10-11 presents the effect of lowering the coefficient of
thermal expansion of the molding compound as a function of the crystalline
silica, a-alumina, and fused silica volume percentage.
Most polymers undergo shrinkage during the polymerization and
cross-linking process. Incorporation of fillers reduces shrinkage by simple
bulk replacement of resin with an inert compound which does not partici-
pate in the cross-linking process.
Filler addition results in increased viscosity and improved toughness.

20

G
b
00
16

12
", .....
Filler contents (vol%)
8
o 62
• 65

4
100 50 o angular
o 50 100 spherical
Filler figure ratio (angular/spherical)

Figure 10-9. The Effect of the Filler Figure Ratio (angular: spherical) as
Measured by the Molding Compound Spiral Flow Length for Two Volume-
Percentages Filler Loadings (Courtesy of National Semiconductor.)
11-414 PLASTIC PACKAGING

~ 70
u
...r:s
""
<.J
c... 60 QFP 54pin 20xI4x2mm
u
c Leads alloy 42
C.
c.::c
0""'"
50 "" Chip: 6x6mm

"
coQi:! VPS: 215°Cl90sec.
.5~
~on
""
•"
coo
;:SU 40
"'0
J20n
U oo
e.:
":; en
c ...
o ::s
30
""
e.., 5
.- 0

0
20 "
.&l
...r:s
£
~
10

0
• 65

100 50 0 angular
0 50 100 spherical
Filler figure ratio (angular/spherical)

Figure 10-10. Moisture Ingress Susceptibility of the Mold Compound as Mea-


sured by the Popcorn Effect

"-
G
":'
~
.......
6

c:
0
":';;
5
c:
x'"
C.
u 4
c;
§
<.J

-=0
..... 3
AI
C
<.J
.~ 1
to: _Cu
'-
g -Au
U -Fe

-Si
0
0 20 40 60 80 100
Filler content (\"ol~)

Figure 10-11. Effect of Lowering the Coefficient of Thermal Expansion of


the Molding Compound as a Function of the Crystalline Silica, a-alumina,
and Fused Silica Volume Percentage
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-415

Investigations have shown that the incorporation of particulate fillers such


as silica, glass microspheres, and alumina trihydrate can increase the
toughness of various epoxy formulations.
A filler can be used to its best advantage if the adhesion between
the polymer and itself is good. In particular, the filler particle-polymer
interface will not be stress bearing and therefore provides a point of
mechanical weakness. Coupling agents are used to increase the adhesion
between fillers and polymers by linking them with covalent bonds. Com-
monly used coupling agents include silanes, titanates, aluminum chelates,
and zircoaluminates. Interfacial adhesion enhances mechanical strength
and the heat resistance of the encapsulant. Although, coupling agents also
improve processability, some adversely affect releasability from the mold
if not correctly administered. This adhesion of fillers to the polymer matrix
through coupling agents also extends to the device and leadframe, reducing
delamination failure.
The main source of radiation in plastic packages comes from fillers.
However, the use of synthetic high-purity silica fillers in recent years has
brought down the alpha emission rates (AERs) of molding compounds
below the detection level of most proportional counters. Ditali and Hasnain
[21] studied AER sources in an electronic part and that outside the major
alpha-source contamination of the die itself, the leadframe material is
more contaminated (per unit surface area) than the plastic encapsulant,
and ceramic package materials are 300-400 times more alpha-emitting
material contaminated than plastic encapsulants.
An approximate conversion factor of AER to soft-error rate is 0.001
AER == 0.1 % Single Event Upset (SEU) per 1000 operating hours for
the contaminated material in contact with the die active surface and alpha-
particle energies in the range of 4-6 MeV. An AER of about 0.001 can
be obtained from 10 ppb of U-238 (4.2 MeV alpha) or 4 ppb of Th-230
(4.7 MeV alpha).

10.2.1.2 Curing Agents and Accelerators

The binder system of a plastic-encapsulant product consists of an


epoxy resin, hardener or curing agent, and an accelerating catalyst system.
The conversion of epoxies from the liquid (thermoplastic) state to tough,
hard thermoset solids is accomplished by the addition of chemically active
compounds known as curing agents. Epoxidized phenol novolac and epoxi-
dized cresol novolac are most often used. The specific composition of
the resin and the hardener system, therefore, is optimized for the specific
application. For example, incorporation of phenol novolacs in the matrix
resin can be used to increase cure speed. In general, the glass-transition
curing temperature is maximized at an epoxide to hydroxyl ratio of
about unity.
11-416 PLASTIC PACKAGING

Curing agents and accelerators perform the primary function of


setting the extent and rate of polymerization of the resin. Consequently,
the selection of the agent and cure chemistry are just as important as
the choice of resin. Often, the same agent is used for both cure
and acceleration.
The cross-linking reactions of epoxy resins are affected by aliphatic
or aromatic arnines, carboxylic or its derivative acid anhydrides, and
complex phenols, whereas homopolymerization is affected by Lewis acids
or bases and their organic salts. The most widely used curing agents are
amines and acid anhydrides. Although aliphatic amines react rapidly at
room temperature, aromatic amine curing agents impart higher thermal
stability and improved chemical resistance. Depending on the curing agent
used, the reactions proceed through different chemistries. Because this
reaction is virtually quantitative, resin and curing agent are mixed in the
ratio of one amino hydrogen to one epoxide. In the case of anhydrides
and phenols as curatives, cross-linking occurs through pendant hydroxy Is.
Table 10-7 lists the curing properties of different classes of curatives
used with epoxy resins. Because of their excellent moldability, electrical
properties, and heat and humidity resistance [22], phenol novolac and
cresol novolac hardeners have become the dominant curatives for micro-
electronic packaging.
To promote cure within a reasonable period of time, it is necessary
to use accelerators. Accelerators reduce the in-mold cure time and improve
productivity by catalytic activity. Typical accelerators include aliphatic
poylamines of tertiary arnines, phenols, nonyl phenol, resorcinol, or semi-
inorganic-derived accelerators such as triphenyl phosphite and toluene-
p-sulforic acid.
For transfer molding of microelectronic devices, epoxy resins are
B-staged to produce storage-stable thermosetting epoxy slugs known as
prepregs. The B-staged resin is formed by reacting terminal epoxides
with aminohydrazides at 70-80°C to yield B-staged resins with excellent
storage stability.

10.2.1.3 Stress-Release Additives

The toughness and stress-relaxation response of epoxy resins can


be enhanced by the addition of flexibilizers and stress-relief agents. Stress-
release additives lower the thermomechanical shrinkage that can initiate,
as well as propagate, cracks in the molding compound or in the device
passivation layer. In terms of molding-compound property modification,
stress-release agents lower the elastic modulus, improve toughness and
flexibility, and lower the coefficient of thermal expansion.
Inert flexibilizers, like phthalic acid esters or chlorinated biphenyls,
remain as a separate phase. Reactive flexibilizers remain in the epoxy
Table 10-7. Pure DGEBA Epoxy Resin Curing Agents Used in Molding Compounds

Typical Cure Typical Postcure Heat


Concentration Pot Life Deflection
Range Temperature Time Temperature at 25°C Temperature
Curative Type (Parts! 100 Parts of Resin Time .L9 .@ .L9 .@ .L9
Aliphatic tertiary amines and
deri vati ves
(room-temperature cure) 12-50 4-7 days 25 0-1 150-200 0.25-3 55-124
Cycloaliphatic amines
:!: (moderate-temperature cure) 4-29 0.5-8 h 60-150 0-3 150 0.25-20 100-160
....
Aromatic amines
(elevated temperature cure) 14-30 4h 80-200 2 150-200 5-8 145-175
Carboxylic and anhydrides'
(elevated-temperature cure) 78-134 2-5 h 25-150 3-4 150-200 24-120 74-197
Lewis acids and bases'
(elevated-temperature cure) 3 7h 120-200 4 -200 >250 175
Latent curing agents
(elevated-temperature cure) 6 1h 175 175 135

'Two-step cure.
Source: From Ref. 17.
11-418 PLASTIC PACKAGING

matrix as a single-phase material and lower the tensile modulus and


improve the ductility of the encapsulant. They also reduce the exotherm
and, in certain circumstances, reduce shrinkage. Flexibilizers can also
improve adhesive joint properties such as lap shear, peel strength, impact
strength, and low-temperature crack resistance.
Stress-release additives lower the thermomechanical shrinkage
stresses that can initiate as well as propagate cracks in the molding com-
pound or in the device passivation layer. In terms of molding-compound
property modification, stress-release agents lower the elastic modulus,
improve toughness and flexibility, and lower the coefficient of thermal
expansion.
In epoxy molding compounds, the major stress-relief agents used are
silicones, acrylonitrile-butadiene rubbers, and polybutyl acrylate (PBA).
Silicone elastomers, with their high-purity and high-temperature proper-
ties, are the most favored stress-relief agents. Silicone elastomers interface-
modified with polymethyl methacrylate (PMMA) possess uniform domain
sizes (1-100 Ilm) and inhibit passivation layer cracking, aluminum line
deformation, and cracking.

10.2.1.4 Flame Retardants

Flame retardants in the form of halogens are added to the epoxy


resin backbone because epoxy resins are inherently flammable. One of the
more important flame retardants is the homologous brominated DGEBA
(diglycidyl ethers ofbisphenol A). When used with normal nonhalogenated
resin, these give self-extinguishing properties, flame retardancy being
achieved by the bromine liberated at the decomposition temperature.
Approximately 13-15% (wt.) of bromine is required to make an unfilled
epoxy flame retardant [17].
Used as a filler, antimony trioxide used as a filleris another commonly
used flame retardant. However, the cost is higher. Homologous brominated
DGEBA [18] epoxy and heterogenous antimony oxides can also be
used together.
Other ways of incorporating flame retardancy in epoxy resins include
using a nonreactive phosphorus-containing diluent, chlorinated waxes,
aluminum hydrate, and various phosphorus derivatives. These are less
popular methods because they can lead to deterioration of physical proper-
ties of the system and limit the choice of the curing agent and filler.
Although flame retardants are a source of contamination in encapsu-
lant formulations, the use of nonhydrolyzable compounds and ion getters
can control their corrosion-inducing effects. However, studies at Plaskon
Electronic Materials [23] show improved high-temperature storage perfor-
mance of Plaskon 3400 epoxy molding compound when flame retardants
are eliminated, as per Figure 10-12.
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-419

100

90
High temperature storage 200°C
80
7400 NAND GATE. NO BIAS
~
';; 70
8
.~ 60
""0
""0
.!! 50
·co
u.
40

30
Improved PLASKON
20 3400 with no flame
10

o
I 1'1 I I I I I I I I I I I I I I I I I I I I I I I I I I
o 400 800 1200 1600

Hours of exposure

Figure 10-12. Improved High-Temperature Storage Performance of


PLASKON 3400 Epoxy Molding Compound When Flame Retardants Are
Eliminated. (Courtesy of PLASKON Electronic Materials.)

10.2.1.5 Mold-Release Agents

The excellent adhesion of epoxy resins to all types of surfaces is


the major reason for the wide application of release agents in microelectric
packaging. But these same properties make release from compounding
and molding equipment difficult. Consequently, mold-release agents are
needed that do not degrade the epoxy adhesion to the package components.
This is achieved by controlling the release-agent activity as a function
of temperature.
Mold-release agents are usually in the form of microplates that vary
from liquids to pasty solids to finely divided powders. In general, the
release agent should be insoluble in the resin mixture, should not melt at
the curing temperature, and should be applied in a continuous film. The
selection of the release agent is determined by the material of construction
of the mold and the type of encapsulation selected.
Different release agents in epoxy molding compounds are used for
compounding (-100°C) and molding (-175°C) operations. Room-temper-
11-420 PLASTIC PACKAGING

ature adhesion of epoxy molding compound to leadframe is provided by


the higher temperature (-175°C) active mold-release agent solidification
and consequent loss of effectiveness. The release agent used for com-
pounding, if not completely degraded, could become active at near 100°C
use temperature and lead to some delamination in the package. Mold-
release agents for epoxy molding compounds include silicones, hydrocar-
bon waxes, inorganic salts of organic acids, and fluorocarbons. Of these,
hydrocarbon waxes, such as carnauba wax, are the most common for
molding compounds used in microelectronic encapsulation. Silicones and
fluorocarbons have poor functional temperature selectivity, whereas or-
ganic acid salts can corrode metallic package elements.

10.2.1.6 lon-Getter Additives


The objective of ion-getter additives is to reduce the conductivity
of any accumulated water at the metal-encapsulant interfaces inside the
package, and thus to retard any electrolytic corrosive degradation pro-
cesses. The incorporation of alkali- and halide-ion gettering agents into
the epoxy molding compound makes residual Na+, K+, Cl-, and Br- ions
in the epoxy unavailable for dissolution into any diffused, accumulated
water in the package. Ion getters in epoxy molding compounds are hydrated
metal oxide powders with particles several microns in diameter. These
materials react with highly active alkali and halide ions to release OR-
and H- ions inside the molding compound and form nearly water-insoluble
alkali and halide compounds. In about 5 wt. % concentration, ion-getter
additives have been reported to reduce aluminum cathodic and chemical
corrosion by l.5-3 times [23]. The main application of such formulated
molding compounds is in packages where the encapsulant thickness is
around a few mils, as in ultrathin outline packages.

10.2.1.7 Coloring Agents


Coloring agents are normally used to distinguish different device
types in packages, reduce photonic activity of the device, and eliminate
device visibility through the normally pale yellow epoxy encapsulant.
The colors are produced by the addition of thermally stable organic dyes
or pigments. Carbon black is used in most plastic-encapsulated silicon
integrated circuits, even though it does add slightly to the electrical conduc-
tivity of epoxy and can reduce moisture resistance. The concentration of
carbon block is usually less than 0.5% to avoid problems related to
moisture absorption and impurities.

10.2.2 "Low-Stress" Molding Compounds


On the other hand, reducing the TCE of the encapsulation down to
a value matching that of silicon is a difficult and challenging task. Positive
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-421

and negative TCE values for a number of polymers, metals, and ceramics
are listed in Table 10-8. Some general observations can be readily made:

1. TCEs for polymers generally lie in the range from 500xlO-7/oC for
polyphenylene sulfide to 2,OOOxlO-7fOC for polyethylenes.
2. Metals, on the other hand, except for Invar and SuperInvar (TCE ""
lOxlO-7fOC), display TCEs ranging from 50xlO-7/oC for molybde-
num and tungsten to close to 240xlO-7/oC for tin and aluminum.

Table 10-8. Thermal Expansion of Some Representative Materials at 25°C

Coefficient of
Material Thennal Expansion

POLYMERS
Polyethylene 1,500-3,000
Polyester 954
Nylon 820
Epoxies (cast rigid) 500
Polyphenylene sulfide 540
Polyphenylene oxide 684

METALS
Aluminum 235
Boron 48
Copper 170
Gold 141
Invar 2
Molybdenum 48
Nickel 129
Silver 187
Tin 235
CERAMICS
AI,03 54
MgO 104
Fused silica 9
Quartz 18
TiO, 75
Si3N4 8
Supertemp (Pycobond) graphite pitch -45
material
Coming 9617
Nb,Os (sintered) -20
~-eucryptite (Li,O . Al,03 . 2SiO,) -60
SiO,· 12Li,O· 8AI,03· 1 GeO, -28
~ . quartz (64.68SiO, . 19.9ZnO . 14.9Al,03) -3
Li,O· Ta,Os -110
TaI6Wls094 -50
11-422 PLASTIC PACKAGING

3. Finally, the TCEs of ceramics are much wider, spanning over three
orders of magnitude. For instance, Coming 7971 has a TCE of
0.3xlO-7/oC, while magnesium oxide has a TCE of 100xlO-7fOC.

One important conclusion to draw from Table 10-8 is that some


materials (e.g., Invar and Coming 7971) can exhibit zero TCE at some
temperatures and negative TCE over another, smaller range of tempera-
tures. Studies of fused silica, for instance, indicate that the TCE is negative
at low temperatures, becomes positive at higher temperatures, and turns
negative again at still higher temperatures. The same behavior is observed
with Ti02-Si02 mixtures. At high Ti02 concentrations, an extended region
of low TCE is obtained-around 0.3xlO-7fOC for Coming 7971[24,25].
Obtaining even lower TCEs in single-phase materials seems to be a very
difficult problem since current theories of thermal expansion do not provide
any guidance for developing such materials. The TCE of such materials
seems to be structure insensitive. Small quantities of second phases and
oriented grains do have some effect-typically less than 30%. Porosity,
dislocation density, grain-boundary area, and stoichiometry, however,
have been found to exert no major influence on the TCE.
It appears that the best approach to achieving low or near-zero TCE
materials is through the development of polyphase materials. For instance,
composites made of a highly anisotropic second phase inserted into an
isotropic matrix can have zero or ultralow TCE for a range of temperatures.
Low-expansion glasses, graphite epoxy, or metal composites are good
examples. However, the anisotropic thermal expansion rules out their
usage in quite a few applications. In PC board applications, for instance,
the z axis expansion of Teflon® glass is worse than that of FR-4 epoxy
glass laminate, being almost three times that of FR-4 at the latter's glass-
transition temperature (120°C). Only particle-reinforced systems can yield
a material with isotropic ultralow thermal expansion.
Research into the development of mirror substrates for high-altitude
large optics (HALO) and high-energy laser optics (HELO) has pointed
out an approach that might be extended to the encapsulation of chips.
Requirements of high thermal conductivity, high microyield strength, low
density, and distortion of less than 0.1 ~m1m in the temperature range
of 1500 K to 300 0 K were imposed. Only a metal matrix composite can
potentially satisfy such constraints. The two experimental systems se-
lected, Ni - Nb20 5 and SuperInvar-Nb20 5, exhibited TCEs higher than
desired for HALOIHELO applications. It was found that a two-phase
metal composite will possess ultralow or zero TCE when the matrix has
a low positive TCE while the filler should display a large negative TCE
[26]. Furthermore, the volume ratio of matrix to dispersed phase requires
very close and accurate monitoring. Good interphase bonding has been
assumed implicitly.
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-423

10.2.3 Molding Formulation Process

Molding compounds are created in automated processing plants that


take raw materials and blend them into a molding powder with the proper-
ties required for molding les. Usually, the formulation process is proprie-
tary. A representative process flow is given in Figure 10-13.

CD - Epoxy Novolac
® - Hardener
® - S~03
Etc.

Material Storage Hoppers

Filler Filler
Storage Storage

1
Extruder Nozzle Auto Weight
Pro portioners
Doctor Blade Resin Ribbon

Part
B
Hopper
Ribbon
Fines Former
Resin Part-A

Ribbon
Breaker
Part-A Resin Part-B

r---'" Disgard Fines

Into Cold Storage & Ship

Figure 10-13. Molding Compound Formulation Process


11-424 PLASTIC PACKAGING

Table 10-9. Properties of Semiconductor Molding Components (Part 1 of 2)

Compound Thermal Thermal Flexural Flexural


Supplier No. conductivity expansion strength modulus
W/m/°C 1O-7/ oC OPa OPa

Sumitomo EME-ll 00-T 0.67 200 0.14 15.66


EME-ll OO-H 0.67 200 0.14 15.66
EME-ll 00-K 0.67 200 0.14 14.19
EME-6200 0.67 200 0.13 12.72
EME-6300 0.42 170 0.13 12.23
EME-91oo-XK 0.67 200 0.14 14.19

EME-9200 0.67 200 0.13 12.72

EME-93oo 0.42 170 0.13 12.23

EME-5000LS 0.67 200 0.15 14.68

Plaskon 3100 1.0 270-750 0.14 14.68


3100LS 0.71 210-600 0.14 14.68

3200 1.0 270-750 0.15 14.68


3200LS 0.71 220-600 0.15 14.68
3300 1.0 270-750 0.15 14.68
3300LS 0.71 210-600 0.15 15.81
3300SH 0.71 210-600 0.15 15.81
Hysol MGl5f 0.71 250-700 0.12 13.76
MG35F 1.46 260-750 0.13 14.44
MG36f 0.75 170-700 0.13 13.76
Nitto HCIO-2 0.63 200-700 0.15 13.70
HC20-2 0.92 240-700 0.15 13.70
HC30-2 1.46 260-700 0.15 14.68
HC50-2 1.97 250-700 0.15 14.68
MP1l9 0.63 180-650 0.13 12.23
MPX-75 0.71 200-650 0.16 14.19
MP150 0.71 160-740 0.13 11.74
MPI50-164 0.71 160-740 0.13 11.74

Blending is a batch process and is lot-controlled. Shipping lots range


from 2,000 to 18,000 kg. A small sample is generally taken from each
lot and sent ahead to the customer for lot-control testing. Resin, hardener,
and accelerator are in close proximity after final grinding, and slow curing
begins at room temperature. To retard this reaction, the molding compound
must be refrigerated until needed at the molding press. A list of Ie grade
molding compound suppliers is given in Table 10-9.
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-425

Table 10-9. Properties of Semiconductor Molding Components (Part 2 of 2)

Extractable Harne
Chlorine Rate Tg Remarks
ppm UL OC
50 94-VO 155 General purpose molding compound for DIPs with small chips
20 94-VO 155 Same as above, but better humidity resistance
20 94-VO 155 Better temperature-cycle performance
25 94-VO 165 Low stress M. C. for large DIPs & PLCCs
25 94-VO 165 Very low stress compound for SO & large thin packages
20 94-VO 155 High purity, low stress, low alpha particle count for molding
DRAMs
25 94-VO 165 Lower stress than 9100-XK low alpha particle count for
DRAMs in PLCC & SOl
25 94-VO 165 Lower stress & low alpha particle count for DRAMs in thin
packages
20 94-VO 135 Low-stress molding compound for SOPs that are solder
dipped
30 94-VO 150 Standard molding compound for lCs & discretes
30 94-VO 150 Low-stress version of 3100 for resistor networks & MOS
devices
20 94-VO 150 General purpose IC molding compound
20 94-VO 155 For large chips in DIPs or PLCCs
20 94-VO 150 For SMT packages & larger DIPs
20 94-VO 155 For gate arrays & microprocessors
20 94-VO 155 For large SMT packages (68-, 84-PlN PLCC) & QFPs
20 94-VO 190 Anhydride cured. Used for high-voltage chips (50-90V)
10 94-VO 160 For high-power discrete
10 94-VO 160 Low-stress compound for large chips in small packages
20 94-VO 162 For small DW, thin flat packages
20 94-VO 161 28-42-lead chips
20 94-VO 162 Discrete devices high voltage lCs > 30 volts
20 94-VO 170 Power transistors, power packages
10 94-VO 170 Low stress and low ionics for SOIC packages
25 94-VO 151 Good adhesion to leadframe for SMT packages
15 94-VO 150 Very low stress for SOIC & PLCCs
15 94-VO 158 Same as MPI50 except low alpha particle count

10.2.4 Lead Frame Design, Materials, and Processes

The leadframe is the backbone of a molded plastic package. Fabri-


cated from a strip of sheet metal by stamping or chemical milling, it
serves first as a holding fixture during the assembly process, then, after
molding, becomes an integral part of the package. The functional require-
ments of a leadframe are:
11-426 PLASTIC PACKAGING

1. A holding fixture that indexes with tool-transfer mechanisms as the


package proceeds through various assembly operations.
2. A dam that prevents plastic from rushing out between leads during
the molding operation.
3. Chip attach substrate.
4. Support matrix for the plastic.
5. Electrical and thermal conductor from chip to board.

10.2.4.1 Materials
Lead frame material selection depends on factors such as cost, ease
of fabrication, and the functional requirements of the frame. There are
three categories of materials to choose from when designing leadframes:
nickel-iron, clad strip, and copper-based alloys. Many alloys, specifically
formulated for plastic IC packages, exist within these categories, as indi-
cated in Table 10-10.
Nickel-Iron Alloys: The most widely used metal for leadframe fabri-
cation is Alloy 42 (42% Nickel-58% Iron). This material was originally
formulated as a glass-seal alloy for use as terminal pins on light bulbs
and vacuum tubes. The first chip-attach method on hermetic as well as
plastic modules was a gold/silicon eutectic alloy bond. Silicon/gold eutec-
tic melts at 370°C, and when solid, has a high elastic modulus. A close
match of thermal-expansion coefficient between silicon and substrate is
mandatory with this bonding system to avoid chip fracture. Alloy 42,
the vacuum tube terminal pin material, was chosen since its expansion
coefficient is close to that of silicon (45 vs. 26xlO-7/ 0 C). Alloy 42, further-
more, can be heat-treated to obtain optimum tensile strength and ductility
for subsequent stamping and lead forming operations. The alloy can also
be easily electroplated or solder-dipped directly without a nickel bar-
rier plating.
Low-thermal conductivity is the most important drawback of nickel-
iron alloys. Since the leadframe is the main conduit by which heat flows
from the chip to the printed circuit board, this can have a profound effect
on the package thermal resistance after prolonged device operation. Clad
materials and copper-based alloys have been developed to reduce thermal
resistance [27].
Clad Materials: A layered composite strip, such as copper-clad
stainless steel, was developed to emulate the mechanical properties of
Alloy 42 while increasing thermal conductivity. Cladding is accomplished
by high-pressure rolling of copper foil onto stainless steel strip and anneal-
ing the composite to form a solid-solution weld. Properties are given in
Table 10-10 [28]. More detailed discussions on this subject can be found
in Chapter 18, "Coated-Metal Packaging."
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-427

Copper Alloys: Copper would be an idealleadframe material from


an electrical and thermal conductivity standpoint, but it has properties
that must be modified before it can be successfully used in plastic pack-
ages [29].

10.2.4.2 Strength
The tensile strength of copper alloys is lower than that of nickel-
iron or copper-clad stainless steel. The addition of iron, zirconium, zinc,
tin, and phosphorus serves to improve the heat-treating and work-harden-
ing properties of these alloys. The resulting tensile strength and ductility
are adequate for leadframe fabrication and function.
Copper-alloy frames can be specified with tempers from one-half
hard to spring temper. Half-hard material is less likely to crack when
bending, but full-hard or spring temper is preferred for modules that are
automatically inserted. Hard copper alloys approach iron-nickel in bending
characteristics so that automatic inserters can be set up to run both alloy
types without constant adjustments to cut and clinch mechanisms.

10.2.4.3 Thermal Conductivity


The ability of the leadframe to channel heat from chip to board is
related to the thermal conductivity of the frame material. The main advan-
tage of copper alloys is their high coefficients of thermal conductivity
(see Table 10-10). The largest reduction in package thermal resistance is
achieved by the use of copper alloys as shown in Figure 10-14 [30].

10.2.4.4 Thermal Expansion


Copper alloys have high thermal-expansion rates with respect to
silicon but nearly match the expansion rate of low-stress molding com-
pounds. This property affects the selection of chip-bonding materials.
Silicon/gold eutectic bonding cannot be used with copper frames because
its high elastic modulus couples thermally induced bending stress to the
silicon, which can cause potential chip fracture. Silver-filled epoxies and
polyimide chip-attach adhesives have been developed and are flexible
enough to absorb the strain developed between chip and copper.

10.2.4.5 Mechanical Design


Considerations in leadframe mechanical design are:

1. Ease of fabrication.
2. Indexing features for package assembly.
3. Chip-attachment substrate and gold wire span.
4. Coplanarity for wirebonding.
Table 10-10. Leadframe Materials (Part 1 of 2)

CAR'P OLIN HUSSEY KOBE


PART NO. A-42 CDA-151 CDA-194 CDA-195 CDA-155 KPC KLF-l KLF-2 KLF-5
TYPICAL Fe-58 Zr-O.l Fe-2.3 Fe-1.5 Ag-O.034 Fe-O.l Ni-3.2 Sn-O.l Sn-2.0
ELEMENTS A-42 Cu-bal P-O.03 Co-O.8 P-O.058 P-O.058 Si-O.7 Fe-O.l Fe-O.l
(%) Zn-O.l Sn-O.6 Mg-O.ll Cu-bal Zn-O.3 P-0.03 P-O.D3
Cu-bal P-O.l Cu-bal Cu-bal Cu-bal Cu-bal
Cu-bal

.... Melt P (0C) 1425 1000 1009 1090 1002 1083 1090 1083 1068
ClO
'" Spec. Grav. 8.15 8.94 8.8 8.92 8.91 8.9 8.9 8.9 8.9
E (GPa) 144.83 120.37 120.37 118.89 117.43 120.37 124.28 127.22 120.37
Thermal Conductivity 15.89 359.8 261.5 196.65 347.27 435.14 219.66 376.56 133.89
(W/mJOC)
Linear Expansion 43 177 163 169 177 170 170 170 165
(lO-'/°C)
Electrical Conductivity 3.0 90 65 50 86 92 55 82 35
(% lACS)
Temper H 3/4H H H H H H HH
Tensile Strength (GPa) 0.64 0.35 0.41 0.47 0.30.30.5 0.39 0.54 0.39 0.54
Elongation 10 7 min 3 min 3 min 222 4 min 7 min 4 min 7 min
Vickers HD 210 104 135 147 95 113 137 125 180 140 180
Table 10-10. Lead Frame Materials (Part 2 of 2)

MIT'B TAMAGAWA T.!. OLIN PAN


PART NO. MF-202 TAMAC-2 TAMAC-5 4 lOSS 19750 19700 PMC-IC

TYPICAL Sn-2.0 Sn-0.15 Fe-0.75 Cr-12.0 Fe-0.6 Fe-0.6 Ni-1.0


ELEMENTS Ni-0.2 P-0.006 Sn-1.25 Fe-bal Mg-0.05 Mg-0.05 Si-0.2
(%) -0.15 Cu-bal P-0.03 P-0.02 P-0.02 P-O.03
Cu-bal Cu-bal Sn-0.23 Cu-bal Cu-bal
Cu-bal

.... Melt P (0C) 1065 1083 1075 1085 1086 1090


'"'" Spec. Grav. 8.8 8.9 8.8 7.8 8.82 8.84 8.9
E (GPa) 112.54 117.43 117.43 19.96 120.37 118.41 127.22
Thermal Conductivity 154.81 376.56 138.07 24.27 261.5 320.4 9 259.41
(W/m/°C)
Linear Expansion 169 177 167 110 169 168 169
(lO-'/°C)
Electrical Conductivity 30 92 40 30 65 80 60
(% lACS)
Temper H H H S H H
Tensile Strength (GPa) 0.59 0.34 0.48 0.62 0.56 0.45 0.55
Elongation 7 min 4 min 4 min 5 min 1.9 1.5 4.8
Vickers HD 185 105 150 220 160 144 160
11-430 PLASTIC PACKAGING

Effect of Molding Compound Effect of Leadframe Alloy

£'
~ ~ -
....-~::.~
.......--.:.~.::.::::
:::::.:.-----
~
~
.!::: 10 .!::: 10
~
a. - - - LowK
•••••.••••• Med. K
§ - - - Alloy 151
•.••.•.•••• Alloy 194
s:. s:.
U U
II: ----- High K II: ----- Alloy 42
1~~~--~~~~~~
30.5 61.0 122.0 305.0 30.5 61.0 122.0 305.0
Air Velocity (M/Mln) Air Velocity (M/Min)

Overall Thermal Performance of 68-Lead PLCC - Material Effects

Effect of Molding Compound Effect of Leadframe Alloy

-... :.::.::: .....


--~-----
~..
....
....
~
- -.:..~

- - - LowK - - - Alloy 151


.••••••••••• Med. K ............. Alloy 194
----- High K ----- Alloy 42
1~~u--~--~~~~~
30.5 61.0 122.0 305.0
Air Velocity (M/Min)

Overall Thermal Performance of 84-Lead PLCC - Material Effects

Figure 10-14. Overall Thermal Performance of 84-Lead PLCC-Material Ef-


fects.

5. Lead-locking and moisture-inhibiting configuration.


6. Stress relief.
7. Plastic support matrix.
8. Terminal lead and standoff configuration.

Ease of Fabrication: Figure 10-15 illustrates design rules for consid-


erations 1, 3, and 4. Specifying a single temper for strip feedstock avoids
proliferation of expensive progressive-stamping dies.
Indexing Features: Design and leadframe strips must consider the
package-assembly system. Tooling holes are located along frame-strip-
edges. These mate with the transfer-mechanism pins, which are part of
the assembly equipment, and include chip bonders, wirebonders, molds,
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-431

Cross Hatched
Area Silver 2.54 mm Max. Wire Span
Plate

Chip Support
Platform

Chip
-J I-e- 0.38 mm Min.

~~Plastic0' Edge

~R'
~ I-- Bonding Finger

~=0.8t

t =Material Thickness
Bonding Finger Coplanarity

Figure 10-15. Lead Frame Clearances, Wire Span, Plating, Limits, and Copla-
narity.

auto-inspection stations, trim and form equipment, and marking machines.


In-process storage and transport-magazine-design depend on the frame-
strip length.
Lead Locking and Moisture-Inhibiting Configurations: The bond
between leadframe and plastic is mechanical. Therefore, any protrusion
or hole on the frame that serves to lock the terminal leads within the
cured molding compound is desirable. Several mechanical features have
been developed to inhibit moisture penetration along the plastic leadframe
boundary. These, along with lead-locking configurations, are shown in
Figure 10-16.
Stress Relief Lead frames cause two types of stress that can be
11-432 PLASTIC PACKAGING

Platform Support
Bar

Tool Index Holes

Preferred Design

Fine Pitch Lead


Staggered Anchor

Figure 10-16. Typical Dip Lead Locking Design.

reduced by design. The most severe one is due to the inherently large
thermal coefficient of expansion (TeE) mismatch between the plastic and
the metal. In the severe cases that may occur during thermal-shock testing,
bending of the package can reach sufficient proportions to impair the
operation of the device. Such bending stress on the chip surface can be
minimized by locating this surface on the neutral bending axis of the
package. The chip-attach platform is offset downward to accomplish this
effect, as shown in Figure 10-17.
The other stress condition encountered is caused by stress concentra-
10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-433

Neutral Bend
Axis (NBA) Offset

t
Chip Support
Platform
Lead Frame Offset

Punch Radiu
Old Design Plastic

Stress Crack

Chip Support Platform


Section A-A

Improved Design Crack Suppression

Coined Chamfer
Dimpled Platform Eliminates Stress
Locks Plastic to Lead Frame Concentration Point

Figure 10-17. Lead Frame Offset-Crack Suppression.

tion points located on the bottom edges of each stamped 1eadframe, as


depicted in Figure 10-17. When a stamping punch drives through the
sheet metal to form a radius, it tends to pull material with it by friction.
Conversely, on the exit side some metal is extruded, forming a burr. Burrs
are sharp and act as stress concentrators. Once cracking occurs, the cracks
tend to propagate from the edge burrs on the chip-attach platform toward
the bonding figures, forming in the process a water-ingress channel. These
11-434 PLASTIC PACKAGING

internal cracks are insidious since they cannot be detected by conventional


x-ray techniques. Burrs are generally removed by adding a coining station
to the progressive die set. The burr is in effect forged into an edge chamfer.
Plastic Support Matrix: The leadframe functions as a reinforcement
matrix for the molding compound, literally something for the plastic to
hang onto while forming the module body. Top and bottom sections of
a plastic module hold together because the plastic is contiguous between
sections. As the metal area is increased, the force needed to shear top
from bottom decreases to a point where fracture occurs along the metal!
plastic boundary during lead-bending or soldering operations.
The ability to retard water penetration along the metal!plastic inter-
face is a function of plastic shrinkage around the metalleadframe members.
Generally the more plastic coverage, the greater will be the compressive
forces. An ideal plastic-to-metal ratio for leadframe design is given by

[10-1]

where Am = metal area and Ap = plastic area.


Terminal Leads: There are four lead configurations used on molded
plastic packages: integral standoff, gull wing, J-Bend, and butt joint, as
illustrated in Figure 10-18. Dual-in-line modules use integral-standoff
leads. They are designed to give clearance between the module bottom
and the printed circuit board surface for cleaning. Integral standoffs should
be designed to provide a clearance between the module and the board of
0.38 mm minimum when placed in a 1.09 mm maximum diameter plated-
through-hole. Lead configurations for surface mounting require multiple
bends. Therefore, tempers ranging from one quarter to one half hard are
chosen to avoid cracking at the bend radii.

Gull Wing J-Lead Butt Joint

Lead Configurations
Integral Standoff

Figure 10-18. Lead Configurations.


10.2 MOLDING COMPOUNDS AND LEAD FRAME MATERIALS 11-435

10.2.4.6 Lead Frame Fabrication

Lead frames are either chemically milled or mechanically stamped


from rolled strip stock. Typical strip thickness is 0.25 mm with thinner
material (0.20 mm) used on high lead count packages such as the 84-
pin PLCC and quad flat pacs. Chemical milling is a process that uses
photolithography and metal-dissolving chemicals to etch a pattern from
metal strip. The process begins by punching registration holes in the strip,
then coating both sides with a photoresist material. The resist is exposed
to ultraviolet (UV) light through a photo mask, developed, and cured.
The leadframe is now delineated with cured photoresist and is ready for
etching. Chemicals such as ferric chloride or ammonium persulfate are
sprayed onto both sides of the strip to etch away exposed metal-leaving
the leadframe intact. Chemically milled leadframes cost more than stamped
frames, although the corresponding tooling costs are much lower. Further-
more, chemical milling provides a faster turnaround time (typically, one
month compared to 6 to 12 months for stamping). This process is ideal
for packages under development, short production runs, and frames too
complicated for stamping.
Stamped leadframes are fabricated by mechanically removing metal
from strip stock with tools called progressive dies. The energy required
to shear metal is directly proportional to the length of shear. Lead frames
have large shear lengths per unit area. Therefore, a large amount of energy
is required to stamp a full frame with one press stroke. Progressive dies
are usually made of tungsten carbide and are arranged in stations. Each
station punches a small area of metal from the strip as it moves through
the die set. Dies for complex leadframes can have up to 16 stations. This
is shown in Figure 10-19.
Progressive stamping dies are expensive, currently about $140,000
for a 40-pin DIP leadframe, and take 6 to 12 months to design and build.
Trouble shooting the die before introducing it to the production line is
usually required. Fortunately, stamping rates can be rather impressive,
with modem presses able to tum out 900 to 1,000 frames per minute.
Stamped frame costs are a factor of 10 lower than etched parts, so tooling
costs can be amortized rapidly.

10.2.4.7 Plating

Some leadframes in the late sixties were fully plated with gold or
silver. Cost pressures over the years have caused the plating to recede,
until only the chip-attach platform and bonding finger ends remain plated
with silver. Silver spot plating is done with conventional plating baths.
Spot dimensions are controlled by masking and should be specified so
the silver limit is 0.38 mm (minimum) from the outer surface of the
11-436 PLASTIC PACKAGING

SfATION15

SfATION1

Figure 10-19. 40-Pin Dip Lead Frame, 1S-Station Stamping Die Proof. (Cour-
tesy of American Microsystems, Division of Gould Industries.)
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-437

module. Silver may be plated directly onto iron-nickel alloy surfaces, but
a nickel underplate is recommended for copper alloys.

10.3 CHARACTERIZATION OF MOLDING·COMPOUND


PROPERTIES
The molding compound is characterized by a set of properties that
determine the suitability of the molding compound for a given application
and process. From a manufacturing perspective, viscosity and flow charac-
teristics and curing times and temperatures are important factors in deter-
mining which encapsulant materials should be used. From a performance
viewpoint, key characteristics include electrical, mechanical, and thermal
properties to chemical and humidity resistance, the water solubility of
contaminants, and the adhesion to chip-component surfaces. Key proper-
ties include the following:

• Mold filling characteristics, resin bleed, hot hardness, and mold


staning
• Shear viscosity at low shear rates
• Cure time
• Flow resistance
• Glass-transition temperature
• Thermomechanical properties
• Crack Sensitivity
• Hydrolyzable ionic purity
• Moisture solubility and diffusion rates
• Adhesion strength to leadframe materials
• Moisture sensitivity of viscosity and glass-transition temperature
• Alpha-particle emission (if applicable)
• Popcorn resistance for surface-mount technology

Technical information on the above properties and information on


optimum processing conditions for the molding compound is provided
by the mature suppliers. Table 10-11 gives an example characterization
for a typical molding compound. Properties and characterization methods
are given in the following subsections.

10.3.1 Shrinkage Stresses


Shrinkage stresses in a molded plastic package result from polymer-
ization and the disparities in the coefficients of thennal expansion among
the various materials that are contained within the package and are in
11-438 PLASTIC PACKAGING

Table 10-11. Typical Characteristics of Molding-Compound Properties

Coefficient of thermal expansion 16 x lO-6rC


Glass-transition temperature 155°C
Aexural modulus 13,700 MPa
Aexural strength 147 MPa
Thermal stress 4.4 MPa
Thermal conductivity 16 x HI' caVcm soC
Volume resistivity 7 X 1016 n cm
Hydrolyzable ionics <20 ppm
Uranium content 0.4 ppb
Spiral flow length 76 cm (30 in.)
Gel time 23 s
Hot hardness 85 (Shore D)

intimate contact. The following relationship can be used to calculate the


magnitude of the thermal shrinkage stress, cr, as the package goes through
a thermal excursion from T, to T2:

[10-2]

where
a = the temperature-dependent coefficient of thermal expansion
E = the modulus of elasticity
c = a design-dependent geometric constant
the subscript p refers to the molding compound and the SUbscript i refers
to either the semiconductor device or leadframe, depending on the de-
sired calculation.

Because liE; « lIEp and the encapsulant is compliant above the


glass-transition temperature, Tg, the above integral can be approximated by

[10-3]

to provide the stress parameter cr*. The stress, or the stress parameter, is
only a crude approximation of the stress level in the material and does not
account for stress concentration points or other geometric and interfacial
features that influence delamination, bending, or cracking. The thermo-
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-439

mechanical property values needed for calculations are normally provided


by suppliers.
The ASTM F-IOO test is designed to determine the thermal-stress
parameter of Eq. [10-3] for the molding compound. ASTM F-100 is a
photoelastic experiment where the molding compound is molded around
a glass cylinder. The deformation of the glass at the test temperature is
determined by counting the fringe patterns produced by polarized light.

10.3.2 Coefficient of Thermal Expansion and


Glass-Transition Temperature

The coefficient of thermal expansion for any material represents a


change in dimension per unit change in temperature. The dimension can
be either volume, area, or length. However, the rate of expansion varies
from material to material and with the temperature. Therefore, the fact
that different materials expand differently with the same increase in tem-
perature necessitates that elements attached together have the same or
similar coefficients of thermal expansion to avoid the possibility of bend-
ing, cracking, or delamination. The glass-transition temperature is an
inflection point in the expansion-versus-temperature curve above which
the rate of expansion (and therefore the coefficient of thermal expansion)
increases significantly.
The coefficient of thermal expansion (TCE) and the glass-transition
temperature are two properties measured by thermomechanical analysis
(TMA) and reported by most suppliers. The tests are described in ASTM
D-696 or SEMI G 13-82 standards. The ASTM D-696 uses the fused
quartz dilatometer to measure the TCE. The specimen is placed at the
bottom of the outer dilatometer tube with the inner one resting on it. The
measuring device, which is firmly attached to the outer tube, is in contact
with the top of the inner tube and indicates the variations in length of the
specimen with changes in temperature. Temperature changes are brought
about by immersing the outer tube in a liquid bath or other controlled-
temperature environment maintained at the desired temperature. Typically,
a graph of expansion versus temperature is plotted. The coefficient of
thermal expansion is the slope of the plot, and the glass-transition tempera-
ture is the intersection point between the lower-temperature coefficient
of thermal expansion (exl) and the higher-temperature rubbery region
coefficient of thermal expansion (az). The glass-transition temperature
separates the glassy temperature region from the rubbery temperature
region of an amorphous polymer. The glass-transition temperature, being
a manifestation of the total viscoelastic response of a polymer material
to an applied strain, depends on the rate of strain, the degree of strain,
and the heating rate. Because improper molding and postcure conditions
can affect both the coefficient of thermal expansion and the glass-transition
11-440 PLASTIC PACKAGING

temperature, most PEM manufacturers remeasure these parameters on a


predetermined quality control schedule.

10.3.3 Mechanical Properties


It is generally recognized that lowering the stress parameter leads
to improved device reliability and can be accomplished by lowering the
modulus, the coefficient of thermal expansion, or the glass-transition
temperature. The flexural strength and flexural modulus are derived from
the standardized ASTM D-790-71 and ASTM D-732-85 and reported by
suppliers. ASTM D-790 suggests two test procedures to determine the
flexural strength and flexural modulus. The first method suggested is a
three-point loading system utilizing center loading on a simply supported
beam. This procedure is designed principally for materials that break at
comparatively small deflections. In this procedure, the bar rests on two
supports and is loaded by means of a loading nose midway between the
supports. The second procedure involves a four-point loading system,
utilizing two load points equally spaced from their adjacent support points
with a distance of either one-third or one-half of the support span. This
test procedure is designed particularly for large deflections during testing.
In either of the cases, the specimen is deflected until rupture occurs in
the outer fiber. The flexural strength is equal to the maximum stress in
the. outer fiber at the moment of break. It is calculated using

[10-4]

where
S = the flexural strength
Prupture = the load at rupture
I = the support span
bbeam = the width of the beam
Flexural modulus is calculated by drawing a tangent to the steepest initial
straight-line portion of the load deflection curve and is given by

Pm
EB = 4bbeamd 3 [10-5]

where
m = theslope of the tangent to initial straight line portion of the load
deflection curve
Eb = the flexural modulus
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-441

The tensile modulus, tensile strength, and percent elongation are derived
from ASTM D-638 and D2990-77 test methods.
Tensile properties of molding compounds, determined according to
ASTM D-638, uses "dog-bone" shaped molded or cut specimens with
fixed dimensions and held by two grips at the ends. Care is taken to align
the long axis of the specimen and the grips, with an imaginary line joining
the points of attachment of the grips to the machine. They are incrementally
loaded to obtain stress-strain data at any desired temperature. A typical
curve is shown in Figure 10-20. The tensile strength can be calculated
by dividing the maximum load (in N) by the original minimum cross-
section area of the specimen (in m2). The percentage elongation is calcu-
lated by dividing the extension at break by the original gauge length and
this ratio is expressed as a percentage. The modulus of elasticity is obtained
by calculating the slope of the initial linear portion of the stress-strain
curve. If Poisson's ratio for the material is known or separately determined
from tensile strain measurements, the shear modulus of the molding com-
pound can be estimated. It is important to note that the stresses encountered
in PEMs are actually a complex mixture of stresses.
The evaluation of the cracking potential of the molding compound
is particularly important for devices where a relatively small amount of
molding compound surrounds a relatively large die (e.g., memories, SOPs,
and ultrathin packages). In the absence of any standard procedure for
such evaluation, ASTM D-256A and D-256B Izod impact test procedures

L
1150 -----------~----------
'I
I I

920 I" I
,... I
~
I

-e'"
Q,
~ I
690
'"
Vi
460 I
I
I
ElongatioDl Elongation
230 at yield I at break:
I

K Q R! M S!
0.04 0.08 0.12 0.16
Strain (mrnlmm)

Figure 10-20. A Typical Curve of Incrementally Loaded Specimens from


Stress-Strain Data.
11-442 PLASTIC PACKAGING

are commonly followed. The specimen is held as a vertical cantilever


beam in test method ASTM 0-256A and is broken by a single swing of
the pendulum, with the line of initial contact being at a fixed distance
from the specimen clamp and from the centerline of the notch and on the
same face as the notch. A variation of this test is the ASTM 0-256B,
where the specimen is supported as a horizontal simple beam and is
broken by the single swing of the pendulum, with the impact line midway
between the supports and directly opposite the notch. These are overstress
tests for the cracking potential of epoxy molding compounds even under
extreme thermomechanical stress conditions and do not test their important
region of viscoelasticity. However, they are simulative of trim and form,
and handling impact-induced cracking susceptibility.
A fracture test that models the strain history of the package for
thermomechanically induced failure is that previously mentioned ASTM
0-790-71 three-point flexural bending test used to determine flexural
modulus. Here a 0.05-mm (2-mil)-diameter center-notched rectangular
specimen is center strained at a rate that simulates the manufacturing
cycle [i.e., 20%/min for liquid-to-liquid thermal shock and 0.1 %/min for
device on-off operation in air (1°C/min)]. The area under the stress-strain
curve is proportional to the energy to break at the test temperature. The low-
temperature data are usually the discriminating factor between molding
compounds because the molded body experiences the greatest stress at
lower temperatures, far removed from the molding temperature.
Because the melt viscosity of a molding compound is shear-rate
dependent and a typical mold subjects different shear rates at different
points of the molding-compound flow channel, the expected shear rates
for a specific molding tool need first be calculated under no-slip boundary
conditions. The general ranges of experienced shear rates are hundreds
of reciprocal seconds in the runner, thousands through the gate, and tens
in the cavity. Considerations must also be given to the temperature and
time dependence of the shear-rate-dependent molten molding-com-
pound viscosity.
Selection of molding compounds based on shear dependence of
viscosity should identify one that has the lowest viscosity at low shear
rates and high cavity temperatures for wire sweep and/or paddle shift
prone devices and in multicavity molds where complete filling before
gelation is a concern [31]. A material whose viscosity is more temperature
insensitive performs better in all suboptimal tool designs. The time depen-
dency of molten molding-compound viscosity originates from two oppo-
site phenomena. Although the epoxy curing process will increase the
average molecular weight and hence viscosity with time, the decrease of
viscosity with increasing molding temperature in the tool will overwhelm
it in the early phases. However, ultimately, both the molecular weight
and the viscosity go to infinity at gelation. Flow-induced stresses, particu-
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-443

lady in distant cavities, could thus be very significant at the latter stages
of mold filling. Consequently, molds with longer flow lengths and longer
flow times need molding compounds with longer gel times at the 150-
160°C mold filling temperature. This requirement is a trade-off for
higher productivity.

10.3.4 Spiral Flow Length


The ASTM D-3123 or SEMI G 11-88 test consists of flowing molding
compound through a spiral coil of semicircular cross section until the
flow ceases. Although it is not a viscometric test, it is a measure of fusion
under pressure, melt viscosity, and gelation rate. The spiral flow test is
used both to compare different materials and control molding-compound
quality. However, it cannot resolve the viscous and kinetic contributions
to the flow length. Higher viscosity and longer gel time could compensate
each other to provide identical flow lengths. The molding tool used in
this test is shown in Figure 10-21.
A "ram-follower" device, specified in SEMI G 11-88, is used to
separately record ram displacement versus time and, consequently, sepa-
rate the viscosity-induced flow time from the gel time in the total spiral
length formation time of different molding compounds. The molding tool
used in spiral flow test covers the several hundred per second shear-rate
range; thus, the test results have no impact on yield and productivity.
However, as an indicator of gel time or maximum flow time of thermoset
molding compounds, the results of this test can help define mold flow
lengths and flow times compatible with the particular molding tool.

10.3.5 Molding Compound Rheology from Molding Trial


This is a test for rheological compatibility of a molding compound
with a device to be packaged in a molding tool by a trial molding operation.
Rheological incompatibility can cause wire sweep, die paddle shifting,
or incomplete filling of the mold cavity resulting in voids. X-ray analysis
of the molded packages and cross-sectioning of the molded bodies through
the paddle support are the primary methods of evaluation for these trials.
Long wirebond spans (> 2.5 mm) and oversized paddle supports are
normally used to create worst-case scenarios of wire sweep and paddle
shift.
The mold filling characteristics are controlled by the pressure drop
through the gates where the molding compound experiences maximum
shear stress. Incomplete fill problems originate from molding compounds
that have higher viscosities at these high deformation rates in the gates.
This flow closely approximates flow through a sudden contraction or a
converging channel and has both shear and extensional components.
Statistically confident sampling of all stochastic phenomena, such
11-444 PLASTIC PACKAGING

Front

42.7 42.7
(1.68) ( 1.68)

~-- 69.8 ---i'"


(2.75)

,...o ( l f - - - - - - 139.7
(5jO)
------+1

Figure 10-21. Molding Tool Used for the Spiral Flow Length Test from ASTM
D-3123

as gate clogging by gel or filler particles, requires a large number of molded


packages. The analysis of mold trial results should consider molding-
compound density and the consequent package-sectioning analysis for
porosity appraisal. Figure 10-22 shows a typical !!lold map displaying a
variety of runners.
Incomplete filling of the mold (because of low packing pressure) in
just a few cavities effectively ruins all the devices loaded because it leads
to high porosity in the encapsulant that would lead to excessive moisture
penetration. It is more common in molding tools with a large number of
cavities (> 150) and with packages of large volume such as PQFPs and
large chip carriers or with package designs with four-sided leads requiring
comer gating.
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-445

Figure 10-22. Molding Mapping Showing the Runners

Resin bleed and flash are molding problems where molding com-
pound oozes out of the cavity and onto the leadframe at the parting line
of the mold. Whereas flash is caused by the escape of the entire molding
compound, resin bleed includes only the strained-out resin. Although root
causes of these two problems can be traced both to the processing condi-
tions of molding, and mold design and mold defects, resin bleed is consid-
ered to be more molding-compound related. Resin bleed occurs more
often with formulations containing low-viscosity resin and large filler
particles, and with process conditions where excessive packaging pressure
is maintained after cavity fills and the use of too Iowa clamping pressure
on the mold halves.
SEMI G 45-88 is a standardized test for assessing a material's
potential for resin bleed and flash. It is a transfer-molding experiment
that measures the flow of molding compound in a shallow-channel mold
(6-75 ~m) and simulates flash and bleed in production tools. Propensity
of resin bleed and flash from improper molding-compound properties is
indicated by long spiral flow lengths obtained in that test.
Moisture has a profound effect of decreasing the viscosity of epoxy
molding compound, and the degree of this effect is a function of the
additives and curing agents used in different formulations [32]. Compared
to dry conditions, this decrease of molten molding-compound viscosity
can be 40% or more at -0.2 wt. % water or higher.
The effect of moisture on the shear-thinning behavior, shown in
Figure 10-23 indicates that viscosity is simply lowered with little change
in shear-rate dependence and power-law index. Not withstanding the fact
that moisture-induced melt viscosity lowering is beneficial in overcoming
flow-stress and mold filling problems, excessive moisture content can
cause excessive resin bleed and voids. Consequently, moisture uptake
sensitivity, determined by the shear-rate-dependent viscosity test, is a
necessary factor in the selection of a molding compound.
11-446 PLASTIC PACKAGING

Isothermal: 125 ·C
......
U
en Time: 30 sec
]; 10,000

i-
.~ 1000
:>

y
Shear rate, (l/sec)

Figure 10-23. EtTect of Moisture on the Shear-Thinning Behavior. (From


Ref. 33.)

10.3.6 Characterization of Molding-Compound Hardening


Hardening of a molding compound is the progressive thermal poly-
merization of the matrix resin that leads to gelation in the latter stages
of the molding process, an ejectable state of hardness during cure, and
complete chemical conversion during postcure. The productivity of plastic
package molding depends on the rate of this chemical conversion.
Mold filling can occur in as little as 10 s at l50-l60°C (specified
by the supplier), and the cure time required before the parts can be ejected
from the mold can range from 1 to 4 min. The cure time is about 70%
of the molding cycle time. Shorter cure times will generally have shorter
flow times into mold cavities before gelation. Multiplunger machines are
designed to handle these short flow times and cure times to provide high
molding productivity. Most molding tools require molding compounds
that flow for 20-30 s and then cure to an ejectable state in less than one
additional minute. Standard evaluation tests for these characteristics of
molding compounds are discussed below.

10.3.7 Gel Time


The gel time of a thermoset molding compound is usually measured
with a gel plate. In gel-time evaluation with a gel plate, a small amount
of the molding-compound powder is softened to a thick fluid on a precisely
controlled hot plate (usually set at 170°C) and periodically probed to
determine gelation. The gel time is a qualitative point in the process where
the material cannot be smeared into a thin coating. The SEMI G 11-88
standard recommends using the spiral flow test as a comparative evaluator.
Gel times indicate the productivity of a molding compound. Shorter gel
times lead to faster polymerizations rates and shorter times for mold cycle,
increasing production.
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-447

10.3.8 Adhesion

Poor adhesion of a molding compound to the die, die paddle, and


leadframe can lead to such failures as "popcoming" during assembly,
corrosion, stress concentrations, and subsequent thermomechanical failure,
package cracking, chip cracking, and chip metallization deformation. Con-
sequently, adhesion is one of the most important discriminating properties
of a molding compound to be chosen for a particular physical and materials
design of a package. The theory and practice of adhesion of integrated-
circuit molding compounds to package elements have been treated by
Kim [33] and Nishimura et al. [34]
A common adhesion test for leadframes to the molding compound
is to mold plugs of the molding compound on the properly surface-treated
leadframe tab, and then determine adhesion strength in a tensile tester.
The molding process used in this test must be identical to that used
in production. For maximum simulation of manufacturing conditions, a
custom-designed leadframe is used in a production mold to generate the
adhesion test specimens. Determination of adhesion strength to other
package materials such as silicon die, polyimide, and silicone overcoat
is accomplished by molding the compound on the flat surface of the
material and measuring the force required to pull them apart. Needless
to say, very careful interpretation of the adhesion test data is required to
differentiate among most state-of-the-art molding compounds.

10.3.9 Polymerization Rate

The novolac epoxy polymerization reaction includes several compet-


ing reactions among three or four reactive species. The chain segments
that form are complicated and difficult to predict. Thus, thermal analysis
methods, which assume that the fraction of the total heat of reaction
liberated is proportional to the fraction of complete chemical conversion,
are preferred for these types of highly filled opaque systems. Several
different empirical forms have been offered to fit conversion data for the
epoxy molding compounds. They do not reflect the molecular dynamics
of the reaction but are, instead, phenomenological in that they assess the
engineering behavior of the reaction without a theoretical basis for the
reaction mechanism or reaction order. Hale et al. [35-37] developed one
of the most noteworthy forms:

[10-6]

where the four fitting parameters for the conversion of epoxide groups,
X, as a function of reaction time are as follows: m" and n, are the pseudo-
11-448 PLASTIC PACKAGING

reaction orders and krl and krZ are rate constants. For a typical epoxy
molding compound, mr = 3.33, nr = 7.88, krl = exp(12.672 - 75601T), and
kr2 = exp(21.835 - 86951T) [36].
Figure 10-24 shows isothermal fractional conversion of epoxide
groups with a drop-off in reaction rate near complete conversion. These
conversion constants differ from one molding compound to the other and
thus form the basis of evaluation of polymerization rate of the compound
in question.
Differential scanning calorimetry (DSC) has been used to obtain the
degree of polymerization of filled molding compounds [37]. Measuring
the heat of reaction versus time during an isothermal cure, it expresses
the fractional conversion as a function of time as equal to the fractional
total liberation of heat:

Mlt=IJ =~ [10-7]
Mltotal 100

An extensive analysis of the polymerization kinetics is generally not


required for material selection. The secondary effects of cure kinetics
such as gel time, mechanical properties, and glass-transition temperature
are sufficient to effectively compare different molding compounds.

10.3.10 Hardening
Hot hardness or green strength is the stiffness of the material at the
end of the cure cycle. A certain degree of this hot hardness is required
before the molded strip of parts can be ejected safely from specific molds.

1.0 r-----------------,

0.8
c:
o
.~ 0.6
c:
o Cure
U 0.4

c 140·C
0.2
6 120·C

o 100·C

-I 012 3 4
Log time (min.)

Figure 10-24. A Plot of Conversion Versus Time for an Epoxy Molding Com-
pound During Cure. (From Ref. 35.)
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-449

Ejection of the strip from the molding tool is also dependent on the
characteristics of the mold. These mold characteristics are the draft angle
of the vertical surfaces, the surface finish of the tool, and the number and
size of ejector pins. Different molding compounds attain this green strength
at different points of cure cycle due to either percentage conversion
achieved or low modulus above glass-transition temperature. It is, thus,
a productivity issue and can either be determined by molding trial or
supplied by the vendor. A hot hardness value of about 80 on the Shore
D scale within lOs of opening the mold is considered acceptable.

10.3.11 Postcure
Most epoxy molding compounds require about 4 h of postcure at
170-175°C for complete cure.

10.3.12 Moisture Effects on the Epoxy Hardening Process


Moisture content of molding compounds has a strong impact on the
viscosity, void density, polymerization kinetics, and the properties of the
cured material. The lowering of the glass-transition temperature by the
use of moist molding compounds indicates that the final inalterable cross-
linked network is different under humid conditions. The effect of moisture
on the polymerization rate originates from the fact that moisture binding
to catalyst sites diminishes its effectiveness. Because of the differences
in the formulation of different molding compounds, the effect of moisture
content could vary widely and needs to be evaluated from the viewpoints
of required encapsulant properties.

10.3.13 Thermal Conductivity


Thermal conductivity is an important property of an encapsulant
used for high-heat-dissipating devices or for devices with long duty cycles.
Although they are measured with standard thermal-conductance testers,
the reported values vary widely, depending on the instrument and the
precision used.

10.3.14 Electrical Properties


The state-of-the-art very large-scale integration (VLSI) parts require
close control of several electrical properties of the molding compound or
the cured encapsulant for superior performance. They include dielectric
constant and dissipation factor (ASTM D-150), volume resistivity (ASTM
D-257), and dielectric strength (ASTM D-149). ASTM D-257 suggests
various electrode systems to determine the volume resistivity by measuring
the resistance of the material specimen by a measurement of the voltage
or current drop under specified conditions, and the specimen and electrode
11-450 PLASTIC PACKAGING

dimensions. The test specimen may be in the form of flat plates, tapes,
or tubes. Figure 10-25 shows the application and electrode arrangement
for a flat specimen. The circular geometry shown in the figure is not
necessary, although convenient. The actual points of measurements should
be uniformly distributed over the area covered by the measuring electrodes.
The dimensions of the electrodes, the width of the electrode gap, and
the resistance are measured with a suitable device having the required
sensitivity and accuracy. The time of electrification is normally 60 sand
the applied voltage is 500 ± 5 V. The volume resistivity is given by

[10-8]

where
Aelee = the effective area of the measuring electrode
Rv = the measured volume resistance
t = the average thickness of the specimen

ASTM D-149 requires that alternating voltage at a commercial power


frequency, normally 60 Hz, be applied to a test specimen. The voltage
is increased from zero, or from a level well below the breakdown voltage,
until dielectric failure of the test specimen occurs. The test voltage is
applied using simple test electrodes on opposite faces of specimens. The
specimens may be molded, cast, or cut from a flat sheet or plate. Methods
of applying voltage include a short-time test, a step-by-step test, and a

Electr"Cie No.1
-fgj.

+b=~~:!!L::::::::
:

Figure 10-2S. Electrode Arrangement for Measurement of Volume Resistance


for a Flat Specimen (ASTM D-257 1994)
10.3 CHARACTERIZATION OF MOLDING-COMPOUND PROPERTIES 11-451

slow rate-of-rise test. The second and third methods usually give conserva-
tive results.
Epoxy composites in dry environments and at room temperatures
have similar electrical properties. Deterioration of some materials may
occur after being stored in a moist environment at high temperatures.

10.3.15 Chemical Properties


The contamination level of a molding compound ultimately deter-
mines the long-term reliability of the parts fabricated with it and used
under rugged conditions. The SEMI G 29 standard procedure is used to
determine water-soluble ionic level in epoxy molding compounds. A water
extract is first tested for electrical conductivity and then quantitatively
analyzed by column chromatography. A separate determination of hydro-
lyzable halides (from the resin, flame retardants, and other impure addi-
tives) is particularly crucial in assuring long-term reliability of PEMs.
Long-term (48 h), high-pressure, and sometimes hot-water (up to lOODC)
extraction from the molding compound and subsequent elemental analysis
is needed for such evaluation. Modem molding-compound formulations
contain as little as lO ppm of these corrosion-inducing ionics. Atomic
absorption spectroscopy and x-ray fluorescence techniques are used to
determine the content of other undesirable contaminants such as sodium,
potassium, tin, and iron. Encapsulants used for memory devices, where
single-event upsets from alpha-emitting impurities in the filler silica must
be minimized, require determination of uranium and thorium content in
the molding compound.

10.3.16 Flammability and Oxygen Index


Encapsulation compounds and plastic-encapsulated parts must con-
form to Underwriters Laboratory flammability ratings (UL 94 V-0, UL
94 V-I, or UL 94 V-2). Molding compounds are evaluated for flammability
by UL 94 vertical bum and ASTM D-2836 oxygen index tests. In the
UL 94 test, a 127-rrun x 12.7-mm (5-in. x V2-in.) cured epoxy test bar
of a predetermined thickness is ignited multiple times in a gas flame, and
the bum time per ignition, total bum time for lO ignitions (5 specimens),
and extent of burning are recorded for proper UL rating. In the oxygen
index test, the minimum volume fraction of oxygen in an oxygen-
hydrogen mixture that will sustain burning of a 0.6 x 0.3 x 8-cm molded
bar of epoxy molding compound is specified.

10.3.17 Moisture Ingress


All plastic packages prefer low moisture diffusivity for high reliabil-
ity and minimum soldering heat damage. At room temperature, epoxy
11-452 PLASTIC PACKAGING

molding compounds absorb water to approximately 0.5% by weight on


long-term exposure to 100% relative humidity [38]. Moisture absorption
rate to saturation is usually determined by weighing molded dry parts
exposed to a given relative humidity and temperature for some extended
periods [39]. Desorption rates of moisture-saturated parts are similarly
determined by drying at an elevated temperature. The selection of an
epoxy molding compound from moisture ingress evaluation is a very
recent phenomenon and is currently quite subjective even for a specific
part and its application environment.

10.4 THE TRANSFER-MOLDING PROCESS

The transfer-molding process is the most popular integrated-circuit


encapsulation method for essentially all plastic packages. Transfer molding
is a process of forming components in a closed mold from a thermosetting
material that is conveyed under pressure, in a hot, plastic state, from an
auxiliary chamber, called the transfer pot, through runners and gates into
the closed cavity or cavities. Figure 10-26 depicts the transfer molding
press.
One of the main limitations of the transfer-molding process is the
loss of material. The material left in the pot or well and also in the sprue
and runner is completely polymerized and must be discarded. For small
parts, this can represent a sizable percentage of the weight of the parts
molded. As a result, the trend today's to utilize fully automatic transfer-
molding presses which mold a few parts very efficiently. Another limita-
tion of today's transfer-molding process is the need to use standard metal
leadframes. Extension of the process to include package designs such as
TAB is still not clear.

Hydraulic ram
[t TranSfer
plunger
MI.
o dmg pellet

Heaters~~rJgs
---...../
Ejector pins
"-

Figure 10-26. Transfer Molding Press.


10.4 THE TRANSFER-MOLDING PROCESS 1\-453

10.4.1 Molding Equipment


Transfer-molding requires four key pieces of capital equipment: a
preheater, a press, the die mold, and a cure oven. The transfer-molding
press is normally hydraulically operated. Auxiliary ram-type transfer
molds are commonly used in transfer molding. This mold has a built-in
transfer pot in which the molding compound is placed. Both the volume
and the size of the molding-compound preforms have to be appropriately
selected for the press capacity. The mold is then clamped using the
clamping pressure. The transfer plunger is activated to apply the transfer
pressure to the molding compound. The molding compound is driven
through the runners and gates into the cavities.
Since the 1980s, aperture-plate and multiplunger molds have been
the dominant approaches to PEM molding. Table 10-12 compares the
features of these molding methods. Aperture-plate molds are a patented
transfer-molding technology (U.S. Patent No. 4,332,537, 1 June 1982)
exclusively developed for PEMs. An aperture-plate mold is constructed
by assembling a series of stacked plates. The leadframes form the aperture

Table 10-12. Comparison of Molding Tools

Feature Aperture-plate molds Multiplunger molds

Number of cavities Several thousands possible From 2 to 100


Flexibility of package types Extremely flexible Relatively inflexible
Process setup and control Moderate Straight forward for
automated tools
Flow-induced stress Minimum Intermediate
problems
Flash and bleed Minimum Average
Molding compound waste 20-40 10-25
(%)
Package yield per cycle High Very high
Ejection External ejection Ejector pins in each cavity
Flow time and cavity Good Excellent
material uniformity
Packing pressure (MPa) 1.4-2.8 Variable on number of
cavities fed from a pot
(1.4-4.1)
Temperature profile and Requires care Excellent
stability
Automation susceptibility Intermediate Very high
Capital cost Low High (lower for small tools)
Labor cost High Low
Maintenance cost Medium High, due to automation
11-454 PLASTIC PACKAGING

plates. The top and bottom of the body are formed by separate plates.
The bottom body-forming plate contains the runner system, and the top
plate is finished for either laser or ink marking. The gates are positioned
between the runners, parallel to the bottom of the body; the aperture-plate
cavities can be formed anywhere along this intersection and their width
can be any fraction of this length of intersection. This flexibility of gate
positioning, along with the much lower pressure drop across the gate in
an aperture plate, results in negligible wire sweep and paddle shift during
molding. These molds are also highly adaptable for different package
types and pinouts.
Multiplunger molds, also called gang-pot molds, have a number of
transfer plungers, feeding one to four cavities from each transfer pot.
They are highly automated and can be easily set and optimized for a new
molding compound.
Figure 10-27 shows a multiplunger mold used for simultaneous
encapsulation of DIPs and quad flat packs. Figure 10-28 shows a multi-
plunger mold with each pot feeding just one cavity.
In general, the mold consists of two halves: the top and bottom. The
mating surface of these two halves is called the parting line. Platens are
massive blocks of steel used to bolt the two halves to the molding press.
Guide pins ensure proper movement of the two halves. The ejector pins
aides the ejection of the component after the mold has opened. Gates are
located where they can be easily removed and buffed if necessary. Properly
designed gates should allow proper flow of material as it enters the mold
cavity. Gates should be located at points away from the functioning parts
of the molded component. Vents are provided in all transfer molds to
facilitate the escape of trapped air. The location of these vents depend
on the part design and the locations of pins and inserts. The vent is
sufficiently small so that it allows the air to pass through, but a negligible
amount of molding compound can pass through it.

10.4.2 Transfer-Molding Process

Figure 10-29 shows the various stages of a typical transfer-molding


process. In this process, leadframes are loaded (6-12 in a row) in the
bottom half of the mold. For both plate and cavity-chase molds, this is
done at a workstation separate from the molding press. A cavity-chase
mold uses a loading fixture; most molding operations have automated
leadframe loaders. Both the moving platen and the transfer plunger initially
close rapidly, but the speed reduces as they close. After the mold is closed
and clamping pressure is applied, the preform of molding material (which
has usually been preheated to 90-95°C, which is below the transfer
temperature) is placed in the pot, and the transfer plunger or ram is
activated. Preheating of the molding compound is done by a high-fre-
10.4 THE TRANSFER·MOLDING PROCESS 11-455

Figure 10-27. Multiplunger Mold Used for Simultaneous Encapsulation of


DIPs and Quad Flat Packs. (Courtesy of National Semiconductor, 1994.)

quency electronic method that works on a principle similar to microwave


heating. The transfer plunger then applies the transfer pressure, forcing
the molding compound through the runners and gates into the cavities.
This pressure is maintained for a certain optimum time, ensuring proper
filling of the cavities. The mold then opens, first slowly. This step is
11-456 PLASTIC PACKAGING

Figure 10-28. Multiplunger Mold with Each Pot Feeding Just One Cavity.
(Courtesy of National Semiconductor, \994.)
10.4 THE TRANSFER-MOLDING PROCESS 11-457

Figure 10·29. Different Parts of a Transfer Mold. (Courtesy of National Semicon-


ductor, 1994.)

known as the slow breakaway. Sometimes it is desirable to have the


transfer plunger move forward so that it pushes out the cull, or the material
remaining in the pot. Finally the component is ejected using the ejector
system in the mold. In an aperture-plate mold, the plates themselves
are loaded with the leadframe strips, as they shuttle in and out of the
molding press.
The parameters controlled by the press include the temperatures of
the pot and the mold, the transfer pressure, the mold-clamping pressure,
and the transfer time to fill the mold cavities completely. The mold
temperature should be high enough to ensure rapid curing of the part.
However, precautions should be taken to control the mold temperature,
/1-458 PLASTIC PACKAGING

because too high a mold temperature may result in "precure" or solidifica-


tion of the molding compound before it reaches the cavities. Electric
heating is the most commonly used method for heating the molds. Multiple
electric heating cartridges are inserted both in the top and bottom halves
of the molds; positioned to supply heat to all the cavities. The applied
pressure ensures the flow of material into all parts of the cavity.
Clamping pressure applied on the mold ensures the closure of the
mold during polymerization or cure and also against the force of the
material entering the cavities. Packing pressure, usually higher than the
transfer pressure, is applied once the mold has been filled. While it is
being transferred into the mold, the molding compound is reacting. Conse-
quently, the viscosity of the material increases, at first gradually. As the
reacting molecules become larger, the viscosity increases more rapidly
until gelation occurs, at which point the material is a highly cross-
linked network.
The pressure applied by the transfer plunger is critical. It should be
sufficient to force the material through the runners and gates into the
cavity and hold the material until polymerization. For high throughput,
it is desirable to transfer and react the material as rapidly as possible, but
decreasing the transfer time requires an increase in the transfer pressure
to fill the mold. The high transfer pressures, typically as high as 170 MPa
(25 ksi), can cause damage within the package, such as wire-sweep short
or, in extreme cases, wirebond lift-off, shear, or fracture. These problems
can be avoided by careful design of the package size and the transfer-
gate size to minimize shear rate and flow stresses during cavity filling.
In general, factors that tend to exacerbate these problems include high
wire-loop heights, long wirebonds, bond orientation perpendicular to the
advancing polymer flow front, rapid transfer times with corresponding
high transfer pressures, high-viscosity molding components, and low-
elastic-modulus wire.
An important way to decrease flow-induced stresses and improve
molding yield is by velocity reduction. Velocity reduction in epoxy mold-
ing compounds reaches its limit before the need for flow-induced stress
reduction is satisfied. In constant-pressure programmed transfer-rate con-
trol, high flow-induced stresses are created in the cavities nearest to and
farthest from the transfer pot, whereas the middle cavities experience the
lowest velocities and stresses. Mold mapping of flow-induced defects
shows lowest yield in extreme-positioned cavities. A transfer-rate profile
that is slow at the start, then higher over the middle cavities, and slow
again when the last cavities are filling helps in reducing the flow-induced
stresses. Different mold designs and molding compounds require distinc-
tive transfer profiles. Velocity-reducing tool changes are permanent for
a particular molding compound and package design. Mold designs aim
at balanced mold filling by maintaining nearly uniform pressure fronts
10.4 THE TRANSFER-MOLDING PROCESS 11-459

through all segments of the mold; in these, channel cross-sectional areas are
controlled for volume-flow and pressure-drop uniformities in all cavities.
However, in all cases, a velocity surge occurs when the molding machine
switches from the transfer pressure to the packing pressure. These transient
high velocities, along with the rapid compression of any remaining voids,
can cause wire sweep. Using a programmable pressure controller to profile
this pressure transition is the best approach to minimizing this problem.
After 1-3 min at the typical molding temperature of 175°C, the
polymer is cured in the mold. Following curing, the mold is opened, and
ejector pins remove the parts. The molded packages are ready for ejection
when the material is resilient and hard enough to withstand the ejection
forces without significant permanent deformation. After ejection, the
molded leadframe strips are loaded into magazines, which are postcured
in a batch (4-16 h at 175°C) to complete the cure of the encapsulant.
Postcure normally involves holding the part in an oven at a temperature
somewhat lower than the mold temperature but well above the room
temperature for several hours. In some instances, postcuring is performed
after code marking to eliminate an additional heat-cure cycle. The most
important consideration in postmold cure analysis is the development of
thermomechanical properties.

10.4.3 Simulation

The transfer-molding process is more complicated and difficult to


treat analytically than either thermoplastic extrusion or injection molding,
because of the time-dependent behavior of the molding compound, the
irregular cross sections of the runners, and the common presence of inserts
in the cavities. However, once a good quantitative model of the transfer-
molding process has been proven, substantial time and cost can be saved.
Long and expensive experimental runs no longer need to be carried out
by trial and error to debug a mold or qualify a new compound.
Modeling the dynamics of transfer molding have met with moderate
success [31,40]. The approach typically involves formulating a chemor-
heological description of the epoxy molding compound and coupling it
with a network flow model. The gating factor for a successful simulation
is a good characterization of the compound to determine its kinetic and
rheological behavior. Most often, the curing reaction of the epoxy can be
described by an autocalytic expression. The dynamic viscosity, as mea-
sured generally through a plate and cone viscometer, is reasonably well
covered by the Castro-Macosko model [40]. By incorporating this material
information into a flow model that includes the geometrical intricacies of
a multicavity mold, an understanding of the filling characteristics of a
particular compound can be obtained.
With a rheological model in hand a variety of scenarios can be
11-460 PLASTIC PACKAGING

simulated to optimize processing conditions and mold design [42,43]. For


instance, different temperatures, pressures, and packing settings can be
estimated to provide optimized filling profiles. To reduce wire sweep, the
dominant-yield loss concern for fine-pitch packages, an analytical model
of wire deformation can be coupled with rheological and kinetic informa-
tion to improve process conditions and package layout design rules [42].
Once a mold has been translated into its geometrical equivalent, various
combinations of runner sizes, gate locations, gate dimensions, and cavity
layout can be evaluated.

10.5 PACKING AND HANDLING

Handling costs for moisture-sensitive plastic-encapsulated microcir-


cuits (PEMs) may constitute as much as 10% of the total cost. Poor
handling can lead to loss of product yield during assembly of plastic-
encapsulated microcircuits on printed wiring boards. Package cracking
during surface mounting is directly associated with improper packing and!
or handling of components, but it can be avoided.
Packing and handling involves materials, containers, and precautions
that ensure that the quality and reliability of plastic-encapsulated microcir-
cuits are not degraded during shipping, handling, and storage. This neces-
sary step between the manufacture of components and their mounting on
printed wiring boards does not add value to the components. However,
poor packing and handling can have a dramatically negative impact on
board assembly yield, manufacturer's cost, and shipping schedules.
Plastic-encapsulated microcircuits must be adequately protected dur-
ing shipping, handling, and storage. Precautions taken in packing can
provide the requisite protection against high-humidity conditions (which
can lead to moisture ingress), against physical damage (which can cause
bent or broken leads), and against damage due to electrostatic discharge.
Automatic handling of packages is facilitated by providing special features
in the packing.
Although all the reasons for special handling procedures are dis-
cussed in this chapter, the focus is on issues unique to plastic packages
such as dry packing, providing a dry nitrogen environment, and baking
before mounting packages on circuit cards.

10.5.1 Considerations in Packing and Handling

Considerations in the packing of plastic-encapsulated microcircuits


include prevention of moisture absorption, protection of leads from hand-
ling damage, prevention of lead solderability degradation, precautions
against electrostatic damage, and packing for automatic handling.
10.5 PACKING AND HANDLING 11-461

10.5.1.1 Prevention of Moisture


Because plastic-encapsulated microcircuits are nonhermetic, they
are susceptible to moisture-induced failures and latent defects. Molding
compounds can absorb enough moisture to cause cracking in moisture-
sensitive packages when exposed to temperatures typical of furnace-reflow
soldering and wave soldering. The phenomenon is known as popcorning
because the pop of the packages is audible. High-temperature and tempera-
ture ramp rates, prime loads that cause popcorning [41], are experienced
by the packages during surface mounting on circuit cards. Of particular
concern are solder-reflow processes that heat the package body, such as
infrared heating, hot-air furnace, and vapor-phase reflow. Solder-reflow
processes that selectively heat the leads and bond pads but not the compo-
nent body, such as hot-bar reflow, laser soldering, and manual soldering,
are not problematic. Wave soldering may be a concern if components dip
through the solder wave.
Through-hole mounted plastic-encapsulated microcircuits do not ex-
hibit popcorning because the component body is protected from the solder
heat by the board. However, popcorning may occur in through-hole com-
ponents mounted on double-sided boards if the components come in direct
contact with the solder wave.
Possible damage due to popcorning includes delamination of the
encapsulant from adjacent metal structures such as the leadframe and die
pad, damaged wirebonds, cratering of bond pads, and cracking of the
encapsulant. Even if the plastic-encapsulated microcircuit continues to
perform its electrical function after popcorning, its reliability is compro-
mised because the cracked and delaminated encapsulant no longer provides
protection. Cracks may also let contaminants reach the active circuit via
capillary action.
Popcoming in moisture-sensitive plastic packages can be avoided by
controlling moisture content with proper packing and handling procedures.
The moisture content of a plastic-encapsulated microcircuit is defined as

. (wet weight - Dry Weight)


MOIsture content (% wt.) = D. h x 100%
ry welg t
[10-9]

Typical moisture absorption curves after a high-temperature bake are


shown in Figure 10-30 as a function of temperature and relative humidity.
The figure illustrates that temperature is the parameter determining the
moisture absorption rate. The saturation limit is a function of both relative
humidity and temperature, although relative humidity is the major factor
determining the steady-state moisture content of the component. Weight
gain by small-outline packages as a function of exposure time in 85°CI
11-462 PLASTIC PACKAGING

0.4...--------------=------,
85 CC/85%RH

0.3

.....,
~ 85 CC/60% RH
c
0.2

-
'OJ
co
85:C/25% RH
fa
'u 4O CC/85% RH
~
0.1 40 0 C/4O% RH
4O°C/25% RH

0.0
0 100 200 300 400
Time (hours)

Figure 10-30. Typical Moisture Absorption Curves, After High-Temperature


Bake, Shown as a Function of Temperature and Relative Humidity. (From
Ref. 42.)

85% rh and 85°C/30% rh environments is plotted in Figures 1O-31a and


1O-31b, respectively [43]. The plots show that the mold compound chosen
has a significant effect on the amount of moisture absorbed by plastic-
encapsulated microcircuits.
For a given package type, there may be safe moisture content which
the package may not popcorn. The critical value at which the component
exhibits cracking depends on the adhesion and interlocking between the
encapsulant and metal surfaces, and on package size, architecture, and
lead count. Generally, large and thin plastic packages are more susceptible
to popcorning; larger packages generate higher stresses at the corner of
the die pad, and thinner packages offer a shorter path for moisture ingress
through the molding compound, leading to higher moisture concentration
at the interface of encapsulant and metal leadframe.
Dry packing must be used for shipping moisture-sensitive PEMs so
they do not absorb more than the safe moisture content. The dry-pack
process reduces the saturation moisture content in packages by keeping
the relative humidity inside the moisture barrier bag below 20%. Other
methods, such as storing in dry nitrogen and baking, may also be employed.
Three methods can be employed to dry components that have ab-
sorbed moisture: high-temperature bake, low-temperature bake, and stor-
age in dry nitrogen. Generally, a package with a moisture level at or
below 0.05% by weight is considered dry enough for dry packing. Compo-
10.5 PACKING AND HANDLING 11-463

(a) 0.40 . . . . _ - - - - - - - - - - - - - - - . . . ,
Molding compound
B

""":' 0.30 A
~
<f?
'-" c
c
.:0 0.20 D

0.\0
32 leaded small outline package
85 °Cl85% RH

0.00 -+-.......-..--,..........,,..-,.--.----r......---r--r--r-..--....-..---T--.---.--I
o 20 40 60 80 100 120 140 160 180
Time (bours)

(b) 0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02 32 leaded small outline package
0.01 85 0 Cl30%RH

0.00
0 20 40 60 80 100 120 140 160 180
Time (bours)

Figure 10-31. Weight Gain by Small-Outline Packages as a Function of Expo-


sure Time in 85°C/85% rh and 85°C/30% rh environments. (From Ref. 43.)

nents are dried to below the safe moisture content to provide the desired
time for shipping and handling in a humid environment before the compo-
nents are mounted on printed wiring boards.

10.5.1.2 Lead Damage and Solderability


Manual handling, together with drying methods, shock, and vibra-
tion, can cause damage to leads. Shipping containers are designed for
11-464 PLASTIC PACKAGING

automatic handling and for preventing lead bend, loss of coplanarity, and
lead contamination.
Mishandling may occur at various process points, such as trim and
form operations, loading of parts into the tester for electrical sorting, or
removal of parts after burn-in. Lead deformation can also be induced by
such process factors as high molding stresses that result in warpage, or
the buckling of thin leadframe strips prior to singulation [44]. Regardless
of the cause, damage to fragile leads can be extensive. Six typical defect
modes have been characterized: coplanarity, sweep, tweeze, twist, body
standoff, and center-to-center spacing [45]. Coplanarity is the vertical
deviation of a lead from the seating plane of the device and is the single
most important parameter in assembly, packing, and handling.
Packing materials that include shipping trays, reels, tapes, and maga-
zines are made of plastic materials, are often unstable at bake temperature,
and outgas products that may condense on package leads, degrading their
solderability. Outgassing of hydrocarbons and other organic contaminants
from low-temperature stable packing materials is especially of concern
during high-temperature bake. The container materials must be stable
enough at bake temperature that any outgassing products do not degrade
solderability. If solderability is a major concern, then the packages may
be baked outside of the shipping containers.
Intermetallic growth is a time- and temperature-dependent diffusion
process. Long exposure to high temperatures during baking can form
excessive intermetallics and degrade solderability. Often, leads are made
of copper, which forms a CU6SnS intermetallic with tin (Sn). The growth
of CU6SnS is one of the causes of lead-finish degradation leading to poor
solderability [46]. Limiting the time of exposure, especially at high bake
temperatures, is therefore crucial for package lead solderability. If the
bake is repeated, the total time of exposure to high temperatures should
be limited. Moreover, oxidation of lead surfaces can cause degradation of
lead solderability; inert gas flow and a low relative humidity environment is
preferred during bake to alleviate this problem.

10.5.1.3 Electrostatic Discharge Protection

Plastic-encapsulated microcircuits may be sensitive to electrostatic


discharge (ESD). Electrostatic discharge can either cause overstress failure
of the component (which can be exposed by an electrical function test)
or induce damage leading to parametric shift and reduced reliability.
Electrostatic discharge occurs most readily in low-humidity environ-
ments. The low humidity levels inherent in baking increase the risk of
ESD damage. Air ionizers may be used in baking ovens to keep devices
from charging, but they cannot dissipate the charge in areas where the
ionized air cannot reach because of the flow pattern.
10.5 PACKING AND HANDLING 11-465

The probability of electrostatic discharge occurring increases with


increased handling, which may lead to tribologically produced charges.
Charge-dissipative or electrically conductive shipping containers are em-
ployed to protect against electrostatic discharge. Charge-dissipative con-
tainers are made of electrically conductive materials, such as resin impreg-
nated with conducting graphite fibers. Often, an antistatic coating of carbon
is applied on the inner walls of shipping containers; the coating is water
soluble and wears out due to friction as components slide around.
Evaluating ESD-preventive characteristics for both graphite-impreg-
nated and antistatic-coated tubes indicates that either type of ESD protec-
tion is sufficient, even for low-humidity environments. There is no signifi-
cant observed difference in component charging for graphite-impregnated
tubes stored at ambient humidity and those stored at < 10% rho Although
there is a difference in charging of components stored in antistatic-coated
tubes in < 10% rh, the jeopardy is considered low because the charging
is 2.2 picocoulomb (pC)/lead-well below the critical level of 10 pCI
lead [47].

10.5.2 Hierarchy of Packing


Dry packing is required to keep moisture-sensitive plastic surface-
mount components dry during shipping and storage. There are generally
three levels of packing for plastic-encapsulated microcircuits, as shown
in Figure 10-32. The first level involves packing components in suitable
shipping containers. The second level involves packing the containers
along with desiccant pouches, a humidity-indicator card, and dunnage in
a moisture-barrier bag; the third level labels the shipping box with appro-
priate handling and shelf-life information.

10.5.2.1 First Level Packing


The first level of packing requires placing plastic-encapsulated mi-
crocircuits in magazines, trays, or tape reels. ESD-preventive characteris-
tics are built into the containers, either by making the container material
electrically conductive or by coating with an antistatic agent.
Magazines are made of plastic or aluminum. Plastic magazines are
treated with a water-soluble antistatic agent that wears out as a result of
friction caused by package sliding or repetitive use. Magazines come in
two forms, one for horizontal stacking of packages and the other for vertical
stacking. Most types of plastic packages (e.g., dual-in-line packages, small-
outline I-leaded packages, I-Leaded quad fiat packs, and plastic-leaded
chip carriers) can be stored in magazines. Dual-in-line packages are the
only plastic package type shipped exclusively in magazines. Flexible
polyvinyl chloride stoppers are inserted into the ends of the magazine to
hold the packages in place.
11-466 PLASTIC PACKAGING

Desiccant pouches

Dual in-line
Small-outline I-leaded
I-leaded quad flatpack

Plastic quad flat


package PQFP
Gull wing small-outline
package SOl

Figure 10-32. Packing Hierarchy for Handling Plastic Encapsulated Microcir-


cuits (From Ref. 1.)

Trays made from injection-molded plastic, such as polyvinyl chloride


and polystyrene, are used primarily for storage of quad flat packs and
bumped quad flat packs. Antistatic characteristics are provided by impreg-
nating conducting fibers with the injection-molded plastic, or by treating
the trays with an antistatic agent, such as carbon. Trays made of polystyrene
do not need an antistatic coating because sufficient fiber loading can be
added to provide critical conductivity. Polyvinyl chloride trays should be
stored between -25°C and +40°C because their shape and color may
change above +55°C. As many as 100 plastic packages can be packaged
per tray.
Tape reels are used for smaller-size plastic packages, such as small-
10.5 PACKING AND HANDLING 11-467

outline plastic packages and small plastic leaded chip carriers. Adhesive or
embossed tapes are used. The adhesive tape holds the plastic-encapsulated
microcircuits on carrier tape; the embossed tape has depressed pockets
in which the plastic-encapsulated microcircuits reside, and a cover tape
ensures that they remain in place. Many of these smaller plastic-encapsu-
lated microcircuits are difficult to handle manually. Tape reels allow
packing that is amenable to automatic handling equipment. The carrier
tape can have sprocket holes for loading registration on one or both sides,
depending on the size of the plastic-encapsulated microcircuits. Tape
widths come in a variety of sizes, ranging from 12 to 44 mm. Many more
plastic-encapsulated microcircuits can be stored in tape reels than in
magazines and trays. Typical tape reels hold 500-3000 PEMs per reel.
Nippon Electronic Company has a reel that holds 12,000 small transistors.

10.5.2.2 Second Level Packing


The second level of PEM packing involves enclosing magazines,
trays, or tape reels in bags. Moisture-barrier bags are used for dry-packing
moisture-sensitive plastic-encapsulated microcircuits. At the time of dry-
packing, devices should have a moisture content of less than 0.05% by
weight. Not all plastic-encapsulated microcircuits are considered moisture
sensitive, so not all are shipped in dry packs. Table 10-13 lists some
package types from Texas Instruments, the shipping container used, and
whether they are shipped in dry packs.

Table 10-13. Shipping Methods and Quantities

Package' Tube Tape/Reel Trays Dry Pack

56-Pin SSOP 20 500 N/A No


44-Pin PLCC 27 500 N/A No
68-Pin PLCC 18119b 250 N/A Yes
64-Pin SQFP N/A' N/A 50 Yes
80-Pin SQFP N/A N/A 50 Yes
I20-Pin SQFP N/A N/A 50/84d Yes
80-Pin SQFP N/A N/A 50 Yes

'SSOP, skinny small-outline package; PLCC, plastic leaded chip carrier; SQFP, small quad flat
pack.
bEighteen packages can be packed in a single tube when a pin is used as a tap; 19 packages can
be packed when plug is used as a tap.
'N/A, not applicable.
dDepending on tray size.
Source: From Ref. 48.
11-468 PLASTIC PACKAGING

Dry-packing uses a moisture-barrier bag (MBB) in shipping contain-


ers with dried plastic-encapsulated microcircuits, a desiccant, and a humid-
ity-indicator card. The moisture-barrier bag is sealed and labeled with the
seal date.
Moisture-barrier bags are generally made of Tyvek, a multilayered
opaque material. MIL-STD-S1705B [49] classifies moisture-barrier bag
materials into two types. Type I has a water-vapor transmission rate
(WVTR) of less than 0.02 gl100 in. 2 of bag area per 24 h; type II has a
WVTR between 0.02 and O.OS gl100 in. 2 per 24 h. The bags should
meet type I requirements for water-vapor transmission rate, flexibility,
electrostatic discharge protection, mechanical strength, and puncture resis-
tance [41]. Most of the air should be removed from the bag by manually
compressing the outside of the bag or by using suction equipment. How-
ever, the bag should not be completely evacuated, as this would reduce
the effectiveness of the desiccant and could also damage the enclosed com-
ponents.
A desiccant is a moisture-absorbing material that is enclosed in the
moisture-barrier bag to absorb moisture from the air inside the bag and
maintain its relative humidity below 20%. A number of different materials
are used as desiccants. Figure 10-33 shows desiccant moisture absorption
rates as a function of desiccant type.
Desiccant material should meet MIL-D-3464 type II [41]. It must
be ductless and noncorrosive. The amount of desiccant needed per bag
to maintain an interior relative humidity below 20% is a function of bag
surface area, bag water-vapor transmission rate, and storage time.

Molecular sieve
25°C/28%RH Phosphorus pentoxide

Calcium sulfate
Clay
Calcium chloride

Silica gel

o 2 3 4
Time (hours)

Figure 10-33. Desiccant Moisture Absorption Rates as a Function of Desiccant


Type. (From Ref. 42.)
10.5 PACKING AND HANDLING 11-469

A humidity-indicator card (HIC) is a means of measuring the relative


humidity of the air surrounding the card. A card is inserted in the moisture-
barrier bag before sealing to help determine the effectiveness of the dry
pack. An illustration of a humidity indicator card is given in Figure 10-
34. The card has at least three color dots, providing indications from 10%
to 30% relative humidity. A pink dot means the card has been exposed
to at least the given amount of humidity; a blue dot means it has not been
exposed to that much humidity. If the 30% dot starts to change from blue
to pink, the relative humidity inside the moisture-barrier bag is higher
than 20%, and the desiccant should be replaced. A humidity-indicator
card must be dry before it can be packaged in a dry pack; if it is not, it
should be baked dry or put in a desiccant bag to dry. The chemical reaction
that causes the change in color is reversible, so cards are reusable.
A warning label is mounted on each dry-packed moisture-barrier
bag. The label adhesive and markings are water resistant. The labels
include a moisture-sensitivity caution symbol, instructional notes, and a
bag seal date.

10.5.2.3 Third Level


The third level of PEM packaging is the outer box, often made of
corrugated cardboard. Bags and magazines, trays, or tape reels are stacked
into the outer carton, which is then taped shut. Bar-code labels, usually
placed on the outer box, indicate the presence of plastic-encapsulated
microcircuits and the seal date. An example of a desiccant bar-code label
is shown in Figure 10-35 [42].

o
HUMIDITY INDICATOR

EXAMINE
ITEM

o
IF PINK

CHANGE
DESICCANT

o
IF PINK

WARNING
IF PINK

DISCARD IF CIRCLES OVERRUN


AVOID METAL CONTACT
241187-6

Figure 10-34. An Illustration of a Humidity Indicator Card. (From Ref. 41.)


._-_.
11-470 PLASTIC PACKAGING

8. SEAL
DESICCANT INCLUDED
DATE: XXXX
fnlel ProdUC:1 ~um~r

r•••"'B.
XXXXXXXXXXX

• __ .iDI.
xxxxxx
XXJOOQOO{
xxxxxx
xxxx

Figure 10-35. An Example of a Desiccant Bar-Code Label. (From Ref. 42.)

10.5.3 Handling of Dry Packages

Dry plastic-encapsulated microcircuits have a finite floor life, after


which they absorb moisture above a safe level. Component floor life can
be defined as the time period that begins after moisture-sensitive devices
are removed from the moisture-barrier bag and the controlled storage
environment, and ends when they have absorbed enough moisture to be
susceptible to damage during reflow. Floor life characteristics are depen-
dent on many factors. The moisture-barrier bag should not be opened
until the components can be board mounted. Before the bag is opened,
seal integrity should be verified. Often, the shelf life of dry-packed compo-
nents is less than 12 months from the seal date (the shelf life is stated
on the bag warning label). If leakage is found or the shelf life is exceeded,
the plastic-encapsulated microcircuits should be dried before they are used
for any reflow soldering.
The length of time moisture-sensitive PEMs may be exposed to the
manufacturing floor environment after being removed from the moisture-
barrier bag depends on the moisture content of the components and the
floor environment temperature and relative humidity. Figure 10-36 shows
typical component weight gain versus time for various manufacturing floor
environments. For example, plastic leaded chip carriers absorb moisture
beyond the safe moisture content of 0.11 % by weight in approximately
130 h in a floor environment of 30°C/60% rho
10.5 PACKING AND HANDLING 11-471

0.6

0.5

"':'
0.4
~
~
'-'
.: OJ 40 °C/85%RH
&,
.a00 4O oCl60%RH
~ 0.2
30oC/60~ RH

0.1

0.0 -+--~---r--~---'--..,...--;

o 100 200 300 400 500 600


Time (Hours)

Figure 10-36. A Typical Component Weight Gain Versus Time for Various
Manufacturing Floor Environments

10.5.3.1 Classification
Exterior configuration is not the sole indicator of moisture/retlow
sensitivity. The same number of days of ambient exposure can result in
different amounts of moisture being absorbed by the package, depending
on the molding-compound material and package design. The same amount
of moisture can cause different amounts of damage in different packages.
Packages with the same exterior configuration (lead count, thickness, and
so on) may not have the same interior construction, and absorption rate
and sensitivity may differ between packages that otherwise look identical.
Consequently, the percentage weight gain moisture content is not useful
for establishing moisture sensitivity. Different package types, materials,
lead counts, and die attach areas will reach different levels of moisture
absorption before retlow damage occurs.
Packages are now classified into six levels based on their storage
tloor life (out of bag) at the board assembly site. Table 10-14 shows the
test conditions and the storage tloor life of packages at the six levels [41].
The packages are first subjected to the appropriate test conditions based
on the level for which they are being tested. They are then subjected to
a prescribed refiow simulation and are evaluated for moisture/refiow-
induced damage. They are classified into the lowest level at which they
11-472 PLASTIC PACKAGING

Table 10-14. Classification and Floor Life of Desiccant Packed Components

Storage Roor Life (out of Bag) at Preconditioning moisture'


Level Board Assembly Site °C/%rh Time (h)

Unlimited at :::; 85% rh 85/85 168


2 I Year at 30°C/60% rh 85/60 168
3 I Week at :::;30°C/60% rh 30/60 168 + METb
4 72 h at :::; 30°C/60% rh 30/60 72 + MET
5 As specified on label range: 24-71 hat 30/60 Time on label + MET
:::; 30°C/60% rh
6 Mandatory bake; bake before use and 30/60 Time on label
once baked must be reftowed within
the time limit specified on the label

, Preconditioning moisture soak to simulate total exposure time conditions.


bMET-Manufacturer's exposure time; that is, the compensation factor that accounts for the time
after bake that the component manufacturer requires to process components prior to bag seal. It
also includes a default amount of time to account for shipping and handling.
Source: From Ref. 41.

pass the acceptance criteria. The classification technique is discussed in


IPC-TM-650 Test Method 2.6.20A: Assessment of Plastic Encapsulated
Electronic Components for Susceptibility to MoisturelReflow Induced
Damage.

10.5.3.2 Board-Level Rework of Moisture-Sensitive Components


If packages must be removed from the board, localized heating
should be used and the maximum temperature excursion should not exceed
200°C. The component temperature should be measured at the top center
of the package body; if the temperature exceeds 200°C, the board may
require a bake-out prior to rework. If the part is to be reused, it should
be baked dry before remount; new replacements should be kept dry as
well. Localized replacement reflow heating avoids resubjecting the entire
board to reflow temperature profiles. The impact of nearest-neighbor
heating should be evaluated as part of the preconditioning flow for qualifi-
cation. If the packages are determined to be nonmoisture sensitive after
assembly, then controlling the board-level moisture content is not neces-
sary for rework.
If the moisture-barrier bag is opened and closed several times, the
cumulative floor exposure time should not exceed the stated floor life;
the time may be extended by storing the plastic-encapsulated microcircuits
in an environment with humidity less than 20%. The floor life differs
10.5 PACKING AND HANDLING 11-473

from manufacturer to manufacturer. Intel claims that only their plastic


leaded chip carriers and plastic quad flat packs with fewer than 100 leads
may be in a floor humidity environment of less than 20% for extended
periods of time. Hitachi [50] contends that its plastic-packaged devices
can be exposed to the floor environment for more than 1 week of cumula-
tive time outside the moisture-barrier bag before reflow soldering.
Once the moisture-barrier bag is opened, component use or handled
options may be pursued in the following order of preference:

• The components may be mounted within a specified time before


moisture is reabsorbed beyond the safe moisture content level.
• The components may be stored outside the moisture-barrier bag in
an environment of less than 20% relative humidity until future use.
• The components may be resealed in the moisture-barrier bag with
new fresh desiccant added and resealed and exposure time noted;
the total exposure time before mounting must be kept within the
allowed time for the exposure conditions.
• The components may be resealed in the moisture-barrier bag using
the original desiccant; this method does not allow the floor life of
the devices to be extended beyond the time indicated by the seal date.

For lead protection, it is recommended that components not be removed


from their original shipping containers until mounting; contact with the
leads can cause positional or coplanarity problems. It is also recommended
that automatic or semiautomatic handling equipment be used in lieu of
manual handling, especially for surface-mounted packages.
Precautions should be taken against electrostatic discharge, espe-
cially if the relative humidity is below 20%, as it is during baking; air
ionizers can be installed in baking ovens to remove static electricity.
Water leakage can wash out the water-soluble antistatic coating in shipping
containers or cause the coating to peel off and lose its effectiveness.
Storing plastic-encapsulated microcircuits in antistatic-coated containers
beyond 6 months should be avoided because the coating may warp over
time. Antistatic-coated containers should not be reused.

10.5.4 Environmental Considerations


Protecting delicate leads during handling and transit is the main
function of the tubes and trays used in packing. However, this protection
comes at a price. Before "green" manufacturing became fashionable,
packing paraphernalia was treated as regular, disposable "brown" trash;
landfill environmental concerns were never raised. As a result, packing
materials are regularly piled and dumped without much consideration of
11-474 PLASTIC PACKAGING

the ecological implications. With sparse land for landfill sites, European
countries were the first to raise the possibility of recyclability in electronic
manufacturing. Nontoxic, lead-free packaging was one practical outcome
of that initial push.
Environmental pressure has forced recyclability to become a prime
concern. As the current packing materials are made mostly of thermoplas-
tics, recycling is relatively trouble free. Used polyethylene tubes or polyvi-
nyl chloride trash can be ground, pelletized, and remolded. The end
products can be resold as packing grade with the "Made from recyclable
materials" label or similar logo. Several industries have already sprung
up to address this market niche. Used packing materials are collected
from the original equipment manufacturer OEMs or assembly houses,
where parts are unloaded from their packing carriers to be mounted.
Cleaning, grinding, remelting, and molding the materials can be a profit-
able enterprise.

10.6 QUALITY AND RELIABILITY


When a product's performance is no longer acceptable, it is said to
have failed. Failure occurs when some failure mechanism-a chemical,
electrical, physical, mechanical, or thermal process acting at some site in
a structure-induces a failure mode, such as an electrical open, a short,
or parametric drift. Common failure sites for the various failure mecha-
nisms in plastic-encapsulated microelectronic packages are schematically
illustrated in Figure 10-37.
In this section, the emphasis is on understanding the failure mecha-
nisms to avoid failures.

10.6.1 Classification of Failure Mechanisms


Failure mechanisms fall into two broad categories: overstress and
wear-out. Overstress failures are often instantaneous and catastrophic.
Wear-out failures accumulate damage incrementally over a long period,
often leading first to performance degradation and then to device failure.
Further classification of failure mechanisms is based on the type of load
that triggers the mechanism-mechanical, thermal, electrical, radiation,
or chemical.
Mechanical loads include physical shock, vibration (e.g., underneath
an automotive engine hood), loads exerted by filler particles on the silicon
die (because of encapsulant shrinkage upon curing), and inertial forces
(e.g., in the fuse of a cannon shell being fired). Structural and material
responses to these loads may include elastic deformation, plastic deforma-
tion, buckling, brittle or ductile fracture, interfacial separation, fatigue
crack initiation, fatigue crack propagation, creep, and creep rupture.
Thermal loads include exothermic die attach curing, heating prior
10.6 QUALITY AND RELIABILITY 11-475

Filler particles Suess relieve agents

Oe ~00 ~
0 • 0 ClOe $
~ 0 °0
o O0 ClO
°0 0 : 0
0 0

Mold compound
Deformation of
metallization Ball bond fracture

Voids Delamination Package cracks

Figure 10-37. Typical Failure Mechanisms, Sites, and Modes in Plastic-Encap-


sulated Devices (From Ref. 1.)

to wirebonding, encapsulation, postmold exothermic curing, rework on


neighboring components, dipping in molten solder, vapor-phase soldering,
and reflow soldering. External thermal loads lead to changes in dimensions
because of thermal expansion and can change such physical properties as
the creep flow rate or cause burning of a flammable material. Mismatches in
coefficients of thermal expansion often cause local stresses that can lead
to failure of the package structure.
Electrical loads include, for instance, a sudden current surge through
the package (e.g., in the ignition system of a car engine during start-up),
fluctuation in the line current due to a defective power supply or a sudden
jolt of transferred electricity (e.g., from improper grounding procedures),
electrostatic discharge, electric overstress, applied voltage, and input cur-
rent. These external loads may produce dielectric breakdown, surface
breakdown of voltage, dissipation of electric power as heat energy, or
electromigration. They may also increase electrolytic corrosion, current
leakage due to formation of dendrites, and thermally induced degradation.
Radiation loads are principally either alpha particles found in packag-
ing materials from trace radioactive elements, such as uranium and tho-
rium, or cosmic rays that can upset memory devices, degrade performance,
and depolymerize the encapsulant. The problem is more acute in memory
devices that are highly sensitive to alpha particles, in which even a minor
alpha flux through the device can flip the binary state of a memory bit.
Due to the stochastic nature of this failure mode, preventive measures
11-476 PLASTIC PACKAGING

usually are built in. A polyimide die coating of proper thickness is a


simple measure that can block damaging rays. Compounding ultrapure
fillers with low alpha residues in the resin is another way to minimize
erratic device behavior due to ionizing charges.
Chemical loads include chemically severe environments that result
in corrosion, oxidation, and ionic surface dendritic growth. Moisture in a
humid environment can be a major load on a plastic-encapsulated package
because of the permeability of moisture through encapsulants. Moisture
absorbed by the plastic can leach catalyst residues and polymerization
by-products from the encapsulant, then ingress to the die metallization
bond pads, semiconductor, and various interfaces and activate failure
mechanisms that degrade the package. For instance, reactive flux residues
coating the package after assembly can migrate through the encapsulation
to reach the die surface.

10.6.2 Analysis of Failures


This section discusses the mechanisms that result in damage or
failure of a plastic-encapsulated device.

10.6.2.1 Die Fracture


Surface scratching and cracks in dies may form at the thermal pro-
cessing, die scribing, and dicing stages. If, after the manufacture of the
die, a preexisting initial crack is equal to or greater than the critical
crack size (depending on the magnitude of applied stress), the die can
catastrophically fracture in a brittle manner. For example, microcracks
nucleated at the top surface from surface scratching of the wafer may
propagate to an active transistor and cause device failure. Edge cracks
developed from die separation damage are most likely to propagate at
the comer of the die, due to high longitudinal stresses common in that
location when the die-Ieadframe structure is temperature cycled. Voids
in the attachment material may also affect die fracture.
Chiang and Shukla [51] summarized die cracking statistics for sam-
ples that had edge or center voids and were subjected to 10 cycles of
thermal shock. The devices with center voids showed no cracks, whereas
devices with edge voids had nearly a 50% failure rate due to die cracks.
Although cracks can originate from edge damage during sawing, fine
control of saw speed and saw blade quality has eliminated most of the
defects. However, die cracks still originate from flaws on the back side
due to backgrinding operations used to manufacture thin packages. De-
pending on the grit size used, grinding can either reduce or increase the
chance of cracking; typically the size of flaws approaches the grit size.
Thus, all coarse flaws should be eliminated by either fine grinding or
back-side etching.
10.6 QUALITY AND RELIABILITY 11-477

During die attach, each die is ejected from the backing tape by a
transfer plunger. Damage to the die may be introduced at this stage,
because the plunger tip and the loading speed govern the size and depth
of the indentation.
The thermal stresses due to the TCE mismatches among the various
materials in the plastic package can be expressed by

[10-tO]

where
Cs = a geometry-dependent constant
a = the coefficient of thermal expansion
E = the modulus of elasticity
and the subscripts p and i denote the encapsulant material and the material
in contact with the encapsulant material (i.e., the chip or leadframe mate-
rial) [40].

If the thermally induced stress given by Eq. (to-I) is greater than the
allowable stress, the package fails. An estimate of allowable stress can
be obtained from fracture mechanics principles [52].

10.6.2.2 Loss of Chip Passivation Integrity


The chip passivation layer is usually made of brittle glass films,
such as silica (Si0 2), phosphosilicate glass (PSG), or silicon nitride (Si3N4).
The aluminum bond pad under the passivation film can break from the
bonding pressure and cannot sustain the passivation film. Phosphorus,
added to the silicon dioxide passivation layer for ion gettering, stress
relaxes the layer and avoids cracking. An optimum phosphorus content
is 1.8% by weight [53]; a high concentration (> 8%) accelerates corrosion
failures of the die due to the formation of corrosive phosphoric acid with
ingressed moisture. Corrosion of the die metallization can result in various
electrical failures, including device leakage, shorts, or opens. Plasma-
enhanced chemical-vapor-deposited (PECVD) silicon nitride has better
moisture resistance than phosphorus-doped silicon dioxide.
Invariably, a larger distribution of cracks is found near the edges of
the device, rather than in the die center. As mapped out with piezoresistive
test chips, high shear stresses are always observed at the edges. The cracks
are not easily seen under the optical microscope; however, once exposed
to a mild etchant, the affected areas become more visible. The solution
11-478 PLASTIC PACKAGING

diffuses through the microcracks in the passivation, attacking the underly-


ing metal and forming dark areas wherever the metal has been removed.
Cracking of passivation occurs due to shearing stresses imposed by
the molding compound [54]. Passivation layer cracking is often associated
with ball-bond lift-off and shearing because of the close proximity of
these structures to passivation layer damage on the edges of the chip.
Polyimide film can also be used as the passivation layer. The film
is relatively soft and sufficiently ductile to withstand bonding pressure
and high temperatures; although the film distorts under stress, it does not
crack. Inayoshi et al. [55] evaluated passivation cracking and showed that
it can be avoided by using a ductile passivation. Polyimide also offers
good adhesion with the molding compound, enhancing the interfacial
integrity critical to the long-term reliability of the package during thermal
excursions [20].

10.6.2.3 Die Metallization Corrosion

Several factors affect corrosion of the aluminum metallization: sys-


tem acidity, metal composition, encapsulation material, passivation glass,
ionic contamination, temperature, relative humidity, applied voltage, and
moisture resistance of the package [56].
The most severe condition for the plastic package and a main cause
of corrosion is the interaction of moisture with ionic contaminants and
ingredients in the molding compound. Moisture can reach the wirebond
by diffusion through the encapsulant or through such defects in the package
as separation between the leadframe and the encapsulant where the leads
enter the package body. Moisture may react with chlorine, in the form
of chloride ions present in many manufacturing processes imposed on
the package. Bromides and antimony trioxide, both used as flame retard-
ants, can be released from the molding compound when high temperatures
occur in the presence of moisture.
Galvanic corrosion requires some voltage potential. The interface
of the dissimilar metals gold and aluminum sets up an approximately 3-
V potential. In the presence of water and an ionic contaminant, galvanic
corrosion can easily take place. Although most of the die metallization
is covered by a passivation layer, defect sites can be present in the
passivation, allowing corrosion of the metal features. Because there is no
passivation in the bond-pad area, this region is also vulnerable to any
corrosive electrolyte. Complete coverage of the bond pad by the wire is
usually impossible because the bond pad is square and the corresponding
ball bond is circular. Unless high bonding force is used to flatten the ball
to obtain greater coverage or round bond pads are designed, exposed
aluminum will remain a fact of package assembly.
Techniques to passivate bond pads for reliability without hermeticity
are expensive and are not typically employed for commercial packages.
10.6 QUALITY AND RELIABILITY 11-479

Round bond pads have been tried in the past but were abandoned because
the process control vision systems of current wirebonders recognize only
square features. Good passivation coverage (using phosphorus glass or
silicon nitride) is necessary for moisture resistance. Measures to avoid
and control corrosion-related failures include reducing ionic impurities
in the encapsulant, using impurity-ion catchers or ion scavengers in the
encapsulant, increasing the encapsulant-to-leadframe adhesion strength,
using fillers in the encapsulant to elongate the path for moisture diffusion,
and using a low-water-uptake encapsulant.

10.6.2.4 Metallization Deformation


The aluminum metallization on the chip can be plastically deformed
by shrinkage forces imposed by the molding compound. Deformation is
highest at the edges of the chip, where the shear stresses are greatest. In
this case, the encapsulant is constrained from moving over the chip by
its adhesion to the chip surface adjacent to the metal trace; any loss of
adhesion between the encapsulant and the chip increases the potential for
metallization deformation.
The stress profile on the die can be mapped out with piezoresistive
strain gauges embedded in the silicon, which provide a good correlation
between measurements and finite element analysis [57]. For a given die
and package configuration, the average in-plane shear stress is about one-
tenth the magnitude of the corresponding principal stresses. Near the die
edges, this scaling factor approaches one-third. The principal stresses in
the die center increase with die size, whereas the reverse applies to in-
plane shear stresses. Edge stresses are sensitive to die dimensions, increas-
ing with die size.
The amount of free die-paddle space around the die also affects the
stress level. Generally, stresses increase as free space increases; thus,
tailoring a die pad to closely fit die size optimizes the stress level. Orienta-
tion and die aspect ratio also influence edge stresses much more than center
stresses. For example, an asymmetrical die with its axis perpendicular to
that of the package encounters lower stresses than one aligned parallel
to the axis.
The amount of stress imposed on a die depends on the thickness and
volume of the surrounding molding compound. Thin packages experience
much less stress than thick packages and are the current trend in the
packaging industry. The in-plane packaging stresses responsible for metal
deformation are caused by lateral compression of the die. A good working
knowledge of the die stress profile can help in the design of future products
that minimize stress-induced damage.
10.6.2.5 Wire Sweep
Wire sweep usually denotes visible wire deformation, typically a
lateral movement in the direction of the compound flow through the cavity.
11-480 PLASTIC PACKAGING

Under this lateral deformation, ball bonds and sometimes stitch bonds
can develop kinks at the attachment point as the wire is pulled off the
axis of the bond. The kink is most often noted on the side of the chip
nearest the gate through which the molding compound enters the cavity.
Reliability concerns with wire sweep include device shorting and
current leakage. Shorting can be from wire to wire, from wire to lead
finger, or from wire to die edge. Failure can be immediate or may not
show up until the package experiences stress excursions.
Wire sweep can occur from anyone of a number of causes: high
resin viscosity, high flow velocity, unbalanced flows in the cavities, void
transport, late packing, and filler collision [58,59,60].

10.6.2.5.1 Viscosity
During radio-frequency preheating, the pellets of mold compound
reach up to 100°C. Compacting the preheated pellets with a plunger in
the molding pot causes the compound to heat further through a "fountain
flow effect." The material near the wall is hotter than the rest and has a
lower viscosity than the colder core. Thus, the compound can be convected
to the cavities near the pot while the core is relatively cold. The ensuing
high viscosity may cause wire sweep.

10.6.2.5.2 High Flow Velocity


For cavities located far away from the pot, sweep can be induced
by a different mechanism. At this stage, sufficient heat has been convected
to the compound to lower viscosity to a level within the processing
window. The majority of the cavities are already partially filled, resulting
in a reduced flow path. As a result, for the same filling pressure, the
velocity of the flow front in the end cavities can be quite high, raising
the potential for wire sweep.

10.6.2.5.3 Unbalanced Flows in the Cavities


Cavity gates, if improperly designed, may produce flow fronts that
are not balanced, above and below the die-paddle assembly, resulting in
a "racetrack effect." One front precedes the other and blocks the vent line
before complete filling, trapping air in the package. This flow imbalance
creates a rolling bending moment that deflects the leadframe and affects
the stability of the wirebonds.

10.6.2.5.4 Void Transport


The change in interfacial tension as voids are convected across the
die surface by the flow front can move the wirebonds. With a rapid
succession of small voids, the wires can oscillate from their equilibrium
position; such motion can be captured permanently by the rapidly curing
resin. Molding voids tend to be more frequent with conventional molding
10.6 QUALITY AND RELIABILITY 11-481

than with multiplunger (gang-pot) molding, as gang-pot presses provide


a shorter flow path for the mold compound and better control of transfer
speed and pressure [61].

10.6.2.5.5 Late Packing


The filling profile programmed into standard molding presses usually
requires filling 90-95% of the mold first under "velocity control." The
profile is then switched to "pressure control," during which constant
pressure is maintained to ensure that a certain density is obtained. The
switchover must be made while the compound still has a low viscosity;
a highly viscous gel could slide across the die and collapse the wire-
bonds [62].

10.6.2.5.6 Filler Collision


Collision of fillers with fragile wirebonds is another cause of wire
sweep. Wirebonds are usually between 23 and 38 /-lm (0.9 and 1.5 mils),
depending on the current-carrying requirements of the device. Fillers, on
the other hand, can have multimodal distributions for better packing. Sizes
can range from a submicron up to 100 /-lm. By comparison, the thickness
of gates in typical molds is around 200-250 /-lm (8-10 mils). As the
highly filled front flows across the die, collision with any large filler
particles can result in plastic deformation of wirebonds.

10.6.2.6 Cratering of a Wire-Bond Pad


Cratering is a failure mechanism that occurs during manufacturing,
usually as a result of improper bonder setup. Bonding parameters that
affect the quality of a bond include temperature, ultrasonic power, force,
and bond time [63].
In ultrasonic bonding, even though the metal flow is equal in all
directions, stacking faults in the silicon occur perpendicular to the direction
of bonding tool motion, verifying that ultrasonic energy can introduce
defects into single-crystal silicon [64,65]. Cratering is common in ultra-
sonic bonding but not in thermocompression bonding. Similarly, Koyama
et al. [66] found that force and temperature cannot cause silicon nodule
cratering without ultrasonic energy.
Contamination of the bond pad can cause improper setup of the
bonding parameters, including the required bonding energy, because con-
taminated bond pads require more ultrasonic energy and higher tempera-
ture to make strong bonds. Moreover, too high or too Iowa static bonding
force can lead to cratering in wedge bonds [67]. Kale suggests that the
optimum bonding force results in more efficient energy transfer, thus
lowering the total energy requirement for the bond. Winchell [64] notes
the effect of bond-pad thickness on the phenomenon of cratering; the
11-482 PLASTIC PACKAGING

bond pad serves as a cushion protecting the underlying silicon, silicon


dioxide, polysilicon, or gallium arsenide from the stresses of the bond-
ing process.
The tendency to crater predominates in thin metallizations. One
percent silicon is sometimes added to the aluminum metallization in order
to prevent spiking in shallow junctions. Micrometer-size silicon nodules
in aluminum bond pads act as stress raisers and crack the underlying glass
during thermosonic gold ball bonding. Koch et al. [68], Ching and Schroen
[69], and Koyama et al. [66] have generally confirmed the silicon-nodule
cratering effect. After bonding is complete, these nodules decrease in the
bonding region and damage appears in the insulation. Corrective action
includes bonding at higher temperatures (250°C), using lower ultrasonic
power, reducing molding stress, and removing a fracture-prone phosphorus
glass layer from under the bond pad. To avoid cratering, Ching uses a
hard titanium-tungsten underlayer for the pad, a modified bonding profile,
and rapid ball touchdown.

10.6.2.7 Loss of Wire Bond


Impurities often cause a loss of surface bondability and premature
bond failure during the operational life of the device. Impurities and
interdiffusion in wirebonds can cause Kirkendall voids [70]. With gold-
aluminum bonds, the interdiffusion front moves through the gold, forming
intermetallic alloy phases. Due to the lower solubility of the intermetallic,
impurities swept ahead of the front precipitate there and act as sinks for
the vacancies produced during the diffusion process.
Plating baths are a major source of such impurities as potassium-
gold-cyanide, buffers, lactates, citrates, carbonates, and phosphates. Other
impurities, like thallium, lead, or arsenic, are added to the bath to increase
plating speed and reduce grain size. Increasing the plating current density
increases the codeposited impurity level exponentially. Variations in plat-
ing process parameters, such as bath temperature and bath concentration,
also cause bondability problems.

10.6.2.8 Wire-Bond Fracture and Lift-Off


Wire can break at the heel of a wedge bond due to the reduced cross
section of the wire. Cracks in the heel of a ball bond can arise as a result
of excessive flexing of the wire during loop formation, especially when
the level of the second bond is significantly lower than the first; repeated
flexing and pulling of the wire occurs as the device heats and cools during
temperature cycling due to thermal mismatches [71]. Fatigue failure by
crack propagation can occur at the heel [72]. A wire can also break at
the neck of a wirebond, leading to an electrical open. Thallium, a major
source of wirebond neck failures, forms a low-melting eutectic with gold
10.6 QUALITY AND RELIABILITY 11-483

and can be transferred to gold wires from gold-plated leadframes during


crescent-bond break-off [73,74]. Thallium diffuses rapidly during bond
formation and concentrates over grain boundaries above the neck of the
ball, where it forms a eutectic. During plastic encapsulation or temperature
cycling, the neck breaks and the device fails [75]. Ball-bond fracture
causes bond lift-off. The fracture can be a result of either tensile or shear
forces induced by thermal stress or the flow of encapsulant during molding.
The latter, also known as wire wash, happens only sporadically; when it
does, however, it signals a molding compound that has expired or been
improperly conditioned [61]. Generally, bond strength is more a function
of temperature cycling than of steady-state temperatures between -55°C
and 125°C, although bond strength decreases as a function of temperature
above 150°C for gold-aluminum bonds [76] and above 300°C for gold-
copper bonds [77].

10.6.2.9 Wire Bond and Bond-Pad Corrosion


Wires inside packages are susceptible to corrosion in the presence
of moisture and contaminants. Chlorine residues from the molding com-
pound, introduced by the capillary action of the water along the wire, can
concentrate around wirebonds. The aluminum reacts with the dissolved
chlorine to produce AICI3, a process accelerated by the elevated tempera-
tures occurring during bum-in and high-temperature storage. Corrosion
of gold wirebonds on aluminum pads results from the liberation of bromine
from brominated flame retardants, a process accelerated by high tempera-
tures. Bromide ions react with the aluminum in the gold-aluminum inter-
metallic phase (Au4A1), forming AlBr3 and precipitating gold. The alumi-
num bromide hydrolyzes to Ah03, with the liberated bromide ions
providing the driving force until all the interrnetallic AU4AI is consumed.
Bond corrosion may not result directly in failure, but it increases
the electrical resistance of the interconnect until the device becomes
nonfunctional. Often, the molding compound exerts a compressive force
on the die surface and the adjacent wirebonds, and interconnection prob-
lems are not revealed until extensive corrosion has occurred. In mild
cases, subtle changes take place in parametric functionality; in more severe
cases, however, complete metallization attack occurs, resulting in open-
circuit failures.

10.6.2.10 Leadframe Corrosion


Although leads generally have a nickel plating under the primary
finishes of nickel, tin, and tin-lead to protect the base metal (usually alloy
42 or copper) from corrosion, the assembly process for leadframes can
generate high residual stresses and high surface contamination. Any cracks
or open voids in this plating can initiate corrosion in the presence of
11-484 PLASTIC PACKAGING

moisture and contaminants. During assembly and handling, the lead fingers
are bent; cracks can develop and expose any corrodible surface to the
external environment. Stress-corrosion-induced cracking can also occur,
especially in alloy 42 leads.
The rate of galvanic corrosion can be high, because the lead finish
is often a cathodic metal with respect to the lead base material. Although
a leadframe can undergo corrosion anywhere on its external surface, the
most sensitive area is at the interface between the molding compound
and the leadframe.

10.6.2.11 Lead-Frame Deadhesion and Delamination

Moisture ingress, either through the bulk encapsulant or along the


interface between the leadframe and molding compound, can accelerate
delamination in plastic packages [78]. Experiments with moisture sensors
reveal that when the adhesion between the molding compound and the
leadframe is good, the main path of ingress into the package is through
the bulk of the encapsulant [58,79]. However, when this adhesion is
degraded by improper assembly procedures-for example, oxidation from
bonding temperatures, leadframe warpage from insufficient stress relief,
or excessive trim and form forces [80]-delamination and microcracks
are introduced at the package outline, and water vapor can diffuse along
this path.
At each interface, moisture can hydrolyze the epoxy, degrading
the interfacial chemical bonds. However, different molding compounds
respond differently to moisture exposure. Low-stress epoxy compounds,
for instance, with silicone modifiers added for stress reduction tend to be
more susceptible to changes induced by moisture than molding compounds
without silicone. A low glass-transition temperature also adversely affects
moisture absorption.
Surface cleanliness is a crucial requirement for good adhesion. Oxi-
dized surfaces, such as copper-alloy leadframes exposed to high tempera-
tures, often lead to delamination [33,81]. The presence of a nitrogen or
forming-gas shroud helps to avoid oxidation and is recommended during
high-temperature processes.
Low-affinity surface finishes, such as silver-spot plating, enhance
interfacial adhesion. Traditionally, silver plating ofthe die pad is employed
for bias control and to prevent oxidation on leads. Unfortunately, silver
plating adheres poorly to molding compounds [31,57]. Newer leadframe
designs use spot plating to minimize both the amount of precious metal
used and the 1eadframe coverage, which is susceptible to delamination.
Releasing agents and adhesion promoters in molding compounds
have been shown to encourage delamination in plastic packages and must
be delicately balanced [78]. Releasing agents facilitate removal of the
10.6 QUALITY AND RELIABILITY 11-485

molded parts from the mold cavities at the risk of greater interfacial
delamination. On the other hand, adhesion promoters ensure good interfa-
cial adhesion between the compound and the component, but parts may
be hard to pry out of the mold cavity.

10.6.2.12 Encapsulant Rupture


Resistance to crack initiation in a material is strongly dependent on
geometry, material properties, and loading. Kitano et al. [82] developed
a method for measuring the critical stress intensity factor for an encapsulant
ruptured by brittle fracture.

10.6.2.13 Encapsulant Fatigue Fracture


Cyclic stresses in a package occur due to mismatches in thermal
expansion caused by temperature changes and temperature gradient. An
estimate of package life subjected to thermal cycles is obtained using
Paris' law [83] for brittle crack propagation.
The stress intensity factor for a change in temperature is calculated
using finite element methods, and the relationship is expressed as a polyno-
mial,f(a). Employing Paris' law, the number of temperature cycles, Ncyclw
for propagating a crack from length, llo to ac is given by
sac

Ncycles = f c{da(a)m [10-11]

ao

where the material constants C3 and m have been determined for the
encapsulant materials using single-edge-notch specimens [84]. The initial
crack length, ao, is taken to be some small value dependent on manufactur-
ing defects, but the total number of cycles to failure is relatively insensitive
to initial crack length, provided the crack is small.

10.6.2.14 Package Cracking ("Popcorning")

Package cracking results from internal stresses generated by the


reflow solder temperature profile during the assembly of plastic-encapsu-
lated devices on circuit cards. In brief, during reflow soldering, the assem-
bly temperature is rapidly raised above the glass-transition temperature
of the molding compound (approximately 140-160°C) to a temperature
of approximately 220°C or more, which is required to melt the solder.
At these temperatures, the thermal mismatch between the molding com-
pound and the materials adhering to it, such as the leadframe and silicon
die, can be large enough to make the interfaces susceptible to delamination.
11-486 PLASTIC PACKAGING

Moisture absorbed by the molding compound can vaporize into steam,


and the resulting volume change can produce a pressure greater than the
adhesion strength of the interface, causing delamination. Delamination
may occur between the molding compound and the leadframe, or between
the compound and the die surface. In extreme cases, the package may
crack, ball bonds may shear from the bond pads, or the silicon may crater
beneath the ball bonds. The cracks often propagate from the die paddle
to the bottom surface of the package, where they are difficult to see during
visual inspection of soldered boards. Occasionally, depending on package
dimensions, the cracks may propagate to the top of the package or along
the plane of the leads to the package sides, where they become more
visible [85]. Proper handling and use after unpacking the assembly from
the desiccant will deter this failure mechanism.

10.6.2.15 Electrical Overstress and Electrostatic Discharge

Electrostatic discharge (ESD) is the transfer of electrical charge


between two bodies at different potentials, either through direct contact
or through an induced electrical field. Electrostatic discharge can occur
during manufacturing, handling, or service. It is neither unique nor more
prevalent in plastic-encapsulated microcircuits than in hermetic packages.
Its failure mechanism can include gate-oxide breakdown and conductive
path formation, with or without shorts. The failure site is often at sharp
edges, such as corners of bond pads and metallizations, where the electric
field strength is the greatest.
Electrostatic charges are generated by frictional sliding contact be-
tween dissimilar materials, or by induction charging of a conductor from
an external charged insulator. Electrostatic discharge arises primarily from
several sources: human contact, leading to static electricity discharge from
the human body through the device; discharge of a statically charged
device due to sliding against a carrier surface; and discharge of an induced
charge in the device conductor from an external, electrostatically charged
insulator-for example, a coffee cup carried across a carpet. Thus, the
cause is electrical abuse.
Electrical overstress (EOS) is the damage to an electrical circuit due
to thermal overstress caused by excessive electrical power dissipation
during a transient electrical pulse. Like electrostatic discharge, electrical
overstress failures are often caused by misapplication of the device in a
circuit, allowing surges in a power line. Failure modes due to electrical
overstress include junction spiking, latchup (common in MOS and CMOS
devices), melted metallization, metal electromigration, and open bond
wires.
Electrical damage from electrostatic discharge and electrical over-
stress have increased in importance due to the progressive decrease in
10.6 QUALITY AND RELIABILITY 11-487

integrated circuit dimensions. Metal oxide semiconductor (MOS) chips


are more susceptible to electrostatic discharge damage because of small
geometries and thin gate-oxide dielectrics.

10.6.2.16 Soft Errors


Soft errors are temporary upsets in the state of a memory bit due
to electron-hole pairs induced by alpha-particle radiation. Soft errors first
showed up in 16K dynamic random access memory devices (DRAMs);
the mechanism was discovered by May and Woods [86]. These errors
did not show a repeat pattern and were erased with each refresh clock cycle.
The charge difference in memory cells, which distinguishes a "0"
state from a "1," is called the critical charge, which for a 64K DRAM is
about (1-2) x 106 electrons. An alpha particle produces 1.4 x 106 electrons
and an equal number of holes as it penetrates 25 J.lm (1 mil) into a silicon
chip. The holes are electrically rejected by the memory cell, and the
electrons collected can change the state of the memory cell until the next
refresh cycle restores the proper state. As the dimensions of the memory
cells are reduced, the critical charge decreases, compounding the radiation
of alpha particles. Like electrostatic discharge, soft errors are not unique
to plastic-encapsulated microcircuits and must also be addressed with
hermetic packages.

10.6.2.17 Solder-Joint Fatigue


Solder-joint fatigue is the same for PEMs on circuit cards as it is
for hermetic-package assemblies. The drivers of solder-joint fatigue are
primarily the global mismatch in the TCEs of the component and the
circuit card, and the local mismatch in the TCEs of the solder and the
lead, and the solder and the circuit card pads. The global mismatch can
be further increased under operating conditions when there are temperature
differences between the component and the circuit card. It may be advanta-
geous to use a PEM over a hermetic part when the board material coeffi-
cient of thermal expansion more closely matches the PEM than the her-
metic (metal or ceramic) package; this is the case with most of the common
organic board materials, such as FR4, cyanate ester, and bismaleimide
triazine. An excellent overview of the solder-joint failure mechanism can
be found in Frear et al. [87].

10.6.3 Ranking of Potential Failures


Failure distributions differ from device to device and manufacturer
to manufacturer. Table 10-15 presents a summary of failure analysis
results in the form of a Pareto ranking of various field failures in very-
large-scale integration (VLSI) class devices for demanding commercial
11-488 PLASTIC PACKAGING

Table 10-15. Pareto Ranking of Failure Mechanisms of 3400 Commercial


VLSI-Class Devices from Multiple Sources, Including Manufacturing
Fallout, Qualification, Reliability Monitors, and Customer Returns

% of Failed
Failure Mechanisms Integrated Circuits

Electrical overstress and electrostatic discharge 19.9


Unresolved 15.9
Gold ball-bond failure at bond 69.0'
Not verified 6.0
Gold ball-bond failure at stitch bond 4.6'
Shear stress, chip surface 3.5'
Corrosion, chip metallizationlassembly 3.2'
Dielectric fail, poly-metal, metal-metal 3.0
Oxide defect 2.9
Visible contamination 2.7
Metal short, metal open 2.6'
Latch-up 2.4
Misprocessed, wafer fab-related 2.4
Chip damage, cracks/scratches 2.4'
Misprogrammed 2.0
Oxide instability 1.9
Design of chip 1.7
Diffusion defect 1.5
Final test escape 1.4
Contact failure 1.2
Bond failure, nongold 1.2'
Protective coating defect 0.9
Assembly, other 0.9'
PolysiIiconlsilicide 0.8
External contamination 0.7'
Others 5.3

'Possible packaging or assembly-related failures.


Source: From Ref. 88.

applications, such as automotive manufacturing and telecommunications


[88]. The sources of failures include manufacturing fallout, qualification,
reliability monitors, and customer returns. Failures due to electrical overs-
tress and electrical discharge dominate, with electrical overstress due to
misapplication and voltage overstress as primary causes. The major causes
of failures were poor handling and assembly. Of the remaining failure
mechanisms, wirebond failure was significant, and corrosion of die metalli-
zation was a relatively small factor.
Vable and Hanna [89] reported a distribution of early field failures
10.6 QUALITY AND RELIABILITY 11-489

of plastic-encapsulated integrated circuits. Their data indicate that 36.3%


of failures are unverified, 22.2% are of undetermined origin, 17% are due
to wirebond failure, 8.3% result from electrical overstress, 2.3% are from
mask/etch defects, 2.1 % are due to electrostatic damage, and 11.9% are
attributed other causes.
Another study, by Bloomer [90] presented failures in commercial
integrated circuits components (Fig. 10-38). The components analyzed
were supplied by a large industrial user from 1984 to 1988. The pie chart
includes failures during manufacturing, incoming component testing, final
board testing, and field operation. Like other studies, this one found that
the major causes of failure are electrical overstress, electrostatic discharge,
and wirebond weaknesses.
Finally, Pecht and Ramappan [91] collected electronics system and
device failure data and noted some of the reliability problems associated
with electronics systems. They also noted that reliability improvements
related to inherent component stress considerations are generally ineffec-
tive in fault prevention, if the stress levels are within design limits. This
contrasts with Arrhenius models, in which increasingly lower stress
(steady-state temperature) is believed (but is not physically proven) to
provide increasingly improved reliability.

10.6.4 Comparison of Plastic and Ceramic Package


Failure Modes

Integrated-circuit packages encapsulated with the first generation of


epoxy molding compounds in the 1960s and 1970s exhibited rather poor

D EOSIESD - 59%
[ll]] Wire bonds - 15%
mD Lead short/open - 7%
~ Die fracture - 4%
ISS! Conductor failure - 3%
E2d Oxide/passivation failure - 3%
• Electrical testing - 3%
EJ (encapsulation
Other - 6%
defects.
solder failure.
die attachment defects,
die mechanical damage,
and metallurgical failures)

Figure 10-38. Distribution of Failures in Commercial Integrated-Circuit


Components (PEMs) from a Large Industrial End User. (From Ref. 90.)
11-490 PLASTIC PACKAGING

performance when compared with their ceramic counterparts. Corrosion


and thermally induced mechanical stress were two common failure modes.
Ionic residues like chloride, potassium, and sodium exceeded several
hundred parts per million in the early epoxies, compared with the less
than 10 ppm obtained with the latesttechnology. The high level of contami-
nants, coupled with the poorer quality of the device passivation in the
sixties and seventies, accounted for the low mean time to failure of plastic-
encapsulated devices. Furthermore, the technology of compound stress
modification was not recognized; silicone modifiers and stress absorbers
were still in the conceptual stage. Today, with advances in molding com-
pounds and device technology, plastic-encapsulated packages exhibit long-
term reliability comparable to and, for many applications, better than
hermetic packages [92-95].
Failure modes are similar in plastic and ceramic packages. Table
10-13 shows a Pareto ranking of failures observed in hermetic packages
by researchers at Texas Instruments. Comparing the Pareto rankings in
Tables 10-15 and 10-16 shows that electrical overstress, electrostatic
discharge, corrosion, and thermomechanical stress are reliability concerns
for both configurations. However, the relative importance of each mode
varies between package types. For example, even though current epoxies
are better purified and almost free of ionic residues, external contaminants
entrained by absorbed moisture can still reach the die surface in contact
with the epoxy. On passivated dies, corrosion often occurs at the exposed
bond pads. In hermetic packages, corrosion is often due to organic contami-
nants from the die attach process mixed with traces of moisture trapped
within the hermetic confines of the package.

10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES


This section addresses the current status and trends in encapsulated
microcircuit packaging technology. The focus is on trends in encapsulation
materials, package design, and manufacturing methods.

10.7.1 Trends in Materials, Design, and Fabrication


Trends in encapsulation materials, package design, and manufactur-
ing methods are presented in this section.

10.7.1.1 Encapsulation Material Trends


Epoxy molding compounds (EMCs) using transfer molding are ex-
pected to remain the major force in molded packaging technology. How-
ever, it is unclear whether they will be able to meet the yield requirements
of high-pin-count complex packages and the heat dissipation needs of
future products. The expected drop of supply voltage in future integrated
10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES 11-491

Table 10-16. Pareto Ranking of Failure Mechanisms of Failed Hermetic


Devices

% of Failed
Failure Mechanisms Integrated Circuits

Electrical overstress and electrostatic discharge 16.4


Hermeticity rejects 6.6
Not verified 5.8
Polysiliconlsilicide 5.1
Pin rejects 3.8
Visible contamination 3.2
Oxide defect 2.3
Chip damage-cracks/scratches 2.1
Metal open 2.1
Metal short 1.9
Contact failure 1.7
Passivation defect 1.6
Contamination on Porous oxide 1.5
Assembly-other 1.4
Bond failure-nongold 1.3
External contamination 1.0
Dielectric fail, poly-metal 0.8
Waferfab-other 0.8
Misprocessed 0.7
Diffusion defect 0.7
Die attach 0.7
Oxide instability 0.6
Package damage 0.6
Lead adhesion 0.5
Silicon defect 0.4

Source: R. S. Martin, personal communication, 1993.

circuits from 5 to 3.3 to 2.5 V will certainly ease the heat-dissipation


demands on PEMs. Among other uncertainties of enhancements in epoxy-
molding-compound properties are the coefficient of thermal expansion
and viscosity tailoring. Figure 10-39 depicts the current development trend
of EMCs, with their specific intended applications.
The following explains some of the terms used in the tables and
figures above:

• Antisolder: the properties of molding compounds that make the


components robust against potential failures during soldering (e.g.,
delamination and package cracking).
11-492 PLASTIC PACKAGING

Semiconductor Future
production Nino products semiconductor
process process

Diffusion
' -_ _ _ _.....1
1..- photo
Resist stripping tape
sensitive PI coatmg

L..-_B_a_ck_gn~·n_d_....JI"- EJep(NEL) system

Functional test 1..- ASMAT - - - -....~ ...._ _ _---.-....1

Dicing 1..- Elep(NEL) system

Die bonding

Wire bonding ASMAT

Molding 1..- Encapsulating


Material
Burn in test 1..- ASMAT ASMAT--fII--+--I.....~L---..---J

Marking I~ Micro print system


Shipping I~ Carrier tape
Assembly

Figure 10-39. Nitto's Technical Development Plan of Semiconductor-Related


Materials. (From Ref. 1.)

• Recyclable recovery: development of molding compounds that can


be recycled so that the molding compound remaining after the mold-
ing process (e.g., material left in the runners) can be remelted by
the customer and used for the next molding.
• Bottomless markless: Plastic package markings on the bottom surface
are impossible to read unless the package is stripped of the board.
Bottomless-markless technology uses a microprint system to elimi-
nate bottom-side markings.
• Single-device burn-in: Burn-in screens for single devices, rather than
manufactured lots, is becoming a necessity as customers look for
10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES 11-493

known good dies (KGD). For example, Nitto Denko is developing


application-specific materials (ASMATs) which are unique in that
they (1) have polyimide substrates bonded directly to the copper
leadframes without any intervening adhesive, (2) can be used for
extremely fine-pitch applications, and (3) allow bumps to be made
in the leadframe/polyimide substrate to increase the reliability of
the die-Ieadframe bonding during burn-in.
• Dam-barless technology: The leadframe is usually provided with a
darn-bar to constrain the molding compound from overflowing during
the molding process. Companies such as Nitto Denko are developing
thin films which when applied to the leadframe would serve the
same purpose while reducing the leadframe's overall material thick-
ness. These new materials are being developed for fine-pitch thin-
package applications where the leadframes are thin (and may deform
upon removal of the dam-bar).
• Local-stress soft error: The die is susceptible to local stresses due
to TCE mismatch effects between the die surface and the molding
compound. Because of this, the die functionality can be affected,
leading to data-retention errors (soft errors) in memory dies. Pres-
ently, companies such as Nitto Denko provide a gel coat on the die
surface to act as a buffer and are developing cavity molds which
would not require gel coats. Future plans are to develop low-modular
materials for molding compounds to provide buffers for local stress
in noncavity packages.
• Kneading: a process used for forming a mixture from the elements
of the molding compound. During kneading, the molding reaction
is allowed to occur to about 10% completion to form a solid continu-
ous mass of the molding compound.
• Pelletizing: The kneaded mold is compressed into tablets to be used
during the transfer-molding process.
• Pad tilt: During molding, the die paddle (along with the die) some-
times tilts with the flow of molding compound, leading to the die
being exposed outside the package. The goal is to simulate this
process to develop molding compounds that are robust against this
type of failure.

For postmolded plastic packaging, injection-molded liquid-crystal


polymers (LCP) have some promising potential for the future. With a
cycle time of only 20-30 s, they can offer several times higher production
rates than epoxies. The base resin is very pure, with low viscosity; even
with filler particles added for high thermal conductivity, it has a very
low coefficient of thermal expansion. Liquid-crystal polymers also have
exceptional solvent resistance and are amenable to optical fiber attachment.
11-494 PLASTIC PACKAGING

Liquid injection and reaction injection molding processes use low-


viscosity monomers or oligomers; the former is a one-component system,
the latter is two. Chemically, these materials are silicones, epoxies, or
epoxy-silicone hybrids. Even with high filler loadings, they have very
low viscosities and their properties can be specially formulated.
Polyimides with low coefficients of thermal expansion have been
developed with values as low as 4 x 1O-7rc. They may be especially
useful in very-high-precision electronics applications, such as insulation
films and plastic-encapsulation underlayers. Most polyimide precursors
interact with copper (leadframe), necessitating the use of passivation
barrier coatings. However, the polyimide precursor based on ester technol-
ogy is compatible with copper metallurgy. Its unique isomer ratio enables
the development of high solid formulations that allow thick coatings of
up to 20 !lm.

10.7.1.2 Package-Design Trends


Thin, small-outline packages (TSOPs), with their compact profiles,
became a key product of the 1990s. By most estimates, they will also be
used increasingly in surface-mount small-outline packages in the future,
especially for memories, as demand for space-saving packages grows
[96]. Other thin package types include the thin quad flat pack (TQFP), thin
sealed small-outline (TSSOP), and thin-body plastic dual-in-line (PDIP)
packages. Integrated-circuit packages with width and/or length shrink
include the fine-pitch quad flat pack (FPQFP), quarter-size small-outline
package (QSSOP), shrink DIP (SDIP), shrink quad flat pack (SQFP),
shrink small-outline package (SSOP), and very-small-outline package
(VSOP). With these or new shrunken packages, suppliers will continue
to provide more functions in less board space, and boards will become
smaller. This trend to skinnier housings will challenge die thinning, wafer
transportation, chip-pad mounting, leadframe design, lead bonding, board
interconnection, molding, and soldering.
JEDEC has standardized a family of 1.4-mm-thick thin quad flat
pack (TQFP) packages that range in size from 5 to 28 mm square and
32 to 256 leads. TQFP usage is driven by applications that require small,
thin, and lightweight semiconductor chip packaging, including notebook
and subnotebook computers, personal digital assistants, portable consumer
products, PCMCIA cards, and disk drives. These applications call for
chips with larger sizes, greater power dissipation, or higher electrical
performance [97]. The trend is toward even larger 1.4-mm-thick TQFP
packages than currently covered by JEDEC, such as the 32-mm and 40-
mm packages, to handle both higher lead counts and larger chip sizes.
However, the molding process gets increasingly complicated when pack-
age sizes increase at the same thickness. Olin Interconnect Technologies
has developed a high thermal and electrical performance TQFP in which
the plastic mold compound is replaced by an anodized aluminum base
10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES 11-495

and lid adhesively sealed to the leadframe [97]. The package uses the
same infrared or vapor-phase reflow board mounting profile as a plastic
package, weighs the same as a plastic package, and is dimensionally
equivalent to a plastic package.
In the area of ultrahigh-heat-dissipation plastic packages, Power
Quad 2, a 160- and 208-lead plastic quad flat pack from Amkor Electronics,
is a significant development in the 1990s [98]. In this package, thermal
management attributes include a large-area solid copper heat sink under
the die. For improved electrical and thermal performance, an internal
ground plane could be used. With a 8.75 x 8.75-mm (350 x 350-mil) die,
6junction.to.case is less than 0.4 °C/W; with a new line of external heat sink and
airflow, a junction-to-ambient thermal resistance of 8.0°C/W will result.
Finally, various design tools for both single-chip and multichip pack-
ages exist on the market. These tools typically address only some aspects
of the IC production flow. For instance, some tools simulate mainly the
electrical performance of the packages for various configurations and
operations frequencies. Others concentrate only on the thermal manage-
ment side by portioning the functions on either the die or the module;
still others focus only on the mechanical aspect. Such a gap is well
recognized by the CAD and modeling tool vendors, and attempts have
been made at integrating the disjointed modules into one single package.
Understandably, most of the development efforts have been devoted to
the high-value-added packages. All the major CAD vendors now profess
to have design tools that can partition, route, and simulate the performance
of the complex packages, such as multichip modules, in terms of electrical
characteristics, thermal dissipation, and mechanical stresses.
Software tools are also available to ensure that reliability, quality,
and yield issues are addressed during package design. One such tool for
reliability is the Computer Aided Design of Microelectronic Packages
(CADMP II) software developed by the CALCE Electronic Package Re-
search Center at the University of Maryland.

10.7.1.3 Package Fabrication Trends


Evolutionary directions in package fabrication include new cleaning
methods, improved manufacturing process control, fine-pitch interconnec-
tions, and new leadframe fabrication.
Manufacturing technology and the cost of the advanced materials
required for high-quality and reliable plastic packaging determine both
acquisition and lifetime cost, and drive the technology for the widespread
use of PEMs in all markets. Performance criteria naturally adapt to the
cost constraints of the market. At near-perfect yield, the present molding
process productivity of the most common PLCC packages is about 800
packaged devices per hour. High-yield plastic packages of the future
using thinner leadframes, fine-pitch wirebonding, and flip-chip or TAB
11-496 PLASTIC PACKAGING

structures will need low-viscosity molding compounds at a low molding


temperature to reduce shear-rate-induced yield losses. Higher production
rates will be required to compete with PGAs and premolded packages.
Smaller molds with fewer cavities and a total cycle time (including in-
mold cure) of 1-2 min in an automated mode will be needed. A high
level of automation will result in lower-cost, uniform-quality packages
with less damage to fragile high-lIO-count assemblies. A clean manufac-
turing environment will reduce contamination-related failures.
Automation will play an increasingly important role in PEM manu-
facture. One approach is partial automation of various labor-intensive
aspects of the process, such as preform heating and handling. A robotic
arm may place the heated preform into the transfer pot and start the process
sequence of transfer, curing, and ejection of the molded leadframes. In
this automation approach, a single operator can handle four or five transfer
presses. In the total automation approach, the process runs without any
operator assistance, although typically 10-20% of an operator's time is
spent in moving different cassettes and checking equipment malfunctions.
The use of a smaller molding tool with fewer cavities in conjunction with
faster-curing molding compounds will further increase productivity. Such
a packaging system is usually totally enclosed to maximize process cleanli-
ness and personnel safety. The cassettes of molded leadframes feed an
automated trim-and-form press, and then move on to a code-marking
station. The partial or total automation of single-pot systems has been
superseded by multiplunger technology in multiproduct production envi-
ronments. In automated multiplunger systems, 6-12 pots feed 6-24 cavit-
ies, with 1-4 cavities per transfer pot. Because of very short flow length,
they are very effective in minimizing voids and promoting high molding-
compound density. State-of-the-art encapsulants-very fast-curing mold-
ing compounds with a total cycle time of less than 2 min in the mold-
will be used in these manufacturing systems.
The increasing use of analytical techniques and microsensors to fine-
tune materials and control process-induced defects falls into the category
of enhanced process control and quality assurance. X-ray radiography
and C-mode scanning acoustic microscopy/tomography are finding in-
creased use as nondestructive techniques for evaluating plastic delamina-
tion; die metal, die attach, and bonding wire deformation; die metal and
wirebond voiding; leadframe, die passivation, die attach, wire, wirebond
and case brittle fracture; and dendritic growth under bias. Mercury poro-
simetry has been successfully employed to find the number and size of
epoxy and epoxy-metal pores in epoxy-encapsulated packages. Epoxy
pores are less than about 0.2 !lm in diameter, epoxy-Ieadframe voids are
about I !lm in diameter, and surface pores range from 5 to 500 !lm in
diameter. Piezoresistive strain gauges integrated into test chips will con-
tinue to be used to directly measure the mechanical stress induced inside
a PEM by encapsulation, die bonding, and other factors, either during
10.7 FUTURE DEVELOPMENTS IN PLASTIC PACKAGES 11-497

fabrication or under environmental stress testing. Solid-state moisture


microsensors will also be used to measure the moisture content at any
specific location inside a PEM.
Challenged by higher integration levels (particularly in ASICs) with
tighter bond-pad pitches at the die level, leading-edge pitches will be at
0.060-0.075 mm (2.5-3 mils) by the late 1990s. Beyond a O.1-mm (4-
mil) pitch, bonding with gold wire may be done by wedge bonding. The
combination of tight pad pitches and shrink packages demands not only
low wire looping (0.09-0.18 mm) but also different loop shapes. This
requires modifying the dopants in the wire to offer the right set of mechani-
cal properties. With tight pitches, the physical limits of leadframe technol-
ogy will also force the placement of lead contacts at ever-increasing
distances from the package center to control wire sweep. Low looping
also promotes wire-to-die edge shorting when the die has inbound bonding
pads to avoid encapsulation stress concentration zones. Edge shorting can
be prevented by the use of new wirebonder software, that allows an extra
bend in the wire at a predetermined site.
The demands of tightly spaced inner-lead bonding (0.05-0.075 mm)
and outer-lead bonding (0.15-0.35 mm), high pin counts, high-end perfor-
mance-driven applications, and high-volume production could also move
bumped tape automated bonding (TAB) to the forefront of the packaging
industry [99]; a potential for which Japan already has the capability
and infrastructure. Limiting factors for TAB applications have been the
expense of tape, the lead time needed to obtain it, and the capital needed
for the bonding equipment. Bump fabrication has also been an obstacle.
Better tape metallurgy, currently available, now permits bum-in and longer
shelf-life before encapsulation. Laser bonding of fine pitch devices is
particularly suited to TAB.
Trends in leadframe technology have important implications for the
future of molded plastic packaging. Copper alloys are replacing 42 Fel
58 Ni (alloy 42) as the leadframe material for moderate-to-high heat-
dissipating devices, such as processors and logics. But their high coefficient
of thermal expansion and average mechanical properties could prohibit
their use in other applications. It is also becoming increasingly difficult
to manufacture thin (::; 0.15 mm) leadframes with close lead tips for
very-high-Iead-count packages. Although chemical etching has replaced
mechanical punching in these situations, as the leads and lead spacing
get smaller, maintaining a lead aspect ratio near unity for the leads by
leadframe isotropic etching is becoming extremely difficult. Consistent
with the need for a mechanically stable lead geometry, leadframe thickness
has decreased from 2.5 mm (100 mils) for low-pin-count devices to under
0.15 mm (6 mils) for high-lead-count (-200) fine-pitch plastic quad flat
packs. Several different types of lead tip layouts offer different pad sizes,
lead-tip spacings, and wirebond lengths [100]. Thinner leadframes, larger
die pads (for short, close-spaced lead tips), and longer wirebond spans
11-498 PLASTIC PACKAGING

pose such molding problems as flow-induced deformation, paddle shift,


and wire sweep. There are also problems with handling the delicate
leadframe in terms of both inner lead wirebonding yields and outer lead
soldering yields. Thus, it appears that leadframe manufacturing limitations
could hinder the development of a very fine-pitch and high-lead-count
molded plastic packaging.

10.7.2 Trends and Challenges in Circuit Card Assemblies


The trends in printed wiring board (PWB) materials and fabrication
technology are likely to include relatively new materials with lower coef-
ficients of thermal expansion, lower dielectric constants, and greater di-
mensional stability, which can be inexpensively manufactured. In particu-
lar, as components become thinner, the TCE is moving to that of the
silicon. High-temperature capability will make board distortion a nonissue.
Thus, a matched in-plane coefficient of thermal expansion of the PWB
will be needed to help reduce solder-joint fatigue.
The use of solder for connecting PEMs to circuit cards also still has
reliability challenges. In particular, the assembly process has aggravated
popcorn package cracking, particularly in thin and ultrathin package styles.
This trend is toward improved molding compounds that resist delamination
at the paddle interface, minimize deformation at the reflow solder tempera-
ture, and deter cracking when deformed by the vapor expansion.
Because of the ban on chlorofluorocarbons, environmentally accept-
able solvent cleansers are being found for assembled-board cleaning opera-
tions. In addition to mild detergents used in industrial dishwashers, non-
flammable cleaning solvents are being developed for these operations.
Molded plastic package material should thus have good solvent resistance.
Although most molding compounds are impervious to these corrosive
fluxing agents at the required soldering times and temperatures, the interfa-
cial paths along the leads are particularly vulnerable to ingress by these
corrosive materials. Consequently, good sealing and adhesion along these
plastic-metal interfaces is needed to prevent flux penetration and package
contamination during fluxing. If fluxes are not removed from the assembled
board-including the underside-by cleaning operations and effluent test-
ing, temperature cycling eventually opens up more of these interface
ingress routes, and premature corrosive failure of PEMs results. The
current trend is to use very mild fluxes, such as those made from lemon
juice (citric acid), that can be easily washed out during board cleaning.

10.7.3 Standards and Requirements


Methods to ensure high quality and long-term reliability will be
established by short-term correlated and tailorable tests. In many cases,
qualification-by-design will replace qualification tests. Screening will vir-
tually disappear as a procedure for eliminating defective parts. Military
qualification and reliability test standards will be discarded, and critical-
10.8 HEAT TRANSFER IN PLASTIC PACKAGES 11-499

Lrad Fromt

• ~1tt.1 Hut ink


• AI, ·Filltd Molding ompound

Figure 10·40. Advances in Heat Transfer of Plastic Packages

use environment related-test conditions will be modified to reflect the


physics-of-failure processes to be uncovered. Users will probably still
demand manufacturer's qualification data for their particular application
environments, but they will maintain supplier partnerships with only those
vendors that can satisfy parts-reliability criteria from field-failure data,
failure analysis, applicable environmental tests, quality assurance tech-
niques, and part-specific design analysis. Qualification requirement matri-
ces will be drawn up to guide suppliers, relating all elements of design,
manufacture, and part assembly variables with the required needs.
The future will also have to bring together differences between
Japanese and American product standards. For example, the standoff
distance between a circuit board and the most popular PEM (PQFP) is
0.25-0.37 mm (10-15 mils) in the u.S.-manufactured product, compared
with near zero in its Japanese counterpart. In addition, Japanese lead
length requirements on higher pin-count, finer-pitch packages have been
shrinking; some lead lengths are as short as 1.0 mm (0.5 mm from the
edge of the package to the vertical portion and 0.5 mm for the foot itself).
United States manufacturers have reached only 0.8 mm for the lead from
the package end to the vertical bend. Differences in lead length standards
between U.S. and Japanese manufacturers have had a direct adverse effect
on board design, as designers have attempted to accommodate both types of
parts. This has led to lower board yield and increased rework. Interestingly,
JEDEC standard for package outlines are looser than assembly tolerance
requirements, resulting in noninterchangeability of even different U.S.
vendor parts made to the same specifications.

10.8 HEAT TRANSFER IN PLASTIC PACKAGES


Plastic packages are notoriously poor for heat-transfer because of
their very low thermal conductivity. However, a number of advances are
being made to improve the heat-transfer characteristics. Referring to Figure
10-40, these advancements include (1) the use of metal heat sinks such
as the VSPA package (see Chapter 7, "Microelectronics Packaging-An
11-500 PLASTIC PACKAGING

Table 10-17. Chip Bonding Adhesives (Part 1 of 2)

Type Resin Filler Electrically


Supplier No. ~ Type Components Conductive

Epotek H20E Epoxy Silver 2 Yes

H20E-175 Epoxy Silver 2 Yes


H35-175M Epoxy Silver Yes

H4l Epoxy Silver Yes

H44 Epoxy Gold Yes


H8lE Epoxy Gold 2 Yes
P-lOll Polyimide Silver Yes
Amicon C-840-4 Epoxy Silver Yes

C-868-1 Epoxy Silver Yes

C-940-4 Polyimide Silver Yes

C-966 Polyimide Silver Yes

C-990 Epoxy Silver Yes

CT-4042 Epoxy Silver 2 Yes


ME-868 Epoxy Oxide No

Ablestik 71-1 Polyimide Silver Yes


71-1LMl Polyimide Yes
84-1LMl Epoxy Silver Yes

84-1LMI Epoxy Silver Yes


84-1A Epoxy Silver Yes
84-1LMISI Epoxy Silver Yes
941-3 Epoxy Silver Yes

Overview") with copper, copper leadframes and the molding compounds,


encapsulants, and die-attach adhesives with aluminum nitride filler. The
copper heat sinks raise the thermal conductivity of typical molding com-
pound from 0.5 to as much as 396w/moC. Aluminum nitride fillers have
demonstrated thermal conductivity in excess of 7w/m°C. These develop-
ments have enabled the plastic packages to remove as much as 5 wattsl
chip. These developments are discussed in both Chapter 4 and Chapter 7.
10.9 DIE ADHESIVES 11-501

Table 10-17. Chip Bonding Adhesives (Part 2 of 2)


Maximum Lap
Cure Shear
Tem~rature Strength Remarks
°C kg/mm2

175 1.06 Two-component epoxy adhesive with high electrical &


thermal conductivity
180 1.06 Low ionic impurity, dual T, version of H20E
180 1.27 Similar to H20E except single component, no mixing
required
ISO 2.11 Single-component epoxy adhesive with low cure temp &
high lap shear strength
ISO 2.11 Single-component, gold-filled epoxy adhesive
ISO 0.91 Two-component gold-filled epoxy adhesive
ISO 0.50 Low cure temperature polyimide adhesive
200 1.6 Low ionic impurity epoxy adhesives can be pre-cured as
B-stage on leadframes
180 1.6 Low ionic impurity epoxy adhesive with low cure temp.
can be B-staged
270 1.45 Low ionic impurity polyimide adhesive. Excellent heat
resistance to 400°C
180 2.5 Low temp. cure polyimide adhesive with low ionic
impurities
275 1.55 Very rapid cure epoxy adhesive, cures on wirebond
heater block
150 1.6 Low ionics two-component epoxy adhesives
180 1.6 High thermal conductivity, electrical insulation epoxy
adhesive
275 0.70 General purpose polyimide direct-attach adhesive
275 2.04 Very low ionic impurity version of 71-1
150 4.58 Low temp. cure conductive epoxy adhesive with low
levels of ionic impurities
150 4.30 High, thermal conductivity version of 84-ILMI
200 4.23 Rapid cure version of 84-ILMI
200 2.11 Low viscosity, low ionic version of 84-ILMI
175 2.11 13-stageable epoxy chip adhesive for preprinting of
leadframes

10.9 DIE ADHESIVES


Both silicon/gold eutectic and polymer bonding are used to bond
the chip to the leadframe. The operation begins with the loading of the
film ring holder, with sawed wafer attached, onto the index stage of the
bonding machine. Lead frames are then fed from magazines along a track
to a heater block. A small square of silicon/gold alloy (6% Si, 94% Au)
is cut from a feed ribbon and transferred to the chip support platform.
11-502 PLASTIC PACKAGING

An optical scanner detects a good chip on the sawed wafer causing a


probe to push the chip above the wafer plane where a collet picks it up,
and transfers it to the heated chip support platform. Chip and eutectic are
now scrubbed together by the collet, forming a hard alloy bond. Heater
temperature is approximately 420°C. Total cycle time for eutectic bonding
is about 6 to 8 seconds.
Polymer bonding is faster than eutectic with a cycle time of2 seconds.
Feed mechanisms for polymer bonding machines are the same as eutectic
bonders. However, leadframes are not heated. Silver-filled epoxy or poly-
imide adhesive paste is transferred to the chip support platform by a print
head, and the chip is pressed into the paste immediately after printing.
The chip attach adhesive properties are listed in Table 10-17. Chip bonded
leadframe strips are loaded into transport magazines. Eutectic bonded
frames go directly to wirebonding while magazines containing polymer-
bonded frames are routed to ovens for adhesive curing. The curing atmo-
sphere is dry nitrogen, and typical schedules are one hour at 150°C for
epoxies, and 30 minutes at 150°C, followed by 30 minutes at 275°C
for polyimides.
More recently, improved silicone gels were introduced as an alterna-
tive means of fastening the silicon die to the support platform [101]. Pin-
grid arrays (PGA) soldered on standard FR-4 boards were used as test
vehicles. In this case, the chip was initially bonded with the gel to the
epoxy base of the PGA. After wirebonding, the entire PGA cavity was
encapsulated with the same gel. Thus, the chip resided in a substantially
compliant medium, and could readily withstand thermal and environmental
stresses. The higher reliability achieved resulted from using the same gel
throughout. Previous attempts at bonding the die with an epoxy-based
adhesive, followed by die coating with a gel were unsuccessful for two
reasons. First, improper cure conditions can leave low molecular weight
amine residues in the epoxy. The unbound amine leached out under
environmental testing and reacted adversely with the silicone gel, and,
second, although the stiffer epoxy can hold the chip rigidly to the base,
thermal mismatch would generate significantly higher stresses ultimately
harmful to the integrity of the device.

10.10 REFERENCES
1. M. G. Pecht, L. T. Nguyen, and Edward B. Hakim. Plastic-Encapsulated Microelec-
tronics Materials, Processes, Quality, Reliability, and Applications. John Wiley and
Sons, New York, 1995.
2. M. G. Pecht. Integrated Circuit, Hybrid, and Multichip Module Package Design
Guidelines: A Focus on Reliability. John Wiley and Sons, New York, 1994.
3. M. Pecht, R. Agarwal, and D. Quearry. "Plastic Packaged Microcircuits: Quality,
Reliability, and Cost Issues," IEEE Trans. Reliability, pp. 513-517, 1983.
4. N. Sinnadurai. "Advances in Microelectronics Packaging and Interconnection Tech-
nologies-Toward a New Hybrid Microelectronics," Microelectron. J., 16:p. 5, 1985.
10.10 REFERENCES 11-503

5. M. Brizoux, et a1. "Plastic-Integrated Circuits for Military Equipment Cost Reduction


Challenge and Feasilibity Demonstration," 40th Electronic Components and Technol-
ogy Conference, pp. 918-924, 1990.
6. Reliability Analysis Center. A DoD Information Analysis Center Plastic Microcircuit
Package-A Technology Review, Reliability Analysis Center, Rome, NY 1992.
7. L. Condra and M. Pecht. "Options for Commercial Microcircuits in Avionic Prod-
ucts," Defense Electron. pp. 43-46, 1991.
8. M. Nachnani, L. Nguyen, J. Bayan, and H. Takiar. "A Low-Cost Multichip (MCM-
L) Packaging Solution," Proceedings of International Electron Manufacturing Tech-
nology Symposium, pp. 464-468, 1993.
9. G. F. Watson. "Plastic-Packaged Integrated Circuits in Military Equipment," IEEE
Spectrum, pp. 46-48, 1991.
10. M. Priore, and J. Farrell. "Plastic Microcircuit Packages: A Technology Review,"
Report No. CRTA-PEM, Reliability Analysis Center, Rome, NY, 1992.
11. Texas Instruments. FIFO Surface Mount Package Information, Texas Instruments,
1992.
12. K. C. Griggs. "Plastic Versus Ceramic Integrated Circuits Reliability Study," Report
No. WP86-2020, Rockwell International Collins Group, 1986.
13. C. A. Lidback. "Plastic-Encapsulated Products vs. Hermetically Sealed Products,"
Summary Report, Motorola Inc., Government Electronics Group, January 1987.
14. L. R. Villalobos. "Reliability of Plastic-Integrated Circuits in Military Applications,"
ESP Report No. PV 620-0530-1, Motorola Government Electronics Group, Tactical
Electronics Division, 1989.
15. R. J. Straub. "Automotive Electronic Integrated Circuits Reliability," Proceedings
of Custom Integrated Circuit Conference, pp. 92-94, 1990.
16. L. W. Condra, G. A. Krornholtz, M. G. Pecht, and E. B. Hakim. "Using Plastic-
Encapsulated Microcircuits in High Reliability Applications," Proceedings Annual
Reliability and Maintainability Symposium, pp. 481-488, 1994.
17. May, C. A., Epoxy Materials. Electronic Materials Handbook, I-Packaging, ASM
IntI. pp. 825-837, 1989.
18. R. K. RosIer. "Rigid Epoxies," in Electronic Materials Handbook, 1 Packaging, pp.
810-816, ASM International, Materials Park, OH, 1989.
19. P. Proctor and J. Sole. "Improved Thermal Conductivity in Microelectronic Encapsu-
lants," Proceedings of the 41st IEEE Electronics Components Conference, pp. 835-
842, 1991.
20. A. S. Chen, L. T. Nguyen, and S. A. Gee. "Effects of Material Interactions During
Thermal Shock Testing on Integrated Circuits Package Reliability," Proceedings of
the IEEE Electronic Components and Technology Conference, pp. 693-700, 1993.
21. A. Ditali and Z. Hasnain. Monitoring Alpha Particle Sources During Wafer Process-
ing, Semiconductor Inti., pp. 136-140, June 1993.
22. N. Kinjo, M. Ogata, K. Nish, and A. Kaneda. Epoxy Molding Compounds as
Encapsulation Materials for Microelectronics Devices, Advances in Polymer Science
88. Springer, Berlin, 1989.
23. B. Bates. "Molding Compounds Technology for Military Applications," Commercial
and Plastic Components in Military Applications Workshop, 1993.
24. J. B. Austin. ''Thermal Expansion of Nonmetallic Crystals," J. Am. Ceramic Soc.,
35(10): pp. 243-253, 1952.
25. B. Yates. Thermal Expansion, pp. 52-70, Plenum Press, New York, 1972.
26. J. Boustani, "Ultra-Low Expansion Metal Matrix Composition," M.l.T. S.M. Thesis,
1981. Cambridge, MA.
27. W. H. Kohn. Materials and Techniques for Vacuum Devices, Reinhold Publishing
Corp., New York, 1967.
11-504 PLASTIC PACKAGING

28. F. 1. Dance and J. L. Wallace. "Clad Metal Circuit Board Substrates for Direct
Mounting of Ceramic Chip Carriers," Electron. Packag. Prod., 22(1): pp. 228-232,
236-237, 1982.
29. J. T. Breedis. "New Copper Alloys for Surface Mount Packaging," J. of Metals,
A. I.M.E., p. 48, June 1986.
30. K. E. Manchester and D. W. Bird. "Thermal Resistance, A Reliability Consideration,"
I.E.E.E. Trans. on Components, Hybrids, and Manu! Tech., CHMT-3(4): pp. 362-
370, 1980.
31. L. T. Nguyen. "Reactive Flow Simulation in Transfer Molding of IC Packages,"
Proceedings of the 43rd Electronic Components and Technology Conference, pp.
375-390, 1993.
32. L. L. Blyler, H. E. Blair, P. Hubbauer, S. Matsuoka, D. S. Pearson, G. W. Poelzing,
and R. C. Progelhof. "A New Approach to Capillary Viscometry of Thermoset
Transfer Molding Compounds," Polym. Eng. Sci., 26(20): p. 1399, 1986.
33. S. Kim. "The Role of Plastic Package Adhesion in IC Performance," Proceedings
of the 41st Electronic Components and Technology Conference, pp. 750-758, 1991.
34. A. Nishimura, S. Kawai, and G. Murakami. "Effect of Leadframe Material on Plastic-
Encapsulated Integrated Circuits Package Cracking Under Temperature Cycling,"
IEEE Trans. Components Hybrids Manu! Technol., CHMT-12: pp. 639-645, 1989.
35. A. Hale, H. E. Bair, and C. W. Macosko. "The Variation of Glass Transition as a
Function of the Degree of Cure in an Epoxy-Novolac System," Proceedings of SPE
ANTEC, 1116,1987.
36. A. Hale, M. Garcia, C. W. Macosko, and L. T. Manzione. "Spiral Flow Modelling
of a Filled Epoxy-Novolac Molding Compound," Proceedings of SPE ANTEC, pp.
796-799, 1989.
37. A. Hale. Epoxies Used in the Encapsulation of Integrated Circuits: Rheology, Glass
Transition, and Reactive Processing, Thesis, University of Minnesota, Department
of Chemical Engineering, 1988.
38. R. Gannamani and M. Pecht, "An experimental study of popcorning in plastic-
encapsulated microcircuits," IEEE Trans. Compo Packaging Mfgrg. Tech. Part A
Vol. 19 No.2 June 1996, pp. 194-201.
39. R. Munamarty, P. McCluskey, M. Pecht, and C. Yip. "Popcorning in Fully Populated
and Perimeter Plastic Ball Grid Array Packages," Soldering and Surface Mount
Technology, No. 22 Feb. 1996, pp. 46-50.
40. L. T. Manzione, J. K. Gillham, and C. A. McPherson. "Rubber Modified Epoxies,
Transitions and Morphology," J. Appl. Polym. Sci., 26: p. 889, 1981.
41. P. Yalamanchili, R. Gannamani, R. Munamarty, P. McCluskey, and A. Christou.
Optimum Processing Prevents PQFP Popcorning, Surface Mount Technology, pp.
39-42, May 1995.
42. Intel Corporation. "Recommended Procedures for Handling of Moisture Sensitive
Plastic Packages," in Intel Corporation Packaging Handbook, Intel Corp., 1993.
43. S. Altimari, S. Goldwater, P. Boysan, and R. Foehringer. "Role of Design Factors
for Improving Moisture Performance of Plastic Packages," Proceedings of the 42nd
Electronic Components and Technology Conference, pp. 945-950, 1992.
44. L. T. Nguyen, K. L. Chen, and P. Lee. "Leadframe Designs for Minimum Molding-
Induced Warpage," Proceedings of the 44th Electronic Components and Technology
Conference, 1993.
45. F. Linker, B. Levit, and P. Tan. "Ensuring Lead Integrity," Adv. Packaging, pp.
20-23, 1993.
46. C. L. Alger, D. E. Pope, P. M. Rehm, and N. Subramaniam. "Solderability Require-
ments for Plastic Surface Mount Packages," Proceedings of the 7th IEEE-CHMT,
IEMTS, 1990.
10.10 REFERENCES 11-505

47. E. Pope. "Moisture Barrier Bag Characteristics for PSMC Protection," Technical
Proceedings. SEMICON-East. pp. 59-69, 1988.
48. Texas Instruments. Texas Instruments Military Plastic Packaging. Preliminary Hand-
book. Texas Instruments, 1992.
49. MIL-B-81705B. Military specification. Barrier Materials, Flexible, Electrostatic-
free. Heat sealable. U. S. Department of Defense, Washington, DC, 1989.
50. Hitachi. Surface Mount Package Users Manual Hitachi, 1991.
51. S. S. Chiang and R. K. Shukla. "Failure Mechanism of Die Cracking Due to Imperfect
Die-Attachment," Proceedings of the IEEE Electronic Components and Technology
Conference. pp. 195-202, 1984.
52. D. Broek, Elementary Engineering Fracture Mechanics. 4th ed., Kluwer Academic,
Boston, 1991.
53. P. P. Merrett. "Plastic-Encapsulated Device Reliability," in Plastics for Electronics.
ed. Martin T. Goosey, Elsevier Applied Science Publication, New York, 1985.
54. S. Okikawa, M. Sakimoto, M. Tanaka, T. Sato, T. Toya, and Y. Hava. Stress
Analysis of Passivation Film Crack for Plastic Molded LSI Caused by Thermal
Stress, Proceedings International Symposium on Test and Failure Analysis pp. 275-
280,1983.
55. H. Inayoshi, K. Nishi, S. Okikawa, and Y. Wakashima. "Moisture-Induced Aluminum
Corrosion and Stress on the Chip in Plastic-Encapsulated LSIs," Proceedings of the
17th Annual International Reliability Physics Symposium. pp. 113-117, 1979.
56. L. 1. Gallace, H. J. Khajezadeh, and A. S. Rose. "Accelerated Reliability Evaluation
of Trimetal Integrated Circuit Chips in Plastic Packages," Proceedings of the 14th
Annual International Reliability Physics Symposium. pp. 224-228, 1978.
57. L. T. Nguyen, S. A. Gee, and W. F. Bogert. "Effects of Configuration on Plastic
Packages," J. Electron. Packaging. 113: pp. 397-404, 1991.
58. L. T. Nguyen. "Moisture Diffusion in Electronic Packages, II: Molded Configurations
vs. Face Coatings," 46th SPE ANTEC. pp. 459-461, 1988.
59. L. T. Nguyen and F. J. Lim. "Wire Sweep during Molding of Integrated Circuits,"
IEEE Electronic Components and Technology Conference. pp. 777-785, 1990.
60. L. T. Nguyen, A. S. Danker, N. Santhiran, and C. R. Shervin. "Flow Modeling of
Wire Sweep During Molding of Integrated Circuits," ASME Winter Annual Meeting.
pp. 27-38, 1992.
61. L. T. Nguyen, R. L. Walberg, C. K. Chua, and A. S. Danker. "Voids in Integrated
Circuits Plastic Packages from Molding," ASMElJSME Conference on Electronic
Packaging. pp. 751-762, 1992.
62. L. T. Nguyen. "Reactive Flow Simulation in Transfer Molding of Integrated Circuits
Packages," IEEE Electronic Components and Technology Conference. 1993.
63. S. A. Gee, L. T. Nguyen, and V. R. Akylas. "Wire Bonder Characterization Using
a P-N Junction-Bond Pad Test Structure," MEPPE FOCUS 91. pp. 156-170, 1991.
64. V. H. Winchell. "An Evaluation of Silicon Damage Resulting from Ultrasonic Wire
Bonding," Proceeding of the 14th Annual International Reliability Physics Sympo-
sium. pp. 98- 107, 1976.
65. V. H. Winchell and H. M. Berg. Enhancing Ultrasonic Bond Development. IEEE
Transactions on Components. Hybrids. and Manufacturing Technology CHMT-I,
pp. 211-219, 1978.
66. H. Koyama, H. Shiozaki, I. Okumura, S. Mizugashira, H. Higuchi, and T. Ajiki. "A
Bond Failure Wire Crater in a Surface Mount Device," Proceedings of the 26th
Annual International Reliability Physics Symposium. pp. 59-63, 1988.
67. V. S. Kale. "Control of Semiconductor Failures Caused by Cratering of Bond Pads,"
Proceedings of the International Microelectronics Symposium, pp. 311-318,1979.
11-506 PLASTIC PACKAGING

68. T. Koch, W. Richling, J. Whitlock, and D. Hall. "A Bond Failure Mechanism,"
Proceedings of the 24th Annual International Reliability Physics Symposium, pp.
55-60, 1986.
69. T. B. Ching and W. H. Schroen. Bond Pad Structure Reliability. Proceedings of the
26th Annual International Reliability Physics Symposium, pp. 64-70, 1988.
70. C. W. Horsting. Purple Plague and Gold Purity. Proceedings of the 10th Annual
International Reliability Physics Symposium, pp. 155-158, 1972.
71. G. G. Harman. Reliability and Yield Problems of Wire Bonding in Microelectronics,
ISHM,1989.
72. D. O. Harris, R. A. Sire, C. F. Popelar, M. F. Kanninen, D. L. Davidson, L. B. Dun-
can, Kallis, and J. Hiatt. "Microprobing," Proceedings of the 18th Annual Interna-
tional Reliability Physics Symposium, pp. 116-120, 1980.
73. N. C. McDonald and P. W. Palmberg. Application of Auger Electron Spectroscopy
for Semiconductor Technology, p. 42, IEDM, 1971.
74. N. C. McDonald and G. E. Riach. "Thin Film Analysis for Process Evaluation,"
Electron. Packaging Production, pp. 50-56, 1993.
75. H. K. James. "Resolution of the Gold Wire Grain Growth Failure Mechanism in
Plastic-Encapsulated Microelectronic Devices," IEEE Trans. Components Hybrids
Manuf. Technol., CHMT-3: pp. 370-374, 1980.
76. J. L. Newsome, R. G. Oswald, and W. R. Rodrigues de Miranda. "Metallurgical
Aspects of Aluminum Wire Bonds to Gold Metallization," 14th Annual Proceedings
of the IEEE Electronics Components and Technology Conference, pp. 63-74, 1976.
77. P. M. Hall, N. T. Panousis, and P. R. Manzel. "Strength of Gold Plated Copper Leads
on Thin Film Circuits Under Accelerated Aging," IEEE Trans. Parts, Hybrids,
Packaging, PHP-II(3): pp. 202-205, 1975.
78. S. S. Kim. "Improving Plastic Package Reliability Through Enhanced Mold Com-
pound Adhesion," IEEE International Reliability Physics Symposium Tutorial, Topic
2, pp. 2d.I-2d.17, 1992.
79. L. T. Nguyen. "Surface Sensors for Moisture and Stress Studies," in New Character-
ization Techniquesfor Thin Polymer Films, ed. H-M. Tong and L. T. Nguyen, Wiley,
New York, 1990.
80. L. T. Nguyen. "Reliability of Postmolded Integrated Circuits Packages," SPE RETEC,
pp. 182-204, 1991.
81. O. Yoshioka, N. Okabe, S. Nagayama, R. Yamaguchi, and G. Murakami. "Improve-
ment of Moisture Resistance in Plastic Encapsulants MOS-Integrated Circuits by
Surface Finishing Copper Leadframe," Proceedings of the 39th IEEE Electronic
Components and Technology Conference, pp. 464-471, 1989.
82. M. Kitano, A. Nishimura, and S. Kawai. "A Study of Package Cracking During the
Reflow Soldering Process (I st & 2nd Reports, Strength Evaluation of the Plastic by
Using Stress Singularity Theory)," Trans. Japan Soc. Mech. Eng., 57 (90): pp.
120-127, 1991.
83. P. C. Paris, M. P. Gomez, and W. E. Anderson. "A Rational Analytical Theory of
Fatigue," Trend Eng. 13: pp. 9-14. 1961.
84. A. Nishimura, A. Tatemichi, H. Miura, and T. Sakamoto. "Life Estimation for Integrated
Circuits Plastic Packages Under Temperature Cycling Based on Fracture Mechanics,"
IEEE Trans. Components Hybrids Technol, CHMT-12(4): pp. 637-642,1987.
85. S. Ito, A. Kitayama, H. Tabata, and H. Suzuki. "Development of Epoxy Encapsulants
for Surface Mounted Devices," Nitto Technol. Rep. pp. 78-82, 1987.
86. T. C. May and M. H. Woods. A New Physical Mechanism for Soft Errors in Dynamic
Memories. Proceedings of the 16th Annual International Reliability Physics Sympo-
sium, pp. 33-40, 1978.
87. D. Frear, H. Norgan, S. Burchett, and J. Lau. The Mechanics ofSolder Alloy Intercon-
nects, VanNostrand Reinhold, New York, 1994.
10.10 REFERENCES 11-507

88. R. B. Ghate. Industrial Perspective on Reliability of VLSI Devices, Texas Instru-


ments, 1992.
89. R. W. Vahle, and R. J. Hanna. Proceedings of the International Congress on Transpor-
tation Electronics, Society Automotive Engineering (October 1990) 225.
90. C. Bloomer, R. L. Franz, M. 1. Johnson, S. Kent, B. Mepham, S. Smith, R. M.
Sonnicksen, and L. S. Walker. "Failure Mechanisms in Through-Hole Packages," in
Electronic Materials Handbook, I, Packaging, ed. by M. L. Minges, pp. 969-981,
ASM International, Materials Park, OH 1989.
91. M. G. Pecht and V. Ramappan. Are Components still the Major Problem: A Review
of Electronic System and Device Field Failure Returns. IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology, Vol. 15, No.6, pp. 1160-1164,
Dec. 1992.
92. C. H. Taylor. "Just How Reliable Are Plastic-Encapsulated Semiconductors for Mili-
tary Applications and How Can the Maximum Reliability Be Obtained?" Microelec-
tron. Reliability. 15: pp. 131-134, 1976.
93. C. H. Taylor and B. C. Roberts. Evaluation of a U. K. Specification for the Procure-
ment of Plastic-Encapsulated Semiconductor Devices for Military Use. Microelec-
tronics and Reliability, 18 pp. 367-377, 1978.
94. P. V. Robock, and L. T. Nguyen. "Plastic Packaging," in Microelectronics Packaging
Handbook, ed. R. R. Tummala and E. J. Rymaszewski, Van Nostrand Reinhold, New
York, 1989.
95. L. T. Nguyen and J. A. Jackson. "Identifying Guidelines for Military Standardization
of Plastic-Encapsulated Integrated Circuits Packages," Solid-State Technol., pp. 39-
45, 1993.
96. R. Iscoff. "Thin Outline Packages: Handle With Care!" Semicond. Int. pp. 78-82, 1992.
97. P. Hoffman, D. Liang, D. Mahulikar, and A. Parthasarathi. "Development of a High
Performance TQFP Package," pp. 57-62, 1994.
98. R. Iscoff. "Amkor Develops Competitor to Multilayer Ceramic, MQUAD Packages,"
Semicond. Inter. p. 34, 1992.
99. R. Iscoff. "Micro SMT Package Avoids Traditional Bonding Methods," Semicond.
Int. p. 40. 1992.
100. W. E. Jahsman. "Lead Frame and Wire Bond Length Limitations to Bond Densifica-
tion," J. Electron. Packaging, 111: pp. 289-294, 1989.
101. K. Otsuka, Y. Takeo, H. Tachi, H. Ishida, T. Yamada, and S. Kuroda. "High
Reliability Mechanism of New Silicone Gel Sealing in Accelerated Environment
Text," Proc. I.E.P.S., pp. 720-726, 1986.

ADDITIONAL READINGS
C. Bloomer, R. L. Franz, M. J. Johnson, S. Kent, B. Mepham, S. Smith, R. M. Sonnicksen,
and L. S. Walker. "Failure Mechanisms in Through-Hole Packages," in Electronic
Materials Handbook, I, Packaging, ed. by M. L. Minges, pp. 969-981, ASM Interna-
tional, Materials Park, OH 1989.
S. Han and K. K. Wang. A Study of the Effects of Fillers on Wire Sweep Related to Semiconduc-
tor Chip Encapsulation, ASME Winter Annual Meeting pp. 123-130, 1993.
S. Mizugashira, H. Higuchi, and T. Ajiki. "Improvement of Moisture Resistance by Ion-
Exchange Process," IRPS IEEE, pp. 212-215, 1987.
Nitto Denko Corporation, personal communication, 1993.
G. F. Watson. "Interconnections and Packaging," IEEE Spectrum, pp. 69-71, 1992.
SIA 1993
G. Wolfe. "Electronic Packaging Issues in the 1990s," Electron. Packaging Production, pp.
76-80, 1990.
11-508 PLASTIC PACKAGING

BIBLIOGRAPHY
ASTM Annual Book ofASTM Standards, American Society for Testing and Materials, Philadel-
phia, 1993.
L. W. Condra, G. A. Krornholtz, M. G. Pecht, and E. B. Hakim. "Using Plastic-Encapsulated
Microcircuits in High Reliability Applications," Proceedings Annual Reliability and
Maintainability Symposium, pp. 481-488, 1994.
L. W. Condra, S. O'Rear, T. Freedman, L. Flancia, M. Pecht, and D. Barker. "Comparison of
Plastic and Hermetic Microcircuits under Temperature Cycling and Temperature Humid-
ity Bias," IEEE Transactions on Components, Hybrids, and Manufacturing Technology,
Vol. 15, No.5, pp. 640-650, Oct. 1992.
L. W. Condra, G. Wenzel, and M. Pecht. "Reliability Evaluation of Simple Logic Microcircuits
in Surface Mount Plastic Packages," ASME Winter Annual Meeting, New Orleans,
Nov. 1993.
A. Gallo, and R. Munamarty. "Popcorning: A Failure Mechanism in Plastic Encapsulated
Microcircuits," IEEE Trans. on Reliability, Sept. 1995.
R. Gannamani and R. Munamarty. "Techniques to Qualify PEMs against Popcorning," Elec-
tronic Materials and Packaging, pp. 24-26, Nov. 1995.
R. Gannamani, and M. Pecht. "An Experimental Study on Popcoming in PEMs," IEEE Trans.
on Components, Packaging, and Manufacturing Technology-Part A, vol. 19, no. 2,
pp. 194-201, June 1996.
N. Kelkar, A. Fowler, M. Pecht, and M. Cooper. "Phenomenological Reliability Modeling of
Plastic Encapsulated Microcircuits," International Journal of Microcircuits and Elec-
tronic Packaging, vol. 19, no. I, March 1996.
R. Munamarty, P. McCluskey, M. Li, P. Yalamanchili, R. Gannamani, and L. Yip. "Delamination
and Cracking in PBGAs during IR Reflow Soldering," BGA Conference, Berlin Ger-
many, 1995.
L. T. Nguyen. "Wirebond Behavior During Molding of Integrated Circuits," Polm. Eng. Sci.
28(4): pp. 926-943, 1988.
L. T. Nguyen and C. A. Kovac. "Moisture Diffusion in Electronic Packages. I. Transport
Within Face Coatings," SAMPE Electronics Materials and Processes Conference, pp.
574-589, 1987.
L. T. Nguyen. "On Lead Finger Designs in Plastic Packages for Enhanced Pull Strength," Int.
J. Microcircuits Electron. Packaging, 15(1): pp. 11-33, 1991.
L. T. Nguyen, A. Danker, N. Santhiran, and C. R. Shervin. "Flow Modeling of Wire Sweep
During Molding ofIntegrated Circuits," ASME Winter Annual Meeting, pp. 27-38, 1992.
L. T. Nguyen, R. H.Y. Lo, and J. G. Belani. "Molding Compound Trends in a Denser Packaging
World, I: Technology Evolution," IEEE International Electronic Manufacturing Tech-
nology Symposium, 1993.
M. G. Pecht. "A Model for Moisture Induced Corrosion Failures in Microelectronic Packages,"
IEEE Transactions on Components, Hybrids, and Manufacturing Technology, Vol. 13,
No.2, pp. 383-389, June 1990.
M. G. Pecht, and V. Ramappan. "Are Components Still the Major Problem: A Review of
Electronic System and Device Field Failure Returns," IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology, Vol. 15, No.6, pp. 1160-1164, Dec.
1992.
C.G.M. Van Kessel, S. A. Gee, and J. R. Dale. "Evaluating Fracture in Integrated Circuits With
Acoustic Emission," Acoustic Emission Testing, 5, 2nd ed., vol. 5, pp. 370-388, ed. G.
Harman, American Society for Non-Destructive Testing, 1987.
P. Yalamanchili, P. Gannamani, R. Munamarty, P. McCluskey, and A. Christou. "Optimum
Processing Prevents PQFP Popcoming," Surface Mount Technology, pp. 39-42, May
1995.
11
POLYMERS IN PACKAGING
G. CZORNYJ-.aM
M. ASANO- Toray
R.L. BELIVEAU-DuPont
P. GARROU-Dow
H. HIRAMOTO-Toray
A. IKEDA-Asahi
J.A. KREUZ-DuPont
O. ROHDE-Ci&a Geigy

11.1 INTRODUCTION

The technologies in computer electronics are very diverse and span


the entire spectrum of engineering and scientific fields. Metals, ceramics,
and polymers are at the heart of all leading-edge semiconductors and
packages used in the microelectronic Industry [1,2]. Today, dramatic
changes are occurring in the applications of semiconductor products which
will challenge existing materials and processes in the 21st Century [3].
Key application trends, including scaling down dimensions to 1 gigabit
technologies, are expected to affect both package and on-chip intercon-
nection, both requiring major changes in materials [4-10]. The material
challenges for semiconductor devices utilizing polymeric materials are
primarily for the formation of on-chip interconnection by extremely fine-
11-510 POLYMERS IN PACKAGING

line lithography [11-15] and for passivation and chip protection [16,17].
The role of packages, beyond simply passive containers in the past, is
viewed as setting limits to the ultimate performance of computers. In this
role, the packages of the 21st Century need to interconnect, power, cool,
and protect devices in such a way that leading-edge performance and
reliability result [18,19]. The materials to perform these functions include
high-conductivity metals, low-dielectric-constant ceramics, and thermally-
stable polymers.
Information processing is performed by the use of semiconductor
devices interconnected, powered, and cooled by packages. Both semicon-
ductors and packages depend significantly on materials, particularly thin-
film metal, oxides, nitrides, thick-film ceramics, and thin-film polymers.
Even though similar polymeric materials are used for semiconductor de-
vice on-chip interconnections, this chapter reviews the status and chal-
lenges of polymeric materials used for packaging applications, primarily
focusing on the use of polymers for interlayer thin-film dielectrics in
single and multichip Packages.

11.2 HISTORICAL PERSPECTIVE


Polymers have played and continue to play an important role in
the fabrication of thin-film wiring layers on both first- and second-level
packages and in semiconductor dielectric applications because of their
excellent thermal, mechanical, and solvent resistance properties. Table
11-1 shows the many technology areas where polymers are utilized in
semiconductor and packaging applications. The first commercial semicon-
ductor device chip where a polyimide was used as part of a dual dielectric
insulating layer in conjunction with quartz was the SAMOS (Silicon
Aluminum Metal Oxide Semiconductor) 64Kbit memory chip introduced
by IBM in 1978 as shown in Figure 11-1 [20]. The main advantages of

Table 11-1. Microelectronics Applications of Polymers

Organic dielectric insulators Solder fluxes


Passivation barriers Sealants
Alpha barriers Adhesion promoters
Stress buffers High-temperature adhesives
Lift-off materials Ceramic binders
Rie barrier materials Encapsulants
Resists Heat-transfer materials

Source: Data from Refs. 21-35.


11 .2 HISTORICAL PERSPECTIVE 11-511

Burlington Chip (SAMOS)

Pb/SN

1.2u S10 2 Al/Cu

Figure 11-1. Cross Section of a Commercial FET Chip with Polyimide

the polyimide in this application over standard quartz dielectric was a


lower dielectric constant of 3.5 versus 4.1 and ease of processing into
uniform coatings on silicon wafers. This type of polyimide was utilized
subsequently in many electronic applications. These applications include
alpha-particle barrier coatings, stress buffer coatings on integrated circuits
(ICs), and passivation coatings as well as interlayer dielectrics on ICs
and high-density thin-film interconnects for multichip modules (MCM-
Ds). NEC is credited with the first use of polymers as dielectrics in 1985
for use in mainframe packaging applications.
IBM announced in 1993 the development of a "thin-film" multichip
module for workstations using a silicon substrate. The main attributes of
this technology are high performance, high wiring density, reliability
without hermeticity, and ability to use semiconductor processes to fabricate
thin-film wiring in a polyimide dielectric to achieve low cost. At about
the same time, multilayer ceramic modules were fabricated using thin-
film polyimide dielectric layers for providing high-density surface wiring
typically called "redistribution" in mainframes. The ES9000 mainframe
system developed by IBM, for example, utilizes a novel poly imide [formed
from a poly(amic ethyl ester) of PMDA-ODA] specially developed to
provide ease of processing with excellent mechanical properties [33,34].
A cross section of the high-density thin-film redistribution layers built
with this poly imide is shown in Figure 11-2 [35]. In fact, the development
of the next generation of higher-density memory and logic chips as well
as thin-film package wiring depends to a great extent on the development
and availability of improved polymeric dielectric insulators. Today, there
are many polymeric systems to choose from, including the newly devel-
oped photosensitive polyimides which offer potential reduction in process-
ing steps during thin-film fabrication. In this chapter, these and other
1/-512 POLYMERS IN PACKAGING

---Pb-Sn SOLDER
--t--FAoN OUT UNE

PAD

GLASS-CERAMIC SUBSTRATE tsO<~~j4-+-C~I-V1

Figure 11-2. Thin-Film Redistribution Showing Glass-Ceramic/Copper Sub-


strate, Polyimide, and Metal Interfaces

materials will be reviewed, properties contrasted, and then summarized


for use in selecting the best material for a specific thin-film wiring intercon-
nection, whether it be for packaging or device applications.

11.3 POLYMER THIN-FILM REQUIREMENTS FOR PACKAGING


AND SEMICONDUCTORS
To develop high-quality thin films with respect to defects and reliabil-
ity, one must select materials that match the functional and reliability
requirements. The material selection criteria that one needs to consider
when selecting high-performance polymers for use in thin-film structure
fabrication are summarized in Table 11-2.
The polymeric requirements for high-density interconnects, on-chip
and in thin-film packages, have many similarities, as shown in Table 11-2.
The main differences that drive these requirements are that semiconductor
applications need thinner dielectric films in the range of 1-4 ~m and they
utilize plasma processing for patterning of finer lines (0.5-3 ~m in width).
The packaging applications on the other hand utilize thicker films, typically
4-6x thicker, and the dimensions are in the 1O-30-~m range.
The packaging applications, to a large extent, will specify the ther-
mal-stability requirement for the polymer whereas the number of poly-
meric layers determines the requirements for high-end mainframe applica-
11.3 POLYMER THIN-FILM REQUIREMENTS FOR PACKAGING 11-513

Table 11-2. Polymer Requirements for Thin-Film Interconnections

Function Requirement

Lithographic dimensions Aspect Ratio = (HeightlWidth)


Semiconductor device Aspect ratio of 1-2 in l-4-l.l.m films
Thin-film package Aspect ratio of 2-3 in 1O-25-l.l.m films
Thermal stability
Semiconductor device Withstand metal deposition and anneal temperatures
Thin-film package Withstand metallization, cure, and chip attach temperatures
Mechanical integrity Structure and substrate dependent
Minimize stress to enhance reliability
Process compatibility No cracking or crazing
Dimensionally stable
No delamination of thin films (good adhesion)

tions. In the case of semiconductors, the annealing of the metal


interconnects may impose temperatures higher than 400°C, but for shorter
time periods, and thus set the thermal-stability requirement for choice of
polymer dielectric to this temperature. The mechanical properties and low
stress requirements come mainly from the need to build multilayer stacks
of films with metals (Cu, Ni, Cr, Ti, Au, W, etc.) on top of low-strength
materials like silicon or ceramics. In the case of semiconductors, the total
film thickness will be much smaller than in the thin-film packages, so the
total stress thickness load imposed on the substrate will be smaller for
semiconductor applications. The stress-induced substrate-bow scales up
with the number of layers in the interconnect structure and the thickness
of each ofthe layers. Also, semiconductors can benefit more than packages
from a lower dielectric constant in reducing signal propagation delays.

11.3.1 Functional Compatibility and Reliability


The robustness of the systems can best be ascertained by subjecting
the materials to the processes utilized in the fabrication of the thin-film
structures on the substrates. In the fabrication of multilayer thin-film
wiring structures such as shown in Figure 11-2, one needs to ascertain
the adhesion of polymer to polymer, polymer to the various metals used
in the structure fabrication, metal to polymer adhesion, and polymer
adhesion to glass-ceramic, ceramic, silicon substrates, PWB, or glass.
The various interfaces, described above and previously shown in Figure
11-2, must not only hold together at the time of the initial build but also
after all of the stressing to which the structure is subjected [chemical,
11-514 POLYMERS IN PACKAGING

thermal, mechanical, and temperature and humidity (T & H) stressing].


The polymer dielectric layers must not crack, craze, or delaminate during
the thin-film building. They must be compatible with the processing
conditions selected for fabricating the multilayer thin-film structure for
interconnection of the integrated semiconductor chips or for interconnec-
tion on the semiconductor chip.
The key parameters that need to be addressed to assess the reliability
are adhesion of the various interfaces, thin-film stresses imparted by local
defects or by metal lines or vias used for making the interconnections,
mechanical properties before and after thermal cycling, polymer-solvent
interactions, compatibility with chip joining and wirebonding schemes,
and ease of fabrication. The polymer must be easily patterned to cleanly
open up the critical dimensions yielding the proper wall profile angles
needed for the subsequent metallization. T & H stressing has the greatest
potential deleterious effect for degrading adhesion and via continuity.
Once these are enumerated and the correlation of the fundamental physical
properties is made to the functional product requirements, one can improve
the materials and use the optimized polymer for the thin-film structure
or semiconductor dielectric interconnections. In summary, the polymer
requirements are thus obtained by matching the functional product require-
ments to the material physical properties. Table 11-3 summarizes in more
detail some typical generic polymer thin-film material requirements such
as lithography, thermal stability, mechanical property, and process condi-
tions as discussed above for high end mainframe applications.
An important aspect of these dielectrics that is receiving the greatest
attention recently is cost. The cost of typical high temperature polyimide

Table 11-3. Summary of Polymer Requirements

Function Requirements for Polymer Cured to 400°C

Lithography Excellent film-forming properties and pin-hole free 1-25 1J.I1l; ability to
rework film consistently before cure
Thermal stability Excellent thermal stability at 400°C
Mechanical Excellent mechanical properties:
Tensile strength ;: 150 MPa
Modulus ;: 3.0 GPa
Strain at break > 20%
Residual stress < 30 MPa
Self-adhesion (> 50 glmm)
Compatible with metals-No reaction
Process Solvent compatible to survive multiple-layer fabrication
Electrical Dielectric constant ~ 3.5, tanB < 0.03
11.4 POLYMERIC DIELECTRICS 11-515

is about a third to half of the cost of gold. Because of this and other high-
cost processes, tools, and facilities for forming thin-film packages, this
technology has not been commercialized beyond mainframes and super-
computers. There are several ways to address the cost of polymers for
thin-films; (1) by developing more efficient synthetic methods to lower
the cost of the polyimides, BCBs, PSPIs, etc. (2) by selection of low-
cost polymers such as olefins and epoxies in applications not requiring
the attributes (enhanced thermal and mechanical stabilities) exhibited by
preceeding list of materials or by (3) the use of large area processing
such as extrusion, or meniscus coating of the polymers that don't waste
the expensive polymer as in spin coating. These are discussed in more
detail in the thin-film chapter.

11.4 POLYMERIC DIELECTRICS

Today, there are many polymeric materials to select from for both
semiconductor and packaging applications. Some of the more common
and commercialized materials are shown in Table 11-4.
Of these dielectrics, polyimides remain the workhorse interlayer
dielectric for many semiconductor and packaging applications and have
been used the most since the late 1970s. Enhancements have been made
over the last decade to produce lower-stress, lower-dielectric-constant
polyimides, including the development of photosensitive polyimides
(PSPIs).

Table 11-4. Polymer Dielectrics for Thin-FUm Packaging

Polyimides (PIs) obtained from


Polyamic acids
Ester precursors
Acetylene terminated end groups
Low-stress derivatives
Non-Polyimides
Benzocyciobutenes (BeBs)
Polyquinolines (PQs)
Oletins
Photosensitive Polymers (negative and positive)
Epoxies
Preimidized PIs
Polyimides with covalently bonded cross-linking groups
Polyimides with ionically bonded cross-linking groups
11-516 POLYMERS IN PACKAGING

11.5 POL VIMIDES


Polyimides are condensation polymers derived from bifunctional
carboxylic acid anhydrides and primary diamines. Aromatic polyimides
exhibit outstanding mechanical properties with excellent thermal and oxi-
dative stability. It is well known that the mechanical properties, thermal
stability, adhesion, swelling, morphology, dielectric constant, coefficient
of thermal expansion, and residual stress are determined to a large extent
by the choice of the diamine and dianhydride components used as starting
materials. Polyimide materials, which have been synthesized and are
commercially available, have been largely made from the monomers
shown in Table 11-5, but today new backbone chemistries are being
developed to improve material performance and lower the material cost
[36,37].
Of the different combinations of monomer components that are
possible, there are basically three categories of homopolymer polyimides
that one can synthesize from the diamine and dianhydride combinations.
These can be categorized from a practical application consideration ac-
cording to the residual stress levels measured on fully cured films, corres-
ponding to low stress, intermediate stress, and high stress. Typical proper-
ties of fully cured films obtained from polyamic acid precursors of
BPDA-PDA (low residual stress with low TCE), polyamic acid or ester
precursors based on PMDA-ODA chemistry, and acetylene terminated
BTDA-type polyimides are summarized in Tables 11-6A and 11-6B. An
excellent review of the literature featuring the synthesis and properties of
polyimides can be found in reference [37-39]. Comparing commercialized
polyimides from Hitachi, DuPont, Ciba Geigy and National Startch and
Chemical Co.
Polyimides made from BPDA dianhydride have higher thermal sta-
bility than those made from PMDA- or BTDA-based monomers as illus-
trated in Figure 11-3. In a similar comparison, aromatic diamines such as
PDA have greater thermal stability than ODA, MDA, or alkyl-substituted
benzidine derivatives. The thermal stability hierarchy described above
based on monomer choice of dianhydride and diamine is corraborated by

Table 11-5. Monomers Used for Polyimide Synthesis

Dianhydrides Diarnines
- -
3,3',4,4'-Biphenyltetracarboxylic acid dianhydride M- and P-Phenylenediamine (PDA)
(BPDA)
Pyromellitic dianhydride (PMDA) 4,4'-Diarninodiphenyl ether (ODA)
Benzophenone dianhydride (BTDA) 4,4'-diaminodiphenylmethane (MDA)
11.5 POLYIMIDES 11-517

Table 11-6A. Summary of Chemistry Versus Features and Application

Polyimide Chemistry Features Applications

PMDAI ODA Excellent thermal/mechanical Dielectric, passivation


properties
PMDAlODA & BTDAlODAlMPD Copolymer with ease of processing Dielectric, passivation
BTDAlODAlMPD Improved adhesion, dielectric Delectric
constant, and planarity
BTDAlODAlMPD esterified High planarity, high solids, gap Alpha barrier
filling but reduced thermal protection
properties
BPDAlPDA Low TCE, low dielectric constant, Dielectric, passivation
low moisture uptake
6FDAlODA Lower dielectric constant, rellowsl Lift-off layer
cross-links removable
Preimidized Soluble removable coating, long- Die coating lift-off
term room-temperature solution layer
stability

the isothermal weight loss shown in Table 11-6B. This thermal stability
is seen as lower isothermal weight loss at elevated temperatures above
400°C and, more importantly, as retention of the original mechanical
properties after thermal stressing. Table 11-7 summarizes and compares
the mechanical, thermal and electrical properties of polyimides derived

Table 11-6B. Typical Commercial Polyimides

Typical Polyimide Properties


Intermediate-Stress High-Stress
Property Low-Stress Polymide Polyimide Polyimide

Polymer system BPDA-PDA PMDA-ODA-based BTDA-based


polymers and BTDA-based acetylene capped
copolymers polymers
Residual stress on 6-15 MPa 20-40 MPa 35-60 MPa
silicon
Moisture uptake < 1.0% 2-3% 2-3%
Dielectric constant 3.0 3.2-3.5 3.2-3.5
Isothermal weight loss at 0.03%/h 0.04%/h 0.17%/h
400°C in N2 or Air
Mechanical properties
Tensile modulus 9 GPa 2.5 GPa 2.7-3.8 GPa
Tensile strength 390 MPa 150-199 MPa 115-140 MPa
Strain at break 25-34% 90-130% 8-15%

Source: Data from Refs. 22-25.


11-518 POLYMERS IN PACKAGING

Dianhydrlde Component
o 0 o 0 000
II II II II II II II
o· C'rry_~.c
, c .l8J~..
c
'0
> o::~©:::o > o:c~©.c
c
,©:c: c
O

II II II II II II
o 0 o 0 o 0
BPDA PMDA BTDA

Diamine Component

H2N- Q>-o- Q>- NH2 >


m-PDA ODA

Q>-CH Q>-
H2N -

MDA
2- NH2
> DMBenzidine

Figure 11-3. Effect of Dianhydride and Diamine Components on Thermal Sta-


bility of Polyimides (BPDA>PMDA>BTDA and PDA>ODA>MDA) [40,41]

from a variety of chemistries to meet todays demanding electronic semi-


conductor and packaging applications. During manufacture of thin-film
structures, whether for semiconductor chip or package, the removal of
solvents from the solvent-based coatings followed by numerous high-
temperature heating and cooling cycles generates residual stresses. The
residual stresses produced by the difference in the thermal-expansion
coefficients of the materials and the moduli of the various material layers
can lead to film delamination, cracking of the metal or polyimide, and
bowing of thin substrates, resulting in poor photolithographic image trans-
fer. These are the main reasons why low-stress materials are highly desir-
able in thin-film fabrication [37].

11.5.1 Poly(amic alkyl ester) Development for Thin-Film


Application
In 1991, IBM introduced a glass-ceramic substrate with thin-film
redistribution as the first-level package utilizing polyimide-copper thin-
film redistribution to match the wiring grid of the glass-ceramic substrate
to the wiring and interconnection grid of the semiconductor chip as shown
in Figure 11-2 [34]. For this application a unique and novel polyimide
(PI), formed from a poly(amic ethyl ester) ofPMDA-ODA, was developed
and called M-PAETE [42].
Table 11-7. Comparison of Polyimide Film Properties Prepared from a Variety Monomer Chemistries Commercialized
by DuPont

Copolymer
Cured Film Homopolymer PMDAlODA &
Properties PMDAlODA BTDAlODAlMPD BTDAlODAlMPD BPDAlPDA
Units PI-2545 PI-2570 PI-2555 PI-2610D/llD

Mechanical
Tensile Strength kglmm2 10.5 10.5 13.5 35
Elongation % 25--40 40 15 25
Density gI cm3 1.42 1.42 1.45 1.46
Modulus kglmm2 140 140 245 845
Moisture Uptake % 2-3 2-3 2-3 0.5
U1
cD Stress 1 MPa 22 15 36 6
Thermal
Glass Transition Temperature °C >400 >400 >320 >400
Melting Point °C None None None None
Decomposition Temperature °C 580 580 550 620
Weight Loss (500°C in air, 2 hrs.) % 3.6 3.6 2.9 1.0
Coefficient of Thermal Expansion (0C}-I x 1Q-6 20 20 40 3
Coefficient of Thermal Conductivity cal/cm sec °C 37 x 10-5 37 X 10-5 35 X 10-5
Specific Heat callg/oC 0.26 0.26 0.26
Electrical
Dielectric Constant (@ 1kHz, 50% RH) 3.5 3.4 3.3 2.9
Dissipation Factor (@lkHz) 0.002 0.002 0.002 0.002
Dielectric Breakdown Field Vlcm >2xl(J6 >2x106 > 2 x 1(J6 > 2 x 1(J6
Volume Resistivity Q-cm >10 16 >10 16 >10 16 >10 16
Surface Resistivity Q >10 15 >10 15 >10 15 >10 15

Note: IStress was measured using a Tropell® Autosort Interferometer. Ref. DuPont product literature.
11-520 POLYMERS IN PACKAGING

For this application, the polyimide must withstand the processes


enumerated in Ref. 34 and 35: wirebonding, soldering with C4 (controlled
collapse chip interconnection), and deletion of signal connection with
laser energy. In addition, the polyimide must be compatible with and
survive the following processes: laser ablation, oxygen ashing, metalliza-
tion, planarization process, lift-off stencil fabrication and removal, and
metal etching.

11.5.1.1 Background
As previously described in section 11.5 it is well known that the
mechanical properties, thermal stability, adhesion, swelling, morphology,
dielectric constant, coefficient of thermal expansion, and residual stress are
determined to a large extent by the choice of the diamine and dianhydride
components used as starting materials. It is now known that the derivative
esters enhance the processing of the polyimides and that specific isomers
further enhance the solution properties. IBM took advantage of all these
characteristics to develop a unique class of PMDA-ODA derivatives and
their isomers. These poly(amic alkyl esters) exhibit hydrolytic stability
due to the elimination of the "monomer-polymer" equilibrium associated
with the poly(amic acids) which are commercially available and used by
others. In addition, the processing has been greatly improved because the
ester derivatives possess a broad and higher characteristic imidization
temperature regime, improved solubility characteristics, and enhanced
adhesion and mechanical properties.

11.5.1.2 M-PAETE Synthesis


IBM developed the polyimide chemistry and the scale-up synthesis
[42]. The preparation of poly(amic alkyl esters) is shown in Figure 11-
4. The diester diacid is obtained from the reaction of the aromatic dianhy-
dride with a suitable alcohol. The meta-isomer is recrystallized from the
prepared diester diacid. Then, the diester diacid is converted into a diester
diacyl chloride which reacts with an aromatic diamine via low-temperature
solution polycondensation in a polar, aprotic solvent like NMP (N-methyl-
pyrrolidone) to yield the desired poly(amic alkyl ester) [42,43]. The poly-
mer is then washed, dried, formulated, and filtered to produce a polyimide
solution which will meet the thickness requirements for thin-film applica-
tion. IBM decided to use the predominantly meta-isomer of PMDA-ODA
ethyl ester due to its better processing. This meta-isomer of the ethyl ester
of PMDA-ODA exhibits excellent coating properties (planarization, low
viscosity at high solids), and superior chemical properties (solution stabil-
ity, nonreactivity to copper metal). After curing, the physical properties
are comparable with those of fully cured polyimide derived from PMDA-
ODA polyamic acid solution. These properties are summarized in Table
11-8.
11.5 POLYIMIDES 11-521

0 0
• IO-C)©tC-OI
0 0

o~o
II II

"-<)©t'-OI
II

~ «(00)2

IO-C C-OI
~
a-c c-a
0 0 II 1/ II II
0 0 0 0
Dianhydride Die. DiIcid Ilesta' Diacyl 011_

o 0
1/ II

::)©t:: + I,.-<O)-O-@-'"
II II
o 0

n
PoIy(amic alkyl CSkr)

Figure 11-4. Typical Synthesis Scheme for Making Poly(amic alkyl ester)
Polymers

11.5.1.3 M-PAETE Solution Properties


In order to meet the film thickness requirements, polymer solutions
in the range of 24-31 wt. % in NMP were formulated from the dry powder
and filtered to 0.2-Jlm levels. The intrinsic viscosity of 47-64 ml/g of
the polymer corresponds to a molecular weight of 30,000-45,000. To
ensure quality, the polymer solutions were analyzed for trace levels of
contaminants, especially chloride ions and covalently bound chlorine. The
isomer ratio and solvent purity were controlled and monitored to ensure
solution stability for thin-film coating of substrates.

11.5.2 Characterization Techniques Used for Evaluation of Thin-


Film Polymeric Materials
The experimental techniques developed to fully characterize thin
films (10-20 Jlm thick) of polyimide to ensure manufacturability and
reliability of the thin-film redistribution will be reviewed next, illustrating
11-522 POLYMERS IN PACKAGING

Table 11-8. Summary of M-PAETE Properties

Meta-poly(amic Ethyl Ester) Precursor Cured to 400°C

Excellent film-forming properties


Planarizeslfills voids
Self-adhesion >100 g/mm
Compatible with Cu-No reaction
Excellent thermal stability at 400°C
Excellent mechanical properties
Tensile strength 160-190 MPa
Modulus 3.0 GPa
Strain at break >70%
Residual stress -30 MPa
Dielectric constant 3.5
Thermal-expansion coefficient (35-40) x 1O.6/" C
Solid powder-can reformulate
Shelf-life stability
Easier to remove alcohol by-product than water

Source: From Ref. 33

the methodology and material characterization used to develop and evalu-


ate the M -PAETE polyimide for the thin-film packaging application. These
techniques were developed for the characterization of thin films as defined
above and are generally used to characterize polymers for these type
of applications [33,38-45]. The same techniques have been applied to
characterize polymers for thin film packaging applications.

11.5.2.1 Mechanical Properties


The strength properties of the polyimides were obtained from stress-
strain diagrams, which describe the behavior of a homogeneous specimen
of uniform cross section subjected to uniaxial tension. The following
tensile stress-strain parameters characterize the tensile mechanical proper-
ties of polymeric materials: Young's modulus, yielding strength, yielding
strain, strain hardening point, stress at break, and strain at break. All these
measurements were initially performed on the meta-poly(amic ethyl ester)
precursor of PMDA-ODA, as it was being developed, to find the molecu-
lar-weight range which produced the best quality films, without crazing
or cracking when subjected to processing conditions.
The samples for mechanical testing were prepared by spin-coating
the solution of the polymer on silicon wafers, baking for 30 min at 85°C
on a hot plate in air, and curing in a convection oven under flowing N2
11.5 POLYIMIDES 11-523

Table 11-9. Mechanical Properties of Fully Cured Polyimide

Stress at Break Strain at Break


Cure Schedule (MPa) ~

A 400°C/30 min 156 54


B 400°C/60 min 186 79

with a four-step curing schedule (150°C, 230°C, 300°C, and 400°C). The
final cured film thickness was 10 11m.

11.5.2.2 Tensile Characterization


The average Young's modulus on the tested samples was approxi-
mately 3 GPa. The average ultimate tensile strength increased from 156
MPa to 186 MPa as the cure time at 400°C was increased from 30 to 60
min. This shows that the best mechanical properties are obtained when
the cure cycle is optimized. These results are summarized in Table 11-9.

11.5.2.3 Process Stressing


A summary of the effects of process conditions A to E on the tensile
properties, as determined by the changes measured in the strain at break,
are shown in Table 11-10. The stressing simulates chip attach, flux clean-
ing, lift-off, and other process treatment conditions. Mter thermal aging
at 400°C for 16 h, the modulus decreased by 10 MPa units and strain at
break decreased by 2 x. The stress and strain at break decreased slightly
upon T & H stressing. The films treated with NMP (85°C) for 60 min
and subjected to thermal aging exhibited about a 2 x decrease in the
strain at break. The strain at break of the specimen subjected to stressing
condition E exhibited about a 4 x reduction in strain at break.

Table 11-10. EtTect of Process Conditions on Tensile Properties

Stressing Condition Strain at Break

A. No stressing (control) Normalized to 1


B. 16 h at 400°C Decreased 2x
C.3OOhatT&H Decreased < V4X
D. NMP+B Decreased 2x
E. NaOH + NMP + xylene + B Decreased 4x
11-524 POLYMERS IN PACKAGING

The T & H exposure and 16 h of thermal annealing are accelerated


stress conditions used for testing purposes only. Because the tensile me-
chanical properties and toughness of the virgin polymer are so high, even
the changes noted in Table 11-10 and described above provide a large
safety margin (robust process) for the thin-film build from a mechanical
point of view.

11.5.2.4 Adhesion of Thin-Film Interfaces


Adhesion of the polyimides at the various interfaces shown in Figure
11-2 is important because this is what holds the structure together and
provides for the thin-film mechanical integrity.
Basic adhesion is the interfacial bond strength, which depends on
the interfacial properties or interactions (electrostatic, chemical, or van
der Waals) and mechanical interlocking [46]. As this adhesion cannot be
measured mechanically, one must substitute another method (e.g., practical
adhesion, which includes the basic adhesion modified by other factors).
Peel strength measurements are one method of measuring practical adhe-
sion and were used as a simple guide to assess interface strength.
The adhesion measurements were made at a peel rate of 5 mmlrnin
at a 90° angle as illustrated in Figure 11-5. Peel measurements were made
on the virgin films as well as on those that were exposed to various thin-
film processing conditions. Table 11-11 shows the adhesion values of
the virgin interfaces as measured by peel strength. These values are
considered excellent.

11.5.2.4.1 Polyimide-to-Polyimide Adhesion


According to Brown et al. [44], PMDA-ODA adhesion to itself is
controlled primarily by the diffusion of the precursor poly(arnic acid) into
a cured or partially cured polyirnide layer. This interdiffusion is controlled
by the cure temperatures of the two layers, with good adhesion resulting
from a low cure temperature of the first layer. Utilizing this concept, a
range of intermediate cures were found between 230°C and 260°C that
exhibited excellent adhesion (100-140 g/mm). The meta-poly(amic ethyl
ester) precursor is more easily partially cured, because it has a higher
onset temperature for irnidization and broader temperature regime where
imidization takes place [45]. Little or no degradation of self-adhesion is
observed after T & H exposure for up to 1000 h as shown in Table
11-12. This is important when one desires to build multiple polyirnide
layers.

11.5.2.4.2 Polyimide-to-Glass-Ceramic Adhesion


The adhesion of polyimide to glass-ceramic was improved by using
surface treatments. After the glass-ceramic was cleaned by oxygen plasma
11.5 POLYIMIDES 11-525

90deg

diced peel strips


2-4mmwlde

PEEL FORCE (F)


PEEL STRENGTH ...
PEEL STRIP WIDTH

Figure U-S. Schematic Diagram Showing Adhesion Peel Strength Measure-


ment Being Performed at a 90 Degree Angle from the Test Sample

ashing, an adhesion promoter (silanol amine) was applied prior to overcoat-


ing with the polymer film. The adhesion promoter bonds to the oxide
surface of the ceramic; this bonding can occur between the oxide and the
silanol group, or the amine group can react with the oxide surface. The
resulting adhesion is 95-110 g/mm.
11.5.2.4.3 Polyimide-to-Metal Adhesion
In the thin-film structure shown in Figure 11-2, the polyimide is
in direct contact with chromium. Polymer solution was spin coated on
chromium and then cured to 400°C, resulting in a peel strength of 80-90
g/mm. For polymer-to-copper adhesion, it is well known that poly(amic
acid) reacts with Cu, forming a polymer-Cu complex during the spin-
coating process. Copper oxide particles are precipitated in the polyimide
upon curing of the poly(amic acid) (imidization process) [47-49]. This
can cause an increase in the dielectric constant, thus favoring the use of
ester-type derivatives of the polymer backbone.
11.5.2.4.4 Metal-to-Polyimide Adhesion
The metal (chromium) to polymer peel strength is 76-90 g/mm. The
metal in this case is deposited over fully cured polyimide. The adhesion
11-526 POLYMERS IN PACKAGING

Table 11-11. Adhesion Values of Interfaces

Peel Strength

Polyimide-polyimide 100-140 glmm


Polyimide-glass-ceramic 95-110 glmm
Polyimide-Chromium 80-90 glmm
Chromium-Polyimide 76-90 glmm

is most affected by coating imperfections which can be reduced by modify-


ing the surface of the polyimide. Because the polyimide has excellent
coating properties, T & H stressing has shown the greatest potential effect
to reduce adhesion at this interface. Therefore, it is important to bake out
any residual water from the films to ensure reliability of this interface.

11.5.2.5 Dynamic Stress Measurement


Residual stress arises from the deposition processes and the mismatch
in the thermal coefficient of expansion of the polyimide dielectric layers
in the thin-film structure, which consists of metals and glass-ceramic. To
assess the development of residual stress in the polyimide, residual stress
measurements were dynamically performed in nitrogen ambient during
thermal curing of the dried films on wafers through the cure process and
subsequent cooling at I.O°C/min rate. The details of the stress measure-
ment technique, and calculations are reported in Ref. 50.
Films were prepared from the precursor solution on wafers, soft-
baked at 80°C for 30 min in a convection oven, and then thermally cured
by a four-step cure process: I50°C/30 min, 230°C/30 min, 300°C/30 min,
and 400°C/1 h. The thickness of fully cured films was about 10 Jlm.
Dynamic stress measurements are shown in Figure 11-6 over the
temperature range of 25-400°C. Upon heating, thermal imidization starts
with loss of solvent, but the residual stress continues to decrease with an

Table 11-12. Self-Adhesion as a Function of T & H Stressing

T & H Stressing Time Peel Strength

Oh 115-129 glmm
500 h 114-112 glmm
lO00h 105-122 glmm
11.5 POLYIMIDES 11-527

~~--------------------------~
PI/Si(1OO)

~ 30
-
~
(/)
(/)

~ 20
rn
+'

i
:J
'C
'in 10
II)
It

o 100 300

Figure 11-6. Residual Stress of Poly(amic alkyl ester) of PMDA-ODA as a


Function of Temperature upon In-situ Heating and During In-situ Cooling
of PI Film

abrupt jump in stress to 11 MPa at - IS0°C. This suggests that the effective
softening temperature of this soft-baked film is - IS0°C. This residual
stress then decreases and at -180°C gradually continues to decrease as
more solvent is removed in spite of imidization occurring over this large
temperature regime. The residual stress decreases to -1 MPa at 300°C
and with continued heating to 400°C, the stress is unchanged, being < 1
MPa. The stress is very low over a large temperature regime during the
curing process. This is a desirable condition for the polyimide in contact
with the other materials, comprising the thin-film structure.

11.5.2.6 Dynamic Mechanical Properties


Dynamic mechanical thermal properties (storage modulus E' and
loss modulus E") were measured at a heating rate of 1O°C/min and a
frequency of 10 Hz in nitrogen ambient over the range of 2S-S00°C.
Figure 11-7 shows the change in the dynamic storage modulus from room
11-528 POLYMERS IN PACKAGING

10.0,----------------.

-
m- E'
Q.
•
"-=..
W9.0
C)

.9

0)
."" 111111 I 71"1.11111 .SII I I I.
...
S 10Hz

7.0
0 100 200 300 400 500
Tee)
Figure 11-7. Variation of Dynamic Storage Modulus E' and Loss Modulus
E" as a Function of Temperature from RT to 500°C for a Fully Imidized
Film Showing a Tg

temperature to 500°C. The storage modulus gradually decreases from


room temperature to -3S0°C and shows a broad glass-transition region
with a Tg of -400°C. This high- Tg provides the dimensional stability
needed to fabricate the thin-film structure and withstand the chip attach
C4 thermal cycles.
M-PAETE was designed to imidize at significantly higher tempera-
tures than its poly(amic acid) counterpart for ease of processing and
is devoid of hydrolytic instability. M-PAETE possesses excellent film-
forming properties, adhesion, thermal stability, mechanical properties,
superior solution properties, and the other processing attributes summa-
rized in Table II-S.

11.6 COMMERCIAL PREIMIDIZED POL YIMIDE FILMS


Whereas fibers can be considered to be unidirectional by virtue of
length, self-supporting films are bidirectional with the added dimension
11.6 COMMERCIAL PREIMIDIZED POLYIMIDE FILMS 11-529

of width. Both structures, of course, have a thickness dimension, and


flexible films for use in microelectronics are generally in the range of
25-125 /..lm. A combination of physical toughness, chemical and thermal
resistance, as well as excellent electrical insulating properties make many
films such as those from all-aromatic polyimides, attractive as dielectric
freestanding film substrates in microelectronic applications.

11.6.1 Desired Properties for Microelectronic Substrates

Advances in microelectronics toward higher circuit density have


required the dielectric substrate to be more compatible with active and
passive components, so as to achieve functionality during end use. One
of the key physical properties is the coefficient of thermal expansion
(TCE) , which must be close to the value of the component with which
it is in contact. Another is the coefficient of hygroscopic expansion (CRE),
which must be as low as possible to avoid stresses during relative humidity
changes. Also, electrical properties must prevent signal distortion or loss.
In general, the dielectric should be thought of as a passive medium to
support and protect circuitry, but otherwise it is inert.
Additionally, the dielectric substrate must have properties that allow
it to sustain the rigors of processing, as well as to abet processing. Conse-
quently, freedom from melting or severe distortion are required. The
material should not shrink during manufacture, otherwise it will induce
stresses or misregistration of circuits in multilayers. Physical strengths
such as high moduli and high tenacities are important to prevent distortion
under tension. Curl and flatness are essential to enable access to equipment
without jamming or poor fitting, or to facilitate further processing such
as installation of semiconductor chips. Chemical milling, typically with
caustic solutions, is an advantage, so as to remove unwanted areas in the
final circuit or to provide holes (personality holes or vias) for electrical
or mechanical connections.
Many other detailed requirements of dielectric substrates for micro-
electronics are available, and often requirements vary significantly depend-
ing on the specific application. For example, in tape automated bonding
(TAB) caustic etchability might be important, but for flexible intercon-
nects, MIT fold endurance might be a demanding need.
Another requirement of the dielectric substrate is that it must adhere
well to adjacent materials, whether they be active or passive components,
otherwise voids will develop during processing with resulting deterioration
in mechanical and electrical integrity. Adhereability is often ignored in
stating properties, and although it is a property that will be highly depen-
dent on the application, it is often critical to functionality. In summary,
the reader might consult other references for appreciation of needs of
dielectric substrates. A number of articles are available that describe these
11-530 POLYMERS IN PACKAGING

Table 11-13. Properties of an Ideal Substrate for TAB Applications

~ Current Requirement Ideal Requirement

Shrinkage, 200°C (%) 0.01 Nil


TCE, 35-250°C (ppm/0C) 17 As required
Modulus, 23°C (GPa) 5.2 10.0
H20 absorption % at 90 rh 1.5 0.0-0.1
Chemically millable Yes Yes
CHE, 30-70 rh (ppm/rh) 10 Nil
Adhereability (kg/cm) 1.4 2.1

needs. Examples are the account by Arnold et al. [51], and for specific
needs of TAB, the comprehensive descriptions by Holzinger [52,53], and
Monisinger [54].
Because polyimide films have an excellent balance of thermal dura-
bility and physical/dielectrical integrities over wide ranges of temperature
and humidity, they have found increasing applications in microelectronics.
The properties of these films are undergoing continual improvement be-
cause of the ever-increasing demands on them as substrates. A summary
of the current property requirement levels for the time frame of the next
few years is given in Table 11-13. Along with these immediate goal levels
is given an ideal requirement list that might only be attainable with
dielectric substrate structures that are considerably altered from what we
now know of as linear aromatic polyimide films.

11.6.2 Manufacturing Methods


Commercial preparations of aromatic polyimide films are essentially
variations of a two-step sequence expanded from adaptations of laboratory
syntheses. They involve (1) preparation of a polyamic acid in a dipolar
aprotic solvent (e.g., N-methylpyrrolidone) and (2) either chemical or
thermal methods of imidization [55]. If the glass transition of aromatic
polyimides is under 300°C, it might be possible to alter the film-forming
step by direct melt extrusion or even alter both steps by melt polymerization
and melt extrusion. However, significant progress toward viable manufac-
turing processes utilizing melt polymerization/melt extrusion have not
been made and have not as yet afforded films for microelectronic pack-
aging.
The chemical method of manufacture involves the addition of an
anhydride, usually acetic anhydride, and a tertiary amine, such as pyridine
or triethylamine, to the solution of polyamic acid. The film must be formed
11.6 COMMERCIAL PREIMIDIZED POL YIMIDE FILMS 11-531

rather rapidly, because imidization proceeds quickly at room temperature


to give an insoluble polyimide/polyamic acid. The polyimide film must
then be heated at high temperature under restraint to complete imidization
and drying.
The thermal method consists of casting the polyamic acid onto a
support and evaporating the solvent with gradual heating. Finally, the
partially converted film of polyimide/polyamic acid is separated from the
support and heated at high temperature under restraint to complete the
curing process of imidization and drying.
Properties of films from the chemical method of manufacture can
be different than those from the thermal method of manufacture, despite
the fact that the backbone structures are the same. It is outside the context
of this handbook to delve into this subject in more detail, except to point
out that processing is just as important as polymer backbone in defining
properties. Orientation of polymer chains, isotropy, and stress can have
profound influences on thermal shrinkage, in-plane TCE, and directional
physical strengths [40,41].

11.6.3 Commercial Sources, Types, and Chemical Compositions


of Preimidized Polyimide Films

The availability of commercial polyimide films stems from 1965,


when DuPont de Nemours & Co. Inc. began large-scale production of
Kapton@ polyimide film. Since that time, Kanegafuchi Chemical Industry
Co., Ltd. and Ube Industries, Ltd. have each commercialized polyimide
films in the time frame of the 19S0s.

11.6.3.1 Kapton® Polyimide Films

The DuPont products are listed under the trademark Kapton@ poly-
imide film. Detailed property information and availability can be obtained
from DuPont High Performance Films and DuPont Japan Ltd.
Films available for microelectronic applications are types H(N),
V(N), E(N), and K(N). The optional letter (N) designates the presence
(N included) or the absence (N excluded) of an inorganic slip additive
in the thousand part per million range to lower the coefficient of friction
and to aid winding, such as roll to roll transfer. For most applications,
the slip additive is present and the "N" designation is noted. The type
H(N) and V(N) films consist of the raw materials, pyromellitic dianhydride
(PMDA) and 4,4' -diaminodiphenylether (or 4,4'-oxydianiline; ODA) as
illustrated in Figure II-S. The difference between H(N) and V(N) is that
the V(N) has superior dimensional stability [i.e., less shrinkage than H(N)
after heating as high as 300°C].
11-532 POLYMERS IN PACKAGING

~o-O-1:
o 0 n
Figure 11-8. PMDAlODA-H(N); V(N); AV

The type E(N) films also contain PMDA and ODA, but other raw
materials are also present. These consist of the dianhydride 3,4,3',4'-
biphenyldianhydride (BPDA) and 1,4-diaminobenzene (or paraphenylene
diamine; PPD), as shown in Figure 11-9.

~!i-Q---~
o 0 0
o-O-~ n

Figure 11-9. PMDAlBPDAlPPD/ODA-E(N) Random distribution of dianhydridesl


diamines ratios of dianhydrides and diamines unequal

Type E(N) films are tetrapolyimides. The effects of various monomer


mole percentages on properties have been discussed, but precise percent-
ages in the commercial film have not been divulged [56]. The type E(N)
films are designed to provide higher modulus, lower TeE, and lower
moisture absorption than those of types H(N) or V(N). Advantages of
type E(N) films are insignificant distortion of thin films during processing,
more compatibility with conductive layers like copper, and less suscepti-
bility to blistering during component attachments. The E(N) films have
also been structured so that they will caustic etch.
Type K(N) films consist of terpolyimide that is based on the raw
materials of PMDA, ODA, and PPD, but the mole percentages of the
diamine components have not been published. The intent of the K(N)
product is to provide a film that has a modulus between types H(N) or
V(N) and E(N), a TeE that matches copper, and the facility to etch easily
in caustic. Moisture absorption is higher than types H(N) V(N), and E(N).
A general structural formula for the type K(N) films is given in Figure
11-10.

~tr-Q-~crO-~
o 0 0 0 n

Figure 11-10. PMDAlPPD/ODA-K(N); NP Ratios of diamines unequal


11.6 COMMERCIAL PREIMIDIZED POL YIMIDE FILMS 11-533

11.6.3.2 Apical® Polyimide Films


Kanegafuchi polyimide films are commercialized under the trade-
mark Apical® polyimide films. These products were originally made in
Japan, but in the last few years an additional manufacturing facility has
been established in the United States. Detailed product information and
availability can be obtained from Allied-Apical Co.
Apical® polyimide films that are available and would have potential
use in microelectronics are Types AV and NP. The type AV is generically
the same polyimide backbone as is contained in Kapton® types H(N)
and V(N), which is shown in Figure 11-8.
The type NP is similar in structure to Kapton® type K(N), which
is given by Figure 11-10, but the exact mole percentages of 4,4' -diaminodi-
phenylether and paraphenylene diamine have not been disclosed. It is
speculated that the polyimide film designated by "NP" has an ordered
structure that can be considered as a somewhat blocked terpolyimide,
but no current documentation is available. The similarity in structure of
Apical® type NP as compared to Kapton® type K(N) cause speculation
that the properties will be similar and the actual properties, given below
(vida infra), supports this.

11.6.3.3 Upilex® Polyimide Films


Upilex® poly imide films are made by Ube Industries, Ltd. in Tokyo,
Japan and they are marketed in the United States by Uniglobe Kisco, Inc.
The film manufactured by Ube for use in microelectronics is desig-
nated as Upilex® type S. It is based on 3,3',4,4'-biphenyltetracarboxylic
dianhydride (BPDA) and paraphenylene diamine (PPD) and the structural
formula is given in Figure 11-11. The backbone is very rigid and, conse-
quently, the film therefrom is characterized by a high modulus and low
TCE. The film also absorbs very little water, and although it can be
hydrolyzed back to the component raw materials with strong caustic at
high temperature, it is for all practical purposes nonetchable.

Figure 11·11. BPDAlIPPD-S

11.6.3.4 Properties of Commercial Polyimide Films


All of the polyimide properties described in this section are based
on films of 25 !lm andlor 50 !lm thickness [57]. All manufacturers supply
their products in thicknesses of 12.5 !lm, 25 !lm, 50 !lm, 75 !lm, and 125
11-534 POLYMERS IN PACKAGING

/lm, and most typical properties are not grossly affected by thickness of
the films. A striking exception to this behavior is the dielectric strength,
and the values for various thickness are given, if values were reported
by the manufacturer. For complete property information over the entire
thickness range of films available, the reader is directed to the manufacturer
and the likelihood that specific property/gage data are available.
The property data shown in Tables 11-14 through 11-17 are taken from
the various manufacturers' property bulletins or from papers by manufactur-
ers' where their products are described [56,57]. No attempt has been made
to misrepresent a product by excluding property data. For data not present
herein, the reader is advised to contact the specific producer.

Table 11-14. Typical Properties for 25- p.m-Thick Films

Kapton® Apical® Upilex®


H(N) +
Property V(N) K(N) E(N) AV NP S Method

Melting point (0C) None None None None None None ASTM·E794
Specific heat 1.09 1.09 1.13 DSC'
J/gOK (cal/g/°C) (0.261) (0.27)
Tgb ("C) 410 >400 359 410 >500 DMA'
Shrinkage (%) IPC-TM650; 2.2.4A
30 min/150°C 0.17 H(N) 0.05
2 hl250°C 0.03 V(N)d 0.1 0.37 JIS C2318
2 hl4oo°C 1.25 1.1 0.39 ASTM-D5214
TCE (ppm/°C)
20-1OO°C 18 12 8.0 TMA'
l00-200°C 31 14 10.0 TMA'
200-300°C 48 16 16.0 TMA'
300-400°C 78 24.0 TMA'
20-250°C 35 17 14 14.0 TMA'
Aammability 94 v-o 94 V-O 94 V·O UL-94
Limiting O 2 Index 37 48 66
Smoke gen. D=<1 0.4 NBS-Kapton
JIS-DROI-Uplilex

Note:
Values for coefficients of thermal-expansions (TCEs) are only meaningful after films have been
preexposed to as high a temperature as the TCE is desired to measure and under no restraint. The
purpose is to remove shrinkage.
'DSC = differential scanning calorimetry.
"Tg = Glass-transition temperature.
'DMA = dynamic mechanical analysis.
dV(N) films are thermally stabilized to give low shrinkage.
'Thermal mechanical analyzer-exact methods unknown. The Kapton films were initially heat sta-
bilized; it is not known if Upilex films were preheated.
Table 11-15. Mechanical Properties-Typical Values at 23°C-25°C for 25-JA-m-Thick Films

Kapton® Apical® UPILEX®

Property Unit H(N2+V(N2 K (N2 E (N2 AV NP ~ Method

Tensile strength MPa 231 306 165 227 392 ASTM-D882


psi 3.4 x 104 4.5 X 104 2.4 X 104 2.3 x 10' 5.7 x 104 ASTM-D882
Stress at 5% elong MPa 90 254 ASTM-D882
psi 1.3 x 104 3.7 X 104 ASTM-D882
<11
Elongation % 72 50 45 60 50 30 ASTM-D882
w
<11 Tensile modulus OPa 2.5 3.4 5.1 8.6 ASTM-D882
psi 3.7 x lOS 5 X 103 7.5 X 10' 1.28 X 106 ASTM-D882
Tear strength
Graves N 7.2 7.9 ASTM-DlOO4
(Initiation) lbf 1.6 1.8 ASTM-DlOO4
Elmendorf N 0.07 0.07 0.07 0.08 ASTM-DI922
(Propagation) lbf 0.02 0.02 0.02 0.02 ASTM-DI922
MIT folding Cycles 2.9 x 10' 1.5 X 104 >lxlO4 ASTM-2176
endurance
Density glcm3 1.42 1.44 1.46 1.47 ASTM-DI505
Coeff.-Kinetic 0.48 0.40 ASTM-Dl894
Friction
(film-film)
Table 11-16. Electrical Properties-Typical Values at 23-25°C for 25-125-f.1.m-Thick Films

Kal1ton® Al1ical® JlI1.ilex®b


Property; Test
(Unit) Gauge Condo H~N} + V~N} K(N} E(N} AV NP §. Method

Dielectric strength (kVImil)' 60 Hzb


25 fJ.m 7.7 7.0 6.0 6.0 6.8 ASTM-DI49
50 fJ.m 6.1 6.5 5.0 5.0 5.1 ASTM-D149
75 fJ.m 5.2 4.5 4.5 ASTM-DI49
U1 125 fJ.m 3.9 3.0 3.0 ASTM-DI49
w
0> Dielectric constant I kHz
25 fJ.m 3.4 3.4 3.2 3.9 3.9 3.5 ASTM-DI50
5OfJ.m 3.4 3.4 3.2 3.9 3.9 3.5 ASTM-DI50
Dissipation factor I kHz
25 fJ.m 0.0018 0.0018 0.0018 0.0035 0.0035 0.0013 ASTM-D150
5OfJ.m 0.0020 0.0020 0.0020 0.0035 0.0035 0.0013 ASTM-DI50
Volume resistivity (Q/cm) 100 V DC
25 fJ.m 1.5 x 10 17 10 17 10 17 10 12 10 12 1017 ASTM-D257
50 fJ.m 1.5 x 10 17 10 17 10 17 10 12 10 12 10 17 ASTM-D257

'Various thicknesses shown for dielectric strength, because this property is very dependent on thickness.
bData on Upilex S obtained at 50 Hz.
Table 11-17. Chemical Properties of Kapton Type H(N) and V(N); 25-fLm

Strength Elongation
Property Retained (%) Retained (%) Test Conditions Test Method

Chemical Resistance
Isopropyl alcohol 96 94 10 min at 23°C IP TM-650
Tuolene 99 91 Method 2.2.3B
Methyl ethyl ketone 99 90
Methylene chloride!
trichlorethylene (I: I) 98 85
2 N Hydrochloric acid 98 89
U1
2 N Sodium hydroxide 82 54
!:l Fungus resistance Non-nutrient IPC TM-650
Method 2.6.1
Moisture absorption 1.8% 50% rh at 23°C ASTM 0-570-81 (1988)'
2.8% Immersion for 24 h at 23°C (73°F)
Hygroscopic coefficient 22 ppml% RH 23°C (73°F), 20-80% rh ASTM 0-1434-82 (1988)'
of expansion
Permeability
Gas mUm', 24h, MPa cm3!(l00 in' 24 h atrn) 23°C (73°F), 50% rh ASTM 0-1434-82 (1988)'
Carbon dioxide 6,840 45
Oxygen 3,800 25
Hydrogen 38,000 250
Nitrogen 910 6
Helium 63,080 415 ASTM E-96-92
Vapor glm'!24 h gI(lOO in' 24 h)
Water 54 3.5
continued
Table 11·17. (Continued)

Strength Elongation
Property Retained (%) Retained (%) Test Conditions Test Method

Resistance to
10% Sodium hydroxide 80 60 95 Immersion at 25°C for 5 days ASTM-D882
Glacial acetic acid 100 95 100 Immersion at 110°C for 5 weeks ASTM-D882
p-Cresol 90 90 95 Immersion at 200°C for 3 weeks ASTM-D882
Water pH = 1.0 95 85 100 Immersion at 100°C for 2 weeks ASTM-D882
pH = 4.2 95 85 100 Immersion at 100°C for 2 weeks ASTM-D882
pH = 8.9 95 85 100 Immersion at 100°C for 2 weeks ASTM-D882
pH = 10.0 95 85 100 Immersion at 100°C for 4 days ASTM-D882
Water absorption 1.2% Immersion in water at 23°C for 24 h ASTM-D570
'"w
00 0.9% Equilibrium at 60% rh, 50°C ASTM-D570
Gas permeability
Water vapor 1. 7 g/m'/mil At 38°C, rh 90% for 24 h ASTM-E96
Oxygen 0.8 mllm'/mil
Carbon dioxide 1.2 mllm'/mil At 30°C, I atm for 24 h ASTM-DI434
11.7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS 11-539

11.7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS


Materials not based on PI chemistries are summarized below.

11.7.1 Benzocyclobutenes
Benzocyclobutenes are being used commercially as insulators for
chip passivation and stress buffering, area array bumping, liquid crystal
display and single-chip and multichip thin film packaging applications.
Benzocyclobutenes, BCBs [58-60] are a family of thermoset resins
commercialized by Dow Chemical under the tradename Cyclotene™. The
chemical structure of generic BCB (I) is shown below. Altering the bridg-
ing "R" group results in a multitude of possible BCB structures. The first
BCB to be commercialized is siloxy containing DVS-BCB (II). The resins
are supplied partially polymerized, "B-staged", to obtain appropriate vis-
coelastic handling properties. The Cyclotene™ 3000 series is formulated
for dry etchability, the 4000 series is formulated for photo definition (see
section 11.12). The 3000 series require no refrigeration. They are stable at
room temperature, showing no change in viscosity after months of storage.
The polymerization (curing) of BCB proceeds through a two-step
process: a thermally driven ring opening followed by a 4 + 2 Diels Alder
reaction producing a structure similar to III. Curing is typically carried
out at 21O-250°C, several hundred degrees lower than is possible for
other thin-film dielectric materials.

R~R'
III
~R"
11.7.1.1 Electrical Properties
The dielectric constant (e r) of Cyclotene 3022 is 2.65 ± 0.05 and is
essentially invariant with temperature and frequency [59] as shown in
Figure 11-12.
Detailed electrical analyses of CulBCB interconnect structures have
been published [61-64]. High-frequency transmission line studies at 26
GHz [61] and 50 GHz studies on HFET ansd HBT amplifiers [64] have
been reported. Electrical properties are summarized in Table 11-18.
11-540 POLYMERS IN PACKAGING

3.4 -.-Tl---r"T1rrrrrrTTtTlrr-,--,--r---r-r--r--T--,

3.0

2.8

Temperature (0C)
30 Frequency (Hz)

Figure 11-12. Dielectric Constant of CycIotene™ 3022 [59]

11.7.1.2 Mechanical Properties


The mechanical properties of 25 Jlm BCB thin films are shown in
Table 11-19.

11.7.1.3 Optical Properties


The optical properties of fully cured films of BCB reveal low birefrin-
gence and low transmission loss [65]. Loss in passive wave guides, fabri-
cated from BCB is more related to feature roughness due to fabrication
technique (etching technology) than to the inherent optical properties
of the BCB dielectric. Interfaces created in photodefinable BCB reveal
smoother surfaces and thus lower losses than the dry etchable grade.
Optical properties are listed in Table 11-20.

11.7.1.4 Water Absorption


The bakeout cycles, normally employed with other dielectric materi-
als having significant moisture absorption in order to avoid blistering and

Table 11-18. Electrical Properties of CycIotene™ Resins

Dielectric Constant (I KHz-I MHz) 2.65


(I GHz-20 GHz) 2.5
Dissipation Factor (I KHz-I mhz) 0.0008
Breakdown Voltage (V/cm) 3 x 106
Volume Resistivity (ohm/cm) I X 10 19
11.7 HIGH TEMPERATURE NON·POLYIMIDE DIELECTRICS 11-541

Table 11-19. Physical and Mechanical Properties of Cyclotene

Cyclotene Tg' Tensile Modulus Tensile Strength Elongation


Series rg (GPa) (MPa) ~

3022 52 >350 2.9 87 8

'Dependent on % cure.

delamination, are unnecessary with BCB. Water absorption for DVS-BCB


is limited to < 0.2% after exposure to high humidity or boiling water [66].
Figure 11-13 shows moisture absorption and desorption rates when
humidity is increased to 81 % and then decreased to 6% at ambient temper-
ature.

11.7.1.5 Thermal and Thermo-oxidative Stability


BCB's are inherently more susceptible to oxidation than fully aro-
matic polyimides [58]. To stabilize BCB to thermo-oxidative conditions,
Cyclotene resins are formulated with an antioxidant. The thermal stability
(determined under nitrogen) and the thermo-oxidative stability (deter-
mined in air) have been studied extensively [67]. Oxidation occurs with
an activation energy of 25 ± 1.5 Kcal per mole.

11.7.1.6 Planarization
Thin film coatings of BCB exhibit excellent planarization properties,
thus eliminating photolithiographic resolution issues caused by the genera-
tion of relief when poorly planarizing materials are used to fabricate

Table 11-20. Optical Properties of CycloteneTM 3022

Optical Property Cyclotene™ 3022

Birefringence <0.002
Loss @ 835 nm (dB/cm) 0.08
Refractive Index
(633 nm) 1.5584
(835 nm) 1.5473
(1320 nm) 1.5398
11-542 POLYMERS IN PACKAGING

.WaterUp~~r-~~~__-+__-+__-r__~__+
(%)
0.12

0.10
1
0.08

0.06

0.04

0.02

o 200 400 600 800 100 120 140


Time (sec) 0 o 0

Figure 11-13. Water Absorption I Desorption for Cyciotene™ 3022 [66]

multiple levels of interconnect. The degree of planarization is dependent


on the film thickness versus the feature thickness, the number of coatings,
the feature width, height and spacing [68]. Typical planarization versus
feature width is shown in Figure 11-14.
In conformal via structures, BCB dielectrics not only planarize under-
lying conductor layers but also planarize the conformal vias, so that the

1.5

.,cu
0 1.0
0::
.,c:cu
0
0.5
N
·c • 4 Ilm Polymer Film
cu
c: • 111lm
cu 0.0
0:: A 161lm

-0.5
1 10 100 1000
Aluminum Line Width (!-1m)
Figure 11-14. Cyciotene™ 3022 Planarization Over 4 f.Lm Thick Isolated Alu-
minum Lines [68]
11.7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS 11-543

next orthogonal layer of conductor can be routed directly over the vias.
Under identical conditions Cyclotene 3022 showed 65-70% planarization
of via features versus 10-15% for typical polyimide dielectrics [69].
Planarization issues are becoming especially important in the manu-
facture of Si and GaAs multilevel metal structures [70,71].

11.7.1.7 Adhesion
A siloxy containing adhesion promoter such as [3-Aminopropyltrie-
thoxysilane, APS] or [3-methacryloxypropyl-trimethoxysilane, MOPS] is
recommended to enhance the adhesion of BCB to various inorganic inter-
faces. BCB adhesion to silicon oxide (wafer), alumina, aluminum nitride
and to itself has been evaluated by ASTM tape tests, "stud pull" testing
after water boil, PCT (pressure cooker testing) and microindentation test-
ing. Such interfaces are reported to show reliable adhesion (zero failures
in a cross hatch tape peel test) after PeT. Adhesion of Cyclotene to
chromium is reported to be similar to both dry etch and photosensitive
PIs [72]. Adhesion to gold or copper is enhanced by a < 500 A layer of
Ti, but not required for copper structures in order to pass reliability tests
[64]. Adhesion to copper is also enhanced by texturing the surface with a
microetching or black oxide treatment similar to PWB processing [73,74].

11.7.1.8 Thin Film Stress


Tensile stress developed during oven or belt furnace (RTC, "rapid
thermal curing"), curing of BCB films is typically measured at 28-32
MPa for oven curing and 32-38 Mpa for belt furnace curing [75,76].

11.7.1.9 Metal Migration


Copper diffusion is not observed when BCB is deposited directly
on copper metallization [77,78] even after 1500 hours at 85% RHl85°C
in contrast to results with polyamic acid based polyimides [78].

11.7.1.10 Cyclotene 3000 series Processing:


Processing of the dry etch 3000 series will be covered here, process-
ing of the photoimageable 4000 series will be covered in section 11.12.

Deposition: The Cyclotene™ 3000 series have traditionally been


applied by spin coating in thicknesses of <1 to 20+ /..lm. Shrinkage on
cure is typically <5%, thus the ultimate thickness needed for each coating
is achievable in one spin coat operation by choice of the appropriate
resin concentration. This can significantly affect high density substrate
fabrication costs [79]. High efficiency coating techniques including spray
11-544 POLYMERS IN PACKAGING

coating, extrusion coating, and miniscus coating have been successfully


evaluated on glass, aluminum and laminate substrates [74,80].

Curing: Extensive kinetic studies have revealed a curing activation


energy, Ea, of 36 ± 0.9 Kcallmol [81]. No volatiles are evolved upon
cure thus shrinkage is low « 5%). Curing is best monitored by FJ'-
IR spectroscopy [82]. Curing must be carried out in an inert (nitrogen)
atmosphere « 100 ppm O2 recommended). Curing can take place in a
typical box oven, on a nitrogen purged hotplate or in a controlled atmo-
sphere belt furnace. Time/temperature conversion curves (Fig. 11-15)
which describe the curing process have been developed from the kinetics
of the reaction. At vitrification (material solidification) the reaction slows
down an order of magnitude but continues in the solid state. Processing
to partial (ca. 70-80%) cure is recommended to achieve the best layer
to layer adhesion [72,77].
Full cure can be obtained on a time scale of minutes to hours by
controlling the curing temperature. In a typical convection oven "soft"
curing is achieved in 30 minutes at 210°C and a full "hard" cure is
obtained in ca. 1 hour at 250°C. Typical cure schedules with ramp up

350

% Cure

--
00
300

!:s
250
!CD
a.
E
CD
I-
200

150~~~--1-~--~-+~T-~--~~

3.6 Sec 36 Sec 6 Min 1hr 10hr 100hr

Time
Figure 11-15. Time / Temperature Conversion Derived from Curing Kinetics
11.7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS 11-545

and cool down take ca. 4 hours to complete. Rapid thermal curing (RTC),
Figure 11-16, in a belt furnace decreases typical cure times from 4-5
hours to <15 minutes and shows no deleterious effect on stress, adhesion
or planarization [75]. RTC on the hot plate of a track coater (under
nitrogen) has also been described [76].

Patterning: Vias can be formed in films of the BCB dielectric by


any of the following techniques:

Photodefinition: Photosensitive BCB formulations are described in


section 11.7.12

Dry Etching: Due to the siloxy content of Cyclotene plasma or


RIB etching requires fluorinated plasmas such as 02/CF4 or OzlSF6 [83].
Masking can be accomplished with photoresist soft masks [86] or metal
or Si02 [83] or Si3N4 [85] hard masks. For soft masked vias, via side wall
angles are controlled by the degree of baking of the photoresist; for hard
masked vias the side wall angles and degree of undercut are controlled
by the plasma conditions.

Laser Ablation: The use of scanning eximer laser ablation through


a hard metal mask has been successfully used to create vias in DVS-BCB
[69,86] with throughputs comparable to RIB based processes. 5 micron
vias with aspect ratios of 0.3 have been reportedly generated in BCB

300 100

80

-=
.-.
200 ~
60
.f!e
40 =
=
e
100 U

20
IRFurnace
Convection Furnace
o 0
0 1 2 3 4 5 6
Time (minutes)

Figure 11-16. Rapid Thermal Curing in Infrared or Convection Belt Fur-


naces [75]
11-546 POLYMERS IN PACKAGING

using this technique [72]. Figure 11-17 shows the ablation rate of Photo-
BCB material as a function of laser fluence [87].
The residue (carbonaceous debris) which was left on the surface
surrounding the ablated features can be removed using an oxygen/fluorine
plasma and/or isopropanol immersion rinsing (with ultrasonication).

0.80 -r------------.
0.70 Nd+3 YAG
,-..
~ 0.60
'E!
j 0.50
';;' 0.40
~=0.30
o
'iI 0.20
:a
« 0.10

0.00 -T4Ih-Ir-T""1r-r-r-r-r-r-r-r-r-r-I""T""I""T""r-f

o~~§~§~~§~§
-----N
Laser Fluence (mJ/cm2)
Figure 11-17. Ablation Rate of Photo-BCB vs Laser Fluence [87]

11.7.1.11 Incorporation of Passive Components


Thin film TiN and NiCr resistors and AIISiONlTiN capacitors have
been fabricated as part of thin film BCB structures [61,88,89].

11.7.1.12 Applications
Figure 11-18 shows a Commercial 486 Computer Module System
fabricated by Saab/Combitech using Cyclotene 3022. It contains an In-
tel486™ microprocessor, 1 Mbyte flash PROM, 0.5 Mbyte SRAM and
ASIC devices for memory management and 110 handling. It has been
qualified by MIL H 38534 [89,90].

11.7.2 Polyquinolines
Polyquinolines (PQs) are linear-chain high-molecular-weight poly-
mers characterized by moderate to high-class transition temperatures,
excellent thermal and thermooxidative stability, low dielectric constants,
and good planarizing characteristics. They were invented by Stille and
11 .7 HIGH TEMPERATURE NON-POLYIMIDE DIELECTRICS 11-547

Figure 11-18. Combitech ElectronicsAB 486 Computer Module System

his coworkers [91] who showed that under certain acid-catalyzed condi-
tions and use of model reactions lead them to use difunctional monomers
to prepare high molecular weight polyquinolines. The generalized structure
of polyquinoline is:

A major distinction in terms of processing thin films of polyimides


compared to PQs are that PIs are usually fabricated through a partially
cured prepolymer which is subsequently cyclized to the polyimide, soluble
polyquinolines are fully cyclized polymers, and therefore require no
post-cure.
Polyquinolines, having no carbonyl groups and do not suffer from
significant moisture uptake. Furthermore, they are less polar and thus have
1/-548 POLYMERS IN PACKAGING

lower dielectric constants (2.6-2.8) which do not increase appreciably with


changes in humidity [92,93].

11.7.3 Olefin Dielectrics


BFG is currently developing a family of dielectric materials marketed
as A vatrel™ Dielectric Polymers for use as interlayer dielectric thin films
based on cyclic olefins [94]. The generic chemical structure of these
polymers is shown below. This polymer family, based principally on
polynorbomene, is produced via a transition metal catalyzed polymeriza-
tion which enables BFG to control the polymerization of bulky, cyclic
monomers to form saturated polymers with high Tg• This mechanism
allows tailored molecular weights to be designed to specific applications.

The family of polynorbomenes are amorphous and exhibit high Tg


(>350 0c) and excellent thermal and electrical stability. These highly
transparent polymers are readily soluble in a range of common solvents.
Of particular interest are decalin and mesitylene. Polynorbomene polymers
have a high glass transition temperature due to the very stiff polymer
backbone. The group "R" represents specific functional groups which
provide adhesion and sites for crosslinking.
Initial studies of polynorbomene polymers reveal (i) excellent ther-
mal and electrical performance, (ii) adhesion to conductors and substrates
without the use of adhesion promoters, (iii) very low moisture content,
(iv) low dielectric constant, and (v) simple solution processing. The hydro-
carbon nature of polynorbomene yields dielectric thin films with very
low moisture absorption (0.1 %) and low dielectric constants (2.6 at 10
kHz). These polymers appear to be quite isotropic from birefringence
measurements using a prism coupler technique. Films which are nominally
2 f..1m thick after being spun onto oxidized silicon wafers exhibit birefrin-
gence on the order of 0.0031 ± 0.0001, more than 20 times lower than
that reported for a typical polyimide (PMDA-ODA). The refractive index
11.8 PHOTOSENSITIVE POLYIMIDES 11-549

Table 11-21. Summary of Polynorbornene Properties

Dielectric ConstantlLoss 2.6/0.0007


Breakdown Voltage >29 MV/cm
Moisture Absorption 0.1 wt%
Elongation 10-20%
Tensile Strength 50-60 MPa
Coefficient of Thermal Expansion 60 x 1~/oC
Modulus 1.4 GPa
Planarization 70% (2 mil pitch, 3.5 11m height, 6 11m film thickness)
Decomposition Temperature 435°C (5 wt% loss, lOoC/min, inert gas atmosphere)

was measured to be 1.505, and free standing films are clear and colorless.
Films are chemically resistant and have revealed no cracking or crazing
when exposed to common processing solvents such as: photoresist, aque-
ous base, PAN etch (phosphoric, acetic, and nitric acid mixture), acetone,
and isopropanol. Patterning has been demonstrated using RIB techniques
(OiCHF3 /Ar plasma) with both soft (photoresist) and hard (silicon diox-
ide) masks.
Polymer solutions of the polynorbornene are applied by spin coating
onto substrates, dried and typically cured above 200°C. These films pass
the adhesion Scotch™ tape test on bare copper, gold, silver, aluminum,
silicon (and oxides) without any adhesion promoter and in the absence of
any metal barrier/adhesion layer (such as chromium, titanium or tantalum).
These films remain adherent even after being placed in boiling water for
2 hours. Likewise, metal sputtered on top of the polymer survives these
adhesion tests, as do additional layers of the polymer. These films also
survive thermal shock tests, 100 cycles from liquid nitrogen to boiling
water. Since the polynorbornene polymers are fully polymerized before
processing, the base polymer properties are not significantly dependent
on the processing conditions and temperatures. Light cross linking is
used to enable multi-layer processing. Table 11-21 summarizes the film
properties of polynorbornene polymers.

11.8 PHOTOSENSITIVE POLVIMIDES


11.8.1 Introduction
The application of polyimides as a protection layer or an insulation
layer has become an important technology in the field of microelectronics
packaging because of their excellent thermal, mechanical, and electrical
properties [95,96]. Historically, Japanese workers at Hitachi were the first
11-550 POLYMERS IN PACKAGING

to implement wet-etching techniques capable of patterning fully cured


polyimides [97,98,99]. The current trend, as reported in the literature, is
to use photosensitive polyimides because of their ability to directly produce
patterns in them using standard photolithography techniques. The advan-
tage of using PSPI is the reduction in the number of process steps required
to form a patterned polyimide layer for making interconnections as shown
in Figure 11-19 and to better control via wall profiles. This development
of photosensitive polyimide has contributed to the progress in fabricating
high-density devices and packages in the microelectronics field [100-102].
The PSPIs utilized in commercial applications are mostly negative working

Non-Photosensitive Polyimide vs. Photosensitive Polyimide

r" ., ' .':,;:, . ,,'!,!H'Ph :::::,':'::':+::':":: :,:, Coating


. SUBSTRATE andPrebake SUBSTRATE

PhotoreSist
andPrebake

UVExposure

L'::::::: _ ~~: ....::::_ :::..~:::~


.~ ',:"

SUBSTRATE SUBSTRATE

Development

Postbake

Polyimide Etching

Photoresist Removal

I==J L':JJ
I ~SU'k/RA~
1"""":""""1 ~
4 Polyimide
Curing

Figure 11-19. Schematic Diagram Comparing Patterning of Non-Photosensi-


tive PI to PSPI
11.8 PHOTOSENSITIVE POLYIMIDES 11-551

precursor types. Figure 11-20 depicts the current main stream approaches
that are being considered for development of photosensitive polyimides.
They are classified into three types, covalently bonded ester-type PSPIs,
ionic bonded-type PSPI, and fully preimidized soluble polyimides.
Most PSPIs have been made today by using the polyamic acid,
which has an ionic bonded amine-linked cross-linking group, or a polyamic
ester which has a covalently linked ester cross-linking group to produce
a photosensitive precursor as schematically shown in Figure 11-21. The
photosensitive polymer precursor is then formulated with a photopackage
consisting of initiators, sensitizers, additives, and solvent to produce a PSPI
solution. The chemistry, formulation, processing, lithography, physical
properties, and applications of the ester-type PSPIs, ionic bonded-type
PSPIs and preimidized polymers are described in this section.

11.8.2 Historical Background


The first PSPI was reported by Kerwin et al. of Bell Laboratories
in 1971 [103]. It was a mixture of polyamic acid and potassium dichromate,
which showed negative working photolithographic properties, but had
many practical difficulties, such as very poor shelf life and high metallic
contamination.
In 1976 Rubner et al. of Siemens invented the first practical PSPI,
to be used for microelectronics applications. This invented technology
was based on using a precursor polymer backbone, in which cross-linkable
functional groups R* are bonded to carboxylic groups through ester link-
ages, as shown in Figure 11-22 [104,105]. In this sense, this system is
called an "ester-type PSPI." Their efforts were focused on the formulation

Novel Positive
PSPI

~l·'
hv
...
Polyamlc Acid hv Polyamic Ester
Amine Salt Linked -.c.----j~~ Ester Linked
Photo·Package
Photo·Package

Polyamlc Acid Precursor Polylmlde-Preimidize(/


(Intrinsically Photosensitive)

Figure 11·20. Technical Approaches to Development of PSPIs


11-552 POLYMERS IN PACKAGING

Photosensitive PI'S

o PSPI Obtained by Adding Photopackage

Photopackage--> ( X-linker , initiator, sensitizer, additives)

/
Diamlne _
Dianhydride _

o Upon Curing Photoactlve Group Is Released and


Volatilized out of PSPI Film

o Desire Exposure Wavelength where PI Backbone does not absorb Energy

- Since most PI's absorb at i-line I desire g-line exposure


for Improved sensitivity

Figure 11-21. Schematic Diagram Showing PSPI Formulation

of a practical composition and a process for polyimide pattern formation,


by using methacrylate groups to impart differential solubility in exposed
versus unexposed regions of the spun film. Their precursor polymer was
based on a pyromellitic dianhydride (PMDA) and oxydianiline (ODA)
backbone, which is the same as that of Kapton@, and is illustrated in
Figure 11-24. The photosensitivity was improved by adding photoinitiators
and co-monomers such as maleimide to the precursor polymer solution
[106]. Siemens licenced this fundamental technology to Asahi Chemical
in Japan, DuPont in the United States, and Ciba Geigy and Merck in
Europe [107].
Each of these chemical companies set out to further develop and
improve the licenced technology, particularly, by enhancing the photosen-
sitivity and increasing storage stability of the solution and several other
polyimide characteristics in order to produce more advanced PSPIs for
commercial use in manufacturing applications [102]. Merrem et al. of
Merck focused their efforts on improvement of the photospeed. They used

o 0
II "

R.o-~fi~HN-o-O-O~~~;:O~~:HO-O~
~O-o-NH_;~~_OR'
o 0
0 0
R*: allyl group, methacrylate group, N-methylmaleimide group

Figure 11-22. Schematic Showing Prototype of Rubner's Ester Type PSPI


11.8 PHOTOSENSITIVE POLYIMIDES 11-553

pentaerythritriallylether as a pendant group [108]. Pottiger et al. of DuPont


examined the precursor based on a benzophenone tetracarboxylic dianhy-
dride (BTDA), ODA, and meta-phenylene diamine (m-PDA) backbone.
This structure was chosen because of its better solubility than Siemens'
original PMDAJODA polymer and offered a number of advantages over
the Siemens' prototype, including better shelf life, higher photosensitivity,
and better adhesion characteristics [99,109]. Rohde et al. of Ciba Geigy
improved the shortcomings of the prototype such as lower photospeed,
excessive volume contraction, and shelf life limitations. By use of novel
high-quantum-yield photosensitizer, tailored to the mercury g-line, it was
possible to lower exposure times remarkably and to reduce the amount
oflayer shrinkage during final cure [110]. They reported the achievement
of high-resolution patterns even in 70-~m-thick layers after development
[111]. Asahi Chemical also focused its endeavors on the improvement
of Siemens' products. Main targets were improvements of the purity,
photosensitivity, and stability of the polymer solution during storage
[112,113].
These licenses have commercialized their improved products, the
trade names of which are PIMEL (Asahi Chemical), Probimide (Ciba
Geigy), Pyralin PD (DuPont), and Selectilux HTR (Merck), respectively.
Toray developed a unique negative working photosensitive polyimide
precursor, which is composed of a polyamic acid and tertiary amine having
an acryloyl group. This photosensitive PI was given a trade name of
"Photoneece."

11.8.3 Thin-Film Processing of PSPls


11.8.3.1 Polyimide Pattern Formation Process using Ester-type PSPI
An ester-type PSPI is normally supplied in liquid form composed
of a precursor polymer and a photopackage including photoinitiators,
photosensitizers, adhesion promoters, and stabilizers, as schematically
shown in Figure 11-21. Illustrated in Figure 11-23 is the process for
polyimide pattern formation using an ester-type PSPI and the correspond-
ing chemical structure in each step. The process steps used for depositing
polymer solutions, exposing, developing, and curing are briefly de-
scribed below:

Coating: Normally the PSPI solution is coated by spin coating onto


a substrate such as silicon wafers, glass, or ceramic substrates,
followed by partial drying to form a polyimide precursor layer.
Exposure: Then, UV light is exposed to the layer through a photo-
mask. Photochemically generated radical species from the pho-
toinitiator starts polymerizing double bonds of the pendant
groups to form a cross-linked structure.
Development: The relief pattern is then obtained by developing the
11-554 POLYMERS IN PACKAGING

PIMEL

CD C0C!ting,
Drying
Il - ._ _ _
!
S_u_bs_t_ra_t_e_ _ _-'

1 ttl
UV-light

, '-.L
photomask

~xposure I WnuuUhl tW"mzA

@
Developme.-n_t_.JY/;:?2!'H?Z,'/j,?ZY.1Zi'/4Z!Y.a.~-1t!17lll7llII7llZ1'a_-.

@) PI MEL pattern
Thermal curing ~ _ _ _ _ _ _~_ _ _ _E~--. !

Figure 11-23. The Process for Forming Polyimide Patterns Using PSPI

exposed image with an organic solvent which can dissolve


preferentially the unexposed area, leaving the exposed area
intact. This process is similar to the one used to pattern negative
working photoresist.
Cure: The relief pattern obtained by the development process is
heated up to 300-400°C in the cure process, in order to imidize
and eliminate the pendant groups from the patterned layer. The
solvent is vaporized and photopackage components as well as
decomposition by-products are removed from the film.
11.8 PHOTOSENSITIVE POLYIMIDES 11-555

11.8.3.2 Comparison with Non-PSPI


In the case of the process using non-PSPI [96,97,98], a conventional
photoresist process is applied. At first, the non-PSPI solution, which is
polyamic acid or ester, is coated on a substrate and then fully or partially
cured. Then a photoresist layer is patterned on the polyimide layer to
form the etching mask. The polyimide layer is etched with alkaline solution
and the overcoated photoresist is removed with chemical stripper and/or
dry process, postcure is followed, if necessary. This process with a non-
PSPI is considered to have some drawbacks. First, the total process is
quite long and, second, the etching step has some problems. Harmful
etchants such as hydrazine and ethylene diamine were used in the early
days of this process but less toxic (Et4 N+OH- in water) solvents are used.
In addition the shape of the designed pattern cannot be reproduced as a
polyimide pattern because of sidewall etching due to the isotropicity of
wet chemical etching. A partial curing process is often used to avoid
usage of hazardous etchants; however, the process margin is narrow
because the etching rate depends on the curing condition and on the degree
of imidization. Contrary to the non-PSPI process, the process using PSPI
is simpler and a fine polyimide pattern with good resolution can be obtained
because the PSPI is functionally photolithographic.
Figure 11-24 shows the comparison of the chemical structures of
non-PSPI and PSPI, including ionic bonded-type PSPI. Non-PSPI is a

Photosensitive Polyi.ide

• Ester type PSPI

* Salt type PSPI


Curing ~ 0

,CO~'OO
-N,
C
I to. h ~ h N
....::; C'
N 0 I
\
g g
Polyillide
Non-photosensitive Polyi.ide
o 0
-fHN.~~S'NHOo-
Ho-C~C-OH
U II
o 0

Figure 11-24. The Comparison of the Chemical Structure Between PSPI and
that of Non-PSPI
11-556 POLYMERS IN PACKAGING

polyamic acid itself, which can be thermally converted to a polyimide


structure through elimination of water molecules. As mentioned above,
there is another type of PSPI called the ionic bonded-type, the structure
of which is a mixture of a polyamic acid and a vinyl monomer with an
alkyl amino group. It is considered that a photochemically initiated vinyl
polymerization occurs to form a poly ion complex comprised of polyamic
acid and polyamine in the exposure step. Pattern formation can be achieved
by the use of the solubility difference between the exposed area and the
unexposed area, because the poly ion complex shows rather poor solubility
toward some organic solvents. The reaction in the curing step is an
imidization and a decomposition of the polymer of the amine monomer.
The ionic bonded-type PSPI may have several practical processing limita-
tions arising from the ionic properties which will be discussed.

11.8.4 Covalently Bonded Ester-Type PSPI

11.8.4.1 Chemistry and Photopackage: Preparation of


Precursor Polymers

Rubner prepared originally the precursor polymer using the "acid


chloride method" [106]. In this process, diacid is converted to acid dichlo-
ride, which is then reacted with diamine to give an ester-type polyimide
precursor, as shown in Figure 11-25. Hydrogen chloride and sulfur oxide
are involved during the reaction, so the film contains about 10 ppm of
the chloride ion, which may subsequently cause corrosion of metallic
material in the IC device. Improvements using a "chlorine-free condensa-
tion reagent method" have solved this problem.
Siemens and Asahi Chemical independently conducted research on
a novel preparation method of ester-type PSPI precursor using carbodii-
mide as a non-chlorine condensation reagent [101,114]. By the addition
of the photoactive alcohol to the aromatic dianhydride, a tetracarboxylic
acid diester is produced, which is then polycondensed directly with
the diamine in the presence of carbodiimide to yield an ester-type
PSPI precursor. Asahi has developed other chlorine-free condensation
reagents, of which the phosphine-disulfide system showed the best
results [115].
DuPont has reported a unique condensation reaction, in which a
part of the aromatic dianhydride is capped with 2-hydroxymethacrylate
(HEMA), and aromatic diamine is then added to the reaction mixture
to form a capped polyamic acid. It is converted to the corresponding
polyiminolactone intermediately by the reaction with trifluoro-acetic anhy-
dride (TFAA), and a treatment of the polyiminolactone with excess HEMA
produces the esterified polyimide precursor [109]. This preparation method
is referred to as the "iminolacton method" in Figure 11-25.
11.8 PHOTOSENSITIVE POLYIMIDES 11-557

o 0

/0
'c~c'
1
\
c Iminalaclone Method

~ J 2~"oH ~
~ ~ ? ~
HOCOCOH ,C:OCOH
. I. 0, I ~. •
~c ~C~ C~
c; 0 c; 0

• Add " " ' ' ' . / : : . j II 9

o 0
b Condensation
0, V
,c~c,
0 + Oiamine
CI ) )_~
" CCI Reagent Method ~ S'

.
. I ' 0 0

ROC ~ COR'
0"

jTriA~.ti.
o
Condensation "",,..,do
Reagent
+
Oiamine
Polyiminolacton

/~
o 0

C!4t=C-COCH,CH,O-C
0,
+HN'~~~-NHOO~
~\d'"rt.
~
C-O·CH,CH,OC-C==CH,
"C
... 0II 0II 0II 0II ct •" .,

Figure 11-25. The Preparation Methods of Ester Type PSPIs

11.8.4.2 Alternate Approaches to the Preparation of


Ester-Type PSPls
There are some ester-type PSPIs other than Rubner's structure. Min-
nema of Philips developed a PSPI precursor by esterification of the poly-
amic acid with isourea, which is the adduct of N-hydroxyethyl-N-methyl-
methacrylic amide and N,N'-diisopropylcarbodiimide [116]. Rubner and
Ahne prepared other ester-type PSPIs by the addition of glycidylmethacry-
late to the polyamic acid [117]. Davis of General Electric studied a
photosensitive polyimide siloxane. He prepared the precursor polymer by
the addition of isocyanatoethyl methacrylate to the polyamic acid contain-
ing silicone diamine [118].
11-558 POLYMERS IN PACKAGING

11.8.4.3 Processing
11.8.4.3.1 Solution Preparation and Properties
PIMEL has three characteristics which are essential for polyimide
coatings used for microelectronic applications: (1) polymer solution purity
and stability, (2) photolithographic characteristics capable of producing
high photosensitivity and high resolution, and (3) excellent thermal, me-
chanical, and electrical properties of the polyimide film obtained after
full cure. Polymer solutions are prepared by dissolving a precursor polymer
in a polar solvent such as N-methylpyrollidone (NMP), together with
additives (e.g., photopolymerization initiators, sensitizers, adhesion pro-
moters, and stabilizers) and filtering through filters having nominall-J..lm
pores. The polymer solution is formulated to different concentrations
resulting in viscosities ranging from 17 P, as shown in Table 11-22. These
formulations are designed for spin coating on manufacturing tools to
produce a range of film thicknesses from 5 to 10 J..lm when fully cured.
Since ester-type PSPI precursors have very good solubility in polar
solvents, it is practical to formulate more concentrated solutions which
can be used to spin coat in one-application films thicker than 50 J..lm.
Because the precursor is chemically stable, the viscosity and photosensitiv-
ity barely change even after 1 month storage at room temperature, as shown
in Figure 11-26, where the photosensitivity is defined as the remaining
thickness after development. This excellent shelf-life stability attribute is
one of the reasons why ester-type PSPls have been widely employed in
large-scale integration manufacturing production lines.
In contrast, non-PSPIs and ionic bonded-type PSPIs are known to
show a decrease in viscosity during storage, which is considered to arise
from the autocatalytic decomposition of polyamic acid and often resulting
in lowering of the molecular weight of the starting polymer precursor.

Table 11-22. Process Conditions and Liquid Properties of PIMEVM

Properties Unit G -7621A G-7613N TL-530

Solution Viscosity P 40 75 17
Development Method Spray Spray Puddle
Resolution (aspect ratio) 1.5 1.5 2
Curing Temperature °C 350 350 450
Film thickness After prebake 11m 10 20 10
After cure !lID 5 10 5
Typical values of ionic impurities ppm CI-<0.5, Na<O.lO, K<O.lO,Cu<O.1O
Fe<O.lO
ppb U<O.03
11.8 PHOTOSENSITIVE POLYIMIDES 11-559

50 .. ___ ..... _~. ___ ._ .... __ ~_ .. ________ .~_. ___ .______ L_._--------t--_·_---_···: ----.-----..... -.. -----

!I -+- kept at 23"C


-+-kept at -15"C

o o~----~------~~--------~
10 20 30 60 90 120 150 180
STORAGE PERIOD (day)
Figure 11-26. Shelf Life of PlMEL

The purity of metal and chloride ions of PIMEL are summarized in Table
11-21 which shows that PIMEL has a sufficient purity for microelectron-
ics applications.

11.8.4.3.2 Photolithographic process


PIMEL is processed in the same manner as photoresists by spin
coating on a silicon wafer, ceramic substrate or glass substrate, then the
substrate is dried on a hot plate or in a clean oven. Figure 11-27 shows
the relation between the rotation speed and the film thickness after prebak-
ing. One advantage of the polyimide coatings over inorganic thin film
deposition is a step coverage or planalization capability, which is needed
to planarize the rough surface topography of actual devices. Degree of
planarization (DOP) is defined by Eq. [11-1].

DOP = (I - DPa) x 100 [11-1]

Di and Da are expressed in Figure 11-28. Ahne reported that the ester-
type PSPI showed better DOP than non-PSPI [107].
PIMEL G-7600 series are designed to yield high photosensitivity to
the wide range of wavelengths of light irradiation from the mercury g-
line to the i-line, as is shown in the spectral photosensitivity curve of
Figure 11-29. Thus, PIMEL enables lithography engineers to achieve high
throughput with proximity aligners, g-line steppers, and mirror projec-
tion aligners.
The exposure step is composed of two kinds of reactions, which are
11-560 POLYMERS IN PACKAGING

[urn]

, I
;
, :
. ,_ ~ .. i ......
.:;. ;
I

.:._L:....:.! .. ~ .. _.' .
"! I

1000 2000 3000 4000

ROT A T ION S PEE D [ rpm]


Figure 11-27. Relation Between the Rotation Speed and the Film Thickness
after Prebaking

the generation of free radicals by decomposition of photoinitiator and a


radical polymerization of vinyl groups. The former reaction is usually
accelerated in the presence of a photosensitizer, and the total reaction rate
is affected by both oxygen and temperature. Loisel and Abadie studied
the photochemical reaction of ester-type PSPIs, DuPont's Pyralin 2732
and 2701, using differential photocalorimetry. The enthalpy of the poly-
merization, the induction time, the photospeed, and the activation energy
were measured under various temperatures and atmospheres [119].
The functional principle for a negative working photoresist to pro-
duce patterns is based on the difference in the solubility between unexposed

--~I~~W}~~;sr::l:
lo- :::::.:::
S' (' no,
A
Dt
o r ' - - i-

Figure 11-28. Parameters for Degree of Planarization (DOP)


11.8 PHOTOSENSITIVE POLYIMIDES 11-561

i-line g-line

400 500
Wavelength (nm)

Figure 11-29. The Spectral Photosensitivity Curve of PIMEL G-series

and exposed areas. An ester-type PSPI has good solubility to specific


organic solvents, it exhibits high development speed, whereas the cross-
linked structure of the exposed portion has good solvent resistance to the
developer. Figure 11-30 shows the changes in thickness of exposed and
unexposed areas during the development. The wide margin on develop-

100
A

--
~
"U
80

Q)

-
C
cu
Q)
a:
E
u:: 20

.2 4 6 8 10 12
Development Time(sec)

• : Exposed Film 0: Unexposed Film


Figure 11-30. Development Characteristics of PlMEL
11-562 POLYMERS IN PACKAGING

ment is achieved and a fine-resolution pattern can be obtained due to this


large difference in solubility.
Figure 11-30 indicates another feature of ester-type PSPI; that is,
the small amount of film thickness loss in the exposed portion during
development, which facilitates precise film-thickness control for manufac-
turing. Figure 11-31 is a typical scanning electron microscopic (SEM)
photograph of a highly resolved PSPI pattern. Figure 11-31 is a typical
scanning electron microscopic (SEM) photograph of a highly resolved
PSPI pattern. The resolved photoimages of PSPIs have an aspect ratio of
1 normally, but, it has been improved up to - 2 by modification of the
development process.

11.8.4.3.3 Pime/® Curing


The PSPI relief pattern, produced after ultra violet (UV) exposure
and development, is converted to the corresponding polyimide upon curing
to elevated temperatures, where two kinds of chemical reactions occur.
First, a o-(alkoxycarbonyl)-substituted benzamide structure is thermally
cyclized to form the imide ring, accompanying the elimination of the
HEMA monomer. The second reaction is the decomposition and vaporiza-
tion of cross-linked portion. The imidization reaction of a precursor poly-
mer has been investigated by thermal scanning Fourier transform-infrared
(FT-IR) spectrometry. Figure 11-32 shows the plots of the degree of
imidization versus temperature of curing, measured at the rate of 5 and
15°C/min. This plot indicates that the imidization reaction occurs between

Figure 11-31. Highly Resolved Releaf Pattern of PIMEL after Development


--
11.8 PHOTOSENSITIVE POLYIMIDES 11-563

~ 100 o .~y

'\
0,
c
80
...cu
0 0'

N 60 0'
"C
E
40 o,
't-
0
,
Q) 20 0,
Q)
0'
•
100
CJ
Q) 0
Q
100 200 300 400
Temperature ("C )

0: 5 "C/min • : 1 5 "C/min
Figure 11-32. Plots of the Degree of Imidization Versus Temperature of
Curing

about 200°C and 300°C. This temperature is a bit higher than that of non-
PSPI or ionic bonded-type PSPI. [120]

11.8.4.3.4 Pimel Properties


Polyimides are being widely used in microelectronics applications
due to their excellent properties as previously discussed. Therefore, PSPls
need to have the same characteristics as those of the corresponding polyim-
ides after they are fully cured. Table 11-23 shows the thermal, mechanical,
adhesion, and moisture uptake of cured PIMEL film. Figure 11-33 is a
typical thermogravimetric curve of PlMEL G-7600 series film cured at
350°C. The starting onset temperature of the decomposition is nearly at
400°C and the 5% decomposition temperature is at about 550°C, which
corroborates the excellent thermal stability of the polyimide layer as being
suitable for semiconductor applications. Electrical properties are as good
as those of non-PSPls. PlMEL has a dielectric constant of 3.3 which is
suitable as an interlayer dielectric for interconnect structures on multi-
chip modules.
Mechanical properties are also very important in these applications.
Lead-on-chip (LOC) technology was widely applied in packaging large
integrated memory devices such as 16 Mbits DRAMs, because it can
11-564 POLYMERS IN PACKAGING

Table 11-23. Properties of Cured PIMEL Film

Polyimide Types
G-7600 Series TL-530A

Properties High Elongation Low TCE


Typical dielectric Low dielectric
constant constant
Thennal properties
5% Weight-loss temp. (0C) >500 >600
Coefficient of thennal expansion (ppm/0C) 40-50 10-20
Internal stress (MPa) 40-50 10-20
Mechanical properties
Tensile strength (MPa) >150 >200
Elongation (%) >30 >15
Young's modulus (GPa) 3.3 6.0
Dielectric constant (1 kHz) 3.3 3.0
Tan 0 (l kHz) 0.003 0.002
Adhesion strength after PeT on Si
(133°C at 3 atm)
Tape peel test (h) >400 >400
Pull strength (MPa) >70 >70
Moisture Uptake (%)
(23°C, 50% rh (%), 24 h) 0.8 0.8

",' . , 'j

*---
100.00
'0
Q)

-
C
"n; 90.00

-
Q)
a:
.J::
0) 80.00
'w
3:
70.00
0 100 200 300 400 500 600

TemperatureCC)

Figure 11-33. Thermogravimetric Curve of PIMEL G-7000 Film Cured at


350°C (heating rate 10°C/min)
11.8 PHOTOSENSITIVE POLYIMIDES 11-565

dramatically contribute to the reduction of the chip size. However, the


buffer-coating layer is exposed to severe mechanical stress during the
leadframe mounting process of LOC. It is believed that the mechanical
properties of the buffer coat, especially, the elongation at break is an
important parameter for LOC packages.
Figure 11-34 is a typical stress-strain (SIS) curve of the cured film
of PIMEL G-7000 series, which has been proven to meet the mechanical
characteristics for LOC application.

11.8.4.3.5 Functional PSPls


Polyimides have been developed from a large variety of backbone
chemistries to meet the very diversified and demanding recent develop-
ments in microelectronics and other fields. PSPls also have been developed
with functional properties other than photosensitivity. Some examples are
shown in this chapter.

11.8.4.3.6 Low-Thermal-Expansion PSPls


The internal stress is the most important factor to consider in multi-
layer substrates using a Cu wiring layer and a polyimide insulation layer
for interconnection. The internal stress is generated in the polyimide layer
during cooling after cure. If the internal stress is too high, it will lead to

20

r""""I

('II~ _ ~
............ _ •••••••••••• - •••••••••••• ••••••

~
......
)(

til
til

~ /
t
20 30 40
STRAIN [%]
Figure 11-34. Stress-Strain of PIMEL G-7000 Grade
11-566 POLYMERS IN PACKAGING

bending and/or cracking of the substrate, and to delamination of the


polyimide layer from the lower layer. The internal stress, 0', is sometimes
expressed as shown in Equation 11-2:

aCt) = K Jr
g E(t)[al (t) - <lz (t)] dt. (11-2)

where
K = constant
Tg = glass transition temperature
E = Young's modulus of film
ahal = thermal coefficient of expansion (TCE) of polyimide and sub-
strate, respectively.

It is concluded from Equation (11-2) that one solution to reduce the


internal stress is to apply a polyimide with a low TCE, because substrates
are generally inorganic materials such as silicon wafers and/or ceramics,
which have rather low TCEs. The basic concept for the design of low-
thermal-expansion polyimides was proposed by Numata and Kinjo of
Hitachi [121]. They concluded that the introduction of rigid components,
namely rodlike structure, results in lowering the TeE. Furthermore, the
TeE of a polyimide film is influenced by the curing process and conditions.
Numata reported that the TCE of the polyimide film cured fixed in a
frame was lower than that of the polyimide film without tension, and
concluded that it was due to the difference in the degree of orientation
in the polyimide molecules [122].
The effect of the molecular weight of PSPI polymers on TCEs have
been investigated. Condensation products of pyromellitic acid-l,4-di(2-
methacryloyloxyethyl)ester and a mixture of meta/para-phenylenediamine
(PDA) [TP], a mixture of ODA and para-PDA [TA] , and a mixture of
para-PDA and 3,3'-dimethylbenzidine [TT] were prepared.
Figure 11-35 shows the value of tensile strength (Ts) and modulus
(E) of polyimide films as a function of a viscosity number (v.n.) of a
polyimide precursor, which corresponds to the molecular weight. It was
found that the tensile strength is almost independent of the molecular
weight of precursors and that the Young's modulus increases with an
increase in molecular weight. The TCE of the polyimide is plotted against
the viscosity number of the precursor in Figure 11-36 for various thickness
of polyimide film. The TeE becomes lower as the molecular weight of
the precursor becomes higher and the polyimide film thickness becomes
thinner. Figure 11-37 shows the relation between the TCE value and
Young's modulus in the case of TP as a polyimide precursor and a film
11.8 PHOTOSENSITIVE POLYIMIDES 11-567

-
C\I
1

E
~~----------~O~----~crr
o -
C\I
E
E E
.......
.......

-
0>

-
C)
.:::£ ~
en UJ
J-

v.n.(ml/g)
Figure 11-35. Tensile Strength and Young's Modulus of Polyimide Films as
a Function of the Molecular Weight of the Polyimide Precursor

thickness of - 15 ~m. The plots imply that the polyimide film with a low
TCE has a higher Young's modulus [120].
A large number of PSPIs with low stress have been developed by
many researchers as reported by Numata [121-122]. Asahi Chemical
has also developed the TL-500 series, the characteristics of which are
summarized in Table 11-22. The TCE is 10-20 ppm/DC, that is about a
third of the value of conventional polyimide. Figure 11-38 shows the

-
S-J
........
\0
I
o

-.....
x
W

(J I I

20 30 40
v.n.(ml/g)
Figure 11-36. Thermal Coefficient of Expansion of Polyimide Films as a Func-
tion of the Molecular Weight of the Polymide Precursor in Various Film
Thicknesses
11-568 POLYMERS IN PACKAGING

-E
600
C \j

E
"-
Ol
~
ill

Figure 11-37. Relationship Between Thermal Coefficient of Expansion and


Young's Modulus of the Polyimide Films in the Various Sample Prepara-
tion Conditions

relation between the internal stress in a cured polyimide layer and film
thickness. Curve b shows the internal stress in the five-layered polyimide
film with the PIMEL TL-series on a silicon wafer. The stress is dependent
on the film thickness; however, the internal stress of a multicoated layer
is lower than that of a one-time-coated layer, even if the total thickness

-
N
E
E
......
OJ

....
~
C/)
---- .. -----e----e---
2.0
--
~ - 2 nd 3 rd 4 th 5 th
....
C/)

(1j b Multi-coated
:: 1.0
- (l)

c:

20 40 60
Film Thickness( j.t m)

Figure 11-38. Internal Stress in Polyimide Film with PIMEL TL-Series on a


Silicon Wafer
11.8 PHOTOSENSITIVE POL YIMIDES 11-569

is the same, as shown by curve a. These results can be explained by the


difference in the degree of the orientation of polyimide molecules found
on the top surface versus the bottom side of the film. The thickness
dependence of curve b is explained by this anisotropy.
High resolution is one of the outstanding characteristics of the TL
grade. The SEM photograph of the PIMEL TL-series after development
is shown in Figure 11-39, where the line/space pattern of 15/15 /lm is
achieved in the film thickness of 35 /lm. Another characteristics of the
TL grade is its low dielectric constant of 3.0, which is desirable for an
interlayer dielectrics application for a multilayer substrate [123].
Nader et al. of DuPont prepared a low-stress ester-type PSPI, the
chemistry of which is biphenyltetracarboxylic dianhydride (BPDA) and
para-phenylendiamine (PDA). According to them, the prepared spin-
coated film showed a TCE of 13 ppml°C.
Interestingly, the cast film exhibited a TCE of 25 ppml°C. They
speculated that the differences in the TCE values were due to the difference
in orientation of the poly imide molecules. The other properties of the
cured film that were reported are an elongation of break of 10%, a modulus
of 6 GPa, a tensile strength of 192 MPa, and a dielectric constant of
3.5 [124].

Figure 11-39. SEM Photograph of the Pattern of PIMEL TL-Series after De-
velopment
11-570 POLYMERS IN PACKAGING

11.8.4.3.7 I-Line photoimageable Ester-Type PSPI


Of the many spectral lines of UV light radiation available, g-line
(436 nm) and broad band are utilized for the exposure of conventional
PSPIs. Recently, i-line exposure has been replacing g-line exposure in
photoresist processes used to fabricate large-scale integration (LSI) de-
vices. Thus, i-line photoimageable PSPI has been desired for exposure
of buffer coatings on higher-density devices, since common use of equip-
ment for photoresists and PSPIs is more cost effective in manufacturing.
PSPI precursors show a rather high absorption at the i-line, limiting the
applicable film thickness. Therefore, a polymer backbone structure with
less absorption at the i-line needs to be developed.
Yamaoka et al. proposed an ester-type PSPI containing fluorine.
They studied 2,2-bis {4(4-aminophenoxy)phenyl} hexafluoropropane or
2,2-bis(3-amino-4-methylphenyl) hexafluoropropane as diamine compo-
nent. The i-line transmittance was measured to be more than 60% for a
3-J.lm-thick film of these PSPI precursor polymers. Conventional ester-
type PSPI precursors based on PMDAlODA chemistry exhibit about 30%
i-line transmittance [125]. It was found that there is a substituent effect
on the i-line absorption of an ester-type PSPI precursor. The introduction
of an electron donating group to an acid group decreases the i-line absorp-
tion. On the other hand, a substitution of an electroinductive group for
the diamine moiety shows the same effect. Good linearity was found
between the absorption per amide unit and a Hammett's value [126].
PSPIs have been developed based on this principle for i-line exposure.

11.8.4.3.8 Base-Catalyzed PSPI


Volksen et al. of mM reported a very unique PSPI. It is basically
composed of ethyl ester-type PSPI precursor and an amine photogenerator.
The amine photogenerator is decomposed by exposure to deep UV light
to give an amine, which catalyzes the imidization reaction of the precursor
polymer. It was also reported that pattern formation was successful due
to the large difference in the solubility between the precursor and partially
imidized polymer [127].

11.9 IONIC BONDED· TYPE PSPls


11.9.1 Chemistry and Photopackage
Photosensitive polyimides have continued to gain great attention in
the microelectronics industry due to the simple fabrication process based
on their direct patternability. Hiramoto et al. developed a unique negative
working photosensitive polyimide precursor, which is composed of a
poly(amic acid) and tertiary amine having an acryloyl group (Fig. 11-40)
[128,129]. Toray's photosensitive polyimide coating, "Photoneece" [130],
is developed by this method. The "Photoneece" is used in various applica-
11.9 IONIC BONDED-TYPE PSPls 11-571

OCYYCONH-{ }-O-( )-NHF


{:HOOC~COOH
·
•
I•
.
I
I•

Figure 11-40. Ionic Bonded-Type Photosensitive Polyimide Precursors ("Pho-


toneece")-Negative Working

tions, because it has some distinct advantages. The first is that swelling
does not occur during a development in spite of it being a negative
working type system, producing high resolution. The second is its high
photosensitivity, which is comparable to that of a conventional photoresist.
The last is that the photosensitive group can be readily eliminated during
thermal curing, so as to give a highly reliable polyimide film. These
advantages, as will be described later, are closely related to the photosensi-
tive group endowing method.
The mechanism of photosensitivity of the ionic bonded-type photo-
sensitive polyimide precursor was investigated by Tomikawa et al. [131].
The mechanism of the photosensitivity of the system is not photo-cross-
linking of the acryloyl group in the tertiary amine as shown in Figure
11-41. The first evidence is that the exposed film can be soluble in
NMP completely. The second is that the acryloyl group does not react
significantly during photoirradiation as determined by NMR studies. The
last is that about half of the tertiary amine (dimethylaminoethyl methacry-
late, DMM) remains in the exposed film after chloroform extraction, but
all of the remaining DMM in the film is recoverable intact. Since the
mechanism for the photosensitivity is not photo-cross-linking of the acry-
loyl group, this gives "Photoneece" an advantage to yield high resolution
patterns and be free from swelling during development in spite of being
a negative working system (Fig. 11-42).
In order to improve photosensitivity, the selection of suitable sensitiz-
ers is important. Most photoradical initiators do not work effectively
except for Michler's ketone and dimethylaminoacetophenone. This result
corresponds to the absence of photopolymerization of the methacryloyl
group during photoirradiation. Aromatic amines, such as N-phenyl-dietha-
nolamine, work as effective sensitizers [132]. In some cases, they work
as a cosensitizer, if other photoradical initiators exist in a photosensitive
system (Fig. 11-43).
Table 11-24 shows the typical properties of "Photoneece" UR series
11-572 POLYMERS IN PACKAGING

.......
~ 120
'-"

~ .....
..., ..... .-
...,
~ 100 ..., v
Q
....0
...c:: 80
....c::
4)

0
u 60
u
U " 0 300 600
I

Exposure Energy (mJ/cm 2 at 365nm)

100
.......
~
'-"

~
~
....0
0
50
.....
c:: eTotalDMM
:s OChloroform Extracted DMM
0
e From Exposed Film
« DChlorofonn Unextracted DMM
a From Exposed Film

o 500 1000 1500


Exposure Energy (mJ/cm 2 at 365nm)

Figure 11-41. Photoreaction of Aminoalkyl Methacrylate (DMM) (From Ref.


128)

produced by Toray. These are composed of different polyimide back-


bone structures.

11.9.2 Processing IIPhotoneece (Exposing, Developing, and


ll

Curing)
As previously shown in Figure 11-19, the use of "Photoneece"
significantly reduces the number of steps needed for pattern generation
compared to the conventional processing using nonphotosensitive poly-
11 .9 IONIC BONDED-TYPE PSPls 11-573

Figure 11-42. Fine Patterns of Photosensitive Polyimides

imide coatings. In general, spin coating is used. Coating thicknesses can


vary widely from 1 /lm to 40 /lm, depending on its applications. As the
film thickness is reduced during developing and curing, the final thickness
of film will be approximately 50% of the prebaked film thickness.

Prebake: Coated substrates should be prebaked on a hot plate or in


a convection oven (in air or nitrogen atmosphere).
Exposure: A contact aligner, mirror projection aligner, or g- or i-
line stepper can be used (see Fig. 11-44).
Developing and Rinse: "Photoneece" can be developed by immer-
sion, immersion with ultrasonic assistance or the spray-puddle
method, using a "Photoneece" developer (DV series), and
rinsed with isopropanol and isobutanol.
Curing: The standard condition for curing "Photoneece" with a
convection oven is a three-step curing at 140-180°C and 200-
300°C for 30 min each, then 350-400°C for 60 min. It is also
11-574 POLYMERS IN PACKAGING

IOOr---------------------------------------~

~ N-phenyl-diet.hanolamine
I: 80
"'"'
I: 0
.~

t1I
I'l
~ 60
UI
UI
OJ
I:
~
40
o Michler's ket.on
.~

.c:
E-t
I'l 20
.....
"'r.."'

Exposure Energy (mJ/cm 2 ,at. 365nm)

Figure 11-43. Photosensitivity of "Photoneece"

possible to ramp up the temperature from room temperature


by 3-5°C/min and hold for 1 h at 350°C.
Removal of "Photoneece" (Rework): Before curing, "Photoneece"
film can be striped by using N-methyl-2-pyrrolidone or alkaline
developers used for positive photoresists. Oxygen plasma etch-
ing is also applicable. After curing, hydrazine-hydrate-type
etchants or oxygen plasma removal systems should be used.

11.9.3 Lithographic Properties and Characterization


"Photoneece," in spite of being a negative working photosystem,
does not swell during development and, therefore, produces very fine
patterns. The characteristic curves of "Photoneece" are shown in Fig. 11-
45. The film thickness remaining after development is more than 95% at
an exposure energy in excess of 200 mJ/cm2 at 436 nm, in both 20 J.1m
and 40 J.1m prebaking film thicknesses.
The relationship between the resolution and the film thickness under
constant developing conditions is shown in Figure 11-46 [133]. "Photo-
neece" has a high resolution, with an aspect ratio of more than 2.0. SEM
photographs of patterning profiles after developing and curing are shown in
Figure 11-47. The patterning profiles of "Photoneece" are not overhanging,
even in 40-J.1m imaged film [133].
11.9 IONIC BONDED-TYPE PSPls 11-575

Table 11-24. ''Photoneece'' UR Series Polyimide Properties Measured on


Solutions and Cured Films

UR-3l00 UR-3800 UR-4l00 UR-S100


Solution Properties Series Series Series Series

Typical grade UR-3l40 UR-384O UR-4l44 UR-S100


Features Standard High- Low Young's High
sensitivity modulus resolution;
low thermal
expansion
Solid content (%) 17 20 20 22
Viscosity (P/2S0°C) SO SO 30 100
Metallic ion content Impurity levels same for all grades
(measured by ICP mass Na < 0.1 ppm, K < 0.1 ppm, Fe < 0.1 ppm, U < O.OOS
spectrometry) ppb, Th < O.OOS ppb
Solvent system NMPI NMPI NMP NMP
N-methyl-2-pyrrolidone (NMP) methyl methyl
cellosolve cellosolve
Film Properties
Mechanical
Tensile strength (kglmm2) 14 14 10 20
Elongation (%) 30 30 10 30
Melting point (0C) None None None None
Coefficient of thermal expansion 40 40 45 45
(ppmfOC)
Flammability All are self-extinguishing
Electrical
Dielectric constant (l kHz, 3.3 3.3 3.3 3.3
25°C)
Dissipation factor (I kHz, 2S°C) 0.002 0.002 0.002 0.002
Volume resistivity (0 cm) >10 1• >10 1• >10 1• >10 1•
Surface resistivity (0) >10 1• >10 1• >10 1• >10 1•
Dielectric strength (kV/mm) >300 >300 >280 >300

Low-temperature removal of photoreactive groups is another feature


of "Photoneece." Most of the photoreactive groups can be removed at a
curing stage lower than 300°C. Therefore, the properties of a cured film
are primarily determined only by the polyimide backbone structures. This
is a very important and desirable characteristic to design into photosensi-
tive polyimides. Figure 11-48 shows the thermogravimetric analysis of
an ionic bonded type ("Photoneece") and a covalently bonded ester type
[130]. Both have the same polyimide backbone consisting of pyromellitic
dianhydride (PMDA) and 4,4'-diaminodiphenylether (DAE). In the case of
the ionic bonded type in which dimethylaminoethylmethacrylate (DEM) is
11-576 POLYMERS IN PACKAGING

I I I

,
365 405436

" PHOTONEECE®
UR-3800,UR-4100,UR-5100

PHOTONEECE®
UR-3100

Hg vapor
emission line

350 400
Wave Length (nm)

Figure 11-44. Sensitivity Spectrum of "Photoneece"

used, a large weight loss occurs between 1OO°C and 200°C, followed by
small weight loss up to 300°C. A solvent (NMP), water (imidization gives
rise to it) and photosensitive group, DEM, are completely removed until
300°C. On the other hand, the weight loss of the ester type starts at higher
temperature than that of the ionic bonded type and is still observed up to
500°C. Thermogravimetric curves (TG) shown in Figure 11-49 indicate that
the cured films of the ionic bonded type are thermally more stable than
those of the ester type and show the same TG curve behavior as exhibited
with the nonphotosensitive polyimide of the same structure.
Comparative investigations of the properties of thermally imidized
polyimides from different types of photosensitive polyimide precursors
were performed by Kojima et al. They studied the thermal, mechanical, and
adhesion properties of the three kinds of polyimide precursors [poly(amic
acid), ionic bonded-type photosensitive polyimide precursor, ester-type
photosensitive polyimide precursor [134]]. As shown in Table 11-25, the
weight loss characteristics and mechanical properties of fully cured films
prepared from the ionic bonded-type polyimide precursor are comparable
to those prepared form poly(amic acid). However, these same properties
measured on the ester-type photosensitive polyimide precursor are de-
11.9 IONIC BONDED-TYPE PSPls 11-577

*
100 r-----------------------------------------~
'0
Q)
0-
o
......
Q)
>
CIl
'0
k
Q)

....
~

<:
......
bO
.~ 50
....s::til
El
Q)
k
en
en
OJ
s::
-It:
....
t)

.<=
~

El
..........
t:.. 0
50 100 500
Exposure Energy (g-line,mJ/cm 2 , at. 436nm)

Figure 11-45. Characteristic Curves of "Photoneece"UR-5100 (From Ref. 133)

Conditions
Prebaking :9Omin at 70c in convection oven
20 :g-line Stepper
700mJ/cm2 :Puddling and Spray 7min
6OOmJ/cm2
500mJ/cm2
400mJ/cm2
t:: 15 300mJ/cm2
o
......
+oJ
::s
..-t
o(/)
Jg 10

20 30 40
Film thickness after developed(um)
Figure 11-46. The Relation Between the Resolution and the FIlm Thickness
of "Photoneece"UR-5100 (From Ref. 130)
11-578 POLYMERS IN PACKAGING

(a)

(b)

Figure 11·47. SEM Photographs of Patterning Profiles After Developing and


Curing of "Photoneece"UR-5100. (a) Coated film thickness after developing, 40
11m; (b) coated film thickness after curing, 20 /lm. (From Ref. 133)

graded when compared to the films prepared from the poly(amic acid),
even after imidization at 400°C. This is thought to be due to the film's
ability to eliminate the photosensitive groups. The photosensitive group
of the ionic bonded type is eliminated faster than that of the ester type.
The photosensitive group of the ester type may also not be completely
eliminated after being imidized at 400°C because the photosensitive group
is attached by a covalent bond, and it reacts with UV exposure.
11,9 IONIC BONDED-TYPE PSPls 11-579

..--..
0 ..... ~
~~
Heating rate: 5·c Imin
----"Photoneece" type
'*en
........
\.~ PMOA/OAE/OEM
en ~~ ---- Ester type

-........
0
.....I ~~~.... PMOA/OAE/HEMA
+oJ
..c ~,

CD ~~
'0) \,
3: 40
~~
....... -......
~---------NMP'----~

100 200 300 400 500

Temperature CC)
Figure 11-48. Thermogravimetric Analysis of an Ionic Bonded Type ("Photo-
neece") and a Covalently Bonded Ester Type (From Ref. 130)

0 ........
..--.. ...... ......... ...
'* ......, 450·C "" ...~"
en
1O·c Imin, N2 '... "\
en ' ...... 400·C-'\ \
a 350·C~, \\
.....J
..c
+oJ 10 ",, ,\
,\
\\
C)
'Q) "Photoneece" type ......\ "",I
3: Ester type
\ I'
\
, ,\
,\
Postbaking Time: 30min
20~____~______~____~~____~~
\ ,\
, " \ I'
200 300 400 500 600
Temperature CC)
Figure 11-49. Thermogravimetric Curves of Cured Films of an Ionic Bonded
Type ("Photoneece") and a Covalently Bonded Ester Type (From Ref. 130)
11-580 POLYMERS IN PACKAGING

Table 11-25. Initial Weight-Loss Temperature and Mechanical Properties of


Cured Films Formed from Polyimide Precursors

Mechanical ProQerties
Initial Weight-
loss Temperature Elongation Tensile Strength
Curing conditions IT! ITcl (kgf/mm')

A 350°C-60 min 450 10 12


400°C-60 min 450 11 12
B 350°C-60 min 355 <1 <1
400°C-60 min 385 <I <I
430°C-60 min 390 <I <I
450°C-60 min 400 <I <I
C 350°C-60 min 395 10 12
400°C-60 min 440 10 12

'Polymer A = Poly(amic acid); polymer B = ester-type photosensitive polyimide precursor; poly-


mer C = ionic bonded-type photosensitive polyimide precursor. (From Ref. 134).

11.10 PREIMIDIZED PHOTOSENSITIVE POL VIMIDES


11.10.1 Chemical Characterization

PSPIs have been offered by OCG under the trade name Probimide®
400 (Probimide 412 and 414). The descriptions "inherently photosensitive"
or "autophotosensitive" have been used synonymously in the literature.
The chemical structure of these special PSPIs is

with R =

h All,
--vCH2~
Et Et

to produce a variety of PSPI related structures.


11.10 PREIMIDIZED PHOTOSENSITIVE POLYIMIDES 11-581

They are prepared by standard PI synthesis, first by the corresponding


polyamic acid and then by chemical imidization in solution. These solvent
soluble polyimides are photosensitive through photocrosslinking by them-
selves and without the addition of any photosensitizers. The photosystem
is an integral part of the PI unit, and the photostructures obtained shrink
only slightly upon curing (350-400°C).
The chemistry was first published in 1985 [135], and a review
together with new and improved experimental materials was given in
199111992 [136]. The polyimides are high molecular weight (Mw by light
scattering = 40,000-60,000), and they are offered as ready-to-use solutions
with 4-butyrolactone as the main solvent. The selection of monomers
is such that an optimum of photospeed, process ability, and mechanical
properties (Tg, thermal stability, elongation, TCE) is obtained. The key
data of these formulations are listed in Table 11-26.

11.10.2 Features
Purity: Due to the simplicity of both the polyimide synthesis and
the formulation processes, the high demands with respect to
metal-ion contamination can be met, in particular with respect
to Na, K, Fe, CI - (<1 ppm).
Storage Stability: Ready-to-use solutions are storage stable in com-
parison to a variety of other PSPI systems, with a typical
change of viscosity of only about + 2% per year of storage at
room temperature.
Spectral SensitivitylPhotospeed: The material is primarily sensitive
in the i-line (365 nm) and is compatible with exposure in the
contact or proximity mode, with projection tools and with i-
line steppers.
Thickness Range: The application range is from 1 to 15 11m (after
cure). The best throughput, given the exposure and the develop-
ing time, is achieved in the film thickness range of I-811m.

Table 11-26. Probimide 400 Series General-Purpose Formulations

Probimide % Solids Solvent Viscosity (cs) Filtration (/J.m) Thickness (J,1m)

Probimide® 412 12 GBL 3500 1.0 3.5-12


Probimide® 414 14 GBL 8200 1.0 4-20

Note: Shelf-life stability at room temperature is 18 months for Probimide 400 series.
11-582 POLYMERS IN PACKAGING

Resolution!Aspect Ratios: Using contact or i-line stepper exposures,


aspect ratios of 1.0 are obtained for films over 5 ~m, and 1.5
for films 1-5 ~m thick.
Thermal Cross-linking: A thermal cross-linking process (350-
400°C, 0.5-2 h typically) renders all photostructures insoluble
and inert. It is important for the understanding of these inher-
ently photosensitive polyimides that photo-cross-linking and
thermal cross-linking are independent of each other, and obvi-
ously follow different pathways. The influence of photo-cross-
linking and thermal curing on final material properties has been
a subject of intensive studies (see Section 11.10.3 below).
Nonshrinking upon Cure: In contrast to other PSPI systems, inher-
ently photosensitive polyimides of the Probimide 400 type
exhibit only minimal shrinkage upon cure (350-400°C).
Shrinkage typically is only 7-10%. Thus, photopattems are
preserved and geometrical distortions are minimal. This is illus-
trated by the SEMs of Figure II-50, showing 7 ~m lines before
and after hard bake [137].
Cu Compatibility: Due to their preimidized nature, Probimide 400
polyimides are fully Cu compatible, unlike polyimides derived
from poly amic acids which can react with Cu.
Glass-Transition Temperature: The glass-transition temperature
of Prohimide 400 is relatively high when compared to other
PIs and PSPIs. The exact value depends only slightly on the
precise curing conditions (350-400°C) and the method of deter-
mination. Thus, DSC gives a value of 369°C, whereas (depend-
ing on curing conditions) thermomechanical values from 350°C
through 410°C have been reported [137,138].
Reduced Outgassing: Probimide 400 polyimides do not contain
thermally instable components/additives like acrylates and pho-
tosensitizers that may give rise to outgassing problems.

Figure 11-50. SEM Micrographs Showing the Side Wall Profiles of Soft-baked
(left) and Hard-Baked (right) 7 f.Lm Lines
11.10 PREIMIDIZED PHOTOSENSITIVE POLYIMIDES 11-583

11.10.3 Origin of Photo-cross-linking


The photomechanism of Probimide 400 related polyimides has been
studied by Lin et al. [139], Scaiano et al. [140,141], Shindo et al. [139],
Yamashita et al. [143-145], Horie et al. [146], and Jin et al. [147]. Irradia-
tion experiments from polyimide solutions, as well as from solid films,
were reported and the photo-cross-linking efficiency was calculated from
Gel Permeation Chromatographic measurements of the irradiated products
as well as from gel doses. Also, spectroscopic studies on model compounds
were carried out to identify the transient active species:
It is now generally accepted that the photo-cross-linking process of
Probimide 400 related polyimides is initiated by H abstraction (photore-
duction) of the triplet state of the benzophenone unit according to the
following chemistry.
Quantum efficiencies of photo-cross-linking as obtained from irradia-
tion of solid films have been calculated to be only (2.0-3.5) X 10-3, and
less than 50% of all benzophenone units are reactive in the solid. These
studies, however, considered only BTDA polyimides with "ortho-mono"-
alkylated diamines. Probimide (400-type polyimides are made with ortho-
di-alkylated diamines, and are much more photosensitive than polyimides
made from (ortho-mono)-alkylated diamines [136,148].

Role of Alkyl Substituents: The two alkyl substituents next to the


imide bond have multiple functions:

• Photospeed is improved to a useful range.

• Solubility of the polyimide is greatly improved.

• Glass-transition temperatures are greatly enhanced through


locking of the diamine rings in the orthogonal position rela-
tive to the tetracarboxylic acid rings.

• Transparency of polyimide is improved such that absorptions


at 365 nm (i-line) are ideally suited for lithography in the
5-1 O-~m thickness range.

Oxidative Stability: Probimide 400 polyimides do not provide ther-


mooxidative stability above about 320°C. Exclusion of O2 «5
ppm), however, allows curing at 350-400°C in nitrogen. It has
been shown that thermogravimetric weight loss does not relate
to alkyl content when oxygen is excluded [149].
Morphology: Although preimidized polyimides composed ofBTDA
and ortho-dialkylated diamines may give rise to highly ordered
11-584 POLYMERS IN PACKAGING

structures which may crystallize from solution, the composition


of Prohimide 400 polymers is such that an absolutely amor-
phous material is obtained. This has been confirmed by Ree
et al. [138]. With other experimental products, crystallization
has been observed [149].
Processing Conditions: The key processing data of ProbimideX
408, 412, and 414 and the new self-priming developmental
product, XB 7003, are listed in the summary in Table 11-27.
Adhesion Promoter: Typically an aminosilane adhesion promoter
(QZ 3289/3290) is applied prior to coating. The new develop-
mental product, XB 7003, is fully self-priming and does not
need a spun-on adhesion promoter. Its photospeed and resolu-
tion capability are the same as with Probimide 414; the process-
ing conditions are practically identical.
Soft Bake: Hot-plate drying is preferred to oven drying, as superior
adhesion properties are achieved.
Exposure: The exposure tool must have output at the i-line, 365
nm. i-Line steppers are preferred. The photospeed is at - 100
mJ/cm2 per micron thickness, thus requiring 600 mJ/cm2 for a
5-~m-thick coating and 1500 mJ/cm2 for a 15-~m-thick coating
(both after hard bake).

Table 11·27. Process Summary for Probimide 400 Series Photosensitive


Polyimides

Process and Parameters Probimide 412 Probimide 414 XB 7003

Adhesion Promoter 1:9 4000/20 4000/20 Self-priming


QZ 3289/3290 (rpm/s)
Spin coating (rpm/s) 2400/25 1500/25 3200/25
Soft bake (OC/min)
Oven 1I0/30 1I0/30 90/30
Hot plate 110/10 1I01l5 90/10
Exposure energy (i-line stepper) 600 1500 600
(mJ-cm2)
Postexposure bake (oC/min) 110/10 IlOilO
1I0/3 110/3
Spray Development
Developer QZ 3501 (rpm/s) 8001100 800/240 800/80
Overlap (rpm/s) SOO/lO SOOl10 SOO/5
Rinse QZ 3512 (rpm/s) 800/20 800/20 800/5
Spin Dry (N2) (rpm/s) 2000/15 2000/15 2000/10
Hard bake caC/min) 350/60 350/60 350/60
Final film thickness (J.1m) 5 15 7
11.10 PREIMIDIZED PHOTOSENSITIVE POL YIMIDES 11-585

Developing: Developing is by spray-on of developer QZ 3501lrinse


QZ 3512 (both xylene free). The development rate is at 16-20
s!llm.
Cure (Hard Bake): Standard hard bake is 350°C, 60 min in high-
grade N2 « 5 ppm O2),
Mechanical Properties: The mechanical properties of Probimide
400 cured films are listed in Table 11-28 [142].
There have been extensive studies to monitor the change!
conservation of material properties toward a variety of process-
ing parameters [137,138,145,151]. The most important parame-
ters analyzed and their influence on materials properties are
as follows:
Photo-Cross-Linking: Photo-cross-linking with the required expo-
sure doses « 2000 mJlcm2, i-line) has little influence on final
mechanical properties. A change of the dynamic storage modu-

Table 11-28. Measured Properties of Probimide 400 Cured Film

Properties Units Measured

Physical
Tensile Strength GPa 2.9
Modulus MPa 147
Elongation % 56
Refractive index (633 nm) 1.66
Density g/cml 1.20
Moisture uptake (50% rh) % 2.0
Stress MPa 48
Thermal
Glass-transition temperature °C 357
Melting point °C None
Decomposition temperature °C 527
Weight loss (400°C in N,) %/h 0.07
Thermal coefficient of expansion ppml°C 37
Thermal conductivity cal/cmls/oC 4.1 x 10.3
Specific heat cal/gl°C 0.32
Electrical
Dielectric constant (I MHz, rh 4%/50%) 2.9/3.7
Dissipation Factor (I MHz, rh 4%/50%) 0.006/0.010
Dielectric breakdown field v/Ilm 348
Volume resistivity Q-cm >6 x 1016
Surface resistivity Q >1 X 10 '5
Adhesion
90° tape peel test, 72 h in boiling water 5
(ASTM D-3359-838)
11-586 POLYMERS IN PACKAGING

Ius E' has been seen at the glass-transition region [137], when
a hard bake of 350°C, 30 min was applied (lower limit of hard-
bake conditions). At higher hard-bake temperatures, however,
thermally generated cross-links dominate over photochemically
formed cross-links. Also, mechanical stress-strain behavior is
primarily dependent on thermal history, not on photochemi-
cal history.
Oxygen Content of Curing Atmosphere: Due to thermal oxidation
of alkyl side groups, the mechanical properties of Probimide
400 may degrade when cured at 350-400°C in the presence
of > 100 ppm O2, Mechanical properties are preserved when
O2 is excluded. O2 contents of <10 ppm but preferably <5 ppm
are recommended.
Hard Bake-Curing temperature (350°-400°C): Hard-bake tem-
perature is the most important and most thoroughly analyzed
processing parameter. If thermal cross-linking is not complete
at a hard-bake temperature of 325°C, a minimal hard-bake
cycle of 350°C, 0.5 h, preferably 1.0 h, is recommended. On
the other hand, a change of mechanical properties is seen at
curing at 400°C, 2 h. Thus, the curing window of 350-400°C
has been thoroughly analyzed. Below is a list of material proper-
ties analyzed that wilVwill not be affected by the choice of the
particular curing temperature within this curing window (1 h
cure time):

• Residual stress is 48-52 MPa and independent of cure tem-


perature.

• Elongation at break changes from about 60% (350°C) to


about 50% (400°C).

• The E modulus remains constant at 2.8-2.9 GPa.

• Tg remains constant from 350°C to 375°C cure, then rises


slightly.

• Solvent swelling (e:g., by N-methylpyrrolidone, NMP) de-


creases sharply as the hard-bake temperature advances
through the Tg (365-375°C), and then remains relatively con-
stant.

Hard-bake-curing time 0.5-10 h: Prolonged curing in Orfree


nitrogen up to 10 h is possible at a cure temperature of 350°C.
There are only minor changes in residual stress, elongation at
11.11 POSITIVE PHOTOSENSITIVE POLYIMIDES 11-587

brake, E modulus, and Tg• Curing at 400°C for 10 h will,


however, noticeably degrade material properties.
Further Experimental Products: Substantial work has been de-
voted to further improve photospeed, spectral sensitivity, pro-
cessing conditions, and material properties of Probimide 400
materials. A summary of such work has been given in Ref.
136 extending the working principles of BTDA-bond Probim-
ide 400 to polyimide backbones based on other dianhydrides
[152,153].

11.11 POSITIVE PHOTOSENSITIVE POLYIMIDES


Photosensitive polyimides offer cost advantage of eliminating the
need for photo-resists, thereby reducing the number of processing steps
in the fabrication of microelectronic devices as previously shown in Figure
11-19. Based on photo reactivity, photosensitive polyimides can be classi-
fied into two broad categories: (1) negative resist types or (2) positive
resist types. Most of the commercialized photosensitive PIs are of the
negative resist type and are derived from polyamic acid precursors which
have either covalently bonded ester or ionic-bonded photosensitive groups
attached to the backbone polymer chain previously shown in Figure 11-
21. These types of polymers were discussed extensively in the previous
sections. Conversely, positive resist types undergo molecular rearrange-
ment such as chain scission that enhances the solubility of the exposed
area. Positive working PSPls are being developed to overcome the limita-
tions [154-155] of localized swelling often found in negative working
systems which may lead to degradation of patterning resolution and use
of less toxic organic solvents for fine line/via patterning. The application
of positive working PSPls are limited today due to the narrow film thick-
ness range available but they are deemed preferable for semiconductor
and packaging applications [156,157]. They offer better resolution and
are developed in aqueous base solutions. In positive PSPls, adjusting the
dissolution rate is critical to avoid film delamination during development.

11.11.1 Positive Working PSPI


There are two approaches to developing positive PSPls: (I) incorpo-
rating a photosensitive group into the backbone or as a side branch of
the base polyamic acid or (ii) by blending a photosensitive compound
with the polyamic acid precursor, a system analogous to polymer blends.
The development of positive PSPls by the first approach is more
difficult and limited success has been achieved. Preparation of PSPls with
the photosensitive moieties in the main chain are relatively complicated
since the introduction of such groups in the main chain may alter the
molecular architecture of the basic polyamic backbone. A positive type
11-588 POLYMERS IN PACKAGING

PSPI prepared from a maleic anhydride dimer exhibited positive resist


like patterning [158]. The cyclobutane ring in the polymer chain undergoes
cleavage upon photo-irradiation which imparts differential solubility in
the exposed areas to produce a pattern after development leaving the
unexposed region unchanged.
Hayase et al. [159] developed a positive working high molecular
weight PSPI from polyamic acid esters with phenol moieties which were
synthesized from diarnines and dicarboxylic acids, polyarnic acid and
napthoquinone diazide. The phenol moieties in the polyamic acid ester
were made through ester linkages. The positive acting PSPI was base
developable and the phenol groups were removed during imidization to
produce a thermally stable polyimide film after curing to 350°C. A 4
micron line and space pattern was produced by using this PSPI, however
as the polyamic acid content was increased to make the resist dissolution
rate faster, the resolution degraded to 100 microns in a 3 micron thick
film. Therefore controlling the dissolution rate is critical to achieving fine
line resist patterns in thicknesses that are required for thin films.
Kubota at Mitsubishi Electric [160] reported a positive working
polyimide based on a polyamic acid backbone with o-nitrobenzyl ester
side chain. The nitrobenzyl group decomposed into a carboxalic acid upon
UV radiation. The exposed part of the film thus became soluble in aqueous
alkaline solution.
Takano and Hayase at Toshiba [161,162] showed that a photosensi-
tive PI can be achieved by blending in a photosensitive ester compound
such as napthoquinonediazide-sulfonyl ester in the polyamic acid matrix.
The mixture reacted as a positive working photoresist and was developed
in a aqueous alkaline solution.
Omote and Yamaoka [163] reported a positive working system in
a mixture of polyamic acid with a 1,4 dihydropyridine derivative. The
solubility in the unexposed area in aqueous alkaline solution was inhibited
by the hydrogen bonding, while the exposed area promoted dissolution
due to the absence of hydrogen bonding.
Sumitomo Bakelite of Japan [164] recently reported a positive work-
ing photosensitive polyimide system consisting of a mixture of polyimide
with and o-napthoquinonediazide. These authors also showed that a posi-
tive photoresist consisting of a mixture of polybenzoxazole (PBO), 0-
napthoquinonediazide, and silicone modified polyimide had better adhe-
sion (due to silicone functionality) and higher thermally stability (due to
PBO) than the polyimide precursor system.

11.11.2 Properties of Positive Working Photosensitive PI

The properties of photosensitive PI are primarily governed by (i)


the photosensitization techniques and (ii) by the modification of the base
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-589

polyamic acid precursor. Hence, the positive working PI generally exhibit


similar properties with those of negative resists since the photosensitization
side groups are usually eliminated once the reaction is completed.
Mechanical properties of photosensitive PIs are in general somewhat
lower than found in non-photosensitive polyimides with the same back-
bone [165]. Takechi et al [163] compared properties of (i) a non-photosen-
sitive, (ii) a photosensitive with ester side group, (ii) a photosensitive
with ionic side group, and (iv) a positive working polyimide. These
polyimides exhibited good mechanical properties and adhesion to a wide
variety of substrates except for the ester-type. The ester-type photosensi-
tive PI showed relatively inferior mechanical properties which were attrib-
uted to the incomplete removal of the photoreactive group [167]. However,
the adhesive and mechanical properties also depends on the constitution
of the polymer backbone. Sashida showed that adhesive properties were
significantly improved with the incorporation of siloxane groups [168].

11.11.3 Outlook for Positive Photosensitive

The positive resists are expected to become more attractive due to


environmental and safety concerns [157,158]. Positive resists tend to
reduce the swelling behavior in the exposed area that are predominant in
the negative working systems. Positive resists offer better resolution and
can be developed in aqueous solution but they have limitations in film
thickness and often cause film delamination from the substrate (1-3) during
fabrication of thin film structures.

11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE

Photodefinable benzocyclobutene (Photo-BCB), is commercialized


by Dow Chemical under the tradename of Cyclotene™ (4000 series).
Photo BCB chemistry [169-172] and processing [172] has been reported.
The use of Photo-BCB has been reported by several companies, including:
NEC [173], Siemens [174], Motorola [175] and MMS [176].

11.12.1 Physical and Chemical Properties

Photodefinable-BCB resins are negative acting materials (unexposed


areas are removed during solvent development) which are sensitive to
365 nm radiation (i-line and/or broadband exposure). They are currently
available as two formulations which cover film thickness ranges of 3 to
81lm (Cyclotene 4024) and 7 to 13llm (Cyclotene 4026). The base resins
are supplied as partially polymerized (B-staged) solutions in mesitylene.
The Photo-BCB resins are formulated with thermally sensitive addi-
tives to impart photosensitivity. Thus, these photosensitive products re-
11-590 POLYMERS IN PACKAGING

Table 11-29. Typical Physical and Chemical Properties of Photo-BCB

Property CycIotene-4024 CycIotene-4046

Solids (% Resin to Solvent) 40 46


Film Thickness Range (mm) 3.5-8.0 7.0-13.0
Viscosity (cSt @ 25°C) 350 lIOO
Density (gm/cc @ 25°C) 0.97 0.95
Ionics (ppm) <1.0 <1.0
Extractable Halogens (ppm) <5.0 <5.0

quire long term storage at -15°C (6 months) and have a pot life of about
1 week at room temperature.
Typical properties of Cyclotene 4000 series products are shown in
Table 11.29.

11.12.2 Electrical and Mechanical Properties of Photo BCB


The electrical and mechanical properties of Photosensitive-BCBs
are very similar to the dry-etch version Cyclotene®-3022 as shown in
Table 11.30 and 11.31.

Table 11-30. Electrical Properties of Photo-BCB

Dielectric Constant (I KHz-20 Ghz) 2.65


Dissipation Factor (1 KHz-I MHz) 0.0008
Breakdown Voltage (V/cm) 3 x 1()6
Volume Resistivity (ohm/cm) I x 10 19

Table 11-31. Mechanical Properties of Photo-BCB

CTE (ppm) 52
Elongation (%) 8
Tensile Strength (Mpa) 87
Tg >350°C
Tensile Modulus (Gpa) 2.9
Poisson's Ratio 0.34
Stress (Mpa) 28-32
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-591

11.12.3 Processing Photosensitive-BeB

The processing sequence for Photo-BCB closely resembles that of


most common negative acting photoresists. There are seven basic process-
ing steps as shown in Figure 11-51.

11.12.3.1 Surface Preparation

The interface between a deposited dielectric film and other layers


(i.e. metal, polymer, and substrate) usually requires some type of surface
treatment or preparation to ensure reliable bonding. The adhesion pro-
moter, 3-Aminopropyltriethoxysilane (3-APS), 0.5% in water (or metha-

Surface
1)
Preparation

"
Coating

.
2)

3)
Pre-Exposure
~~
Bake
J

4) Exposure

S)
Solvent
Develop
- Rework

6) Cure

I
7) Descum - Metallization
Processing
r--

"
Systems
Assembly

Figure 11-S1. Photo-BCB Process Flow


11-592 POLYMERS IN PACKAGING

nol), is used to enhance the adhesion of Photo-BCB to copper (power/


ground planes) and silicon-oxide (substrate) [177].
The adhesion between multiple layers of Photo-BCB (polymer-to-
polymer adhesion) is enhanced by partially curing the underlying layer,
thus leaving unreacted sites for the next layer ofPhoto-BCB to chemically
bond to [178]. This practice produces a monolithic interface between
successive Photo-BCB layers. In general, the use of 3-APS is beneficial
between all layers where there is a high ratio of metal to polymer, i.e.
on top of power and ground planes.

11.12.3.2 Coating
Spin Coating For wafer applications such as IC secondary passiv-
ationlbuffer coating, wafer redistributionlbumping and MCM-D fabricated
on silicon photosensitive-BCB resins can be applied to substrates on a
track coater. A 10 sec/500 rpm relaxation spin is used to spread the
polymer followed by a 30 sec spin at high speeds (2600 and 2400 rpms)
to achieve the final (post cure) polymer thicknesses of 5 ~m or lO~m.
At the beginning of the spin cycle, a mesitylene backside rinse is employed
to clean the substrate backside and to suppress polymer fiber formation
in the spinner bowl. A second backside rinse at 800 rpms is used immedi-
ately after polymer coating for edge bead removal.
Spin curves for the two Photo-BCB formulations are shown in Figure
11-52. The top curve for each represents the film thickness after deposition
and pre-bake (see pre-bake section). The lower curves are after cure (see
cure section).
Spray coating, extrusion coating [179] and meniscus coating [180]
have been used to deposit photo BCB on 400 mm glass, aluminum and
PWB panels. Material utilization for spray cpoating is about 75%, for
extrusion and meniscus it is in excess of 90%. Film uniformity across
the boards was measured to be in the range of <5% (3 sigma) for meniscus
and extrusion coating and <10% for spray coating.

11.12.3.3 Pre-Exposu re Bake


After coating the substrates were baked to drive off excess solvent
and stabilize the films before going to photo-lithography. This operation
is normally carried out in-line on a hot plate station located sequentially
after the wafer coating station. Alternatively, a convection oven has been
used (off-line) with equally good results. The pre-exposure bake conditions
need to be tightly controlled because they effect several subsequent pro-
cessing operations, such as exposure dose and develop time. Film proper-
ties, such as via resolution, film retention, and film quality are also affected.
Figure 11-53 shows the effect of pre-exposure bake on via resolution.
The optimum pre-bake conditions for processing are selected based
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-593

w~------------------------,
(10)
15

I·s
.....,
10

(I) 5

i
(I)

0
(5)
~
-... 10 Pre-Baked
S

~:::~
~

0
0 1000 2000 3000 4000 5000 6000 7000
Spin Speed (rpm)

Figure 11-52. Spin Curves for 5 and 10 Jl.m Photo-BCB on 4 inch Silicon
Wafers. High Efficiency Coating Techniques

60 ....-:(1:-::0):--------------,
50 5o,.un I Cracks I>
40
30
20
10
O~oooooooooBHI--IJ.Llif_tl~----------...:s__8__I
:)U (5)
Cracks
40
30
20
10
O~~~~~~~~--~~~~~
40 50 60 70 80 90 100 110 120 130
Pre-Bake Telll>erture (degC)

Figure 11-53. Via Resolution vs Prebake Temperature (hot-plate) for 5 Jl.m


and 10 Jl.m Photo-BCB
11-594 POLYMERS IN PACKAGING

on obtaining maximum via resolution without sacrificing structural proper-


ties. At low prebake temperatures the films are washed away (liftoff)
during the subsequent solvent development process. At high temperatures
the films are susceptible to crack formation during development. The hot
plate pre-bake conditions of 75°C for 90 sec (5 11m films) and 90°C for
90 sec (10 11m films) are typical. This gives a robust processing window
for each resin.
Oven bakes of 75°C for 20 minutes were found to work well for
both formulations (5 11m and 10 11m).

11.12.3.4 Exposure
One of the most important factors in obtaining quality vias with the
desired sidewall profile is the exposure process. In general, for photosensi-
tive polyimides, the mask dimensions define the dimensions at the bottom
of the vias while the top dimensions change (blowout) during development
and cure [181] . For the Photosensitive-BCBs, the mask dimensions define
the dimensions at the top of the via; and the bottom via dimensions are
determined by the slope of the sidewall and the extent of the "foot" (region
of gradually sloping polymer at the bottom of the via). All of the via
resolution values reported in this paper refer to the dimensions between
the foot which are totally clear of polymer (see Figure 11-54).
Exposures can be performed using: (a) a projection scanner and (b)
printers in hard contact, soft contact, and proximity modes. In general,
contact and projection techniques produced more vertical side wall profiles
than proximity modes because of the reduction in focus effects [182].
The use of contact exposure, however, results in yield loss and an increase
in the number of mask cleaning operations required.

Top of Via
(mask dimensions)

Via
Opening

Foot

Figure 11-54. Schematic Diagram of a Via in Photo-BeB where e is the Via


Sidewall Angle
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-595

Films of the Photo-BCB resins are highly sensitive to i-line radiation,


but can be exposed using broad band tools. The absolute dose level
required to image Photo-BCB is related to the optical cross-section of
the photo-additives contained in the resin, relative to the thickness of
the film.
Figure 11-55 is a plot of via resolution versus exposure dose and
Figure 11-56 shows via resolution vs exposure gap using a Karl-Suss
MA-150 proximity exposure tool. 11 is clear that resolution increases with
a decrease in exposure gap. Figure 11-57 is a plot of via resolution versus
exposure dose using a Perkin-Elmer scanning projection printer.
At extremely low doses, Photo-BCB films are insufficiently cross-
linked at the bottom of the film, and swelling or liftoff of the films results
during the development process. At excessive exposure doses, the vias
begin to close up due to light scattering off the underlying surfaces.
Exposure doses of 175 to 200 mJ/cm2 is typical using the 5 /lm formulation,
and 750 to 900 mJ/cm2 is typical for the 10 /lm formulation.

11.12.3.5 Solvent Develop


The unexposed areas of the film can be developed using a variety
of different solvents. Stoddard Solvent had been used to define the vias with
good results [169,170]. Isopar-L®lProglyde-DMM® (isopar/proglyde)

60
(10)
50
,.... 40
~ 30 35J.1ll1
e 20
.~ 25
'-" 10
~
.!ia 0
(5)
~ 'u
~ 40
~ 30
20
10
0
0 250 500 750 1000 1250 1500

FJcposure Dose (mJ/cm1

Figure 11-55. Via Resolution vs Exposure Dose for 10 ILm and 5 ILm Photo-
BCB Using a Karl Suss MA-150 (@ 10 ILm print gap)
Figure 11-56. Via Resolution vs Exposure Gap for 10 J.tm and 5 J.tm Photo-
BCB Using a Karl Suss MA-150

60
(10)
50
,...... 40
= 30
III

a
·e......, 20
10
~
.S! 0
~ (5)
'u
~ 40
35J.11ll
~ 30
20 25J.11ll
10
0

Exposure Dose (mJ/cm1

Figure 11-57. Via Resolution vs Exposure Dose for 5 J.tm and 10 J.tm Photo-
=
BCB Using a Perkin-Elmer Projection Printer (focus depth 0, aperture 4)=
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-597

(Dow DS2100) and tri-isopropyl benzene (Dow DS 3000) improve the


via resolution and exhibit wider processing windows.
In general photo PI looses thickness (about 50+%) due to shrinkage
during imidization when water and the cross linking agent are evolved.
BCB, however, looses very little thickness during curing (about 5%) but
does loose thickness during development due to the disolution of lower
molecular weight (Mw), uncrosslinked polymer chains. This loss is depen-
dent on variables such as exposure dose, but is typically 10-20%.
Vias can be opened using "puddle" and "immersion" techniques
with isopar/proglyde as the developer. The puddle method consists of
placing the exposed substrates onto the stationary chuck of the spin coater
and dispensing a layer of solvent (puddle) on the surface. The solvent
was left on the film for a predetermined period of time to allow complete
dissolution of the non-exposed areas. At the end of the development period
the devices are rinsed by spinning, while a stream of the rinse solvent
(developing solvent or water based surfactant) is sprayed over the surface.
The key to a successful development is to allow adequate puddle
time for completion of the develop process and rinsing briefly (while
spinning) to remove the dissolved polymer from the substrate surface.
When determining the process windows, the development "end-point" is
visually detected as a series of colored interference fringes which disap-
pears when the film is completely dissolved from the non-exposed areas.
An additional 10-30 sec of over-development is typically used to ensure
complete dissolution (see Figure 11-58).
The develop end-point, is dependent on the film thickness, solvent
composition, delay time between prebake and development, and tempera-
ture of the solvent. Figure 11-59 illustrates the affect of film thickness
on development time. For both formulations, thicker films take longer to
develop than thin films.
Figure 11-60 shows the relationship between develop endpoint and
pre-bake temperature.
Figure 11-61 illustrates the rate of develop endpoint increase as a
function of the delay time between prebake and development steps using
the standard 90°C/90 sec (10 11m films) and the 75°C/90 sec (5 11m
films) prebakes.
The combination of these phenomena require that either a set protocol
be used for developing within a specified time window, or an end-point
detection system be used for each wafer (or batch of wafers). In order to
cover a 24 hr delay (typical in a manufacturing environment), a puddle
develop time of 75 sec is recommended for the 5 11m formulation and
125 sec for the 10 11m formulation. This puddle time should provide an
adequate amount of over development and good via resolution while
maintaining an acceptable level of film loss. For delays greater than 24
hrs, additional develop time is required.
11-598 POLYMERS IN PACKAGING

60
Fndpoint (10)
50
40
.......
III 30
e= 20
.~
'-' 10
III
0
=
.S!
III :>V Fndpoint
=
~
40
30
~ 20
10
0
000000000000000
N('f')-.t-or'l\Or--OOO'lO-N('f')-.t-or'l\O
...... .....c .....c .............. .....c ""'"""
Develop Time (sec)

Figure 11·58. Via Resolution vs Develop Time for 10 ....m and 5 ....m Photo·BCB

160
(10)
140
,-..
~ 120
rIl
'-'

=
..... 100
~
~
80

=
eg. 60 (5)
40
~0
20

o 2 4 6 8 10 12 14 16 18 20
Film Thickness (microns)

Figure 11·59. Develop End·point vs Film Thickness for 5 ....m and 10 ....m
Photo-BCB
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-599

150
(10)
125
100
,-. 75
~
en
50 Liftoff
=
'-'

25
l.:I
0
(5)
~
125
-8- 100
~ 75
50
25
0
30 40 50 60 70 80 90 100 110 120 130
Hotplate Temperature (degC)

Figure 11-60. Develop End-point vs Hot Plate Temperature for 5 f1m and 10
f1m Photo-BCB

140
(10)
120
..-.. 100
'-'
~

~
= 8060
'0
I
'g
~
0
=
S 80
(5)

-8'~
0
60
40
20
0 5 10 15 20 25 30
Delay Coat-Develop (hrs)

Figure 11-61. Develop End-point vs Delay Time After Coating for 5 and 10
f1m Photo-BCB
11-600 POLYMERS IN PACKAGING

Figure 11-62 shows the film thickness as a function of over devel-


opment.
An SEM of a 35 !lm via formed by isopar/proglyge development
using the puddle technique in a 10 !lm layer of photo BCB is shown in
Figure 11-63.
Other traditional developing techniques such as tank immersion and
spray developing are applicable with triisopropylbenzene (TIB) developer.

11.12.3.6 Rework
Prior to cure, films which are inspected and found to have defects
or are misaligned, can be stripped. Parts to be stripped are submerged in
a stripper bath solution for about 30 minutes at RT. This process does
not effect underlying layers of metal or soft cured polymer.

11
(10)
10

...- 9
~
e
.Y
8
Endpoint
S 7
v.I
v.I

] 6 (5)

~ 5
:§
~ 4

3 Fndpoint
2
25 50 75 100 125 150
Develop Time (sec)
Figure 11-62. Film Thickness as a Function of Over-Development for 5 and
10 f1m Photo-BCB
11.12 PHOTOSENSITIVE BENZOCYCLOBUTENE 11-601

Figure 11-63. SEM Micrographs of 35 mm Vias in lOJLm Film of Photo-BCB

11.12.3.7 Cure

At this stage of processing, the polymer film was cured to ensure


resistance to subsequent processing operations, e.g. chemical baths, metal-
lizations, and thermal cycling. The cure must be carried out under nitrogen
[183] . Experience has shown that adhesion between multiple Photodefin-
able-BCB layers can be enhanced if each layer of BCB is partially cured
to about 75-85% (soft cure). After all processing is complete the films
were cured to a 95+% level in a final curing step (hard cure). Alternatively,
cure may be performed in a belt oven with equally good results [184] .

11 .12.3.8 Descum

A descum process is required to clean the vias of debris and/or


remove the very thin residual film which is often present « 100 Angstroms)
after the development process. This is typically accomplished with a O 2/
CF4 mixture.

11.12.3.9 Substrate Dicing

Two different dicing alternatives have been examined for Photosensi-


tive BCB: (1) photo pattern and then solvent develop the scribe streets
11-602 POLYMERS IN PACKAGING

at each layer during the fabrication process; or (2) laser ablation [185,186]
through the full dielectric thickness after all processing is complete. Both
alternatives are followed by traditional sawing through the silicon wafer.

11 .12.3.10 Applications of Photo BeB


Figure 11-64 shows a 43 sq mm R-3000 workstation fabricated using
the Cyclotene™ 4000 series by NEC (174). The chip set includes an
R3000 CPU (1 integer Unit, 1 Floating Point Unit) and 1 Mbyte of SRAM
memory. The chips have been mounted using NEC flip TAB technology.
The conductor is 4 Jlm x 20 Jlm plated up copper. No adhesion or barrier
layer metallization is present resulting in a conductivity of 1.75 W/cm.
BCB layers are 9 Jlm thick resulting in an overall impedance of 50 W ±
4%. The modules passed MIL 883 long term reliability tests [500 thermal
cycles from -45 to 125°C; high temperature storage (125°C) for 3200
hrs; 85% humidity and 85°C for 3200 hrs.] Resistance changes in all
cases were <0.2%.
Figure 11-65 shows a commercial MMS Twinstar™ module. The
Twinstar contains two Pentium® processor die mounted on a CulBCB
high density substrate. The substrate is housed in a 347 pin PGA. The
Twinstar is used in high end server applications [176].

Figure 11-64. NEC R-3000 Workstation MCM


11.13 PHOTOSENSITIVE EPOXY 11-603

Figure 11-65. MMS Twinstar CulBCB High Density Module

11.13 PHOTOSENSITIVE EPOXY


Epoxies are usually made photodefinable by mixing the base resin
with photosensitizing agents. Intrinsically photosensitive epoxies are in
the developmental stage and their availability on a commercial scale
are limited.

11.13.1 Photodefinable Cyanate-esters and Epoxies


A negative acting acrylate-triazine based photodefinable epoxy was
formulated by AT & T for the Polymer Hybrid Integrated Circuit (POLY-
CHIC) to provide the dielectric medium between conductors layers
[187,188]. The low dielectric constant (e :: 2.8 for the base triazine)
combined with fine line capability in a controlled impedance environment
provide the fabrication capability for MCM applications requiring high
frequency, high interconnect density with a large number of 1I0's. This
polymer is spray coated to produce films 25-50 Ilm thick, UV patterned
and solvent developed. The formulation of this material is described in
reference 188. A summary of the polymer properties and their correspond-
ing electrical properties are summarized in Table 11-32.
11-604 POLYMERS IN PACKAGING

Table 11-32. Physical and Electrical Properties of Cured Triazine-based


Epoxy (PHP-92)

Property Value Comments

Moisture Absorption -2% @ 100% RH typical of epoxy systems 25% RH-


IlZ/Zo < 5% 85% RH
Thermal Conductivity 0.2 W/m-K
TCE 65 ppml°C TMA (25-lOO°C)
Tg 180°C (I 55-2 10°C) DSC & TMA
Thermal 180°C long term
Stability 300°C spikes (i.e. soldering)
E, (Young's Modulus) 2-7 x 1010 dynes/cm2 mechanical tensile tester
Room Temp. Film Stress 2.2-2.9 x lOS dyneslcm2 warpage test with fused quartz disk
Thermal Cycling >200 cycles -40°C + 130°C
E (for resin) 2.8 GHz range: Zo from reflection
coefficient and propagation
tan /) <.025 GHz range
Dielectric Strength 35V/l!m (900 V/mil) ASTM D149
Volume Resistivity >3.7 x 10 12 Q-cm ASTM D257 (instrument-limited)

Researchers at IBM designed several chemically amplified epoxy


resists systems by combining commercially available epoxy resins and
photoinitiators such as aryldiazonium, diaryliodonium, and triarylsulfon-
ium metal halides [189]. When formulated with the onium salt photoinitia-
tors, epoxy resins undergo cross-linking reactions to produce negative
tone resists. Negative tone resists often encounter swelling of the cross-
linked matrix during development in organic solvents. Swelling results
in distorted images and loss of adhesion to the substrate. Stewart et al
[190] formulated an organic solvent developable UV resist capable of
providing submicron images with reduced swelling. The reaction involved
conversion of optically transparent commercial resins, styrene-allyl alco-
hol copolymers to glycidyl ethers, thereby providing cationically polymeri-
zable functionalities.

11.13.2 Intrinsically photodefinable epoxy


Ciba offers a higher dielectric constant (4.1 at 1 MHz) photodefinable
LMB 7081 epoxy resin with a shelf life of 6 months [191]. In research
studies, this material was deposited as a thin dielectric layer (-20 microns)
on copper cladded laminates using the following processing scheme: (i)
spin coat at 2000 rpm for 35 sec, (ii) tack dry at 60-80° C using a hot
plate, (iii) expose at 365-400 nanometer, -1600 mJ/cm2, (iv) oven heat
at 110°C, 1 hr, (v) develop in y-butyl lactone, 3-5 minutes, and (vi) a
11.14 MICROELECTRONIC APPLICATIONS OF PSPls 11-605

final cure at 150°C for 1 hour. Vias as small as 40 11m in both array
and chain configurations were opened without any smearing of epoxy.
Preliminary results indicate that the LMB 7081 formulation can be a good
candidate material for thin film deposition of MCM-L type application
primarily due to its simplified processing and much lower cost compared
to polyimides.

11.14 MICROELECTRONIC APPLICATIONS OF PSPls


Polyimide coatings or films have been widely utilized in the field
of microelectronics in both chip and packaging applications due to their
excellent thermal and electrical properties. For semiconductor devices,
these applications include (a) protective coatings for Si and GaAs devices,
(b) insulation dielectric layers, and (c) stress-relief buffer coatings. For
thin-film packaging, applications include (a) dielectric insulation layers
and (b) corrosion-protective passivating layers to protect metallurgy in-
structures fabricated for use in the computer and communications indus-
tries. They are also used as protective coatings and dielectric insulation
layers in printer thermal head applications and in photolinesensors. Starting
in the late 1980s, PSPI coatings have been utilized in these applications
because very precise relief patterns could be formed by using conventional
photoresist processes. Many researchers also have proposed various kinds
of applications of PSPIs in other fields. A sampling of some of the
more common applications which use the PSPls described in chapter are
listed below.

11.14.1 Chip Coating Applications


11.14.1.1 Alpha-Particle Protection Layers
In the early 1980s, memory chips had encountered a very serious
problem called soft error, which is the disappearance of memory caused
by penetration of an alpha particle. The source of the alpha particle was
found to be uranium or thorium in the inorganic filler, such as silica or
alumina, in encapsulation materials. One of the preventive measures was
polyimide coating over the chip as an alpha-particle protection layer. A
thick layer of more than 30 11m is necessary to absorb an alpha particle.
The first target of Rubner's PSPI was this application. However, alpha-
particle shielding is not as critical today because fillers of low uranium
content became available. The polyimide coating itself must have on
the order of less than 1 parts per trillion of uranium or thorium
content [ 192].

11.14.1.2 Buffer Coating for Stress Relief


The stress generated under plastic-encapslation processing causes a
substantial amount of damage to the LSI chip by shearing and moving
11-606 POLYMERS IN PACKAGING

alumimun wiring and cracking of the inorganic passivation layer. The


polyimide chip coating before encapslation can prevent these damages
and increase production yields. Therefore, this process is carried out in
many LSI production lines, where PSPI is patterned so that the via hole
for wiring and the dicing line is opened. Khan et al. studied the effect of
buffer coating on the production yields of chips. They applied ester-
type PSPI based on BTDA-ODA-MDP to bipolar interface devices
containing piezoresistive circuits with plastic encapsulation, which would
typically have 10% yield loss due to threshold voltage failure. Acceler-
ated testing with unbiased and biased pressure-cooker storage produced
results in favor of the PSPI-coated devices as compared with the
uncoated control. [193]
In the past, the use of Probimide 400 as a buffer coat has been delayed
since the base of installed i-line steppers required for this technology
was limited. Today, i-line steppers are available in many manufacturing
facilities which may give new opportunities for Probimide 400 use. Again,
for throughput considerations, the thickness range of 2-6 )lm is recom-
mended.

11.14.1.3 Thin-Film Interconnect


One main application of "Photoneece" is in semiconductor devices.
In this application, the photosensitive polyimide is required to adhere well
to silicon, Si3N4, and encapsulating resins. "Photoneece" exhibits superior
adhesion to these materials, as shown in Table 11-33. It is especially
suitable for GaAs substrates which require low curing temperature, below
300°C. Naitoh et al. evaluated "Photoneece" as an insulation layer for
five-layered metallization structures having 20K gates CMOS logic LSI
[194], where the thickness of the polyimide layer was 1.6 )lm and the
via hole size was 3 )lm.

11.14.2 Packaging Interlayer Applications


11.14.2.1 Interlayer Applications in High-Density Multichip Modules
Polyimide has been utilized as insulation layer in interconnect struc-
tures because of the good planarization properties, good thermal perfor-
mance, and excellent electrical characteristics for the propagation of high-
speed signals [195,196].

11.14.2.2 Applications Using Covalently Bonded Ester-Type PSPls


Nowadays, much attention has been given to PSPI in this field, due
to easy patterning of fine via holes for interconnection between signal
layers by a very simple process. Ester-type PSPIs have many advantages
for this application (e.g., wide process margin, etc.). Chakravorty of Boe-
11.14 MICROELECTRONIC APPLICATIONS OF PSPls 11-607

Table 11-33. Adhesion Properties of "Photoneece" to the Molding Compound"

Curing Adhesive Strength (MPa)


Temperature
Polyimides .LQ Initial After PCTb

Poly(amic acid) composed 320 12 0


of PMDA and DAE 350 12 0
350' 15 0.2
350d 24 0.5
400 18 0
"Photoneece" UR-3140 320 30 18
350 30 21
400 39 24
"Photoneece" UR-4144 320 18 12
350 30 21
400 30 15

'TH7006 produced by Toray.


bPCT (120°C, 100% rh, 168 h).
'02 plasma treatment to polyimides.
dAminosilane treatment to polyimides.

ing evaluated Selectilux HTR-3 of EM Industries as a dielectric material


in a high-density thin-film CulPI interconnect structure on a silicone
substrate [197,198]. He investigated the dependence of the PSPI or poly-
imide profile on exposure dose, linewidth, and bake temperature. He
fabricated a high-density multi chip memory module containing two layers
of orthogonal X- Y conductor lines with interconnecting vias, with a density
of close to 800 lines/cm. The signal layers consisted of 25-).tm-wide, 6-
).tm-thick electroplated copper lines fabricated at a pitch of 80 ).tm in a
poly imide dielectric 6 ).tm thick. The module consisted of six 64K SRAM
chips, two 8-bit buffers, and four discrete ceramic chip capacitors. Trans-
mission characteristics and signal cross-talk between lines were also mea-
sured [199-205].
There are many reports in the literature describing the formation of
interconnect structures which use ester-type PSPIs. Takahashi and Kolesar
of Sacramento Air Logistics Center, applied Selectilux HTR 3-200 to the
preparation of a wafer-scale integrated circuit [203]. Myszka et al. of
Motorola fabricated multichip module type-D using PSPI, including the
ester type. They built up polyimide dielectrics and copper conductor
interconnect layers on cofired substrates, and they called this structure a
"grafted multilayer" [207].
11-608 POLYMERS IN PACKAGING

11.14.2.3 Applications Using Ionic Bonded-Type PSPls


Another main application field is an assembly package which requires
a high-density metallization and low-dielectric-constant insulation layer
in order to reduce package size and signal propagation delay [208,209),
A multichip module for a computer is also included in this field [210,211].
Inoue et al. (NEC) reported the multi chip module [212] for the NEC SX-
3 supercomputer using the "Photoneece" shown in Figure 11-66 having 25-
11m-wide, 75-l1m center-to-center spacing high-density wiring, poly imide
insulating layers 20 11m in thickness and with 60 11m square via holes,
four signal layers and 6-ns/cm signal transmissions on a 225-mm square
substrate have been achieved by using "Photoneece."
Takasago et al. (Mitsubishi Electric Corp.) has developed a hybrid
IC which uses "Photoneece" for the insulation layers [213-15]. They use
copper substrates to get good heat dissipation. A multilayer circuit with
polyimide insulation layers and copper heat sink for integrated circuits
are formed on this substrate. Standard film thickness of "Photoneece" is
10 11m.

11.14.2.4 Application Using Nonshrinking Preimidized PSPI


Due to its dimensional stability upon curing, the most appealing
application ofProbimide 400 has been in multichip modules. Very substan-

3rd Signal

al

Poyer & Ground

1/0 Pin Ceramic Substrate

Figure 11-66. Cross Section of the Multichip Module Used for SX-3 Supercom-
puter Using "Photoneece" (From Ref. 212)
11.14 MICROELECTRONIC APPLICATIONS OF PSPls 11-609

tial work has been done through various groups at Boeing Corp. [197-
203]. Thus, in a key paper, a fiber-optic receiver/transmitter set with eight
copper/polyimide layers is described, which has copper lines 5 ~m thick
with a nominal linewidth of 19 ~m. When working out this eight-layer
MCM process, the Boeing groups could also introduce the following
essential processing innovations when working with Probimide 400:

• Probimide 400 has been shown to be Cu compatible, in contrast to


other PSPI (ester type).
• Cure time per polyimide layer could be reduced to 5 minl400°C to
give enough solvent resistance for the next layer to be applied.
• Solvent diffusion on the other hand can advantageously be exploited
for greatly improved planarization.

A cross section of a copperlProbimide 400 interconnect structure with high


planarity is shown in Figure 11-67. A top view of improved planarization
through solvent diffusion in Probimide 400 is depicted in Figure 11-68.

Figure 11-67. Micrograph ('" 200x) of a Polished Cross Section of the Opti-
cal Transceiver
11-610 POLYMERS IN PACKAGING

Figure 11·68. SEM Photographs Showing the Superior Planarization


Achieved Corresponding to a Denser Conductor Configuration in a Copper/
Probimide Interconnect Structure. The fabrication process is outlined in Chakravarty
and Tanielian [197]; see Chakravarty et al. [198] for other pertinent details.
11.15 RESEARCH DEVELOPMENTS/FUTURE POLYMERS 11-611

11.14.2.5 Optical Waveguides

Probimide 400 has been shown to exhibit low optical losses when
used as a waveguiding material [198]. Losses of <0.5 dB/cm have been
obtained. The generation of embedded channel waveguides through the
generation of photolocked regions have been described by Chakravorty
[197-200]. Thus, a dopantlike benzy1dimethylketal is included in the
photo-cross-linkable Probimide 400 matrix and photochemically locked
by exposure to form channels of lowered refractive index.
In this context, it appears particularly appealing to combine syntergis-
tically optical and electrical material interconnecting technologies with
one polymeric material, as described in Ref. 206.

11.14.2.6 Other Applications

As described in the preceding section, the application of PSPI started


in the field of microelectronic packaging. Nowadays, PSPIs are applied
in other fields such as optoelectronics [216], display [113], and micromach-
ines [217]. PSPI evaluation in these new fields are mostly carried out by
using ester-type PSPI, possibly because ester-type PSPI will meet the
diversified requirements of the microelectronics and related industries.

11.15 RESEARCH DEVELOPMENTS/FUTURE POLYMERS


11.15.1 Ultrathin PSPls (LB Membrane)

A Langmuir-Blodgett (LB) membrane is an unimolecular layer,


which recently has attracted a lot of attention from the viewpoint of
functional membranes such as photoresist, conductive membrane, sensor,
and molecular devices. Uekita-et al. of Kanegafuchi Chemical studied a
novel LB membrane of ester-type PSPI and its patterning properties. They
prepared a precursor having the heptadecenyl group by condensation
reaction of diacid chloride of diheptadecenyl pyromellitate and diamino
diphenyl ether. They found that the mixture of the precursor, octadecyl
alcohol, and benzoin ethyl ether gave Y-type multilayer LB films. The
thickness of the monolayer was measured to be 22 Aby ellipsometry. A
film consisting of 31 LB layers deposited on a silicon wafer was irradiated
by deep UV (254 nm) light. After the exposure, the multilayer was devel-
oped in chloroform-dimethylacetoamide and rinsed in ethanol. The photo-
sensitivity, defined as a normalized thickness of 0.5, was 140 mJ/cm2•
Infrared spectra indicated that cross-linking occurred between the dou-
ble bonds.
After the mixed multilayers (21 layers) were exposed through a
photomask, development of the film gave a negative image. A resolution
of a 0.5 mm space and a 1.0 mm line was obtained [218].
11-612 POLYMERS IN PACKAGING

Lu et al. of Southeast University in China prepared ultrathin conduc-


tive films by pyrolysis ofPSPI LB films at 1OOO°C in hydrogen atmosphere.
They found that the simple ester-type PSPI based on PMDAlODAI
HEA (hydroxyethyl acrylate) chemistry could form a LB membrane.
The conductivity of the pyrolyzed LB films increased with the film
thickness and reached the stable value of 180 Scm-I. The pyrolyzed
LB films on silicone dioxide substrate were studied by scanning
tunneling microscopy [219].

11.16 SUMMARY AND FUTURE CHALLENGES

Most negative PSPIs shrink about 50% upon curing to 400°C,


whereas others are preimidized and intrinsically photosensitive. Positive
PSPIs are just starting to be developed and have not yet been commercial-
ized. Typical properties of some PSPIs are shown in Table 11-34. Just
as in conventional nonphotosensitive polyimides, the push is on to develop
low-stress, low-dielectric-constant polymers for the chip and packaging
applications. PSPIs are being developed with low-stress and low-dielec-
tric-constant polyimides which can be made into either positive or negative
acting photosensitive polyimides.
The challenge the technical community faces is to make the backbone
chemistry as transparent as possible to the exposure wavelengths used
for patterning fine lines and vias over a variety of substrates. After formula-
tion of the polymer with the photopackage, the base resin properties
should remain unchanged, especially after thermal stressing and solvent

Table 11-34. Typical Commercially Available Photosensitive Polyimides

BTDA- BTDA- BPDA- BTDA-


Backbone Chemistry X-DA ODA ODA PMDA-ODA

Lithography photopackage i-Line g-Line i-Line g-Line


Thermal stability 0.12% h 0.08% h 0.06% h 0.11% h
(isothermal weight loss
at 400°C)
Residual stress 56MPa 42 MPa 42MPa 42 MPa
Mechanical properties
(cured films)
Tensile strength 137 MPa 192 MPa 178 MPa 151 MPa
Modulus 2.8 GPa 3.4 GPa 4.0 GPa 3.5 GPa
Strain at break 59% 69% 40% 44%
Glass transition 370°C 352°C 310-450°C 355°C
temperature (Tg)
11.17 REFERENCES 11-613

processing. Finally, the photopackage components should cleanly evolve


from the polyimide film before complete imidization, to produce defect-
free films. The challenges in developing low-cost PSPls for both semicon-
ductor and packaging applications with minimum photopackage impact
on the base polymer properties and with minimal film shrinkage are formi-
dable.
Acknowledgment: The authors wish to thank Dr. Swapan Bhatta-
charya of Georgia Tech for sections related to epoxy and positive photo-
polymer.

11.17 REFERENCES
1. R. R. Tummala. "Electronic Packaging in 1990's, A Perspective from America,"
IEEE Trans. Components Hybrids Manu! Technol., CHMT-14: pp. 262-271,1991.
2. R. R. Tummala. "Ceramic and Glass-Ceramic Packaging in the 1990's," J. Am.
Ceram. Soc., 24: pp. 895-908, 1991.
3. Integrated Circuit Eng. Corp. Update & "Worldwide IC Industry Economic Forecast",
pp. 2-18, 1991.
4. "Experts: GaAs Applications Booming", Semicond. Int., pp. 26-29, October 1991.
5. R. Iscoff. "High Speed GaAs: Still a Niche Technology?" Semicond. Int., pp. 60-
66,1992.
6. "Japan Report," Semicond. Int., p. 35, October 199.
7. G. Larrabee and P. Chatierjee. "DRAM Manufacturing in the 90s." Semicond. Int.,
pp. 90-92, May 1991.
8. R. J. Kopp. "Navigating to Advanced Wafer Processing," Semicond. Int., pp. 34-41,
January 1992.
9. Electron. Eng. Times, Issue 687, pp. 1 and 16, April 6, 1992.
10. P. J. Cavill, 1. M. Wilkenson, and P. E. Stapleton. "Wafer-Scale Integration,"
Microelectron. Manu! Technol., pp. 57-58, May 1991.
11. P. Burggraaf. "Lithography's Leading Edge, Part I: Phase-Shift Technology," Semi-
condo Int., pp. 42-47, February 1992.
12. I. A. Shoreef, J. R. Maldanado, Y. Vladimirsky, and D. L. Katcoff. 'Thermoelastic
Behavior of X-ray Litho Masks During Irradiation," IBM J. of Res. Devel., 34: pp.
718-735, 1990.
13. P. Singer. "Trends in CMOS Development," Semicond. Int., pp. 56-60, April 1992.
14. "Wafer Process News," Semicond. Int., p. 32, February 1992.
15. Proceedings, 8th International IEEE VLSI Multilevel Interconnection Conference,
1991.
16. M. M. Khan, T. S. Tarter, and H. Fatemi. "Stress Relief in Plastic Encapsulated
Integrated Circuit Devices by Die Coating with Photodefinable Polyimide," Proceed-
ings of the 38th Electronics Components Conference, Los Angeles, Ca. May 9-
11, 1988.
17. G. Samuelson and S. Lytle. "Reliability of Polyimide in Semiconductor Devices,"
Symposium on Polymer Materials for Electronic Applications, Coatings, and Plastics
Chemical Division, ACS 2nd International Congress, Las Vegas, NV, pp. 1-27, 1980.
18. T. Kwok and P. S. Ho, in Diffusion Phenomena in Thin Films and Microelectronic
Materials, ed. by D. Gupta and P. S. Ho, pp. 369-425, 1988.
19. M. A. Korhonen, P. Borgesen and Che-Yu Li. "Stress-Induced Voiding and Stress
Relaxation in Passivated Metal Lines"; M. Inoue and S. Ogawa. "Japanese Perspec-
11-614 POLYMERS IN PACKAGING

tive: Recent Studies on Stress Induced Phenomena in Metallization in Japan"; P.


Totta. "Interconnection Materials Reliability U. S. Experience," First International
Workshop on Stress Induced Phenomena, 1991.
20. R. A. Larson. IBM J. Res. Devel., 24(3): p. 268, 1980.
21. H. C. Cook, P. A. Farrar, R. R. Uttecht, and J. P. Wilson. OBM Tech. Disc. Bull.,
16: p. 728, 1983.
22. A. W. Lin. Evaluation of Polyimides as Dielectric Materials for Multichip Packages
with Multilevel Interconnection Structure, Proceedings 39th Electron. Components
Conf. (ECC), pp. 148-154, 1989.
23. B. Schwartz. "Multilayer Ceramics," Mat. Res. Soc. Symp. Proc., Vol. 40, pp.
50-59, 1985.
24. E. S. Anolick, J. R. Lloyd, G. T. Chiu, and V. Korobov, in Polyimides: Synthesis,
Characterization, and Applications, ed. by K. L. Mittal, vol. 2., pp. 889-904, Plenum
Press, New York, 1984.
25. "Overcoats Protect RAMs from Alphas," Electron. Rev., pp. 41-42, September
11, 1980.
26. M. F. Bregman and C. A. Kovac. Solid State Technol., pp. 75-80, 1988.
27. R. J. Jensen, 1. P. Cummings, and H. Vora.IEEE Trans. Components Hybrids Manuf
Technol., CHMT-7(4): pp. 384-393. 1984.
28. T. Watari and H. Murano. IEEE Trans. Components Hybrids Manuf Technol.,
CHMT-8(4): pp. 462-467, 1985.
29. H. Tsunetsugu, A. Takagi, and K. Moriya. Int. J. Hybrid Microelectron. 8: pp.
21-26, 1985.
30. R. J. Jensen. Proc. ACS Div. Poly. Mater.: Sci. Eng., 55: pp. 413-419, 1986.
31. L. B. Rothman. Solid State Sci. Technol., 2216-2220, 1980.
32. K. Sato, S. Harada, A. Saiki, T., Kimura, T. Okubo, and K. Mukai. IEEE Trans.
Parts Hybrid Packaging, 9PHP-9: p. 176, 1973.
33. G. Czornyj, R. Chen, S. Kim, G. Prada-Silva, M. Ree, H. Souleotis, and D. Yang.
Proc. of the 1991 Spring Meeting of the Materials Research Society, 1991.
34. T. F. Redmond, C. Prasad, and G. A. Walker. "Polyimide Copper Thin Film Redistri-
bution on Glass Ceramic/Copper Multilevel Substrates, 41st Electronic Components
and Technology Conference, p. 689, 1991.
35. R. R. Tummala, H. R. Potts, and S. Ahmed. "Packaging Technology for IBM's Latest
Mainframe Computers (S/390IES9000), 41 st Electronic Components and Technology
Conference, p. 682, 1991.
36. C. Feger, M. M. Khojasteh, and S. E. Molis, Polyimides: Trends in Materials and
Applications, Proceedings of 5th International Conf. on Polyimides, Society of Plas-
tics Engineers, Ellenville, N. Y. 1994.
37. D. Wilson, H. D. Stenzenberger and P. H. Hergenrother, Polyimides, Blackie & Son,
Ltd., Chapman and Hall, 1990.
38. E. S. Moyer, in Polyimides: Synthesis, Characterization and Adhesion, J. E. McGrath,
L. T. Taylor, T. C. Ward, J. P. Wightman, Eds. pp. 89-187, Virginia Tech Center
for Adhesive and Sealant, Blacksburg, Virginia, 1992
39. P. Garrou and I. Turlik, Chapter 4, Materials of Construction: Substrate, Dielectric,
Metallization. pp. 93-164, Thin Film Multichip Modules, G. Messner, I. Turlik,
J. W. Balde and P. E. Garrou, ISHM 1992.
40. S. Numata, N. Kinjo and D. Makino, Chemical Structures and Properties of Low
Thermal Expansion Coefficient Polyimides, Polymer Eng. & Sci., Vol. 28,
#14, 1988.
41. M. T. Pottiger, J. C., Coburn and J. Edman, The Effect of Orientation on Thermal
Expansion Behavior of Polyimide Films, Journal of Polymer Sci.: Part B: Polymer
Physics, Vol. 32, pp. 825-837, 1994.
11.17 REFERENCES 11-615

42. A. F. Arnold, Y. Y. Cheng, P. M. Cotts, R. D. Diller, D. C. Hofer, M. Khojasteh, E.


Macy, P. R. Shah, and W. Volksen. "Polyimide Coating Compositions Based on
Meta-Dialkyldihydrogen Pyromellitate and Aromatic Diamines", U.S. Patent
4,849,501, 1989.
43. W. Volksen, D. Y. Yoon, J. L. Hedrick, and D. Hofer. "Chemistry and Characteriza-
tion of Polyimides Derived from Poly(amic alkyl Esters)", Materials Science of High
Temperature Polymers for Microelectronics, ed. by D. T. Grubb, I. Mita, and D. Y.
Yoon, p. 23, 1991.
44. H. R. Brown, A. C.M. Yang, T. P. Russell, W. Volksen, and E. J. Kramer, Polymer,
29: p. 1807, 1988.
45. W. Volksen, D. Y. Yoon, and J. Hedrick. "Polyamic Alkyl Esters: Versatile Polyimide
Precursors for Improved Dielectric Coatings," 41 st Electronic Components and Tech-
nology Conference, p. 572, 1991.
46. K L. Mittal. "Adhesion Measurement: Recent Progress, Unsolved Problems, and
Prospects," in Adhesion Measurement of Thin Films, Thick Films, and Bulk Coatings,
ed. by K L. Mittal, pp. 5-17, American Society for Testing and Materials, Philadel-
phia,1978.
47. Y. H. Kim, G. Walker, J. Kim, and J. Park. "Interfaces of Metal to Polyimide," J.
Adhesion Sci. Technol., 1(4): p. 331, 1987.
48. Y. H. Kim, J. Kim, G. F. Walker, C. Feger, and S. Kowalczyk. "Interface and
Adhesion of Polymers to Metals," J. Adhesion Sci.Techno!., 2: p. 95, 1988.
49. S. Kowalczyk, Y. H. Kim, G. F. Walker, and J. Kim, "Polyimide-Copper Interfaces;
Growth Sequence Dependence and Solvent Effects," J Vac. Sci. Technol., A6(3):
1989.
50. M. Ree, T. L. Nunes, K-J. Chen, and G. Czornyj, "Structure and Properties of
BPDA-PDA Polyimide from Its Poly(amic acid) Precursor Comp1exed with an
Aminoalkyl Methacrylate," in Materials Science of High Temperature Polymers for
Microelectronics, ed. by D. T. Grubb, I. Mita, and D. Y. Yoon, p. 211, 1991.
51. F. Arnold, S. Z.D. Cheng, F. W. Harris, and F. Li. Trends Polymer Sci. 1(8): p.
243,1993.
52. S. T. Holzinger. "TAB Types and Material Choices," EXPO SMT'88 Technical
Proceedings, p. 247, 1988.
53. S. T. Holzinger. "TAB: Mechanical, Chemical, and Thermal Considerations for Three
Layer TAB Materials," EXPO SMT International '89 Technical Proceedings, p.
149, 1989.
54. M. Monisinger. "Advantages of Polyimide Film in TAB Packaging," Surface Mount
Technol., 61, October, 1989.
55. C. E. Sroog, A. L. Endry, S. V. Abramo, C. E. Berr, W. M. Edwards, and K L.
Olivier. J. Polym. Sci. A, 3: p. 1373, 1965.
56. J. A. Kreuz, and R. F. Sutton. Institute ofPrinted Circuits 33rd Annual Meeting, 1990.
57. H. Nagano, H. Kawai, and K Yonezawa. Symposium on Recent Advances in Poly-
imides and Other High Performance Polymers, 1987.
58. R. A. Kirchoff, and K. J. Bruza. "Benzocyclobutenes in Polymer Synthesis," Prog.
Polym. Sci., 18: pp. 85-1185, 1993.
59. P. Garrou. "Polymer Dielectrics for Multichip Module Packaging," Proc. IEEE, 80:
p. 1942, 1992.
60. Burdeaux, D., Townsend, P., Carr., Garrou, P., "Benzocyclobutene Dielectrics for
the Fabrication of High Density, Thin Film Multichip Modules", J. Electronic Materi-
als, Vol. 19, 1990, p. 1357.
61. S. F. Gong, J. Strandberg, H. Theide, H. Hentzell, H. Hesselboom, and W. Kamer.
"Investigation of High Speed Pulse Transmission in MCM-D," IEEE Trans. Compo-
nents Hybrids Manu! Techno!., CHMT, 16: p. 735, 1993.
11-616 POLYMERS IN PACKAGING

62. T. Shimoto, K. Matsui, and K. Utsumi. "CulPhotosensitive BCB Thin Film Multilayer
Technology for High Performance Multichip Modules," Proceed. Int. Conf. MCM's,
p. 115, 1994.
63. P. Chinoy, and J. Tajadod. "Processing and Microwave Characterization of Multilevel
Interconnects Using BCB Dielectric," IEEE Trans. Components Hybrids Manu!
Technol., 16: p. 714, 1993.
64. H. Sakai et al. "A Millimeter-Wave Flip-Chip IC Using Micro-Bumpo Bonding
Technology," Proceed. IEEE Int. Solid States Circuit Conf., 1996, p. 408.
65. M. Robertsson, K. Engberg, P. Eriksen, H. Hesselboom, M. Niburg, G. Palmskog.
"Optical Interconnects in Packaging for Telecom Applications," Proceed. 10th Euro-
pean Microelectronics Conference, 1995, p. 580.
66. H. Projanto and D. Denton. "Moisture Uptake in BCB Films for Electronic Packaging
Applications," Proc. MRS Symp., 203: p. 295, 1991.
67. T. Stokich, D. Burdeaux, C. Mohler, P. Townsend, S. Warrington, J. Tou, B. 1. Han,
e. Pryde, H. Bair, and G. Johnson. "Thermal and Oxidative Stability of Polymer Thin
Films Made from DVS-BCB." T. Stokich, D. Burdeaux, C. Mohler, P. Townsend, M.
Dibbs, R. Harris, M. Joseph, e. Fulks, M. McCulloch, and R. Dettman. "Advances
in the Thermo-oxidative Stabilization of DVS-BCB Polymer Coatings," Proc. MRS
Sympos., 265: p. 275, 1992.
68. T. M. Stokich, C. C. Fulks, M. T. Bemius, D. C. Burdeaux, P. E. Garrou, and R. H.
Heistand. "Planarization with Cyc10tene 3022 (BCB) Polymer Coatings," Mater.
Res. Soc. Symp. Proc., 308: p. 517, 1994.
69. T. Tessier. "A Comparison of Common MCM-D Dielectric Material Performance,"
Proceedings 6th SAMPE Electronics Conference, p. 347, 1992.
70. F. A. Sherrima, I. A. Saadat, S. Sekigahama, A. A. Abado, 1. O'Brien, and M.
Thomas. Manufacturing Studies of BCB as the Interlevel Dielectric Material for
Multilevel Interconnect MCM and VLSI Applications," Proc. ISHM, p. 596, 1992.
71. S. Bothra, M. Kellam, and P. E. Garrou. "BCB as an Interlevel Dielectric in a
Multilevel Metal System," J. Electron. Mater., 23: p. 819, 1994.
72. T. Shimoto, K. Matsui, M. Kimura, and K. Utsumi. "High Density Multilayer
Substrate Using BCB Dielectric," Proceedings International Microelectronics Con-
ference, p. 325, 1992.
73. T. G. Tessier and E. G. Myszka. "High Performance MCM-LD Substrate Approaches
for Cost Sensitive Packaging Applications," Proceed. Int. Con! MCM's, p. 200, 1993.
74. A. J. Strandjord, R. H. Heistand, J. N. Bremmer, P. E. Garrou, and T. G. Tessier. "A
Photosensitive BCB on Laminate Technology," Proc. ECTC, p. 374, 1994; IEEE
Trans. Components Hybrids Manu! Technol. (in press.)
75. P. Garrou, R. Heistand, M. Dibbs, T. Manial, T. Stokich, C. Mohler, P. Townsend,
G. Adema, M. Berry, and I. Turlik. "Rapid Thermal Curing of BCB Dielectric,"
IEEE Trans. Components Hybrids Manu! Technol., CHMT 16: p. 46, 1993.
76. T. C. Hodge, B. Landmann, S. A. Bidstrupp, and P. A. Kohl. "Rapid Thermal Curing
of Polymer Interlayer Dielectrics," Int. J. Microcircuits Electron. Packaging, 17: p.
10, 1994.
77. R. Heistand, R. DeVellis, T. Manial, A. Kennedy, T. Stokich, P. Garrou, T. Takahashi,
G. Adema, M. Berry, and I. Turlik. "Advances in MCM Fabrication with Benzocyc1o-
butene Dielectric," Proc. Int. Microelectronics Con!, p. 320, 1992; J. Microcircuits
Microelectron. Packaging, 15: p. 183, 1992.
78. G. Adema, L. Hwang, G. Rinne, and I. Turlik. "Passivation Schemes for Copper!
Polyimide Thin Film Interconnections Used in Multichip Modules," IEEE Trans.
Components Hybrids Manu! Technol. CHMT 16: p. 53, 1992.
".17 REFERENCES 11-617

79. R. H. Heistand, D. C. Frye, D. C. Burdeaux, J. N. Carr, and P. E. Garrou. "Economic


Evaluation of Deposited Dielectric MCM Manufacturing Costs," Proc. Int. Conf
MCM, p. 441, 1993.
80. L. Laursen and P. E. Garrou. "Consortium for Intelligent Large Area Processing,"
Proc. Int. MCM Conf, 1995.
81. T. M. Stokich, W. M. Lee, and R. A. Peters. "Real Time FT-IR Studies of the
Reaction Kinetics for the Polymerization of DVS-BCB Monomer", Proc. MRS, p.
103, 1991.
82. C. E. Mohler, A. J. Strandjord, D. J. Castillo, M. R. Stachowiak:, R. H. Heistand,
P. E. Garrou, and T. G. Tessier. "Micro ATR as a Probe of BCB Layers for MCM-
LD Applications," Proc. MRS Symp, p. 295, 1993.
83. R. W. Johnson, T. Phillips, W. Weidner, S. Hahn, D. Burdeaux, and P. Townsend.
"BCB Interlayer Dielectrics for Thin Film Multichip Modules," IEEE Trans. Compo-
nents Hybrids Manuf Technol., CHMT 13: p. 347, 1990.
84. B. Rogers, M. Berry, I. Turlik, P. E. Garrou, and D. Castillo. "Soft Mask for Via
Patterning in BCB," Int. J. Microcircuits Electron. Packaging, 17: p. 210, 1994.
85. M. Schier. "Reactive Ion Etching of BCB Using a Silicon Nitride Etch Mask," J.
Electrochem. Soc., Vol. 142, 1995, p. 3238.
86. T. Tessier and G. Chandler. "Compatibility of MCM-D Dielectrics with Scanning
Laser Ablation Via Techniques," Proc. ECTC, p. 763, 1992.
87. Strandjord, A., Garrou, P. E., Ida, Y., Cummings, S., Kisting, S., Rogers, B., "MCM-
D Fabrication with Photo BCB: Processing, Sio1derbumping, Systems Assembly and
Test", Proceed ISHM, Los Angeles, 1995, p. 402.
88. W. Radlik, K. Plehnert, R. H. Heistand, C. D. Castillo, and A. Achen. "MCM-D
Technology in a Communication Application," Proc. Int. MCM Conf, p. 402,1994.
89. H. Theide, et al. "High Reliability 4-Layer MCM-D Structure with BCB as Dielec-
tric," Proc. Int. MCM Con!, 1995.
90. J. Carlson. "486 Computer Module System," Proc. Int. MCM Conf, 1995.
91. J. K. Stille, Macromolecules, 14, 870, (1981).
92. N. H. Hendricks, "Development of Thermally Stable, Low Dielectric Films for
Aerospace Applications", Final Report, NASA Contract No. NASI-18832, (1989).
93. N. H. Hendricks, M. L. Marrorco, D. M. Stoakley and A. K. St. Clair. 'Thermally
Stable, Low Dielectric Po1yquinolines for Aerospace and Electronics Applications,"
SAMPE, Vo!. 4, Electronics Matls.-Our Future, Edited by R. F. Albred, R. J.
Martinez, and K. B. Wirchmann, SAMPE 1990. Society for the Advancement of
MatI. and Process Engineering.
94. M. I. Bessonov, M. M. Koton, V. V. Kudryavtsev, and L. A. Laius. Polyimides:
Thermally Stable Polymers, 2nd edn., Plenum, New York (1987), pp. 1-95.
95. Avatre1 Olefinic Dielectric Polymers, BF Goodrich Product Literature.
96. A. Saiki, S. Hirada, T. Okubo, K. Mukai, T. Kimura, J. Electro-chern. Soc: Solid
State Sci. Techno!., 124, 1619 (1977).
97. A. Saiki, K. Mukai, S. Harada, Y. Miyadera, Preprints ACS Org. Coat. Plastic Chern.,
43,459 (1980).
98. Y. Harada, F. Matsumoto, T. Nakakado, J. Electro-chern. Soc.: Solid State Sci.
Techno!., 130, 1, 133 (1983).
99. R. D. Rossi. "Polyirnides in Microelectronics Package," Plast. Electron. Packag.
Trends Technol., pp. 243-251, 1991.
100. R. Rubner, A. Hammerschmidt, R. Leuschner, and H. Ahne. "Photosensitive Organic
Dielectrics: Po1ybenzoxazole versus Po1yirnides," Proc. Int. Symp. on Polymer for
Microelectronics-Science and Technology-(PME'89)," pp. 789-810, 1990.
11-618 POLYMERS IN PACKAGING

101. H. Ahne, and R. Rubner. "A Simple Way to Producing Industrial Polyimide Patterns,"
Siemens Forsh. Entwickl.-Ber., 16(3): pp. 112-116, 1987.
102. M. T. Pottiger. "Second Generation Photosensitive Polyimide Systems," Solid State
Technol., 32 (12) pp. SI-S4, 1989.
103. R. E. Kerwin and M. R. Goldrick. ''Thermally Stable Photoresist Polymer," Polym.
Eng. Sci., 11(5): pp. 426-430, 1971.
104. R. Rubner. "Production of Highly Heat-Resistant Film Patterns from Photoreactive
Polymer Precursors. Part 1. General Principles," Siemens Forsh. Entwickl.-Ber., 5(2):
pp. 92-97, 1976.
105. R. Rubner, W. Bart, and G. Bald. "Production of Highly Heat-Resistant Film Patterns
from Photoreactive Polymer Precursors. Part 2. Polyimide Film Patterns," Siemens
Forsh. Entwickl.-Ber., 5(4): pp. 235-239, 1976.
106. R. Rubner, H. Ahne, E. Kuhn, and G. Kolodziej. "A Photopolymer-The Direct
Way to Polyimide Patterns," Photographic Sci. Eng., 23(5): pp. 303-309, 1979.
107. H. Ahne, H Eggers, W. Gross, N. Kokkotakis, and R. Rubner. "New Electronic
Application of Polyamic Acid Methacrylate Ester" Proc. Second Int. Con! on Poly-
imide, pp. 561-574, 1985.
108. H. Merrem, R. Klug, and H. Hartner. "New Developments in Photosensitive Poly-
imide," Polyimides Synthesis, Characterization, and Applications, pp. 919-931,
Plenum Press, New York, 1984.
109. M. T. Pottiger, D. L. Golf, and W. J. Lautenberger. "Photodefinable Polyimides: II.
The Characterization and Processing of Photosensitive Polyimide Systems," 38th
Electron. Components Conf., pp. 316-321, 1988.
110. o. Rhode, M. Riediker, and A. Schaffner. "Recent Advances in Photoimagable
Polyimides," SPIE, 539: pp. 175-180, 1885.
Ill. O. Rhode, M. Riediker, A. Schaffner, and 1. Bateman. "High Resolution, High
Photospeed Polyimide for Thick Film Applications," Solid State Techno/., (9) pp.
109-112, 1986.
112. A. Ikeda, N. Tsuruta, H. Ai, and T. Isoya. "High Performance Photosensitive Poly-
imide," Polyfile, 27(2): pp. 19-21, 1990.
113. S. Ogitani. "PIMEL: Photosensitive Polyimide Coatings for Electronics," Proc. Int.
Symp. on Polymer for Microelectronics (PME'89), pp. 158, 1989.
114. Y. Matsuoka, A. Ikeda, and H. Ai. "Process for Preparing Polyamides," Japanese
Patent Kokai, 59-193737, 1984.
115. Y. Matsuoka, A. Ikeda, and H. Ai. "Synthetic Methods of Photosensitive Polyimide
Precursor for Electronic Use." Polymer Preprints, Japan, 42(3): pp. 741, 1993.
116. L. Minnema and J. M. van der Zande. "Pattern Generation in Polyimide Coatings
and its Application in an Electrophoretic Image Display," Polymer Eng. Sci., 28(12):
pp. 815-822, 1988.
117. H. Ahne, E. Kuhn. R. Rubner, and E. Schmidt. "Method for the Preparation of Highly
Heat-Resistant Relief Structure and the Use Thereof," US Patent 4,311,784, 1982.
118. G. C. Davis. Photosensitive Polyimide Siloxane, pp. 259-269, American Chemical
Society, Washington, DC, 1984.
119. B. Loisel and M. J.M. Abadie. "Kinetics Studies of Photosensitive Polyimides by
Photocalorimetry (DCP)," Polyimide and Other High-Temperature Polymers., (Euro-
pean Technical Symposium on Polyimide and High-Temperature Polymers (2nd»
pp. 471-492, 1991.
120. T. Sakuma, S. Ogitani, and A., Ikeda. "Study on Thermal Curing in Ester Type
Photosensitive Polyimide," J. Photopolymer Sci. Technol., 8(2): pp. 277-280, 1995
121. S. Numata and N. Kinjyo. "Chemical Structures and Properties of Low Thermal
Expansion Coefficient Polyimides," Polymer Eng. Sci., 28(14): 906-911,1988.
11.17 REFERENCES 11-619

122. S. Numata, S. Ohhara, K. Fujisawa, 1. Imaizumi, and N. Kinjo. ''Thermal Expansion


Behavior of Various Aromatic Polyirnides," J. Appi. Polymer Sci., 31: pp. 101-
110, 1986.
123. Y. Matsuoka, K Yokota, S. Ogitani, A. Ikeda, H. Takahashi, and H. Ai. "Ester-
Type Photosensitive Polyirnide Precursor with Low Thermal Expansion Coefficient,"
Polymer Eng. Sci., 32(21): pp. 1619-1622, 1992.
124. A. E. Nader, K. Imai, J. D. Craig, C. N. Lazaridis, D. O. Murray, M. T. Pottiger,
S. A. Dombchik, and W. 1. Lautenberger. "Synthesis and Characterization of a Low
Stress Photosensitive Polyimide," 9th Int. Tech. Conf. on Photopolymers, pp. 333-
342, 1991.
125. T. Omote, T. Yamaoka, and K. Koseki. "Preparation and Properties of Soluble and
Colorless Fluorine-Containing Photoreactive Polyimide Precursor" J. Appl. Polym.
Sci., 38(3): pp. 389-402, 1989.
126. Y. Matsuoka, Y. Kataoka, Y. Tanizaki, and A. Ikeda. "The Design ofI-Line Photoima-
gable Polyimide Precursor," Polymer Preprints, Japan, 42(7): pp. 2694-2696, 1993.
127. D. R. McKean, G. M. Wallraff, W. Volksen, N.P. Hacker, M I. Sanchez, and 1. W.
Labadie. "Base-Catalyzed Photosensitive Polyimide," Polymer Mater. Sci. Eng., 66:
pp. 237-238, 1992.
128. H. Hiramoto and M. Eguchi. U. S. Patent 4.24S,743, 19.
129. Y. Yoda and H Hiramoto, J. Macromol. Sci. Chem., A21: p. 1641, 1984.
130. H. Hiramoto. Mater. Res. Soc. Symp. Proc., 167: 87, 1990.
131. M. Tomikawa, M. Asano, G. Ohbayashi, H. Hiramoto, Y. Morishima, and M.
Kamschi, J. Photopolym. Sci. Technol., 5: p. 343, 1993.
132. G. Ohbayashi, Umemoto, H. Hiramoto. U. S. Patent 4,547,455, 19.
133. M. Asano, M. Eguchi, K. Kusano, and K. Niwa, Polym. Adv. Tech., 4(4): p. 261,1993.
134. M. KoJima, H. Sekine, H. Suzuki, H. Satou, and D. Makion, Proc. 39th Electronic
Components Conf, p. 920, 1989.
135. J. Pfeifer and O. Rohde. "Polyirnides: Synthesis, Characterization, and Applications,"
Proc. 2nd Int. CosXfi Polyirnides, p. 130, 1985; Recent Advances in Polyimide
Science and Technologx, ed. by W. D. Weber and M. R. Gupta, p. 336, Society of
Plastics Engineers, Brookfield, CT, 1987.
136. O. Rhode, P. Smolka, P. A. Falcigno, and J. Pfeifer. Polymer Eng. Sci. 32 (21): p.
1623, 1992.
137. R. Mo, K. Schlicht, T. Maw, M. Masola, and R. Hopla. in Proceedings of Advances
in Polyimide Science and Technology, ed. by C. Feger, M. Khojasteh, and M. Htoo,
p. 559, Technomic Publishing Company, Lancaster, PA, 1993.
138. M. Ree, K. Chen, and G. Czomyj. Polymer Eng. Sci., 32 (14): p. 924, 1992.
139. A. Lin, V. Sastri, G. Tesoro, and A. Reiser. Macromolecules, 21: p. 1165, 1988.
140. J. Scaiano, A. Becknell, and R. Small, J. Photochem. Photobiol., A: Chem., 44: p.
99, 1988.
141. J. Scaiano, 1. Netto-Ferreira, A. Becknell, and R. Small. Polymer Eng. Sci., 29(14):
p. 942, 1989.
142. Y. Shindo, T. Sujimura, K. Horie, and I. Mita. Eur. Polym. J., 26(6): p. 683, 1990.
143. H. Higuchi, T. Yamashita, K. Horie and I. Mita. Chem. Mater. 3: p. 188, 1991.
144. T. Yamashita, K. Horie, and I. Mita. in Polymers for Microelectronics-Science and
Technology, p. 837, ed. by Y. Tabata, I. Mita, S. Nonogaki, K. Horie, S. Tagawa,
Kodansha, 1990.
145. T. Yamashita, T. Kudo, and K. Horie, in Polymers for Advanced Technologies, ed.
by M. Levin, vol. 4, p. 244, John Wiley and Sons, New York, 1993.
146. K. Horie. J. Photopolymer Sci. Technol., 5(2): p. 315, 1992.
147. Q. Jin, T. Yamashita, and K. Horie, J. Polym. Sci. Part A: Polym. Chern. 32,
503,1994.
11-620 POLYMERS IN PACKAGING

148. O. Rohde. 3rd Annual Int. Con! Crosslinked Polymers, p. 197, 1989.
149. H. Lee and Y. Lee. J. Appl. Polym. Sci., 4: p. 2087, 1990.
150. T. Maw, M. Masola, and R. Hopla. Polymer Mater. Sci. Eng., 66: p. 247, 1992.
151. T. Maw and R. Hopla. Mater. Res. Soc. Symp. Proc., 203: p. 71, 1991.
152. W. Chiang and W. Mei. J. Polymer Sci. A: Polymer Chem. 31: p. 1195, 1993.
153. W. Chiang and W. Mei. J. Polymer Sci. A: Polymer Chem. 50: p. 2191, 1993.
154. J. Sassmannshausen. R. Schultz and E. Bartmann, US Patent, 5,1004,768, April
14, 1992.
155. R. H. Hayase, N. Kihara, N. Oyasato, S. Matake, and M. ~ba, SPIE 1446, Vol 8,
pp. 439-445, 1991.
156. E. Perfecto, C. Osborn, and D. Berger, ICEMM Proceeding, 1993.
157. M. Asano and H. Hiramoto, in "Photosensitive Polyimides, Fundamental and Applica-
tions", Edited by K. Horie and T. Yamashita, Chapter 5, Technomic Publishing
Company, Lancaster, PA, 1995.
158. J. Moore and A. Dasheff, Chemistry of Materials, 1, p. 163, 1989.
159. R. Hayase, N. Kihara, N. Oyasato, S. Matake and M. Oba, Polymeric Materials
Science and Engineering, Proceedings of the ACS Division of PMSE, Vol. 66, pp.
243-244, 1992.
160. S. Kubota, T. Moriwaki, T. Ando, and A. Fukami, J. Macromol. Sci, Chern., A24,
p. 1497, 1987.
161. K. Takano, Y. Mikogarni, Y. Nakano, R. Hayase, and S. Hayase, J. Applied Polymer
Sci., 46, p. 1137, 1992.
162. S. Hayase, Y. Mikogarni, K. Tanako, Y. Nakano, and R. Hayase, Polymer Adv.
Tech., 4, p. 308, 1993.
163. T. Omote and T. Yamaoka, Polymer Eng. Sci., 32, p. 1632, 1992.
164. T. Banba, E. Takeuchi, A. Tokoh, and T. Takeda, Proceedings, IEEE 41st Electronic
Components and Technology Conference, p. 564, 1991.
165. A. E. Nader, C. N. Lazeridis, D. K. Flattery, and W. J. Lautenberg, MCM Proceedings,
Denver, p. 410, 1992.
166. E. Takeuchi, T. Takeda, and T. Hirano Proceedings, IEEE 40th Electronic Compo-
nents and Technology Conference, p. 818, 1990.
167. M. Kojima, H. Sekine, H. Suzuki, H. Satou, D. Makino, F. Kataoka, J. Tanaka and
F. Shoji, Proceedings, IEEE 39th Electronic Components and Technology Confer-
ence, p. 920, 1989.
168. H. Sashida, T. Hirano, and A. Tokoh, Proceedings, IEEE 39th Electronic Components
and Technology Conference, p. 167, 1989.
169. E. W. Rutter Jr., E. S. Moyer, R. H. Harris, D. C. Frye, V. L. St. Jore, and F. L.
Oaks, "A Photodefinable Benzocyciobutene Resin for Thin Film Microelectronic
Applications", Proceedings of 1st International Conference on Multichip Modules,
Denver, p. 394, 1992.
170. E. S. Moyer, E. W. Rutter Jr., M. T. Bernius, P. H. Townsend, R. F. Harris, H.
Pranjoto, and D. D. Denton, "Photodefinable Benzocyciobutene Formulations for
Thin Film Microelectronic Applications: Part II", Proceedings IEPS, pp. 37-50,1993.
171. E. S. Moyer, G. S. Becker, E. W. Rutter, Jr., M. Radler, J. N. Bremmer, M. T.
Bernius, D. Castillo, A. J.G. Strandjord, R. Heistand, P. Foster, and R. F. Harris,
"Photodefinable Benzocyciobutene Formulations for Thin Film Microelectronic Ap-
plications. III. 1 To 20 Micron Patterned Films", MRS Symposium Proceedings,
Boston, Vol. 323, pp. 267-276, 1994.
172. Strandjord, A., Garrou, P. E., Ida, Y., Cummings, S., Kisting, S., Rogers, B., "MCM-
D Fabrication with Photo BCB: Processing, Siolderbumping, Systems Assembly and
Test", Proceed ISHM, Los Angeles, p. 402, 1995.
11.17 REFERENCES 11-621

173. T. Shimoto, K. Matsui, and K. Utsumi, "Cu I Photosensitive-BCB Thin Film Multi-
layer Technology for High-Performance Multichip Module", Proceedings of the
International Conference on Multichip Modules, Denver, p. 115, 1994.
174. W. Radik, K. Plehnert, M. Zellner, A. Achen, R. H. Heistand, D. Castillo, & R.
Urscheler, "MCM-D Technology for a Communication Application", Proceedings
of the International Conference on Multichip Modules, Denver, p. 402, 1994.
175. Tessier, T., Myszka, E., "Approaches to Cost Reducing MCM-D Substrates", Proceed.
ECTC, p. 570, 1993.
176. M. Skinner, P. E. Garrou, D. Castillo, S. Cummings, K. Liu, D. Chazen, R. Rein-
schmidt, S. Westbrook, C. Ho, B. Rogers, "Twinstar-Dual Pentium Processor
Module", Proceed. Int. Conference on MCMs, Denver, p. 75, 1996.
177. P. H. Townsend, D. Schmidt, T. M. Stokich, S. Kisting, D. C. Burdeaux, D. Frye,
M. Bernius, M. Lanka, and K. Berry, "Adhesion ofCYCLOTENE (BCB) Coatings On
Silicon Substrates", MRS Symposium Proceedings, Boston,., Vol. 323, p. 365, 1993.
178. P. Garrou, "Polymer Dielectrics for Multichip Module Packaging", Proceedings of
IEEE, Vol. 80, p. 1942, 1992.
179. L. Laursen, P. E. Garrou, "Consortium for Intelligent Large Area Processing", Pro-
ceed. Int. MCM Conf., Denver, p. 112, 1995.
180. A. Strandjord, P. E. Garrou, R. Heistand and T. G. Tessier, "MCM-DIL: Large Area
Processing Using PS-BCB", IEEE Trans. CPMT, Vol 18(2), 1995.
181. E. D. Perfecto, D. G. Berger, C. T. Osborn, G. White, "Engineering PSPl's for MCM-
D Applications", Int. J. Microcircuits & Elect. Pkging.", Vol. 16, p. 319, 1993.
182. E. D. Perfecto, C. Osborn, D. Berger, "Factors That Influence Photosensitive PI
Lithographic Performance", Proceed Int Conf MCM's, Denver, p. 40, 1993.
183. D. Burdeaux, P. Townsend, Carr, P. Garrou, "Benzocyclobutene Dielectrics for the
Fabrication of High Density, Thin Film Multichip Modules", J. Electronic Materials,
Vol. 19, p. 1357, 1990.
184. P. E. Garrou, R. H. Heistand, M. Dibbs, T. A. Manial, C. Mohler, T. Stokich, P. H.
Townsend, G. M. Adema, M. J. Berry, and I. Turlik, "Rapid Thermal Curing ofBCB
Dielectric", Proceedings ECTC, San Diego, p. 770, 1992.
185. T. G. Tessier & G. Chandler, "Compatibility of Common MCM-D Dielectrics with
Scanning Laser Ablation Via Generation Processes" IEEE Trans. on CHMT, Vol.
16, p. 39, 1993.
186. T. Shimoto, M. Matsui, M. Kimura, & K. Utsumi, "High Density Multilayer Sub-
strates Using Benzocyclobutene Dielectric", Proceedings 7th IMC, Yokohama, p.
325, 1992.
187. A. V. Shah, E. Sweetman, and C. Hoppes, Proceedings of the National Electronic
Packaging and Production Conference, v 2. Publ by Cahner Exposition Group, Des
Plaines, Illinois, pp. 850-862, 1991.
188. E. Sweetman, Characteristics and Performance of PH-92: AT&T's Triazine-based
Dielectric for POLYHIC MCMs, Microcircuits and Electronic Packaging, Vol. 15,
No. 4,pp. 195-203, 1992.
189. H. Ito, and C. Willson, ACS Symposium Series 242. Publ by ACS, Washington,
DC, USA, pp. 11-23.
190. K. Stewart, M. Hatzakis, and J. Shaw, Technical Papers, Regional Technical Confer-
ence-Society of Plastics Engineers. Publ by Soc of Plastics Engineers, Brookfield
Center, CT, pp. 205-213, 1988.
191. Ciba Geigy Product Literature, LMB 7081.
192. H. Ahne, H. Kmger, E. Pammer and R. Rubner. "Polyimide Patterns made Directly
from Photopolymers" Polvimides Synthesis, Characterization, and Applications. Ple-
num Press, New York, pp. 905-918, 1984.
11-622 POLYMERS IN PACKAGING

193. M. M. Khan, T. S. Tarter and H. Fatemi. "Stress Relief in Plastic Encapsulated,


Integrated Circuit Devices by Die Coating with Photodefinable Polyimide" Electron.
Components Conf., 38: 425-431, 1988.
194. N. Naito, N. Kakuda, T., Wada, IERE Electron Device Letters, EDL-6, 589, 1985.
195. H. Ahne, R. Leuschner and R. Rubner. "Recent Advances in Photosensitive Poly-
imides," Polymer for Advanced Technologies, 4, pp. 217-233, 1992.
196. S. Ray, D. Berger, G. Czornyl, A. Kumar and R. Tummala, "Dual-Level Metal (DLM)
Method for Fabricating Thin Film Wiring Structure," Proc. Electron. Components
Technol. Conf. 43, pp. 538-543, 1993.
197. K. Chakravorty and M. Tanielian, Appl. Phys. Lett. 60 (14), 1670, 1992.
198. K. Chakravorty, in Polymers for Advanced Technologies, 4(4), p. 251, M. Levin
ed., John Wiley, 1993.
199. C. Chien, K. Chakravorty, SPIE 1323, 338, 1991.
200. K. Chakravorty, Appl. Optics, 32 (13), 2331, 1993.
201. K. K. Chakravorty, C. P. Chien, 1. M. Cech, L. B. Branson, J. M. Atencio, T. M.
White, L. S. Lathrop, B. W. Aker, M. H. Tanielian, and P. L. Young. "High Density
Interconnection using Photosensitive Polyimide and Electroplated Copper Conductor
Line," Electron. Components Conf, pp. 135-142, 1989.
202. K. K. Chakravorty, 1. M. Cech, C. P. Chien, L. S. Lathrop, M. H. Tanielian, and
P. L. Young. "Photosensitive Polyimide as a Dielectric in High density Thin Film
Copper-Polyimide Interconnect Structures," 1. Electrochem. Soc., 137(3), pp. 961-
966, 1990.
203. G. L. Takahashi and E. S. Kolesar. "Silicon-Hybrid Wafer-Scale Integration
Achieved with Multilevel Aluminum Interconnects," Proc. IEEE National Aerospace
Electron. Con!, pp. 17-23, 1989.
204. K. Chakravorty, C. Chien, 1. Cech, and M. Tanielian. 3rd Int. SAMPE Electron.
Conf, p. 1213, 1989.
205. J. Cech, A. Bumeff, and L. Knapp, Polymer Eng. Sci., 32(21), p. 1646, 1992.
206. D. Smith, D. Scheider, R. Fu, E. Chan, G. LaRue, T. Williams, and M. Tanielian,
DoD Fiber-Optics Conference, 1990
207. E. Myszka, G. Demet, H. Fuerhaupter, and T. Tessor. "The Development of a
Multilayer Thin Film CopperlPolyimide Process for MCM-D Substrates," Proc. Int.
Conf on Multichip Modules," 1992.
208. T. Ohsaki, T. Yasuda, S. Yamaguchi, T. Kon, Proc. IEEE Int. Electron. Manuf.
Techno!. Symp., 178, 1987.
209. S. Sasaki, T. Kon, T. Ohsaki, Proc 39th Electronic Components Conf., 629, 1989.
210. T. Watari, H. Murano, IEEE Trans. Components, Hybrids, Manuf. Technol., CHMT-
8(4),462, 1985.
211. K. Kimbara. A. Dohya. T. Watari. Mat. Res. Soc. Symp. Proc., 167,33,1990.
212. Inoue, K. Seino, K. Tamura, S. Hasegawa, K. Kimbara, J. Photopolym. Sci. Techno!.,
5, 385, 1992.
213. H. Takasgo, M. Takada, K. Adachi, A. Endo, K. Yamada, T. Makita, E. Gofuku,
Y. Ohnishi, Proc. 36th Electronic Components Conf., 481, 1986.
214. M. Takada, T. Makita, E. Gofdku. K. Miyake, A. Endo, K. Adachi, Y. Morihiro,
H. Takasago, Proc. Int. Symp. Microelectron, 540, 1988.
215. H. Takasago, K. Adachi. M. Takada, J. Electron. Materials, 8(2), 319, 1989.
216. D. W. Hewak and H. Jerominek. "Channel Optical waveguides in Polyimides for
Optical Interconnection by Laser Direct Writing and Contact Printing," SPIE Vol.
1213, Photopolymer Device Physics, Chemistry, and Applications: pp. 86-99, 1990.
217. M. G. Allen. "Polyimide-Based Processes for the Fabrication of Thick Electroplated
Microstructures," The 7th International Conference on Solid-State Sensors and Actua-
tors, pp. 60-65, 1993.
11.17 REFERENCES 11-623

218. M. Uekita, H. Awaji, M. Murata, and S. Mizunuma. "Application of Po1yimide


Langmuir-Blodgett Films to Deep UV Resist," Thin Solid Films, 180: pp. 271-
276,1989.
219. Z. H. Lu, J. Y. Fang, L. Wang, Z. L. Chen, S. J. Xiao and Y. Wei. "Ultrathin
Conductive Films Prepared by Pyrolysis of Photosensitive Polyimide Langmuir-
Blodgett Films," Solid State Comm., 82(9): pp. 711-713, 1992.
12
THIN-FILM PACKAGING

RAO TUMMALA-Georgia Tech


WEIPING LI-Georgia Tech
TED TESSIER-Motorola
TOM WASSICK-IBM

12.1 INTRODUCTION

Thin film refers to a coating layer of thickness typically in the range


of from a few (2-3) atomic layers to a few (1-5) microns. However,
film thicknesses up to 20-50 /lm are frequently regarded as thin-film in
electronic packaging. Thin-film packaging is the technology for conductors
and dielectrics deposition and patterning in package fabrication, which
resembles the technology used for integrated-circuit (IC) chip fabrication.
Thin-film packaging is distinguished from thick-film packaging, mainly
ceramic and printed wiring board (PWB) packaging, in two aspects: (a)
the typical dimensions of conductors and dielectrics are about 2-25 11m
versus 100 /lm and above in thick-film packaging; (b) the typical methods
of thin-film deposition include sputtering, evaporation, chemical vapor
12.1 INTRODUCTION 11-625

deposition (CVD), more recently electro/electroless plating and polymeric


solution coating, and other similar methods, which are all typical sequential
processes, whereas in thick-film packaging parallel processes are more
common, such as lamination and cofiring of ceramic greensheets. Never-
theless, the current trends in electronic packaging are the enhancement of
thick-film technology to fabricate finer feature sizes which are traditionally
considered only achievable by thin-film technology and the adoption of
some of the parallel thick-film processes for thin-film fabrication. Due to
these developments, the traditional division between thin- and thick-film
packaging technologies is diminishing and the definition of thin-film pack-
aging is becoming less obvious. Still, the mainstream of thin-film packag-
ing can be well identified which allows a fairly extensive compilation on
this technology. In this chapter, the need for thin-film packaging, electrical
performance considerations, and typical structures in thin-film packages
are first elucidated. Subsequently, the materials, processes, and cost analy-
ses of thin-film packaging are discussed in detail. Repair is especially
important in thin-film packaging and is dealt with in the context of cost and
yield. Reliability of packages has been discussed in Chapter 5, "Package
Reliability," and reliability issues which are unique to thin-film packages
are identified here. Some major commercial applications of thin-film
packaging are reviewed, followed by a summary of the emerging techno-
logies. Integration of various passive components is becoming imperative
for thin-film packages to enhance electrical performance and packaging
efficiency, which is also included in this chapter along with some applica-
tions and future predictions. Finally, the current development and future
directions of thin-film packaging technology are summarized and pro-
jected.
Packaging must deal with the inexorable trend of semiconductor
technology toward higher levels of integration, as pointed out in Chapter
7, "Microelectronics Packaging-An Overview" a trend that is driven by
an irresistible economic force and the augmentation of human capabilities
by powerful computational facilities [1], and will continue to the limits
of physical laws. The level of integration is measured by the number of
logic gates on an integrated circuit (IC) chip. High levels of integration
have been achieved by miniaturization of the structures on chips. Miniatur-
ization has also conferred high speed on microelectronics; essentially,
miniaturization reduces gate capacitance, which can, therefore, be charged
more quickly, and gate-to-gate interconnection delay. In addition, speed
also depends on device technologies as is illustrated in Figure 7-7 of
Chapter 7, along with the other major device attributes. Translating the
high speeds available on chips to fast, multichip systems requires minimi-
zation of delays attributable to the package. This means placing chips as
closely together as possible to reduce chip-to-chip travel times. Each of
these chips are expected to have up to 5000 I/O connections as discussed
11-626 THIN-FILM PACKAGING

in Table 7-2 of Chapter 7, and all packaging functions must be contained


in the smallest possible area. All these requirements as illustrated in Figure
12-1 and previously in Chapter 7, Figure 7-62 [2] point to the need for
thin-film packaging. It is seen in Figure 12-1 that, with the chip I/O counts
going up to above 600, the pad width and spacing become less that 25
/lm and thin-film packaging, characterized by the application of photolitho-
graphic technology, is necessary. In Figure 7-62, Chapter 7, the packaging
efficiency and wiring dimensions as a function of various thin- and thick-
film packages are plotted, illustrating the advantage of thin-film package
over other packages with maximized packaging efficiency and minimized
size of portable and consumer electronics in the future. For this reason,
thin-film packaging is generally considered to be the ultimate in packaging
and is usually assumed as multichip packaging with minimized intercon-
nection delay between IC chips. Figure 12-2 reaffirms the need for thin-
film packaging. With wiring density going up to 2000 cmlcm2 by the
end of next decade, thin-film technology would probably be the only
choice left [3]. In general, the longest interconnections between devices
within a package dictate the interconnection delay and form the critical
paths. The critical paths restrain the cycle time of a processor such as
for mainframes or supercomputers which, in turn, primarily determines

Wldth=Space=4(L+d)
----rn-
L=Length of Chip
E d=Chip to Pad Distance
.3:
CII
c
I0100 ---------------~---~~-,~~--.----------
~
C

I
II

50

~ 25

10 20 50 100 200 600 1000

Pin Count, n

Figure 12-1. Requirements for Thin Film of 110 Pads. (After Ref. 2.)
12.1 INTRODUCTION 11-627

2,000

1,800
Required
1,IICICI
Silicon
1,400
Efficiency

Ii! 1,210 f. chip area \


~
\board area)
t 1,000
.! 'Assume 9
IIlClCl "cost-performance"
J chips based on
6CIO technology
introduced in the
400 year shown.

210

1915 ,. 2101
Vea,
2004 2007 2010

Figure 12-2. Evolution of Wiring Density on Substrate

the processing speed and throughput. A good example of performance


enhancement with shorter critical paths is the use of the multichip modules
(MCM) by IBM in its mainframes since 1980 to reduce the cycle time
from about 60 ns to less than 8 ns as illustrated in Figure 12-3. It is noted,
however, that ceramic multichip module (MCM-C) packaging rather than
thin-film multichip module (MCM-D) packaging was initially employed.
Other companies such as Hitachi, Fujitsu, and NEC have basically done
the same using somewhat different multichip modules, some with ceramic
or glass-ceramic (Fujitsu) and some with thin film on ceramic or glass-
ceramic substrates (NEC). In the past, high-power bipolar devices have
been used in all of these ultrahigh-performance systems. However, bipolar
devices have continually been and will totally be replaced by CMOS
(complementary metal-oxide-semiconductor) devices as the perfor-
mance of CMOS approaches that of bipolar, as indicated in Figure 12-4,
yet at much lower power and cost, as indicated in Figure 7-7, Chapter 7.
The limitations on space make packaging of large computers a diffi-
cult technology with many facets. A system contains hundreds, and some-
times thousands, of chips. Packaging is also a subject about which it is
difficult to generalize, as there is no unanimity among system designers
as to the best approach to packaging large, fast machines. There is a much
greater variability among the packaging technologies to different systems
than among their chip technologies.
The package communicates electrically with the chip by means of
IIOs that connect lands on a chip to package wiring. Some of the IIOs
are used to supply electrical power to the chip and others are used to
11-628 THIN-FILM PACKAGING

60
3033

50

40
,-.,

5
Q)

§ 30
E Total Packaging
.g ~
~ 20

10
ES 9000 (8 ns)
E,=~.4
E,=5.0
AgoPd l TF Cu .---- E,=?.4 .l Glass-Ceramic/Cu
(1 Chip) (54-133khi s Ai 20 3+Mo : 121 Chi s
1970 1975 1980 1985 1990

Year

Figure 12-3. Performance of IBM Mainframes

1000r---------.----------r----------,

t~
Eel
MAINFRAME
. ...
Hitachi
.........
••• ,
•

u
CQI
~
C7'
• •••
100~----~__~------~~~~------~

•e
••
.. . ee
••
2!
LL. e
ee
.,.j/, eee CMOS
U ee
o • MAINFRAMES
U 10~--~--~~--------+_--------~

+ 1990 1995 2000


----i.~ Year
Figure 12-4. Evolution of Bipolar and CMOS- Based Mainframes and
RISC Systems
12.1 INTRODUCTION 11-629

transmit signals. The number of gates on a chip does not directly interact
with the package. The number of 1I0s on a chip is more relevant to the
packaging of the chip than the number of gates it contains.
The gate arrays in large computers are part of systems containing
a great many chips and hundreds of thousands or millions of logic gates.
The results of logic operations performed on a chip are often needed on
another chip in a distant part of the system, and many 1I0s are necessary
to meet the demand for rapid transmission of information throughout the
computer. The need for many 1I0s in logic chips that are part of a large
system was recognized long ago [4] and is described by a relation known
as Rent's rule, as pointed out in Chapter 7, "Microelectronics Packaging-
An Overview". In fact, the rule is not only applicable to chips, but also
to multi chip modules, boards, or any partition of a large computing system.
It has the form

P=BN [12-1]

Here N is the number of logic gates on a chip or other partition and


P is the number of signal connections that must be made to it. Band s
are constants which depend on the specific IC type and system design.
The general form of this rule is supported by a large number of empirical
studies [5-7], as pointed out in other chapters.
It is certainly possible to fabricate a chip with fewer connections
than the number given by Equation 12-1, but the basis of Equation 12-1
is that it is then not possible to utilize most of the circuits on the chip as
a part of the system. Both II0s and circuits contribute to the cost of a
system, and Equation 12-1 is a compromise aimed at obtaining a certain
functional capability at minimal cost. Hardware designers differ as to the
optimal way to make the compromises involved. Thus, it is not surprising
that different authors disagree as to the best values for the parameters in
Equation 12-1. The exponent s is variously given in the range 112 to
2/3 and the factor B is between 2 and 4.
It is noteworthy that the number of 1I0s grows more slowly than
the number of gates on the chip. It appears, as might reasonably be
expected, that signal destinations are found on the chip of origin more
frequently as the level of integration increases. This fact has also been
explicitly noted [8]: the fraction of logic steps in which the result is
transmitted to a different chip decreases with increasing level of integra-
tion. The reduction of the fraction of off-chip signal paths provides a
motivation, in addition to cost reduction, for increasing integration in
large machines: the off-chip paths are longer and cause longer delays
than paths on a chip; longer series of logic operations on a single chip
improves performance. Indeed, this fact is responsible for a significant
part of the improvement in computer performance in the last decade.
11-630 THIN-FILM PACKAGING

Interconnections have been moved from packages to chips, where they


are shorter, less expensive, and more reliable. Paths that pass from one
chip to another at low integration can be completed on a chip at high
integration [8,9].
One of the reasons that, in spite of the economic force and the
performance advantage pressing toward high levels of integration, the
number of gates on chips in large machines lags far behind the maximum
capability achieved in memory and microprocessor chips is the large
number of 1I0s. Higher levels of integration are achieved by miniaturiza-
tion of features on the chip; the size of the chip are also increasing slowly.
The increased number of connections to the chip must be accommodated
without a corresponding increase in area allotted to chip mounting, and
constant miniaturization of the connections is needed. At present, it is diffi-
cult for technology to provide as many as 400 1I0s on a chip, a number that
can support less than 5000 logic gates according to Rent's rule. However,
the trends in semiconductor technology indicate the need for as many as
1000 1I0s per chip within the next decade. This is only achievable with thin-
film packaging technology, as illustrated in Figure 12-1.
The problem of connections between one electronic package level
and the next does not end at the meeting of the chip and the module. The
multichip modules must be mounted on and connected to large boards.
The boards contain the wiring that connects the modules to one another.
In accord with Rent's rule, very large numbers of connectors are needed on
the module. For example, the original IBM Thermal Conduction Module
(TCM) had 1800 pins to connect to the next higher level of package [10].
Ofthese, 1200 are allotted to logic signals and 600 to the delivery of power.
Another somewhat similarly packaged computer uses 2177 connections to
the multichip module [11]. These connections are constructed with a quite
different technology than the connections to the chip, as they must be
easily separable for purposes of assembly and testing, and for replacement
in the field to allow engineering changes (EC) and repairs. They are
much larger in size than the connections between the chip and the module.
The area per connection in contemporary large computers is about 3 x
1O-3cm2 on chips and 0.06 cm2 for multichip module to board connectors.

12.1.1 Thin-Film Package Wiring

The large number of connections to the chips and to higher package


levels must be joined by a dense network of wires. The IBM Thermal
Conduction Module, a ceramic substrate on which 100 chips are mounted
in an area of less than 100 cm2, can contain 288 m of wire. Therefore,
each square centimeter of the substrate contains well over 100 cm of wire.
The wiring has a pseudorandom character; wire channels are provided
and wires placed in them to interconnect the chips to implement the
12.1 INTRODUCTION 11-631

logical structure of the computer. As the levels of integration increase,


more and more logic devices are placed in such multichip modules and
increasing wiring densities will be required in the package to support
them. A similar trend is evident on chips; chips now contain of the order
of 1000 cm of wire per cm2 of chip area. The high wiring densities on
packages are achieved by fabricating the wiring in a few thin-film layers.
The trend to higher wiring density can be expected to continue. In
fact, greater wiring density is associated with increasing levels of integra-
tion through Equation 12-1. This can be illustrated by a simple model.
Let M chips be placed on a multichip module substrate. Then the total
number of signal IIOs on the module is MBN'. If these IIOs are joined
in pairs, there are (l/2)MBN' wires. The wires have an average length
that is some fraction f of the linear dimension of the module. Measured
in units of the chip-to-chip distance (the chip "pitch"), the side of the
module is M"2, and the average wire length can be written as fM"2.
Multiplying these terms together, the total amount of wire on a module
(in chip pitches) is

[12-2]

The amount of wire that must be supplied under each chip is

[12-3]

The provision for the large amount of interconnecting wire is one


of the factors that limits the space that must be allotted to each chip. Let
wires be placed in K layers. Also assume that wire channels have a width
W. The channel width W must include space for vias that allow one layer
to be connected to another, so that W is considerably greater than the
actual width of the wire itself. Furthermore, because of the pseudorandom
character of the wiring, not all of the channel can be used; the utilization
of the available channel space is likely to be only 50% [10]. Additional
notation to describe these influences will not be introduced; they are
regarded as included implicitly in Wand in the average wire length.
If each chip occupies an area A on the module, then the length of
wire needed to traverse one chip site is the chip pitch, A 1/2, and the area
that it occupies is WA 112. The area of all of the wiring needed under a
chip, now found by using Equation 12-3, must equal the total area available
for wiring, KA:

[12-4]
11-632 THIN-FILM PACKAGING

Solving Equation 12-4 for A yields

[12-5]

The area per chip required for the wiring is inversely proportional to
the square of the number of layers. For example, using numbers somewhat
representative of the IBM TCM [10], if f = 112, K = 16, M = 100,
W = 0.05 cm, and N = 500, and assuming that B = 2.5 and s = 0.6, A
turns out to be 0.7 cm2• Of course, A must also be greater than the area
of the chip plus additional space needed for assembly.
The need for more layers of wiring is also seen by considering the
resistance of the wire. Let Po be the sheet resistance of the wire. (Again,
a factor to account for the fact that the wire is narrower than the channel
appears here and is assumed to be absorbed in Po-) Then, because the
length of the average wire is JMI12A 112, its resistance is

[12-6]

Taking A from Equation 12-5 gives

R =j2MBN Po/2K [12-7]

The resistance is inversely proportional to K. Note also that increasing


W does not directly impact R-W must be determined by limitations of
the fabrication process and by reduction of the distance between chips,
AII2 (Eq.l2-5).
Equations such as 12-5 and 12-7 are often more useful when solved
for K and used to estimate the number of wire layers needed from other
design criteria, for example, the acceptable resistance and module area
per chip_ Even more layers are needed than suggested by the above analysis
of signal wiring; conductors to distribute power and ground planes to
control impedance must be inc1uded_ The numbers quoted above for the
Thermal Conduction Module from IBM are typical of "thick film," in
which the dimensions of conductors are defined by mechanical masking
of the deposition process that is referred to as paste screening in Chapter
9, "Ceramic Packaging."
Increasing integration places ever greater demands on the number
of layers needed in thick-film technologies_ Although multilayer ceramic
(MLC) packaging has managed to cope with a remarkable number of
layers, there are inherent disadvantages. Vias that pass through many
layers are long and have high inductance. Switching currents through the
inductance causes voltage transients that constitute undesirable noise (dI
12.1 INTRODUCTION 11-633

noise); the higher the inductance, the greater the noise, as discussed in
Chapter 3, "Package Electrical Design."
In addition, a certain degree of compatibility between the spacing
of connectors on a chip and the spacing of wires is required. If the spacing
of wires is larger than the spacing between pins on a chip, it becomes
difficult to fit the wires into the pin pattern. The inevitability of an increas-
ing density of pins has already been mentioned and will increase the
difficulty of contacting chips with thick-film techniques.
Thus, for several reasons, the ability of thick-film methods to package
high-speed, highly integrated chips is threatened. These limitations of
thick-film technology have led to the development of ''thin-film'' pro-
cesses for packages, in which lines made of high-conductivity metals are
defined by lithographic methods akin to those used in IC chip fabrications.
Thin-film processes promise to permit the use of much narrower lines on
packaging substrates. Tables 12-1 and 12-2 illustrate the differences and
the need for thin-film technologies in terms of the above wiring model
with representative numbers. Table 12-1 compares the relevant parameters
of thick- and thin-film methods. The principal difference in the context
of the model is the narrower thin-film channel width. According to Equa-
tion 12-7, this translates directly into smaller package areas per chip or
fewer layers of wire.
Table 12-2 uses the wiring model to compare the implications of
the thick- and thin-film processes for the physical parameters of the
package as levels of integration increase. Initially, an area per chip of 1
cm2 is taken as a goal. The number of wire layers needed to permit the
package area in column 3 is calculated and presented in columns 4 and
5 for thick- and thin-film packages, respectively. Thick-film packages
have successfully provided 16 layers of signal wire and may be able to
cope with two or more times this number. However, the large number of
layer-to-Iayer connections required are certain to eventually limit advances
to more layers. If 40 signal layers are regarded as a practical maximum,
thick-film packaging fails for levels of integration around 6000 gates per
chip. This is a prime motivation for the development of thin-film technol-
ogy. Other advantages include better signal transmission by the nature

Table 12-1. Assumptions Concerning Thick- and Thin-Film Technology

Thick Film Thin Film Advanced Thin Film

Channel width (11m) 300 25 12


Conductor width (11m) \00 12 6
Resistance per unit length (Q/mm) 0.1 0.3 1.2
11-634 THIN-FILM PACKAGING

Table 12-2. Comparison of Module Wiring with Thick-Film and Thin-Film


Techniques

Thick-Film Thin-Film
GateslChil1 Chil1s/Module ArealChil1, A Wire Layers Wire Layers
(N) (M) (cm2) (K) (K)

1,000 100 12 2
2,000 100 18 2
4,000 100 27 3
6,000 100 41 4
16,000 100 62 6
32,000 100 95 8
32,000 100 2 67 6
32,000 36 57 4
32,000* 100 4
64,000 100 144 12
64,000 100 101 9
64,000 36 86 6
64,000* 100 6

*Advanced thin film-see Table 12-1.

of low-dielectric polymers and better electrical conductivity provided by


thin-film copper wiring. In fact, even in cases in which both techniques
might be applicable, the difficulties mentioned above favor thin-film
methods.
Part of the need for higher levels of integration has been satisfied
by increasing chip size. Chip areas already exceeded 1 cm2 and would
grow larger in the foreseeable future, and an attempt to reflect the possibly
larger substrate area simply because of an increase in the area of the
chip itself is incorporated in Table 12-2 for levels of integration above
16,000.

12.1.2 Thin-Film Module I/O Connections


Table 12-3 addresses problems that will eventually be encountered
by thin-film packages as integration increases. Rent's rule predicts that
very large numbers of connections to the next higher package level are
needed. However, the area of the substrate does not increase in proportion
to the number of connections to the higher level in the scenario of Table
12-2. Thus, the fifth column of Table 12-3 shows that miniaturization of
such connections will be required, a difficult task, as these connectors
will probably be expected to be mechanically separable in the field. Beyond
12.1 INTRODUCTION 11-635

Table 12-3. Additional Parameters of Modules with Thin-Film Wiring


M = 100 chips/module, A = I cm2/chip
N K NIM PinsIModule Area Per Mod. Pin R
(em') (n)

1,000 2 1.0 X 105 2,500 0.04 3


4,000 3 4.0 X 10' 5,700 0.017 3
16,000 6 1.6 x 106 13,000 0.0076 3
32,000 8 3.2 x 106 20,000 0.0050 3
64,000 12 6.4 x 106 30,000 0.0033 3
32,000* 4 3.2 x 106 20,000 0.0050 12
64,000* 6 6.4 x 106 30,000 0.0033 12

*Advaneed thin film- see Table 12-1.

a certain point, Rent's rule may not be applicable to the substrates of


Tables 12-2 and 12-3, as they contain over a million gates and can
constitute a complete medium-to-Iarge computer.
Although the number of layers is reduced by an order of magnitude
as compared to thick-film packaging, a two order-of-magnitude increase
in the level of integration uses up this advantage. One can be confident
that, as in Ie chip technology, package linewidths can be reduced and
permit the advantage of fewer layers to be retained for a longer time as
is illustrated by the "advanced thin film" in Table 12-2. However, unless
unusual aspect ratios can be achieved in the wires, increases in the
resistance of the long lines on the package will rapidly become unaccept-
able, as illustrated by the "advanced thin film" in Table 12-3. Another
possible direction would be reducing the number of chips on a module,
thereby reducing both the number of layers of wire and the length of wires.
Examples are included in Table 12-2. However, this would complicate the
physical design of the next higher level of package.
A final caveat: Although the characteristics of the chip have been
taken as a starting point, the chip and the package are interdependent.
The nature of the chips that are actually used will depend on the ability
of the package to deal with them, not only for wireability but also
for power distribution and heat dissipation. The latter demands are
discussed below.

12.1.3 Power Supply and Heat Removal


The quest for high speed has prevented the power dissipation per
circuit in gate arrays for large machines from decreasing very much as
11-636 THIN-FILM PACKAGING

the level of integration of the chips has increased. There is a trade-off


between power and delay; the highest speed demands that the gates be
operated at high power. Many power 1I0s are needed to carry the electrical
power to the chips. The package must include conductors that distribute
the power throughout the system. The power-supply wiring is simpler in
the sense that it does not have the pseudorandom character of the signal
network. Electrical resistance in the power supply leads must be carefully
controlled, however. A module may be powered by 100 A of current,
which must reach the chips with less than 5% voltage loss, for example.
The number of wiring levels in the module used to deliver current at the
supply voltage and return it through ground is not much different from
the number needed for signal transmission.
An aspect of the chip that poses a serious problem for the package
is the intermittent delivery of large amounts of current to the signal lines
by driver circuits on the chip [12]. Each such event changes the current
drawn from the power supply by the chip by tens of milliamperes. Some-
times many driver circuits switch simultaneously, causing a sudden large
change in the current drawn from the power supply, a change of a sizable
fraction of an ampere. The possible number of simultaneous switchings
will apparently increase with the level of integration. The inductance of
the power supply leads then produces a significant change of supply
voltage at the chip. For example, if a current change of 0.25 A in 250 ps
is to produce a voltage transient of less than 0.1 V, the inductance must
be less than 0.1 nH. This is approximately the inductance of only 100
!lm of wire. The power-supply transient can be controlled with capacitive
filters or decoupling capacitors, but the package must then provide the
capacitor very close to the chip, if not on the chip. One such arrangement
of an integrated decoupling capacitor is illustrated elsewhere [13] and
discussed in Section 12.9.
The electrical power delivered to the chip is dissipated to heat, and
another demand made on the package by high-speed gate array chips is
the removal of large amounts of heat. Today, a power of 10-40 W, 10 W
in chips being shipped and as much as 40 W in chips being developed,
may be produced on a gate-array chip. High heat fluxes are carried away
from the chip by special cooling structures that transfer the heat to air or
another fluid that carries it out of the machine, as discussed in great detail
in Chapter 4, "Heat Transfer in Electronic Packages." Chip temperatures
must be kept below some maximum value, say about 70°C; only a certain
finite temperature difference is available as a driving force for the heat
flux. The limited heat-transfer coefficients at interfaces require that suffi-
cient area be provided to allow heat flow from one material to another
with an acceptable temperature drop. Apertures through which fluids flow
must be large enough to permit the flow to take place at pressures that
can be provided by inexpensive and quiet apparatus. All of these restric-
12.2 ELECTRICAL PERFORMANCE 11-637

tions limit the rate of heat removal and constitute another lower bound
to the area per chip, particularly with thin-film packaging. Fairly complex
structures are needed to achieve even the present rates of heat removal.
It is likely that the power per gate on logic chips will decrease as
integration increases. A very rough attempt to extrapolate trends in bipolar
chips is presented in Table 12-4 together with the implications for removal
of heat from packages. The table clearly illustrates the need for continued
increases in cooling capability, as compared with about 5 W/cm 2 in current
practice (first row in Table 12-4).
Relief from the heat-transfer limitations may also be found in recent
development to build large mainframe computers from CMOS integrated
circuits. CMOS circuitry dissipates much less power than high-speed
bipolar circuitry. Furthermore, operation at the temperature of liquid nitro-
gen improves performance by a factor of 2. The electrical resistivity
limitation in advanced technology will also be relieved by cryogenic
operation and, additionally, motivates interest in it. However, cooling
efficiency at these low temperatures will present additional challenges.

12.2 ELECTRICAL PERFORMANCE


The physical structure and the materials used in the construction of
a thin-film package are constrained by the electrical properties that are
needed to attain the best performance [12,14]. The electrical requirements
generally relate to:

Table 12-4. Estimated Power Density of the Thin-Film Modules

Gates Chip Area, A Power/Gate Module Power Densit:t


(N) (cm') (mW) (W/cm')

1,000 5 5
4,000 5 20
8,000 5 40
16,000 1 5 80
32,000 2 5 160
32,000 2 5 80
4,000 2.5 10
8,000 2.5 20
16,000 2.5 40
32,000 2.5 80
4,000 1 4
8,000 8
16,000 16
32,000 32
11-638 THIN-FILM PACKAGING

• Point-to-point propagation (velocity, attenuation, distortion)


• Noise containment (coupled, ill, shielding)
• Impedance (matching terminations, resistance)

Transmission-line characteristics are of special importance in high-


speed systems in which short pulse lengths require that high-frequency
line impedance and load impedance be matched to prevent reflected pulses
from appearing as noise, as discussed below.
Chapter 3, "Package Electrical Design" discusses the factors and
trade-offs involved in designing an electronic system. A summary is
repeated here in the context of thin-film packaging technology.
The width and thickness of lines determine the DC resistance,
whereas the spacing between lines is determined by allowable cross-
talk. In a single-sided thin- or thick-film structure with no control of
characteristic impedance, Zo, the cross-talk between two adjacent lines is
determined by electrostatic and electromagnetic coupling between the two
lines. Coupling in this case decreases as the conductor spacing increases.
Cross-talk can be reduced by placing a ground plane closer to the line
than the line-to-line distance, as in Figure 12-5. Further improvements
can be made by having two ground planes, one above and the other below
the line, but both closer to the conductor lines than from one line to an
adjacent line. The close spacing between signal and ground, however,
results in excessive capacitance and affects the characteristic impedance,
defined as [series inductance (L)/capacitance (C)]II2. Very-high-
performance applications require that all lines be terminated in order to
keep the signals that are sent down the lines from reflecting at the far
end and transmitting backward along the lines. A digital signal can also
be partially reflected from a discontinuity in the transmission line. The
reflection coefficient, which gives the fraction of the reflected signal, is
determined by the characteristic impedance and by the load resistance
that terminates the line. If a given transmission line has a characteristic
impedance of lOOn and the load resistance is also lOOn, the signal is
totally absorbed by the load and none of it is reflected back. This is an
.ideal situation. If the load resistance is 200n, however, a third of the
signal is reflected and adds to the initial signal on the line. A load resistance
of 50n also yields a reflection coefficient of one-third, but the signal is
subtracted from the initial one. A characteristic impedance range of 50-
lOOn is commonly used in designing digital computers. In a triplate
structure, the linewidth, related to interconnection density, as described
previously, determines the thickness of dielectric required between a line
and the reference plane for a given characteristic impedance, given a
value of dielectric constant for the insulating layer. The characteristic
impedance Zo is related to the inductance and capacitance per unit length
(L and C) by
12.2 ELECTRICAL PERFORMANCE 11-639

....
~ --s--+l~1

Alumina
€r = 9

a) Conventional Alumina Substrate

I~.-----------s----------~~~I
Polyimide
€r = 3

b) Polyimide with Ground Plane Moved Up

1---1/2 s "I_ 1I2S~

Polyimide
€r = 3

c) Polyimide with Twice the Line


Density, Same Cross Talk
Figure 12-5. Lowering Dielectric Constant Improves Package Density. (a)
Alumina substrate; (b) polyimide dielectric; (c) polyimide dielectric with twice the line
density, same cross-talk.
11-640 THIN-FILM PACKAGING

Zo = (L / C)112 [12-8]

Consequently, the vertical dimensions of a multilayer thin-film (MLTF)


structure depend on the electrical parameters desired.
The requirement that coupled noise levels be kept low enough to
prevent switching a receiver on a neighboring line will demand that the
line spacing be large enough so the mutual inductance of two neighboring
lines will be less than a given value. This requirement will modify the
thin-film interconnection structure so that

1. Parallel runs are kept to a certain maximum length


2. A reference plane is used to reduce the coupled energy
3. Wiring channels are placed appropriately far apart

Clearly, these approaches are examples of the manner in which


electrical requirements modify and restrict the set of thin-film structures
that can be used for an application.
Pulse-propagation velocity and dispersion impose materials require-
ments on the thin-film dielectric. The pulse speed is given by

v = c / £112 [12-9]

where
£ = dielectric constant of the insulating layer
c = speed of light
V = pUlse-propagation velocity

For high-speed systems, it is desirable for the thin-film dielectric to


have a low value of dielectric constant A short pulse is composed of a
range of high-frequency components (up to 1 GHz and above). Conse-
quently, the dielectric constant of the thin-film dielectric should be rela-
tively constant up to these frequencies, as well as close to unity, to ensure
both a minimized pulse broadening and a high velocity of propagation.
Additionally, the lowering of the dielectric constant enhances package
performance in other ways, as discussed below.
A reduction in dielectric constant in Figure 12-5 can allow a lower
thickness of dielectric, for equivalent capacitance, and enable the ground
plane to be moved closer to the line. Additional lines can, therefore, be
accommodated for the same cross-talk, as illustrated. Thus, the effect of
a low dielectric constant, in addition to increased speed of signal, results
in improved wiring density of packages, which further results in improved
systems performance. The effect of the dielectric constant, therefore, is a
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-641

synergetic effect of speed improvement from Equation 12-9 and improved


packaging density.
Finally, line resistance attenuates the signal as it propagates in a
driver line (see Chapter 3). A conductor's conductivity, cross section,
length, and driver/receiver characteristics must be all coordinated in order
for the package interconnections to function properly. Properties of thin
films given later can be used in the design of thin-film packages that
meet not only geometric interconnection functional needs but also those
imposed by the requirements for electrical functionality.

12.3 THIN-FILM VERSUS THICK-FILM PACKAGES

Materials and processes for thin-film technology are distinctly differ-


ent from those used in other packaging technologies, such as multilayer
ceramic or printed wiring boards, as illustrated in Figure 12-6. In the
latter technologies, thick-film processes such as doctor-blading and lami-
nating of greensheets and epoxy-glass cloth are the typical processes by
which insulating structures are fabricated. These techniques are covered
in Chapter 9, "Ceramic Packaging" and Chapter 17, "Printed-Circuit Board
Packaging." They are denoted as thick-film approaches because they utilize
materials with coarse granularity of about 100 Ilm (inorganic greensheets
and epoxy-glass cloth) and involve processes of low resolution and preci-
sion that make them not easily adapted to the fabrication of multilayer
structures with feature sizes less than 100 Ilm. Table 12-5 compares the
characteristics of each of these three technologies for wiring dimensions
and layers.
Materials utilized for thin-film packaging have granularity generally
smaller than I Ilm (photoresist, polymer films, amorphous inorganic struc-
tures formed from sputtered materials, reflowed glass, etc.) and the typical
feature sizes in thin-film packaging are in the order of 25 Ilm and below.
Consistent with the potential offered by such materials for fine-line pattern-
ing, the processes used for thin-film packaging primarily employ photoli-
thography for conductor-pattern definition and via generation through
the dielectric polymer.
The advantages of thin-film technology over thick-film technologies
are better elucidated by specific examples as follows.

12.3.1 Size of Thin-Film and Thick-Film Packages

High-density thin-film packages such as the ones discussed in this


chapter are compared in Table 12-6 with thick-film packages involving
both sequential buildup of multilayers and cofired multilayer structures
discussed in Chapter 9, "Ceramic Packaging." A thorough discussion of
various aspects of all these packages is reported elsewhere [15], including:
11-642 THIN-FILM PACKAGING

(A) (8) (C)


Cofired MLC Thin Film on Si Printed Wiring
MCM-C MCM-D MCM-L ~
r=P r=P

Metallize
_••••?... C
Coopppper Foil
Greensheet Copper-Cladded
Apply Resist Laminate
II II
Vias Punched Etch Metal
Remove Resist Pattern Etching

IE iii
Vias MetaU'
C' H
.........
~Permanent
Lines Screened Spin-Coat , Plating
PoIyimide . . . .IIIIII!.IIlII. Resist Film

til D1m
Form Plating Mask

==
C' r=I
Apply Resist

r
Laminated

III ij ill
==
Drilling
Expose/Develop

r
Coslntered

Etch Polyimide /\
Catalyzer

P Seeding

==
H Electroless
,,---...;;; Copper
Remove Resist
Cure

Electroless Copper
I' Metallize
Via/Conductor

Figure 12-6. Comparison of Thin- and Thick-Film Technologies

• Packaging density
• Package delay
• Package power
• Cost

The analysis, assuming a very-Iarge-scale-integration (VLSI) chip


of 7.5 rnm size, is summarized in Table 12-6. The printed-wiring board
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-643

Table 12-5. Typical Package Parameters of the Three Multichip Module Types

Characteristic MCM-L MCM-C MCM-D

Description High-density Cofired low- Thin film on silicon


laminated printed- dielectric-constant
wiring board ceramic substrate
Max. wiring density 300 800 250-750
(cm/cm2)
Min. linewidth (Ilm) 60-100 75-100 8-25
Line space (Ilm) 625-2250 125-450 25-75
Max. substrate size 700 245 50-225
(mm)
Dielectric constant 3.7-4.5 5-5.9 3.5
Pinout grid (mm) Array 2.54 Array 2.54 Peripheral 0.63
(staggered)
Max. No. of wiring 46 63 4
layers
Via grid (Ilm) 1250 225-450 25-75
Via diameter (Ilm) 300-500 100 8-25

in Table 12-6 is assumed to be multilayer epoxy-glass board with four


to eight X and Y conductor planes spaced appropriately between ground
planes. The conductor widths are assumed to be 125 Jlm and plated-
through-holes (PTH) are on 254-Jlm centers. Thick-film multilayer ceramic
involves sequential deposition and firing of dielectrics and conductors on

Table 12-6. Comparison of Package Figures of Merit-Normalized. After


Neugebauer. Ref. [15].

Overall
Power Size or Relative Packaging Packaging Figure
Packaging Approach Delay Weight Cost Density Delay of Merit

Printed-wiring board 1.00 1.00 1.00 5 7.8


Thick-film multilayer
ceramic 1.08 0.42 1.02 12 7.8 2
Multilayer ceramic 0.34 0.20 0.65 25 5.6 20
Thin-film multilayer
with chips on one
side 0.19 0.14 0.60 36 4.8 60
Thin-film multilayer
with chips on both
sides 0.08 0.07 0.44 72 4.1 400
11-644 THIN-FILM PACKAGING

prefired alumina substrate. The conductor width and dielectric thickness


are assumed to be 250 /lm and 40 /lm, respectively. The multilayer ceramic
stands for cofired multilayer alumina and molybdenum substrate described
in Chapter 9, "Ceramic Packaging." The thin-film multilayer is the same
as the multilevel polyimide (PI) or glass thin-film technology discussed
in this chapter. Linewidth and dielectric thickness are assumed to be 25
/lm and l/lm, respectively. The thin-film hybrid with chips on both sides
involves the same multilevel thin-film technology as before, but chips
are bonded on both sides with appropriate via connections in between.
Referring to Table 12-6, packaging density is calculated by dividing the
chip area by the chip footprint. Package-delay calculations are estimated
simply from the output capacitance, the on-chip resistance of the output
driver, and the buffer delay. Power-density calculations are based on
power requirements per unit area of the package-thin-film packaging
technologies requiring a power supply about 50% more than the multilayer
cofired ceramic, consistent with packaging-density improvements pointed
out earlier. Multilevel thin film in Table 12-6 is clearly shown to provide
as much as a sevenfold improvement over the printed wiring board and
50% over the state-of-the-art multilayer ceramic (MLC) substrate. The
potential of drastic size reduction of thin-film packages is illustrated in
Figure 12-7 for a hypothetical nine-chip multichip package each with
1000 1I0s. In this figure, the printed wiring board is assumed to be on a
50 mil via grid, ceramic on a 3 mil grid, and thin film on a 10 mil grid.
This size reduction is the primary motivation for the development of thin

I Chip I I Chip I I Chip I


Area Reduction
El EI El
I Chip I I Chip I I Chip I
~ EJ EJ
Chip

"""
+48mm+
Chip

MCM-D
Chip

(10 MIl. Via Grid)


I Chip I I Chip I I Chip I

...- - - - - 240 mm
MCM-L (Card)
(50 Mil. Via Grid)

Figure 12·7. Comparison of Sizes of Hypothetical 9-Chip Modules Fabricated


by Different Technologies
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-645

films on printed wiring boards such as the surface laminar circuitry (SLC)
recently developed by IBM Japan as discussed in Section 12.8 [16]. This
technology is also referred to in the industry as MCM-D on MCM-L
(printed-wiring board multichip module) or MCM-DIL.

12.3.2 Cost of Thin-Film and Thick-Film Packages


The relative cost of the three packages discussed above is also
summarized in Table 12-6, as well as in Figure 12-8 as a function of
typical chip IIOs interconnected onto these packages. Cost is by no means
a precise or absolute parameter that describes the package; it is a relative
measure that changes as a function of time, volume, yield, equipment
depreciation, and so on. Nevertheless, it is a good indicator to differentiate
various packages as proposed here. Clearly, the printed wiring board
technology, based on photolithography patterning and lamination followed
by mechanical drilling and plated-through-hole, provides the least wiring
but it is also the cheapest technology of the three. It is also obviously the
technology that meets most of the low 110, consumer, and portable product
requirements. Two recent breakthroughs, one in direct flip chip attach to
the printed wiring board by means of underfill encapsulants, as discussed

$10K~----------------------~

$3K

$30

$10
Technology
Merge
$3 MCM·D/L

100 1K 10K
I/O Count

Figure 12·8. Cost of MCMs as a Function of 110 Counts


11-646 THIN-FILM PACKAGING

in Chapter 8, "Chip-to-Package Interconnections," and the other is one


that add one or more layers of surface thin-film wiring referred to as
surface laminar circuitry (SLC), as discussed later in this chapter as well
as in Chapter 17, "Printed-Circuit Board Packaging," provide additional
motivation for the use of this technology. These developments are illus-
trated in Figure 12-9 but are discussed in more detail in the aforementioned
chapters. The technology merger in Figure 12-8 allows "best of both
world," and thin-film on PWB as the single strategic direction the industry
is expected to take.
The cofired-ceramic technology provides improvements over the
printed wiring board technology in two areas: finer wiring (75 /lm lines
and vias) and more layers (up to 100). The cost per 110 of this technology
is, however, higher than printed wiring board technology. The thin-film
technology discussed in this chapter provides the most wiring, but the
cost per 110 is the highest. The significant thin-film costs involve setting
up a cleanroom manufacturing facility (class 10-1000). A number of
approaches discussed in Section 12.10, however, provide potential solu-
tions to lowering the cost to a level that this technology can be expected
to be almost as inexpensive as today's standard printed-wiring board tech-
nology.

12.3.3 Single-Layer Thin-Film Packages


Thin-film packages have and will continue to be used both for single-
chip and multichip module packages. For single-chip packages such as
the ones illustrated in Figure 12-10, thin film is used to redistribute or
fan-out the chip 110 grid to that of printed wiring board. The chip 110

c:J c:J

PTH Epoxy-Glas Core

c:J c:J
Tsukada, 5I9l 1
Figure 12-9. Surface Laminar Circuitry by IBM, Japan
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-647

(a) Cr-Cu Metal 2

Ceramic

(b)

Figure 12-10. First Generation of Single-Chip Thin-Film Packages by (a)


IBM, and (b) Hitachi

grid is typically about 100-250 11m and that of the board typically about
0.4-2.5 mm.
The simplest type of thin-film package is one in which the conductor
is a single patterned metal thin-film layer on a substrate. The implementa-
tion can assume one of two schemes. One of these is a fan-out or space-
transformer pattern, whereas the other is an interchip connection. In
both cases, the application does not allow crossing of the point-to-point
conduction paths because of the topology of single-layer structure. The
space transformer is usually used to connect the signal and power termina-
tions of a chip to the coarser grid of the second-level package. The
other type of single layer structure that is used in simple applications
predominantly involves parallel transmission of signal and power. Parallel
channels, busses, and power supply are examples of such applications.
The substrate for fan-out may be rigid (e.g., ceramic or silicon) or
flexible as in tape automated bonding (TAB) for chip mounting discussed
in Chapter 8, "Chip-to-Package Interconnections." Thin-film processes
for forming two-layer and three-layer tapes as described in that chapter
employ subtractive or plate-up technologies which are discussed later in
11-648 THIN-FILM PACKAGING

this chapter. In the case of parallel signal transmission, fabrication of the


conductors on a flexible carrier, such as a polyimide, permits a packaging
entity that can be bent during system assembly or repair. This technology
is most commonly used in consumer electronics.
Functional requirements determine the materials and structures of a
packaging entity. The conductivity of the conductor material, together
with its cross-sectional geometry, determines the resistance per unit length.
This parameter must be chosen in the case of signal conductors so that
the ohmic signal-level drop from driver to receiver is within the system's
specifications.
In the case of power-carrying conductors, both insulation resistance
(IR) drop and conductor heating must be controlled by proper choice of
conductor materials and geometry. Coupled noise in simple packages with
single-layer thin-film conductors is controlled in two ways, as discussed
earlier in this chapter and in Chapter 3, "Packaging Electrical Design."
One is to reduce mutual inductance by wider spacing of conductors,
whereas the other is to intersperse signal and power lines. This latter
technique can also provide control of the high-frequency impedance of
signal lines. As in other package geometries, there is a competition between
density and functionality.
High wiring density requires close spacing of narrow conductors,
whereas power capability, and containment of signal-level drop and cou-
pled noise place an upper bound on wiring density. This competition is
usually not an issue in relatively low-performance simple packages, such
as single-chip packages, but as performance demands become aggressive
(e.g., high 110 count or high-speed chips and systems), multilevel, multi-
chip packaging entities provide extra design degrees of freedom, at a yet
higher cost. Furthermore, materials and mechanical effects also place
constraints on the construction of multilayer, multichip package. The
conductors must adhere to the substrate in the presence of mechanical
stress. Such stress may be intrinsic (due to the fabrication processes) or
may be externally applied during system construction and use (as can
occur in the bending of a flexible substrate) or may be caused by thermal
cycling of chips and conductors during power on-off cycles. Severe stress
can also develop during system shipment where temperatures can range
from as low as -40°C to as high as 120°C, as discussed in Chapter 5,
"Package Reliability." More detailed discussion on multilevel, multichip
packaging is found in the following sections.
The single-chip packages illustrated in Figure 12-10 and their pro-
cesses are discussed in greater details in Section 12.7.
12.3.4 Types of Multilayer Thin-Film Structures
Manufacturability, cost, and electrical performance are the three
primary factors that are considered when choosing materials, processes,
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-649

and structures for multilevel thin films (MLTF). From the structure point
of view, one needs to choose between planar and nonplanar thin-film
processes based on cost/performance trade-offs.
The choice of one type of structure over the other also depends
on customer requirements such as wireability, impedance, and global
planarity. Therefore, it is desirable to maintain process, materials, and
tool commonality between the two types of structure wherever possible.
One exception is the mechanical planarization technique that is needed
to build planar structures. The criteria that are chosen to compare the
processes are (1) packaging density, (2) manufacturability, (3) process
and materials maturity, and (4) cost. Multilevel, thin-film interconnect
structures can be essentially classified into two types: (1) unfilled via type
requiring staggering of vias in successive levels and (2) filled via or
"stud" type which permits stacking of vias in successive levels, as shown
in Figure 12-11.
One of the first process considerations in the design of the thin-film
package is the choice of the via structure. The conformal via process
gives rise to a nonplanar structure, whereas the filled via process gives
rise to a planar structure. Planar structures produce the highest wiring
density for a given linewidth and pitch. One has to consider the trade-
offs between the process complexity and tooling costs, and the electrical
advantage of increased wiring density.

12.3.4.1 Planar stacked studs


Planar structures can be formed by first defining the wiring and via
pattern in the dielectric and then metallizing the structure. In this approach,
planarization to remove the excessive metal is done at the wiring level.
Two variations exist for this process. In the first variation, the intralevel
dielectric (between adjacent conductor lines) and the interlevel dielectric
(between two conductor layers, X and Y) are applied separately. After

Metal Level 2

=v_.~
(A) Unfilled Via
Dielectric
Metal Level I

Metal Level 2
Dielectric
Metal Level I
(B) Filled Via

Figure 12-11. Filled and Unfilled Via Interconnect


11-650 THIN-FILM PACKAGING

the application of the intralevel dielectric, the vias are patterned and the
polymer is cured. Similarly, after the application of the interlevel dielectric,
the conductor features and vias are patterned as shown in Figure 12-12.
In the second variation of this approach, the vias can be formed after the
conductor lines in the intralevel dielectric are defined. In both cases, the
whole dual-level structure is then metallized by electroplating or sputter-
ing. The excessive metal is then removed by mechanical or electrochemical
planarization [1]. Because the intralevel dielectric is applied on a planar
surface, the thickness is well controlled. On the other hand, the conductor-
layer thickness is defined by the planarity of the interlevel dielectric layer
and the mechanical planarization process. It is for this reason that the
planarization technique chosen should give local planarity, not global
flatness .
The alternative is to first define the metal wiring and via features
and then applying the dielectric, followed by planarization to remove the
excessi ve dielectric at the stud level. This latter process has been developed
and implemented by many companies and institutions [17] .
In this process, the wiring level and studs are fabricated by pattern

, Prepare surface for MLTF


Substrate

,,,,,,,,,, ,,,
Apply and softbake PSPI

" ,,
Expose

, Develop

Cure PSPI

Repeat PSPI processing for wiring level

Deposit MetaJ

Mechanical Planarization

Figure 12-12. Process for Planar Stacked Stud Interconnect


12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-651

electroplating through a photoresist stencil using a common plating seed


layer. A polymer dielectric is then applied by spin or spray coating. The
excessive polymer is removed by mechanical means. This exposes the
tip of the studs for connection to the next level of interconnect and
planarizes the entire structure. This process gives the greatest flexibility
in the choice of the polymer and the coating technique. The conductor
thickness and width is defined by the uniformity of the electroplating and
seed-layer etching processes. The dielectric thickness control is much more
difficult because it is defined by the mechanical planarization technique.
Figure 12-13 shows the cross sections of (a) a stacked via structure,
and (b) the corresponding planar structure.

12.3.4.2 Nonplanar staggered vias


Nonplanar structures are simpler to process and require no mechani-
cal planarization. They also require fewer dielectric application steps and
their associated cures. The vias are generated through the dielectric layer,
and then the vias and conductor lines are metallized on top of the dielectric
layer. Both via formation and conductor line patterning are by lithographic
techniques. The connections between the conductors below and above
the dielectric layer are made through the via openings in the dielectric.
No planarization is produced in the metallization processes of vias and
conductor lines. Partial planarization of the topography of the underlying
conductor line and via metallization is achieved by the capability of
leveling of the dielectric polymer coating. Without further planarization

Figure 12-13. (a) Planar Stacked Stud MLTF Structure; (b) Planar MLTF
with Flip Chip Connection
11-652 THIN-FILM PACKAGING

steps, a conformal or nonplanar structure resulted and is aggravated with


the later built-up of the multilayer thin-film structure. A disadvantage of
this process is the use of low leveling or self-planarization of commercial
dielectric polyimides, which limits the number of layers. This can be
alleviated by incorporating multiple coatings of polyimide which tend to
increase the degree of planarization (OOP), or by using dielectric polymers
with higher planarization capability.
In the conformal via process shown in Figure 12-14, the via holes
are formed by photolithographic methods in a photosensitive polyimide
(PSPI) dielectric or laser ablation and metallized conformally by elec-
troplating through a thick positive photoresist mask.
The vias could also be fabricated using reactive ion etching (RIB)
or wet etching. Laser ablation and photosensitive polyimide lithography
are robust processes and have been successfully implemented. Metalliza-
tion can be accomplished by subtractive etching, pattern plating, or lift-
off. The choice between subtractive etching and pattern electroplating
will primarily depend on the ground rules of the package and the tolerances
required. The minimum pitch for subtractive etching as a function of

,
Prepare ceramic substrate· define capture
pad by subtractive etching of Cr-Cu-Cr
Ceramic

Apply P PI and expose/develop vias

Sputter Cr-Cu Seed Layer


I
, Apply photoresist and expose/develop lines

Electroplate Cu to Final Thickness

Strip photoresist and etch seed layer

Figure 12·14. Process for Conformal Staggered Via Interconnect


12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-653

linewidth, undercutting and line thickness is shown in Figure 12-46, and


is defined as follows:

P = L + M + 2T + U, [12-10]

where
P = minimum pitch
L = linewidth
M = minimum resolution
T = line thickness
U = undercut

When the ground rules do not allow for subtractive etching to be used
as the metallization process, pattern electroplating has to be used.
Figure 12-15 shows the cross section of a nonplanar structure.

12.3.4.3 Electrical Trade-offs


Two fundamental electrical issues need to be addressed when com-
paring the planar and nonplanar structures. These relate to the differences
in the module power distribution due to stacked studs as opposed to
staggered vias and electrical variations due to the effect of the dielec-
tric nonplanarity.
A typical triplate wiring structure (two signal levels and two ground!
power levels) is shown in Figure 12-16, which consists of two wiring
layers sandwiched between two mesh planes. The cross section assumes
a planar structure that guarantees a uniform dielectric thickness across
the entire module.

Figure 12-15. Cross Section of One Level of Nonplanar Staggered Via Thin-
Film Interconnect
11-654 THIN-FILM PACKAGING

Mesh 2

Wiring

~T ~ Wiring

_ Mehl

Figure 12-16. Triplate Wiring Structure (Planar Process)

The physical parameters in Figure 12-16 such as the mesh width


(Mw), signal linewidth (Sw), dielectric thickness (D), metal thickness (n,
and dielectric constant (E) affect the characteristic impedance Zo of the
line. The variation of Zo as a function of the physical dimensions is shown
in Figure 12-17 assuming a planar process. Because the design of the
mesh plane plays a critical role in the calculation of 20, a parallel mesh
is assumed where all lines are completely shielded with an added restriction
that Mw = 2Sw' The adjacent wiring level adds an additional capacitive
loading which has been included in Figure 12-17.

60r-------------------,

.S

lSL--..I..-...L...---I.-....I....-I....---'--..I..-........-....I....--I
0.2 0.3 0.4 os 0.6 0.7

SjD

Figure 12-17. Characteristic Impedance, Zo, versus Thin-Film Dimensions


(Planar Process)
12.3 THIN-FILM VERSUS THICK-FILM PACKAGES 11-655

The Zo variation shown in Figure 12-17 may not be true for a


nonplanar process due to the dielectric nonplanarity. In a nonplanar pro-
cess, the mesh layout has a larger role in deciding the line characteristics
than in a planar process. Based on extensive modeling of the nonplanar
cross section, the parallel mesh design with all lines completely shielded
represents the optimum layout that minimizes the dielectric nonplanarity
in critical areas, thus providing a controlled impedance environment for
the transmission lines. The cross section of a triplate wiring structure for
a nonplanar structure is shown in Figure 12-18.
As seen in Figure 12-18, the two wiring levels do not look alike,
thus resulting in different Zo for the two levels. The bottom wiring level
is similar to that in Figure 12-17 due to the use of a parallel mesh; hence,
the variation of Zo is the same as in Figure 12-17. However, the top wiring
level has variations due to dielectric nonplanarity, which results in a lower
Zo as compared to the bottom wiring level.
The difference in the power distribution due to the use of stacked
studs (planar) as opposed to staggered vias (nonplanar) is best illustrated
with an example. Consider a module with a pin-grid array supplying
power to the chips through a thin-film triplate wiring structure. Because
the vias or studs supply the current to the chip, they decide the module
voltage drops.
Assuming D = 5.5 11m and T = 4.5 11m, the staggered via scheme
produces an additional 37.5% voltage drop due to the longer current path
as opposed to the stacked via scheme. Hence, it is important that a good
power distribution be used in the nonplanar structure with more vias,
which decreases the module voltage drop through current sharing.
Although only the Zo variation has been considered here, all electrical
parameters are affected by the dielectric nonplanarity. Hence, for a good
design using the nonplanar structure, it is imperative that the designer be
aware of the structural features and the electrical variations so as to create
an optimized layout that meets the application requirements.

Power

.r-- -........., Signal

E~ Signal

_ Ground

Figure 12-18. Triplate Wiring Structure (Nonplanar Process)


11-656 THIN-FILM PACKAGING

12.4 THIN-FILM MATERIALS AND PROCESSES

Having reviewed the structures to which the functional requirements


of systems lead, an overview of materials and processes is in place. Several
multilevel thin-film structures [20,21] that are under development or in
manufacturing are listed in Table 12-7, along with both the conductor
and dielectric materials, as well as their processes. The number of layers
ranges from two to six on substrates ranging from silicon to alumina to
multilayer ceramic to glass-ceramic to sapphire, and more recently to
printed wiring board (PWB). The substrate sizes range up to 100-225 mm
at present and are expected to be in excess of 400 mm in the near future.
Roll-to-roll continuous fabrication as described in Section 12.8.2 is also
being pursued. A list of the properties of common dielectric and conductor
materials for thin-film packaging are given in Table 12-8. The set of
materials and processes described here to fabricate thin-film packages is
reasonably comprehensive. More detailed discussions can be found in
many references [2,17,20-42]. Thin-film packaging is one of the fastest
evolving packaging technologies today and is expected to continue well
into the next century. Consequently, the emergence of novel materials
and processes is anticipated.
The general process of fabricating thin-film multichip or single-chip
packaging involves five basic steps:

1. Dielectric deposition on substrate


2. Via formation in dielectric
3. Metal deposition
4. Metal patterning
5. Bonding pad or terminal metallurgy deposition

12.4.1 Thin-Film Substrates


The various substrates used for high-density thin-film package manu-
facturing are as follows:

• Silicon wafer
• Dry pressed alumina
• Aluminum nitride
• Cofired alumina
• Cofired low temperature or glass-ceramic
• Polished metals (AI)
• Printed wiring board (FR4)
Table 12-7. Some High-Density Multilevel Thin-Film Technologies Reported

Number 1 ~ 3 4 2. 6 7

Metal Aluminum Copper Aluminum Copper Copper Copper Copper


Thickness <11m) 1 5 I 5 5.5 25 2
Width (/-1m) Il 25 25 10 8 100 10
Pitch (J.1ffi) 22 75 100 25 25 250 20
(I) Resistance (mQ) 27 Il 3-4 4 0.1 10
U1
..... Dielectric SiO, Polyimide SiO,lsapphire Polyimide Polyimide Polyimide Polyimide
glass
Dielectric Canst. 3.8 3.5 5.5 3.5 3.5 3.8 3.5
Thickness (/-1m) 1 25 450 8 6 60 10
Deposition Sputter/CVD Spray Sputter/CVD Spun on Spun on Foil Sputter
Substrate Silicon Al,03 Sapphire Silicon Alumina Copper Silicon
Reference [20] [Il] [20] [20] [20] [20] [21]
11-658 THIN-FILM PACKAGING

Table 12-8. Thin-Film Materials Properties

Coefficient of
A. INSULATOR Dielectric Thermal Thermal Method of
Constant EXl1ansion Conductivity Del1osition
1O-7/°C W/mK

Borosilicate glass 4.1 30 5 Spray


Quartz 3.8 5 7 Sputter
Polyimide 3.5 500 0.2 Spray, spin
Polyimide (PIQ-L 120) 3.5 30 0.2 Spray
Aluminum nitride 8.8 33 230 Sputter
Teflon 2.1 200 0.1 Melt
Lead borosilicate 7.8 35 5 Spray
Boron nitride 5.5 30 250 CVD
Diamond 5.7 23 2000 High pressure
Epoxy 3.2-4.3 150-780 0.18-0.88 Curtain, dry film
Benzocyclobutene (BCB) 2.6 350-660 0.2 Spin, extrusion

B. CONDUCTORS Resistivity
lJ.ll·cm

Copper 1.67 170 393 Evap. or sputter


Gold 2.2 142 297 Evap. or sputter
Aluminum 4.3 230 240 Evap. or sputter
Nickel 6.8 133 92 Evap. or sputter

To date, cofired ceramic and silicon wafer are the most common
substrates. Extensive research and development efforts are being devoted
to the enhancement of wiring density of thin-film structures on printed
wiring board substrates typified by the IBM Surface Laminar Circuitry
(SLC) technology or MCM-DIL as described elsewhere in this chapter.
The inherent low-cost nature of printed wiring board substrates is the
biggest driver behind this trend.
Cofired ceramic with wiring inside is referred to as an active substrate
and has several advantages over silicon:

1. Wiring channels can be optimized between the ceramic and thin-


film layers. The thin-film nets can be used for redistribution and for
wiring of cycle-time-determining nets. The rest of the nets, including
multidrop nets which run over long distances, can be wired in the ce-
ramic.
2. Use of both thin-film and cofired ceramic wiring reduces thin-film
complexity and the number of thin-film layers, thereby increasing
thin-film yield.
12.4 THIN-FILM MATERIALS AND PROCESSES 11-659

3. More power planes can be provided when the power consumption


and dissipation of the chips are high.
4. Cofired ceramic allows for module level signal and power area array
connection. This configuration has lower AI noise and inductance
compared to power distribution from the side of substrates.

The major advantage of a silicon substrate is the good thermal


expansion matching that of IC chips. Higher thermal conductivity is
also a favorable property. Furthermore, certain active and passive devices
can potentially be integrated into the substrate. The major drawback is
that no wiring is allowed within the silicon substrate, which is intrinsically
a passive substrate.
Surface preparation of cofired ceramic is a very important process
step before multilayer thin-film fabrication. The substrate should have a
surface roughness of less than 1000 A and should have a surface flatness
of 10 J..lm or less to be suitable for thin-film photolithography [43]. This
is achieved by lapping and polishing the cofired-ceramic substrate. The
surface is then cleaned and modified by chemical or physical means prior
to thin-film metal or dielectric deposition. The first thin-film layer on
cofired ceramic consists of large pads which capture the vias of the ceramic
substrate. The use of a capture-pad layer facilitates the matching of a
regular photolithographic grid to the irregular ceramic grid. The bottom
reference plane is combined with the capture pad level when it is electri-
cally acceptable.
On thin substrates like silicon or cofired ceramic with very few
layers, bowing due to the thermal coefficient of expansion (TeE) mismatch
between the thin films and the substrate leads to focusing problems during
photolithography, whereas this effect is often negligible on thick, rigid
substrates. The bowing can be controlled during exposure by the use of
a vacuum chuck. One drawback of a cofired ceramic is its low thermal
conductivity. In high-power, high-performance systems flip-chip technol-
ogy is used to connect high-II0-count chips, with the additional advantage
of effective heat removal from the back side of the chip.

12.4.2 Dielectric Materials


The dielectric materials for thin-film structures can be classified as
either organic or inorganic. A wide variety of organic materials, all poly-
mers, with adequate electrical, thermal and mechanical properties have
been employed as the dielectric layer in thin-film packaging. Polyimides
are the most common ones in use today. The inorganic materials typically
used are either borosilicate glasses, processed using thick-film techniques,
or high-temperature materials such as aluminum nitride (AIN), boron
nitride (BN), or quartz (Si0 2) deposited by either sputtering or chemical
11-660 THIN-FILM PACKAGING

vapor deposition processes. All these materials with their electrical and
thermal properties were listed previously in Table 12-8. The general
requirements regardless of which type of dielectric is used for forming
multilevel thin-film dielectrics are as follows:

1. Low dielectric constant (less than 4)


2. Thermal-expansion matching to substrate and/or silicon, or good
ductility and/or low modulus to minimize stress development
3. Thermal stability to withstand the temperatures in subsequent chip
attach and pin-joining high-temperature processes (400°C with
97/3 Pb/Sn, 230°C with 60/40 Pb/Sn)
4. Good adhesion to copper or other conductors
5. Self adhesion
6. Ease of processing, preferably with planar structures
7. Low water absorption

Reliability and cost are other important requirements.

12.4.2.1 Organic Materials and Their Properties


The trend in the industry appears to be for organic dielectrics. Thin-
film polymer dielectric layers in conjunction with thin-film copper conduc-
tors offer the greatest electrical performance because of the low dielectric
constant of polymers and the high electrical conductivity of copper metalli-
zation obtained by sputtering or conventional plating technologies. Organic
materials considered for this application include fluoropolymers, polyole-
fins, epoxies, benzocyclobutene (BCB), and derivatives of polyimides in
PMDA-ODA, BPDA-PDA, and other systems. All these materials are
discussed in great details in Chapter 11, "Polymers in Packaging." The
ideal properties of polymers that can be considered for multilevel thin
films are listed below.

• Low intrinsic stress (multilevel structure, reliability)


• High fracture toughness and ductility (processability)
• Excellent self-adhesion and adhesion to metals (AI, Cu, Cr, Ti) and
ceramic substrates (alumina, glass-ceramic)
• Nonreactive with conductor metals (especially copper) (reliability)
• Reworkable before full cure (yield enhancement)
• Compatible with other process solvents (processability)
• Dielectric constant of 3.5 or lower (electrical performance)
• Thermally stable to 400°C with 97/3 Pb/Sn or to 230°C with 60/40
Pb/Sn on FR4 (flip-chip and pin-grid array connection)
12.4 THIN-FILM MATERIALS AND PROCESSES 1I~61

• Low moisture/solvent uptake (reliability)


• Stable viscosity and long shelf life (manufacturability)
• Formation of pinhole-free films with good planarization ability (yield
and manufacturability)
• High solid content and high film retention (process ability)
• Via formation by wet etching or laser ablation (processability)

In addition to the above properties, the residual stress developed as


a result of mismatch in thermal expansion between the polymer and the
substrate to which it is bonded should be considered in selecting the
polymer for multilayer thin-film applications. There are three factors that
determine this stress:

• Thermal-expansion mismatch
• Stress-relief properties of polymers
• Elastic modulus of polymer

A rule of thumb is to multiply the thermal-expansion mismatch with


elastic modulus and arrive at polymer selection based on lowest product
of these. PMDA-ODA polyimides, in spite of their very high thermal-
expansion coefficient (350 x 1O-7/0C) compared to polyimides based on
BPDA-PDA of thermal expansion coefficient around 30 x 1O-7fOC, reflect
residual stresses that are only about five times more than the low-expansion
material. The majority of these differences can be attributed to the differ-
ence in elastic modulus between the two materials.
Referring to Figure 12-19, which shows the thermal expansion of
a typical polyimide as a function of temperature up to the chip-joining
temperature, it is not surprising that such a material with thermal coefficient
of expansion around 450 x 1O-7/oC (room temperature to 400°C) develops
so much stress when bonded to silicon, glass-ceramic, or even alumina
substrate that either the substrate warps to an unacceptable level or the
polyimide film peels off or cracks. Observations supporting these, particu-
larly with thick polyimides, have been reported [44]. A new multilevel thin-
film technology based on the use of low-thermal-expansion polyimides has
been reported by Hitachi. Polyimides of rodlike backbone structures with
thermal expansion coefficients as low as 30 x 1O-7/oC were developed in
recent years. Two such polymers coded PIQ-LlOO and PIQ-Ll20 are
illustrated for their thermal expansion behavior (Fig. 12-20). Their low
thermal expansion is attributed to the restraining of thermal expansion
by rodlike crystallized molecules within intermolecular spaces, somewhat
similar to the effect of glass fibers in fiber-reinforced plastics. If the lower
11-662 THIN-FILM PACKAGING

20

18

16

-
~

S
14

c: 12 PIO
0
·iii
c: 10
to
Q.
X
w
iii
8
...CD
E
6
.s::.
I-
4 Heating Rate: 5°C/min

0 100 200 300 400


Temperature (0C)

Figure 12-19. Thermal Expansion of Conventional Polyimide. After Ref. 44.

thermal expansion coefficient is attributed to this mechanism, one would


expect these low-thermal-expansion films to have a higher modulus in
the same direction. Figure 12-21 illustrating the relation between the
thermal expansion coefficient and modulus for various poly imide films
confirms this proposed mechanism [45].
Deposition of dielectric polymers is usually done by solution-
casting, in which a solution of polymer precursor is coated onto the
substrate, e.g., by spray or spin coating, and then baked and thermally
cured. The two later operations drive off the solvent and cross-link the
polymer. Because of the leveling of the polymer precursor solution, the
topography of the underlying conductor lines and vias is lessened at
the top surface of the cast solution. Despite the fact that the good planariza-
tion of the freshly deposited polymer solution is partly lost after baking
and curing, partial planarization is still obtained, which is desirable in
the construction of a multilayer structure. Such planarization provides a
base for the subsequent buildup of multilayer high-density structures
without topography problems. Accumulation of topography will be eventu-
ally large enough to prevent further buildup of the structure, and a process
of planarization step is necessary when more layers are required. Lamina-
12.4 THIN-FILM MATERIALS AND PROCESSES 11-663

PIQ-L100
4

"~O°C

300°C
: I 350°C

.
,
/
;'
/
/ 4OO"C

o 100 200 300 400 500


Temperature (OC)

Figure 12-20. Low-Thermal-Expansion Polyimide Films Cured Under Vari-


ous Conditions. After Ref. 44.

-
,..
~
I
7
It)
I
0
..-
~
c::
·us0c::
as
0-
x
W
Cii
...E 3

--
Q)
s=
~
0 2
cQ)::
'0
:E
Q)
0
()
0
Elastic Modulus (x10 10 dyn/cm2)

Figure 12-21. Relationship Between Thermal Coefficient of Expansion and


Elastic Modulus
11-664 THIN-FILM PACKAGING

tion of the dielectric polymer dry film is another possible approach but
is less extendible dimensionally and more prone to defects. Generally, it
would not be considered a true thin-film process.
Additional information on the use of polymers for MCM-D is given
in Chapter 11, "Polymers in Packaging."

12.4.2.2 Inorganic Materials and Their Properties


Inorganic materials are generally deposited by one of the three fol-
lowing processes:

l. Thick-film processes (spraying, screening, and sedimentation)


2. Sputtering
3. Chemical vapor deposition

Spraying and screening processes are generally limited to about


15-25 ~m thickness and are similar to the thick-film processes described
in Chapter 9, "Ceramic Packaging." Sedimentation illustrated in Figure
12-22 requires the formation and dispersion of submicron particles in a
suitable organic solvent, followed by centrifuging to obtain thin and
compact films. Chemical vapor deposition involves formation of volatile
compound of the desired metal followed by oxidation of the metal to form
the ceramic. Three examples of inorganic dielectrics [46] are illustrated in
Table 12-9. The most common of these is Si02 + AI, as practiced by
nCHIP described in Section 12.7.
One of the unique processes for the formation of inorganic dielectric
films involves the development of photosensitive materials that can be
mixed with dielectric slurries. A thin-film process based on this unique
development, shown in Figure 12-23, involves formation of Au or other
conductor lines by one of the three metallization processes described later
in this chapter, followed by screen printing and drying of a photosensitive
inorganic dielectric which is exposed to ultraviolet (UV) light through a
mask. The layer is subsequently developed, washing off the undesirable
dielectric, and the dielectric is then fired. It is reasonable to think that
such a concept could be applied to thick-film process as well, enhancing
thick-film technology into the arena of thin-film technology, as discussed
in Chapter 9, "Ceramic Packaging."
Planar structures of multiple thin films of inorganic dielectrics could
be fabricated this way but involves planarization by chemical, mechanical,
or a combination of planarization processes. Such a process, developed
in the early seventies [47], involved slurry deposition of borosilicate glass,
the thermal expansion of which matches that of alumina substrate, but
with the lowest dielectric constant possible (4.2) for that thermal expan-
sion. Such a material is processed in an atmosphere of N2, followed by
12.4 THIN-FILM MATERIALS AND PROCESSES 11-665

Centrifuging
Vial
Before After
Colloidal
Suspension
Particles .... 0.1 fJ)
Sedimented
Glass

Silicon Wafer

..
.. .
l
~
GlaSS
Film - - { 1
SUbstrate----1 1
I
1---25,..m~
I Small
Scale
I
I--- 25f.1m-
I
Rough Smooth

( }:u:::.t.-J=
Glass

~
I
1---25 m m -
I Large
Scale
I
1---25 mm~
I
Uniform But Rough Smooth But Non-Uniform

Low Dielectric High Dielectric


Constant Media Constant Media

Figure 12-22. Sedimentation Process for Deposition of Glass Films

H2 up to 50 a C above its softening temperature, and then switching gas


to N2 at that temperature [47]. Bubble-free films of 20-50 ~m thickness
resulted due to out-diffusion of H2 when its partial pressure became zero
in N2 atmosphere only at temperatures slightly above the softening point
of glass, corresponding to viscosity of 106 Poises. Such a process, illus-
trated in Figure 12-24, involved deposition of the capture pad of chromium-
copper-chromium, followed by the plate-up of copper stud prior to deposi-
tion and firing of glass at about 800 a e. The structure was mechanically
planarized after every glass firing. A total of five glass and five copper
metallization layers have been formed this way. The major disadvantage
of this approach using glass, however, is that every time the new glass layer
is fired, the previous layers underwent softening resulting in movement of
lines and generation of bubbles even at temperatures only slightly above
11-666 THIN-FILM PACKAGING

Table 12-9. Inorganic Thin-Film Dielectrics and Processes

Insulator AIN + Cu BN+Au SiO~ + Al

No. of layers 5 5 3
Material AIN BN Si02
Thickness 0.5 Ilm/Iayer 0.2 Ilm/Iayer 21lm/Iayer
Via hole size 20 11m 20 11m 5-10 11m
Deposition method Sputter CVD PECVD

Conductor

Material Cu Au Al
Linewidth 20 11m 20 11m 10-20 11m

softening temperature. These problems are largely solved with glass-


ceramics available currently. The major advantages, nevertheless, include
total thermal compatibility of a materials system from alumina to glass
as well as the thermal stability allowing joining processes up to 800°e.
The shortcomings of vitreous glass are in fact solved by the use of
crystallized glass in obtaining low-cost multilevel multichip module [48].
The process sequence in building this type of structure is illustrated in
Figure 12-25 wherein thin copper conductors are formed by screen printing
and firing, followed by nitrogen-fireable crystallizing glass with through
holes by screening. The surface smoothness of the glass-ceramic coating
is accomplished by a thin coating of vitreous glass. Subsequently, thin
Ta2NINiCrlAu! films were deposited, patterned, and the wirebond pads
were plated with additional gold.

12.4.3 Thin-Film Processes

A variety of methods have been used or are being developed to


fabricate thin-film interconnections. As mentioned at the beginning of
this section, the fabrication of thin-film packages involve several basic
steps which are summarized again in Figure 12-26. It is noted in this
table that patterning is a process step which is required for both dielectric
and conductor layers. Multichip module (MCM) fabrication is assumed
in this table and most of the following discussions, recognizing the fact
that MCM packaging presents the most challenges for thin-film packaging
technologies, and is the most important application. Nevertheless, the
principles also apply to single-chip packaging. Furthermore, for each of
the processing steps listed in Figure 12-26, a variety of methods are
12.4 THIN-FILM MATERIALS AND PROCESSES 11-667

----;--
Au

L.-_ _ _ _ _ ~
S_u_bs_t_ra_te_ _ __ ___J Thin Film Wiring (AI20a>

;:; - Screen Printing

UV Light

CC::lI-==::::'-_=:::l--C::::::l-C=:J::J - Expose

Solvent Spray

~ ~ I~ ~ ~ ~ ~ ~ ~ ~ ~ ~- Develop

~/V;.HOI.__'
~'"~

1\....--
_ _ _ _I} ----1 1--FIri~
Figure 12-23. Photosensitive Thin-Film Dielectric Process (NEC). Source:
MES '87

available and being used, as listed in Figures 12-27 through 12-31. The
majority of the techniques involved in thin-film package fabrication as
shown in these tables will be discussed in length here. The technological
choice is a very complicated issue, and the optimization of fabrication
process depends on many factors, such as intended application, design
rules, materials selection, manufacturing establishment, and costs, and
varies from time to time, and from manufacturer to manufacturer.
11-668 THIN-FILM PACKAGING

'--------------'~ultiiaYer
Thin Film Capture
Pad Deposition

Ceramic

!
Copper Stud
Plate Up

Repeat
Process

Glass Firing

Mechanical
Planarization

Figure 12-24. Sequential Multilayer Thin-Film Glass/Copper Process. After


Ref. 47.

As examples, Table 12-10 compares the most commonly used tech-


niques for the definition of vias, conductors, and bonding pads [49] in
terms of process flow. The complexity of thin-film package processing
is readily visible. The inherently sequential and in situ nature of thin-
film package fabrication will become more evident in the discussion
that follows.
12.4 THIN-FILM MATERIALS AND PROCESSES 11-669

(1) Copper Thick Film 96% Alumina Substrate


Conductor Line Wa?iWmpa + F?7?7!mZ:!
Formation

(2) Crystalline Glass


Layer Formation

(3) Repeat (1) ~:222i?i<~ I


(4) Repeat (2)

~
(5) Vitreous Glass
Layer Formation

(6) Thin Film Formation

(7) Thin Film Patterning

~ II ~
(8) Protective Resin
(Thin Film) Coating

(9) Au Plating

~i" ~
Figure 12-25. Fabrication of Inorganic Thickffhin-Film Multilayer Sub-
strate. Courtesy of Fujitsu Ltd.

Figure 12-26. Generic MCM Fabrication Processes


11-670 THIN-FILM PACKAGING

Palleming

~
(rtf'n Photo-
Printing lithography

I Photoresist I Resist A si.tw


------.--J1
- : :.iOO

-=::J Oepo ilion

Figure 12-27. Generic Patterning Processes in MCM Fabrication

I Full
Thickness

Figure 12-28. Generic Metal- or Resistive-Layer-Deposition Processes in


MCM Fabrication

Form
Linu

,-_:_t:_~__."lf__~_t7_h_--,I[,-__~_i~_. _--''-''';;';;'==_..J[bll.ion J
Figure 12-29. Generic Conductor-Line-Formation Processes in MCM Fabri-
cation
12.4 THIN·FILM MATERIALS AND PROCESSES 11-671

Deposit
Insulator

Figure 12-30. Generic Insulator-Deposition Processes in MCM Fabrication

In general, thin-film processes are compared on the basis of

• Number of process steps


• Process complexity
• Process robustness
• Process window
• Defect levels
• Planarization requirements
• Reworkability
• Repairability

12.4.3.1 Polymer Processing


12.4.3.1.1 Deposition of Polymer Thin Films
In general, there are probably more than 20 different ways to deposit
thin-film polymers, but this section will only review those that are most
prevalent today or expected to be in the future for thin-film packaging,
which are summarized below (Table 12-11). Spin coating is currently the
standard method for both photoresists and dielectric polymer coating.
Spray coating has been borrowed from the painting industry and finds
wide application in microelectronic parts coating. Roller coating, meniscus
coating, screen printing and extrusion coating are emerging coating meth-

Figure 12-31. Generic Via-Formation Processes in MCM Fabrication


11-672 THIN-FILM PACKAGING

Table 12-10. Comparison of Processes Used for the Definition of Vias,


Conductor Lines and Terminal Metals in MCM-D

Via Definition

Laser Ablation PSPI Lithography RIE Wet Etch


- --
Surface preparation Surface preparation Surface Surface preparation
Apply adhesion layer Apply adhesion layer preparation Apply adhesion layer
Apply PI Apply PSPI Apply adhesion Apply PI
Cure PI Expose features layer Apply photoresist
Laser ablate features Develop features Apply PI Expose features
Plasma clean Cure PSPI Cure PI Develop features
Plasma clean Apply photoresist Wet etch
Expose features Photoresist strip
Develop features Cure PI
RIE
Photoresist strip
Residue removal

WiringlTerminal Metal Definition

Additive Electroplating Lift-Off Evaporation Through


Subtractive Etching (Wiring and Terminal (Terminal Metals Mask
(Wiring Only) Metals) Only) (Terminal Metals Only)

Surface preparation Surface preparation Surface Surface preparation


Sputter Cr-Cu-Cr or Al Deposit plating seed preparation Mask attach/align
Apply photoresist Apply photoresist Apply adhesion Evaporate terminal
Expose features Expose features layer metals
Develop features Develop features Apply soluble Mask removal
Etch features Plate wiring or terminal polymer
Photoresist strip metals Apply RIB barrier
Photoresist strip Apply photoresist
Remove seed layer Expose features
Apply capping layer Develop features
RIE
Evaporate
terminal
metals
Lift off stencil

ods, particularly for large-area-processing-based thin-film package fabri-


cation. Both spin coating and spray coating have been employed for thin-
film deposition in MCM-D fabrication, although there is no up-to-date
literature on the application of the other coating technologies in this area.
Spin Coating: This method has been used for decades in the semicon-
ductor industry to coat photoresists and dielectric polymers on the tradi-
12.4 THIN-FILM MATERIALS AND PROCESSES 11-673

Table 12-11. Deposition Technologies of Thin-Film Polymers

Film
Coating Thickness
Method Range Advantages Limitations

Spin < 15 11m Excellent thickness control and High waste, favors small
reproducibility, good round wafers
uniformity, standard IC fab
process
Spray > 10 11m High throughput, high material Limited planarization, poor
usage, capable of large-area control of thickness and
coating uniformity, poor
reproducibility
Roller High throughput, high material Susceptible to
usage, large-scale continuous contamination and defects,
process poor quality when film is
thin
Meniscus 10-40 11m High material usage, high Susceptible to
throughput, large scale contamination and defects,
continuous process poor quality when film is
thin, not suitable for high
viscosity solutions
Screen > 5 11m Good control of thickness, Poor uniformity when film
printing good reproducibility, capable is thin, batch process,
of patterning susceptible to defects
Extrusion sub-11m to Almost no material waste, good Susceptible to
100 11m control of film thickness, good contamination, defects due
uniformity, large-area coating, to slit inhomogeneity
high throughput

tional round wafers. For dielectric polymer thin-film coating, a liquid


solution of a polymer is puddled onto the wafer surface, usually with a
volumetric dispenser, and the wafer is then spun to produce a uniform
coating. Excess polymer solution is often dispensed onto the substrate to
assure good coverage. Film thickness depends on many variables, such
as solution viscosity and solid content, the time and angular speed of
spinning, and the dispense volume. The most sensitive parameters for
controlling film thickness are the spin speed and viscosity. Film thicknesses
from 1 to 200 !lm are feasible. Thicker films can be obtained with high-
viscosity solutions at low spin speed for a short spin time, which adversely
affects the resultant film uniformity. Therefore, multiple coatings are
preferred for film thicknesses> 15 !lm, with the additional advantage of
reduced defect rate, notably, pinholes. Spin coating is a simple process and
a well-established practice with automated spin coaters widely available. A
11-674 THIN-FILM PACKAGING

large data source exists and the spin process has been modeled and
remodeled widely in the literature.
Spray Coating: The principle behind spray coating is as following:
A nontoxic gas (air or N2) is forced out of a nozzle to atomize a spray
of polymer solution which is swept across the surface of the substrate.
Tiny polymer droplets bond to one another as they hit the surface and
form a film. In one operation, the nozzle may pass several times over the
same spot on the substrate to assure good coverage and uniformity. Process
variables include solution viscosity and solid content, solution flow rate,
nozzle size and distance from substrate, nozzle pressure, atomization
pressure, and sweep speed. It is important to carefully balance between
low viscosity, which facilitates the atomization of the polymer solution,
and solid content, which determine the flow rate of the solution through
the spray nozzle to achieve the desired thickness of polymer on the
substrate [50]. However, too Iowa viscosity in combination with high
surface tension may cause pullback of the polymer solution deposited on
the substrate. The atomization pressure is also critical; too low results in
large droplets coming out of the nozzle which may not flow together to
yield an uniform coating, whereas too high causes excessive foaming of
the solution and bubbles which may still be present after the drying circle.
Spray coating is, in principle, considered a favorable alternative for a
thicker film coating (> 15 /lm) owing to its capability of large-area coating
on a noncircular substrate and high throughput [51]. Uniformity of spray-
coated film drops off when film thickness is less than 10 /lm, as found
for benzocyclobutene (BCB) [28]. It was also found that spray coated
BCB exhibited an undesirable orange peel surface finish under certain
conditions [52]. Few detailed processes of spray-coating have been re-
leased despite some reported applications [33,53]. Thorough characteriza-
tion of spray-coated thin film and much optimization of the process will
be required for spray coating to become a viable coating technique in
thin-film packaging.
Roller Coating: The basic idea is to coat one or two rollers with
the polymer solution and then roll them across the surface of the substrate.
In practice, the substrate is usually dragged underneath the moving roller.
One roller will coat one side of the substrate and two rollers can place
polymer on both sides of the substrate simultaneously. Film thickness is
primarily dependent on the polymer solution properties, such as viscosity,
as with spray coating. At this time, an exact thickness range is unavailable.
Figure 12-32 depicts this process.
Roller coating is widely used in MCM-L fabrication where one or
two wiring layers are placed on each surface of the board. Large-area
boards (20 in. by 24 in.) may be roller-coated. Major concerns with roller
coating include chronic defect problems and process sensitivity resulting
in yield loss.
12.4 THIN-FILM MATERIALS AND PROCESSES 11-675

Polymer
---.r--_...oi1....- ..~----,
Coating

Substrate

Figure 12-32. A Schematic of Roller Coating

Meniscus Coating: Liquid polymer solution is pumped out of a


porous tube over which the substrate slides. Material may be collected
under the tube and recirculated into the center of the tube, so it is much
more efficient in terms of waste than spin coating. Process parameters
include distance from substrate to tube, substrate velocity, surface tension
of the coating polymer solution and the substrate surface, viscosity and
solvent evaporation rate. Film thickness depends mainly on substrate
speed and ranges from 15 to 40 /lm. Uniformity across large boards
(300 mm by 300 mm) was between 2% and 4% for testing done with
BCB at Specialty Coating Systems, Inc. [54]. Figure 12-33 is a schematic
of meniscus coating [55].
Screen Printing: This method features deposition of the polymer
through a screen. A screen is placed on or very close to the surface of
the substrate. Liquid polymer is flowed across the screen and a squeegee
is pulled across the screen to force the liquid through the screen and onto
the substrate. An important part of the machine is called a ftoodbar. It

Stationary Substrate

Moving Coating Module

Figure 12-33. A Schematic of Meniscus Coating


11-676 THIN-FILM PACKAGING

pulls the material back onto the screen after the squeegee pushes most
of the excess off of the screen and out of range. The major factor determin-
ing film thickness and quality is the screen itself. Screen parameters
include mesh density and screen emulsion. Film thickness down to 5 Jlm
is possible with variations of ± 1.5 Jlm under ideal process conditions.
Extrusion Coating: The definition of extrusion is the process of
converting a raw material into a product of specific cross section by
forcing the material through an orifice or die under controlled conditions.
The cross section may vary widely as plastic pipes and spaghetti, which
are examples of extruded products. In microelectronic applications, the
extruded cross section is a narrow rectangle (i.e., a thin, paperlike film
with the cross section consisting of the film thickness and substrate width).
Figure 12-34 [56] illustrates the basic idea behind thin-film extrusion. In
general, the critical output variable of an extruder being used for polymer
deposition is film thickness. It depends on the volumetric flow rate of the
polymer into the die, the substrate width and velocity (the substrate moves
under the stationary extrusion head), the distance from the head to the
substrate, and the shrink factor of the polymer in question.
Film extrusion may be done in two ways: wet or dry. In wet extrusion,
the material is softened with or dissolved in solvents. This allows the
process to be performed at a low temperature. In dry extrusion, the material
is softened and made to flow by the application of temperature and pressure
only. As with the other deposition methods previously mentioned, poly-
mers for microelectronic applications are extruded as a solution in organic
solvent; correspondingly, wet extrusion would be the appropriate method
for thin-film deposition. There are three types of extruders differentiated
by the materials feeding method: ram/cylinder, screw, and pump. The
earliest machines were all ram-type extruders, but the other versions were

Stage

Optic.IRaii

Figure 12-34. A Schematic of Extrusion Coating by F AS Technologies, Inc.


12.4 THIN-FILM MATERIALS AND PROCESSES 11-677

invented to overcome some basic problems associated with ram extrusion.


The most fundamental problem with ram-type extruders is that extrusion
is not a continuous process. Periodically, the process must be halted so
that the cylinder can be refilled. Thus, the screw principle of polymer
feeding was adopted and is still the dominant type in the plastic-forming
industry today. Screw extruders are not considered suitable for thin-film
coating in microelectronics applications, however, due to the propensity
of polymer contamination and other inadequacies. A gas-driven diaphragm
has been the standard fluid-pumping technology for extruders in the micro-
electronics industry for more than 30 years. The dispensing rate and
volume by this pumping mechanism are very sensitive to the polymer
fluid viscosity and the inherent system impedance, and, as a consequence,
reproducibility has been proven to be a tough task.

12.4.3.1.2 Via Generation Approaches


Vias in thin-film package substrates are used to provide vertical
electrical connections between adjacent layers. Despite the many process
options available for via generation, including punching, mechanical dril-
ling, laser ablation, and chemical etching, only four methods as listed in
Table 12-10 are suitable for MCM-D substrate fabrication. Small via size
and the high requirement on alignment primarily determined the method
of choice. The very high routing density achievable with the MCM-D
substrate is due as much to its fine-line capabilities as to its ability to
very precisely locate small via openings in layers of polymer dielectric.
In Figure 12-35, a correlation between MCM-D and MCM-C density is
shown for MCM-D linewidth and spacing combinations of 25 on 75 11m
and 12 on 25 11m versus the number of cofired MCM-C layers needed
to achieve the same interconnect density [27]. Based on these estimates,
the use of thin-film wiring can reduce the number of interconnect layers
required for a given application by as much as 4 : 1 over that achievable
with a cofired-ceramic substrate. Kambe et al. [57] has compared state-
of-the-art PWB, high-temperature cofired ceramic, and MCM-D techno-
logies, and has come to the same conclusion (Table 12-12).
Specific via requirements are typically determined by the metalliza-
tion process that is used in fabrication. For those technologies where
vapor-deposited (sputtered or evaporated) metallization schemes are used,
the highly directional nature of these metal-deposition techniques dictates
the need for sloped via sidewalls in order to ensure adequate coverage.
On the other hand, the need for sloped sidewalls can result in significant
enlargement in upper via dimensions due to the thick layers (8 to 25 11m)
of dielectric commonly used in MCM-D structures resulting in significant
loss of routing density. Figure 12-36 shows the difference in via sidewall
angles as a result of isotropic and anisotropic via etching. Table 12-13
11-678 THIN-FILM PACKAGING

25n5 Microns 12/25 Microns


12
35 140

30 120

Thin-Film 25 • 100
Cofired
Layers Layers
MCM·D 20 80 MCM-C

15 60

10 40

5 20

10 20 30 40 50 60

GateS/Chip in Thousands

Figure 12·35. Comparison of MCM·D and MCM·C Substrate Density Capa·


bilities

demonstrates the effect of decreasing via sidewall angle and increasing


dielectric thickness on the resultant upper via dimension for a nominal
10 !..lm via opening. In the case of an anisotropic via etch process, the
upper via dimension (W2) would be equal to the lower via dimension (WI).
As the sidewall angle decreases and/or the dielectric thickness increases

Table 12·12. Comparison of Wiring Densities of MCM·D Interconnects


Achievable with MCM·C and MCM·L

Printed-Wiring Board Alumina Multilayer Polyimide Multilayer

Linewidth (mm) 0.100 0.100 0.D25


Line pitch (mm) 0.250 0.200 0.050
Via size (mm) 0.150 Dia. 0.100 Dia. 0.025 Sq.
Via pitch (mm) 0.300@ 0.250® 0.D75(£)
Dielectric constant 4.5@ 9.5@ 3.5®

Wiring density
(~r ~ 1.4 (~r ~ 16
Wiring length
®
-~0.8
©
@ ~0.3

(:r ~
@

Propagation delay
~ 1.2
©
@@
(®ya ~0.2
12.4 THIN-FILM MATERIALS AND PROCESSES 11-679

(a)

, ••••• , • I ••

420003 15KV X?00 43uru


(b)
ANISOTROPIC ETCH PROFILE
~..." ,

Figure 12·36. (a) Isotropic versus (b) Anisotropic Etch Profiles


11-680 THIN-FILM PACKAGING

Table 12-13. Thickness and Wall Angle Effect on the Upper Via Dimension

Thickness of Dielectric Layer


Sidewall Angle 10 flm 15 flm 20j.Lm

45° 30.0 11m 45.0 11m 60.0 11m


60° 21.6 11m 27.1 11m 33.1 11m
70° 17.3 Ilffi 20.9 11m 24.6 11m
90° 10.0 11m 10.0 11m 10.0 Ilffi

via enlargements of as much as three to six times the nominal dimension


can occur. Most of the commonly used MCM-D via generation processes
are typically in the 50° to 80° wall angle range. Metallization processes
which are essentially nondirectionallike electroless/electroplating of cop-
per and other metals for typical aspect ratios do not impose any constraints
on the via sidewall angle.
Via formation is widely recognized as being a significant contributor
to the overall MCM-D substrate fabrication costs due to the larger number
of processing steps generally required. As a result, a considerable amount
of effort has been devoted throughout the industry to the evaluation of
alternative via generation processes. As reported in the literature, several
of these techniques have been successfully used for via formation in
MCM-D substrates. A schematic comparison of the most commonly used
alternatives are illustrated in Figure 12-37.
For example, when IBM's metallized-ceramic-polyimide (MCP)
process was being defined in the early 1980s, wet etch was the only
available low-cost via formation technique. Reactive ion etching (RIB)
evolved to maintain compatibility with the semiconductor manufacturing
operations which were expected to be required for manufacturing multi-
layer thin films on a silicon carrier. Laser ablation was developed specifi-
cally for large substrates which required a high-yield process. Photosensi-
tive polyimide (PSPI) lithography was developed as a cost reduction to
reactive ion etching and wet etching. Based on this experience, the four
via formation processes can be compared, considering the following:

• Aspect ratio of the via


• Via sidewall profile
• Availability of tooling

It is clear from Figure 12-37 that projection laser ablation requires


the fewest process steps and has only one critical step, making it the
Inherent
RIEISLA Photosensitivity Wet Etching Projection Ablation

1) Mask Deposition

2) Photolithography 1..'
_~~_....
C>
~

3) Mask Patterning

4) Resist Stripping

5) Etching Er=19 Pr=19


E7:=i
6) Mask Stripping E r=1 9 Pr=19
Figure 12-37. Comparison of Via Generation Techniques
11-682 THIN-FILM PACKAGING

process of choice. The main drawback of the projection laser ablation


process is that laser tools and dielectric masks are not universally available.
Typically, the dielectric mask is also more expensive than the standard
chrome masks used for the other three processes. When projection laser
ablation tools and mask technology are not available, photosensitive poly-
mer lithography is preferable to wet etch and RIE. However, photosensitive
polymer materials are still evolving and are more costly than standard
polymers.
Reactive ion etching is not a preferred method for via formation
due to the large number of critical process steps. It is a preferred method
only when existing semiconductor manufacturing tools need to be used
to manufacture MCM-D to reduce the overall cost. The very steep wall
profile that can be obtained by the trilevel reactive ion etching process
is not required for multilevel thin films. Wet etch, on the other hand, is
suitable only when large vias need to be defined in very thick polymers.
In most applications for high-performance systems, the dielectric is less
than 12 !lm thick. For dielectric films less that 12-15 !lm thick, photosensi-
tive polymer lithography is preferable to wet etch because it has fewer
process steps and critical process steps as well.
Reworkability is a very important factor in the choice of the via
formation process. In both laser ablation and reactive ion etching processes,
the vias are formed after fully curing the polymer and defects caused
during the via formation process cannot be repaired or reworked easily.
In the wet etch and photosensitive polymer lithography processes, the
vias are formed prior to full cure and hence can be reworked for defects.
The large manufacturing database on laser ablation of vias, however,
suggests that very few defects are created during the process, and rework-
ability is not a critical issue for laser ablation.

Laser Ablation: Laser ablation has been widely used to drill through
holes in printed wiring board and ceramic substrates. It is also a very
competitive candidate technology for via formation in multilevel thin-
film packaging as a low-cost, high-throughput, and high-yield technology.
Minimized processing steps and almost no limitations on the polymers
to be ablated are the most prominent attributes of laser ablation via
formation, as can be seen in Table 12-10 and Figure 12-37. Cost modeling
as discussed later in this section has indicated that the lowest cost of
MCM-D substrates can be achieved by using laser ablation for via forma-
tion. Laser ablation is more environmental friendly than photosensitive
polymer technology in that no organic chemicals or solutions are employed
for via formation. Currently, high cost of ownership, complexity of tech-
nology, and equipment reliability concerns are the major impedances for
the wide acceptance of this technology. With further improvement in laser
technology and lowering of equipment cost, laser ablation would certainly
become a more attractive option for via formation.
12.4 THIN-FILM MATERIALS AND PROCESSES 11-683

The basic interaction mechanism between laser light and matter


can be of photothermal and/or photochemical nature [58]. Laser-induced
material melting and vaporization are the most common photothermal
phenomena in laser ablation, whereas plasma formation over the incident
area of the substrate is a non thermal process. With the presence of adequate
gaseous or liquid substance, more complex interactions among the laser
light, the substrate, and the environment may occur, resulting in laser
etching [59]. For via formation in polymer thin films, ultraviolet excimer
laser at wavelengths of 193 nm (ArF), 248 nm (KrF), 308 nm (XeCl), or
351 nm (XeF) is most commonly used [30,31,59-61]. A short wavelength
has the advantage of a low photothermal effect on the substrate but requires
much more complex optics and stringent optical material selection. Typi-
cally, laser ablation is carried out with a succession of high-power-density
pulses. The etching of the surface is a linear function of the number of
pulses [62]. Sufficient light absorption of the polymer at the laser wave-
length and low thermal conductivity of the polymer are necessary to
achieve high ablation efficiency. Either proper selection of the laser radia-
tion source or modification of the polymer can improve light absorption
[63]. A threshold of power density is observed below which no significant
ablation takes place. This threshold is dependent on the wavelength of
the laser and the polymer material being ablated. Direct nonthermal bond
breaking is suggested to be an important mechanism for polymer ablation
where photon energies in excess of energies for molecular dissociation
or band-gap excitation are employed.

Scanning Laser Ablation: Scanning laser ablation (SLA)-based via


generation processes have been used successfully in MCM-D [60,64] and
high-density printed-wiring-board fabrication [65]. The via generation
tool is driven by either a XeCI (308 nm) or KrF (248 nm) excimer laser.
Although contact shadow masking approaches have been successfully
used for other less demanding laser micromachining applications, the via
resolution and placement accuracy achievable is inadequate for most thin-
film applications. Consequently, as is commonly the case for RIB via
generation processes, an aluminum or copper conformal masking layer
has been used to define the via locations. The masked parts are positioned
on a computer-controlled precision x-y table and are moved in a serpentine
pattern under a stationary homogenized excimer laser beam, thereby
exposing the dielectric layer in the open areas of the mask to the ablative
power of the laser light. Unlike RIB processing, which is inherently a
batch processing technology and performed under vacuum, scanning laser
ablation processing is done under ambient conditions. Consequently, one
of the major attractions of laser via fabrication approaches is their compati-
bility with in-line substrate processing.
In establishing a robust, manufacturable SLA via generation capabil-
11-684 THIN-FILM PACKAGING

ity, a process window must be defined which allows for the ablation of
the underlying dielectric layer through the openings in the conformal
metal mask at an acceptable rate (>0.25 ~mJpulse) while maintaining a
ftuence well below the mask damage threshold. Soot accumulation from
the ablation process around the mask via openings has been found to be
an important additional factor in determining the overall robustness of
the conformal metal masks. Localized overheating of the masking layer
around the via sites can result in excessive soot redeposition. A number
of approaches including the use of a He or O 2 process gas has been shown
to significantly reduce the amount of soot generated, thereby enhancing the
mask stability. Minimum recommended mask thicknesses vary somewhat
with the masking metal used; however, thicknesses greater than 3 ~m are
generally recommended for commonly used dielectric ablation ftuences
of 250-350 mJ/cm2• Commercially available SLA tools, with a high-
powered (100-150 W) industrial excimer laser, are capable of throughputs
that are comparable with an RIE-based via generation process. The need
for a much thicker metal mask for SLA via generation than for RIE
processing does however adversely impact the processing cost and mini-
mum achievable feature size associated with the deposition and patterning
of this thicker masking layer.
Effective soot removal is required to ensure reliable adhesion of the
next metal layer. Two main cleaning processes have been used for this
application. Aqueous cleaning using a pumice or brush scrubbing process
has been used to mechanically remove any residue from the ablation
process. Alternatively, contactless, plasma-based, dry soot removal has
also been shown to be successful, as described elsewhere [61].
Both RlE- and SLA-based via processes involve a number of preetch
and postetch processing steps as demonstrated in Figure 12-37. In order
to simplify and thus cost-reduce these via generation processes, efforts
have been underway to reduce the number of required processing steps.
One such approach has been the use of thick, patterned photoresist layers
as erodible RIE or SLA etch masks [66,67]. A successful demonstration
of a photoresist-mask-based SLA via generation process used to etch
through 10 ~m BCB is shown in Figure 12-38 at various stages in the
process sequence [67].

Projection Laser Ablation: Projection laser ablation via generation


involves the use of excimer laser tools that are very similar in concept
and construction to photolithographic steppers used in semiconductor
processing. The via sites to be ablated are defined by openings in a discrete
mask, in a manner analogous to the exposure of a photoresist layer in a
stepper [68]. Changes in tool optics and mask technologies from those
used in conventional phototools are required to withstand the high-power
pulsed UV laser beam necessary to effect the ablation process. Like the
12.4 THIN-FILM MATERIALS AND PROCESSES 11-685

• approx.....Itt to BCB
aelectivlty 011.5:1.
AFTER THE SCANNING
ABLATION PROCESS

Figure 12-38. Via Generation in BCB Dielectric Using a Photoresist Abla-


tion Mask

SLA equipment described in the previous section, these projection tools are
powered by 100-150-W KrF or XeCI excimer lasers. The first commercial
application of a projection laser ablation technology was in the manufac-
ture of MCM-CID substrates for the IBM ES9000 system [61]. A pro-
prietary dielectric on quartz masking technology was developed for this
application. Figure 12-39 [61] schematically illustrates the working princi-
ple of projection laser ablation tool.
The major process variables influencing the ablation rate as well as
the via shape and sidewall angle achievable with a projection ablation
process are the fluence, the number of pulses, and the focus of the image
of the beam. The importance of the focus on the resultant via sidewall
angle was recently reported [61] and is shown in Figure 12-40. A SEM
microphotograph of typical via openings generated using a projection
ablation process is shown in Figure 12-41. The typical via sidewall angle
obtained using a projection ablation process is approximately 65° com-
pared to 75°-85° with a scanning laser ablation process.
Projection laser ablation through a mask using an excimer laser at
308 nm is a very robust process [69], achieving greater than 99% substrate
yields on thin-film structures with greater than 100,000 vias per substrate.
The wall profiles are very uniform and can be controlled from 30° to 65°
by changing the focus conditions [61]. The process is not aspect-ratio
limited for packaging applications, and almost any polymer dielectric can
be used. The carbonaceous debris formed during laser ablation is removed
11-686 THIN-FILM PACKAGING

LaserUV
XeCI308 nm ~ ........ Homogenizer

~7~:,;j Field Lens ~


Mask n

Tool Control

[]

Figure 12-39. A Schematic of Projection Laser Ablation Tool

by plasma cleaning. Defects due to residue are dramatically lower in this


process compared to the RIE process [61].
Despite the obvious advantages of projection laser ablation resulting
from the elimination of all the preetch and postetch processing steps
associated with RIE and SLA processing, until recently, more widespread
use of this via generation technology has been hampered by two major
factors. First, the high application, capital equipment costs and dedicated
functionality of such an ablation tool rendered its use virtually impractical
for all but very high-volume MCM-D production as is the case at IBM.
Additionally, the unavailability of a commercial source for a proven
projection ablation masking technology further discouraged usage of this
technology. More recently, efforts initiated at IBM and elsewhere to
address these fundamental barriers are expected to broaden the user base
of this via generation technology throughout the industry in the future.

Reactive Ion Etching: To date, throughout the industry, reactive


ion etching (RlE) has been by far the most widely used via formation
technology. Several examples of the successful use of RlE processing for
via generation in MCM-D substrate technologies have appeared in the
literature [70,71]. For the most part, this preference has been based on
12.4 THIN-FILM MATERIALS AND PROCESSES 11-687

vtII AMIIIM bough 10 "'" poIytmlde

Figure 12-40. Effect of Focus on the Resultant Sidewall Angle Achieved by


Projection Ablation (Courtesy of Motorola)

PASSIVAnON OPENINGS 90 jIIII OPENINGS In POLYIMIDE


ESgooo TCM CHIP SITE

Figure 12-41. Projection Ablation of Vias in 12-....m-Thick Polyimide (Courtesy


of Motorola)
11-688 THIN-FILM PACKAGING

the evolution of these via etch processes from long-standing semiconductor


fabrication techniques and the associated ready availability of suitable
parallel-plate RIB tools. An 02-rich, fluorinated gas (e.g., CF4, CHF3, or
SF6) mixture has been commonly used in these processes, thereby provid-
ing for etch rates well in excess of 1 /lmlmin. Gas mixture, power, and
pressure are all interactive variables that must be optimized according to
the desired via sidewall angle, mask undercutting, and etch rate. Typical
via etch results are shown in the SEM microphotographs in Figure 12-42.
Reactive-ion-etching via processing requires a conformal masking
layer to define the via sites to be etched. As is seen by the schematic of
a typical RIE process flow in Figure 12-37, the bulk of the processing
steps in this process are for mask deposition and removal. Because the
ultimate resolution capabilities of a RIE-based via generation process
depend on the resolution achievable in the patterning of the conformal
masking layer, a high etch selectivity to the polymer dielectric layer is
desirable so as to minimize the thickness of the masking layer required.
Several alternative RIE masking layers have been successfully used and
fall into either erodible or nonerodible etch mask categories. Erodible
etch masks including Si02, Si 3N4, and photoresist layers are typically
etched at an appreciable rate under the same etch conditions used to etch
the via openings in the underlying dielectric layer. Because most MCM-

Figure 12-42. SEM Photomicrographs of RIE Generated Via


12.4 THIN-FILM MATERIALS AND PROCESSES 11-689

D technologies involve the use of 8-25-~m-thick dielectric layers, non-


erodible, usually metal masking layers have been most commonly used
[70,71]. Because these metal masks are not eroded by the dielectric etching
process, they can be very thin and are typically in the 3000-5000-A-thick
range. Unlike wet etching, the polymer is reactive-ion-etched after full
cure or complete polymerization.
One example of an erodible etch mask system is a simple image
transfer through a positive photoresist stencil. To ensure that all the vias
are open with no residue and to improve process robustness, the polymer
is overetched. To compensate for this overetch, the photoresist is made
slightly thicker than the polymer, as the etch rates of polymer and photo-
resist are nearly the same. The positive photoresist is reftowed prior to
reactive ion etching to obtain a tapered wall profile. After etching, organic
photoresist strippers are used instead of aqueous strippers to remove the
photoresist, because the photoresist is UV hardened during reactive ion
etching. Bilevel processes work best for polymer thicknesses below 5 ~m.
To increase the range of polymer thickness that can be etched and
to improve the wall profile, a trilevel stencil can be used [72]. In this
process, a silicon-containing organometallic layer (RIE barrier) is depos-
ited by chemical vapor deposition between the polymer and the photoresist.
The image is transferred from the photoresist to the organometallic layer
by CF4 plasma etching. An oxygen plasma, which does not etch the RIE
barrier, is subsequently used to transfer the image to the polymer. The
RIE barrier is then removed after the vias are defined. The wall angle
can be tailored by varying the gaseous composition. Unlike the bilevel
process, the photoresist thickness is not critical and is used only to transfer
the image to the RIE barrier. The process window is wider for this process
and the choice of the polymer dielectric is not restricted.
Backsputtering during reactive ion etching gives rise to redeposition
of metallic species in the vias, causing increased defect levels and increased
via resistance [73]. The reactive ion etching residue can be cleaned by
additional wet or dry processes. The use of low-power reactive ion etching
reduces the amount of backsputtering but also decreases the throughput
of the process.

ChemicallWet Processing of Dielectrics: Chemical or wet processing


of dielectrics involves the use of aqueous or organic solvents to etch vias
in MCM-D dielectric layers. Despite the more limited via design rule
capabilities of these "wet" via processing alternatives over those achiev-
able with the dry etch processes described earlier, the process simplicity
and inherent lower capital costs of such approaches make them particularly
attractive for those more cost-sensitive MCM-D applications. To date,
these approaches have fallen into one of two major material/process cate-
11-690 THIN-FILM PACKAGING

gories, namely inherently photosensitive dielectrics and wet etchable mate-


rials.

Photosensitive Polyimide Lithography: Of the inherently photosen-


sitive dielectric materials available, negative-acting photosensitive poly-
imides (PSPI) have been by far the most commonly used. The limited
resolution and resolvable via aspect ratios achievable with photosensitive
polyimides due to the shrinkage (in excess of 50%) that occurs during
post-via etch hard baking as well as their generally poor shelf life and
stability have somewhat limited their acceptance for MCM fabrication.
However, as amply highlighted in Figure 12-37, the process sequence
associated with the use of an inherently photosensitive dielectric material
is considerably simpler and less capital equipment intensive than the other
via generation options. With the recent availability of a wide range of
photosensitive polyimides and others, this method of via fabrication has
become more attractive. The number of process steps required to form
vias is reduced due to the use of a unilevel stencil. The advantages of
this process have been well documented [74]. The chemistry of PSPI is
complicated because it has to meet all electrical, mechanical and litho-
graphic requirements. The PSPI most commonly used is a negative tone
resist which cross-links upon exposure to UV radiation and becomes
insoluble. The unexposed area is dissolved in an organic-based developer.
Photosensitive polyimides are formulated by adding sensitizers, initi-
ators, and other photoactive components to the base polyimide. Upon
curing, these photocomponents and the condensation reaction products
are volatilized, leading to a shrinkage of about 50%. This shrinkage can
be reduced by using preimidized PSPIs, but these PSPIs have a narrower
process window and, in most cases, inferior mechanical, thermal, or sol-
vent-resistance properties. The PSPI lithography process is most cost-
effective compared to wet etch and RIE even though the dielectric material
is more expensive [75]. The process window is very wide, and PSPI
lithography has better resolution and can accommodate higher aspect
ratios compared to wet etch. As a typical example, 8-/lm vias in 6-/lm-
thick cured PSPIs have been defined using this process.
Unlike laser ablated vias, the wall profile of vias formed by PSPI
lithography is not trapezoidal in shape. The PSPI via wall profile is steeper
in the top portion of the via with a very shallow foot. Because the PSPI
is cured after the via is defined, a cusp or lip is formed at the top. The top
dimension of the via for a particular bottom dimension can be controlled by
changing the lithography parameters [76]. Also, the adhesion promoter
that is used between dielectric layers is not required because most PSPI
materials are formulated with an adhesion promoter, thereby eliminating
the need for doing this in a separate step.
Examples of the successful use of photosensitive polyimides in the
12.4 THIN-FILM MATERIALS AND PROCESSES 11-691

fabrication ofMCM substrates have appeared in the literature [34,77-79]


with the bulk of this activity occurring in Japan. Figure 12-43 is a typical
via generated in a photosensitive polyimide. In recent years, reports on
via formation in a photosensitive benzocydobutene (BCB) for use in
MCM-D substrate fabrication has also appeared in the literature [80].

Wet Etch: Wet etching of conventional nonphotosensitive polyim-


ides have also attracted interest as the industry strived to reduce MCM-D
substrate processing complexity and cost. With this type of via generation
approach, a relatively thin (1-5 flm thick) patterned photoresist layer
serves as the mask template to define the sites of via etching in an
underlying dielectric layer, typically when the material is still partially
imidized. Because the via photodefinition process involves the use of a
positive acting photoresist layer, such a process tends to be significantly
less susceptible to particulate-based defects than the negative-acting inher-
ently photosensitive dielectric materials mentioned previously. Prelimi-
nary results describing the potential of wet-etch-based via generation
processes have been reported in the literature in the past few years [81-83] .
The vias are developed by dissolving the exposed areas in an alkaline
solution. Process reproducibility is controlled by monitoring the degree
of imidization prior to wet etch. For a given bake temperature the degree

0(; I) (194 15KV Hoo ' .... 501.111


Figure 12-43. Vias Generated in a Photosensitive Polyimide
11-692 THIN-FILM PACKAGING

of imidization is also a function of the polyimide thickness [84] . Due to


the isotropic nature of the wet etch process and the chemistry involved,
the wet etch process gives a very shallow wall profile. Because the final
curing of the polyimide is done after via definition, shrinkage of the
polyimide will add to the increase in the top via dimension compared to
the bottom. An SEM microphotograph of typical vias obtained with this
type of process is shown in Figure 12-44.
A 2 : 1 aspect ratio (via bottom dimension: thickness) is preferred
for a robust manufacturable process. The minimum feature dimension is
recommended to be greater than 15 !lm. The wet etch process limits the
choice of polyimides for use as the dielectric. The tooling is simple because
it is an aqueous-based process. Defect levels are controlled by (1) using
a negative tone resist to reduce reaction with the underlying polyimide
during stripping of the photoresist after wet etch, (2) implementing a
two-pass photolithography process, which involves two coats of resist
application to eliminate any pinholes in the photoresist prior to polyimide
etching, and (3) introducing a via cleaning process after stripping of
the photoresist.

Via Generation Summary: As is clear from this section, a number


of via generation alternatives exist for thin-film substrate processing. All

Figure 12-44. Via Generated by Wet Etching in Amoco 4212 Polyimide Film
12.4 THIN-FILM MATERIALS AND PROCESSES 1I~93

of these processes have their own advantages and disadvantages. The


simple processing sequences such as the wet etching processes tend to
provide the lowest-cost solution but at the expense of minimum achiev-
able resolution.

12.4.3.2 Metallization Processes

A number of metallization processes have been used for MCM-D


substrate fabrication, with the most common approaches summarized
previously in Table 12-10 and illustrated schematically in Figure 12-45.
Owing to the origins of this multichip module substrate approach in
semiconductor processing, sputter-deposited aluminum metallization has
been the most widely used alternative mentioned in the literature. Addition-
ally, because high-resolution subtractive patterning of 2-4llm-thick sput-
tered aluminum layers down to feature sizes as small as 8-10 Ilm is
achievable with good reproducibility and minimal lateral etching, several
examples of MCM-D substrates based on sputtered aluminum metalliza-
tion have been reported in the literature [85,86]. As MCM-D technologies
have continued to evolve, a general desire to use lower-resistance copper
conductors has developed. This general trend has resulted in the need for
more complex metallization for the tendency of interaction between bare
copper and polyamic acid precursors, resulting in copper diffusion into
the dielectric and poor adhesion at the copper/polyimide interface. In
order to avoid this undesirable interaction, barrier metals, induding Cr,
Ti, and Ni, have been used [87-89]. This tends to further complicate
subtractive metallization patterning processes where the additional etching
of barrier layers is required. Additionally, the sequential etching through
upper barrier, core, and lower barrier layers results in conductors with
exposed copper on the sidewalls. Recent reports have suggested that the
use of polyimide formulations based on polyamic acid ester precursors
[90] and DSV-BCB [91] dielectrics are compatible with unprotected cop-
per conductors.
Evaporationllift-off approaches for MCM-D substrate metallization
also owe their origins in semiconductor processing and have been reported
in the literature [72,92]. In this additive approach, a thick photoresist layer
is deposited and patterned with the desired conductor pattern so as to
achieve a retrograde sidewall profile. The desired copper-based conductor
with appropriate barrier metals is then sequentially deposited by evapora-
tion onto the substrates. The carefully optimized photoresist profile to-
gether with the inherently directional nature of evaporation processes
results in a discontinuous metal deposit layer.
Semiadditive copper electroplating metallization processes have also
been used in MCM-D processing [77,92,93]. In this approach, a thin
seeding layer (3000-5000 Athick) is deposited by sputtering or electroless
11-694 THIN-FILM PACKAGING

(a) leaD
ub lrale 1:' ==== I"".
Sub trale

Apply
Metallize
Multilayer 1ii!~~~!1 CI.on
Resis1 I:: ub trait ubotntt

em
Expow
Deposit
De\'elo~ • Conducting ~~~~~
l:
Apply
Photor 1st
Top Res ..1
Laye r ted LIIyer

Elch Apply Expo eJ


Res; I • • • •• Photo ..,,;'t, Develop
oderla)'tr Expo e and Pillern
Develop

Plote Metol Ion Mill


Pallern f:::::::!:==:lor Eteb Mel al

RrmOVf

-==='=='=::1
Photoresisl trip
Lift OfT I - t:~==~~ ond ub tnoetively Photor.,,; t
PhOloreol
Etth ted Layer
(a) Lin OfT (b) Electroplate (c) ublntt"'. Mttllll izaliOD Prot

(b)

r 1

, r

Subtnclive Seml-Add/tive Fully-Additive

Figure 12-45. Conductor Deposition and Patterning. (a) Process Alternatives; (b)
Line Profiles Obtained with Each Processes.

copper seeding. The bulk of the conductor thickness is then built up


by electroplating through a patterned photoresist template. Subsequent
stripping of the resist layer followed by etching of the thin seeding layer
leaves the isolated conductor layers. Due to the significantly lower capital
costs associated with electroless/electroplated Cu metallurgy and the com-
patibility of this process with high-volume, large-format substrate panels
12.4 THIN-FILM MATERIALS AND PROCESSES 11-695

as indicated by years of use in the PWB industry, more widespread


use of this metallization approach for "low-cost" MCM-D processes is
anticipated in the future. IBM's recently announced SLC technology is
a case in point [16]. Additional information on the SLC process will be
outlined in the following section.

12.4.3.2.1 Wiring-Level Metallization


Aluminum and copper are used as the primary conductor metals for
MCM-D. Gold has been used by some MCM-D manufacturers but suffers
from high cost and poor adhesion to polyimide. Wiring patterns in alumi-
num are formed by subtractive etching. Copper conductor lines are defined
either by additive electroplating or by subtractive etching. Lift-off for
wiring-level definition is not favored because of

• Expensive tooling
• Need for RIB barrier and RIB etch stop
• Concern with repeated exposure of the thin-film structure to solvents

The critical elements in the choice of the process are as follows:

• Aspect ratio
• Tolerance requirement
• Conductor metal
• Availability of required tooling

Subtractive Etching: The wiring-level metals are deposited either


by sputtering or evaporation. Sputtering is preferred over evaporation due
to its higher throughput and its ability to conformally cover the sides of
the vias. The line thickness tolerance for this vacuum deposition process
is better than 10%. Positive resist lithography is used to define the wiring
pattern and the metal is wet-etched in aqueous-based solutions. The linew-
idth is controlled by the subtractive etching process. Due to the isotropic
nature of subtractive etching, the resist mask is sized larger than the
desired final metal feature dimensions to compensate for the undercutting
during etching (Fig. 12-46). This limits the subtractive etching process
to features with low aspect ratios. Subtractive etching of aluminum has less
undercut than copper, resulting in a better line profile and linewidth control.
Other factors that control the linewidth uniformity are the following:

• Nonuniformity in etch rate between edge and center of the substrate


• Electrochemical effects
• Topography
• Etchant chemistry
11-696 THIN-FILM PACKAGING

...
T
--_........

Pitch (P) = Linewidth (L) + Minimum


Resolution (M) + 2x Line Thickness (T) +
Undercut (U)

Figure 12-46. Undercut of Copper During Subtractive Etching

The non uniformity in etch rate between the edge and the center of
the substrate is minimized by using appropriate spin/spray etch tools. As
end-point detection is difficult, differences in the edge-to-center etch rate
is compensated by overetching. This factor becomes more important as
the size of the substrate increases. Also, as the conductor thickness in-
creases, the linewidth tolerance control becomes more challenging, even
if the aspect ratio is favorable.
Multilevel thin-film structures have been fabricated using both cop-
per and aluminum as the conductor metal. To enhance adhesion of copper
to polyimide and to prevent its interaction with pol¥imide, the copper
wiring is sandwiched between two thin layers «300 A) of Cr. Cr-Cu-Cr
is deposited in one sputtering step and the trilevel structure is subtractively
etched in two different solutions. In this process, the sidewalls of the
wiring channels are not covered with Cr. Because polyimide adheres well
to the Cr-coated top surface of the wiring channels, coating of the sidewalls
with Cr is not critical. Aluminum metallization does not require the barrier
metal because it has good adhesion to polyimide without interaction.
Copper conductors are preferred over aluminum conductors due to
its higher electrical conductivity. Aluminum is preferred to reduce cost
if a semiconductor manufacturing line is used for manufacturing MCM-
D. The primary defect type observed during subtractive etching of copper
and aluminum is shorts due to resist flaws and contamination. Overall,
subtractive etching is the lowest cost wiring definition process and is used
when the aspect ratio allows for it (Table 12-14).

Additive Electroplating: Additive electroplating is feasible for both


copper and gold wiring. High-aspect-ratio wiring is possible with additive
electroplating, and the linewidth tolerance depends on the photoresist
12.4 THIN-FILM MATERIALS AND PROCESSES 11-697

Table 12-14. Copper Additive Electroplating and Subtractive Etching


Metallization Regimes

LinewidthlPitch {I:!:m}
Thickness 12125 20/40 20/50 25n5

211m Subetch Subetch Subetch Subetch


311m Electroplating Subetch Subetch Subetch
4 11m Electroplating Electroplating Subetch Subetch
5 11m Electroplating Electroplating Subetch Subetch
6 11m Electroplating Electroplating Subetch Subetch

lithography tolerance. A photoresist developed in IBM gave better than


5% control in linewidth uniformity. The line thickness, on the other hand,
is dependent on the uniformity of the electroplating process. Uniformity
in pattern density, pattern shape, and feature size across the substrate is
critical in controlling the thickness uniformity. By optimizing the elec-
troplating tool and the current density distribution, a thickness uniformity
of 10% or better can be achieved, even when the pattern is nonuniform.
Cr-Cu is used as the blanket conductor layer (seed layer) for elec-
troplating. The seed layer is flash etched after electroplating and resist
removal. Copper features are covered (capped) with a metal or polymer
layer after electroplating; its function is similar to that of Cr in the subtrac-
tive etching process. Because the structure is capped after resist removal
and etching of the seed layer, the sidewalls of the conductor lines are
also capped. The final linewidth uniformity is a function of photoresist
lithography, flash etch, and capping uniformity. Line opens are the primary
type of defect detected in additive electroplating. Unlike subtractive etch-
ing, additive electroplating technology has no restrictions on line thickness
or substrate size.
Table 12-14 shows the preferred conductor-line definition process
for an assumed undercut of 1 : 1 for the subtractive etch process. In the
subtractive etching process, where the metal is sputtered to the final line
thickness, the metal thickness on the sidewalls of the vias will be less
than on the top surface. This might be a reliability problem for very thin
conductor layers on thick polymer dielectric layers. On the other hand,
in additive electroplating, only a thin seed layer is deposited by sputtering.
The metal thickness is uniform on both the top and the sides of the vias
due to the isotropic nature of electroplating. Depending on the reliability
requirements and the substrate size, additive electroplating may be prefera-
ble to subtractive etching even when the aspect ratio and line pitch are
favorable for subtractive etching.
11-698 THIN-FILM PACKAGING

PSPI LithographylPlanarization: Photosensitive polyimide (PSPI)


lithography is one of the ways to define the wiring channels in a planar
structure [32,94]. As was previously discussed, PSPI lithography involves
patterning by photo exposure, development in solvents, and thermal curing.
Blanket sputtering or electroplating of Cr-Cu is used to fill the channels
defined in PSPI, and mechanical planarization is used to remove the
excessive metal. The top layer of the polyimide is also removed during
planarization, which puts additional requirements on the mechanical prop-
erties of the polyimide. Variation in the density of the wiring pattern leads
to variations in the wall angle (profiles) of the PSPI after cure. This is
due to a 50% shrinkage of the PSPI thickness on curing, leading to
unconstrained shrinkage (Fig. 12-47), especially in regions where the
distance between features is smaller than the cured PSPI thickness.
To simplify this complex process, the via and wiring features were
metallized during the same metal deposition and planarization step (re-
ferred to as Dual-level metal, DLM, in IBM). The thickness tolerance of
the wiring channels is controlled by the planarization process and can
exceed ±15%. Figure 12-48 is an example of substrates fabricated by this
process [17].

12.4.3.2.2 Via Metallization


Several different via metallization processes have been successfully
used in the fabrication of MCM-D modules. In most MCM-D technologies
described to date, sidewall metallization is carried out simultaneously with

More Lateral Shrinkage in Isolated Lines

$.dn~ ~h:.' ' ~ Semi-Isolated Lines

t
Constrained Shrinkage Via
Diameter not a function of density

Figure 12-47. Effect of PSPI Shrinkage on Wiring and Via Profile


12.4 THIN-FILM MATERIALS AND PROCESSES 11-699

Figure 12·48. Micrographs of Via and Conductor Line Structures Fabricated


by Dual-Level Metal (DLM) Process

the deposition of the next metal layer, resulting in electrical connection to


the previous metal layer. A cross section of this metallized via structure
is shown in Figure 12-15. Because this type of metallization approach
generates significant topography in the immediate areas of the vias, stair-
stepped via structures are commonly used in these multilayer MCM-D
designs. Although this fact significantly affects the maximum interconnect
density achievable, most of the MCM-D prototype substrate designs de-
scribed in the literature could be accommodated with this via structure.
Several solid-metal via filling processes have been reported in the literature
to satisfy specific, more aggressive MCM design rules where a stacked
via configuration is desired/required and the significantly greater number
of processing steps involved can be justified. In these higher-density
interconnects, substrate planarization is maintained during the whole se-
quential substrate fabrication process. A schematic comparison of stair-
stepped and stacked via arrangements are shown in Figure 12-49. A
number of different via filling approaches have been used, including
electroless nickel via filling processes [92,93] (Fig. 12-50), via fill by
evaporationllift-off [72], and via post formation with subsequent dielectric
deposition and lap back [16,96] (Fig. 12-51).

12.4.3.2.3 Terminal Metals and Their Processing


Terminal metals consist of bonding pads for chip connections and
the engineering change features on top of the signal and power layers
11-700 THIN-FILM PACKAGING

(a)

(b)

Figure 12-49. A Comparison of (a) Stacked and (b) Conformal/Stair-Stepped


Via Arrangements

[97,98]. The tenninal-metal process is similar to the wiring-level process,


but the range of metals that need to be deposited is more extensive. Bulk
and surface composition of the tenninal metals is critical due to the
sensitivity of the chip connection process to defects and contamination.
The tenninal-metal stack ranges from Cr-Cu for flip-chip joining with
limited rework or reflow requirements to Cr-Cu-Ni-<1500 AAu for flip-
chip connection with multiple reflow and rework requirements. Wirebond-
ing, on the other hand, will need Cr-Cu-Ni-> 1 /lm Au. Nickel serves as
a diffusion barrier between gold and copper, thus reducing the formation
of Cu-Au intermetallics. Cobalt can also be used as the diffusion barrier.
The terminal metals can be defined by the following:

• Evaporation through a metal mask


• Additive electroplating
• Lift-off

If only Cu is needed as the terminal metal, as in the metallized-


ceramic-polyimide (MCP) process, subtractive etching is used. The termi-
nal metal process selection depends on

• Tooling availability
• Type of substrate
12.4 THIN-FILM MATERIALS AND PROCESSES 11-701

CROSS-SECTION OF RESULTANT
MULTILAYER STRUCTURE

Figure 12-50. SEM Photomicrographs of Vias Filled with an Electroless Ni


Process (Courtesy of Motorola)

• Size and shape of top surface features


• Cost.

Table 12-15 compares the process options for terminal-metal fabrication.

Evaporation Through Mask: Evaporation through a metal mask is


the simplest method of depositing the terminal metals. Unfortunately, it
is not always applicable for multilevel thin-film fabrication. The metal
mask has to conform to the substrate for intimate contact between it and
the mask. This is possible only when the substrate is flexible, like silicon.
The rigidity of cofired-ceramic substrates prohibits the use of evaporation
through a mask due to shadowing around the features. This method is
suitable for defining bonding pads but not for fine-line wiring which is
needed on the top surface for the engineering changes in certain applica-
tions.

Additive Electroplating: Additive electroplating was chosen as the


terminal-metal process for the multilevel conformal via structures due to
its low start-up and operational cost. The process sequence and limitations
1/-702 THIN·FILM PACKAGING

Polyimide coating
& planarization

Plating interconnect
deposition, resist
(b) patterning, &
copper plating

Resist strip,
(e) resist patterning,
& pillar plating

Resist strip, plating


interconnect strip, &
(d) overcoat

(e) Polylmide coating

Planarization
(f)

Figure 12-51. MCC Via Post Processing Approaches. (a) Polyimide coating (b)
Base metal deposition, resist patterning and plating of copper lines. (c) Resist strip,
base metal strip, and nickel overcoat. (c) Polyimide coating and (d) Planarization with
mechanical polishing.

are similar to that described in the wiring section except that capping of
the features is not required.

Lift-OtT: Lift-off technology is practiced for the terminal-metal


definition in the IBM ES 9000 processor. The lift-off stencil can either
be bilevel or trilevel. Because the terminal metals are evaporated, Cr, Cu,
Ni, Co, and Au can be used. The trilevellift-off stencil is comprised of
soluble PI + RIE barrier + photoresist. The features are defined in the
photoresist and the image is transferred to the RIE barrier and soluble PI
12.4 THIN-FILM MATERIALS AND PROCESSES 11-703

Table 12-15. Comparison of Terminal-Metal Process Techniques

Metallization Number of Critical Unit Choice of


Process Process SteI!s Processes Metals Limitations

Evaporation 4 Mask attach/align Cu, Ni, Co, Au No wiring features


through mask
Subtractive 7 Resist lithography Cu Only flip chip, no
etching reflows
Additive 8 Resist lithography, Cu, Ni, Co, Au
electroplating electroplating
Lift-off 12 Resist lithography, Cu, Ni, Co, Au Requires etch stop
RIE residue (3 additional
removal steps)

by reactive ion etching. After evaporation of the terminal metals, the


stencil is lifted off in an organic solvent. In this process, aspect ratios of
better than 1 : 1 are possible and both engineering change features and
bonding pads can be defined.
For the process to be robust, an RIE barrier is needed to protect
the underlying wiring and dielectric levels. This is accomplished by the
sputtering of blanket Cr-Cu-Cr prior to the stencil build. In the ES 9000
processor process, the Cr-Cu-Cr barrier doubles as a redundant metal
layer to improve the yield of the underlying wiring layer. A lithography
step is used to cover the terminal metals and the redundant features. The
rest of the blanket layer (Cr-Cu-Cr) is subtractively etched. The primary
defect type is shorts due to defects during the subtractive etching of the
RIE etch barrier.
In summary, evaporation through a metal mask is the terminal-metal
process of choice but can be implemented only on silicon substrates. Also,
engineering change features cannot be incorporated into the top surface.
Additive electroplating is a low-cost process and is the preferred process
for cofired-ceramic substrates. The lift-off process requires twice the
number of process steps compared to additive electroplating and has the
highest number of critical processes. Terminal metals by liftoff is more
expensive than either the additive electroplating or the evaporation through
mask process.

12.4.3.3 Planarization
Planarization is a very important issue for the multilevel thin-film
packaging process regardless which coating technique and via generation
and metallization processes are used. It has a great impact on both the
11-704 THIN-FILM PACKAGING

package perfonnance and the manufacturing process. Planarization is


necessary to maintain properties such as characteristic impedance which is
geometry dependent. Variations in impedance may cause signal reflections
which lead to logic errors. Geometry also affects line resistance which,
in tum, dictates power loss and heat generation. Furthennore, as lines
become narrower, the electromigration phenomenon becomes more promi-
nent and thickness variations should be minimized to reduce electromigra-
tion-related failure. Photolithographic resolution depends on the power
absorption and thickness of the polymer thin film, and variations of film
thickness will result in resolution degradation. Another issue is dielectric
breakdown. If a dielectric film is too thin at one point, the high electrical
field at this point could possibly cause a breakdown, destroying the integra-
tion of the package function. Finally, yield and reliability tend to degrade
as wiring density becomes high and feature size becomes small. In brief,
planar structure is always desirable in tenns of electrical perfonnance but,
most of the time, would greatly complicate the processing and markedly
increase the cost of packaging. Challenges of planarization are growing
as aspect ratios increase. More densely packed circuits mean narrower
lines. To keep resistance from getting too high, the cross-sectional area
is maintained by making conductor lines thick and, as a consequence, the
step height is made larger. The goal of planarization is to reduce step
height through application of some sort of polymer coating or through
mechanical polishing.
The capability of planarizing the surface topography of the underly-
ing substrate is a major processing requirement for polymer thin-film
coating. The degree of planarization (DOP) parameter was first introduced
by Rothman [99] to characterize this processing property as is shown in
Figure 12-52. The degree of planarization is defined as the fractional
reduction in surface feature topology realized by the coating; that is,

DOP = (1 -~) x 100


tAl
[12-11]

It is obvious that 100% planarization corresponds to a fully planarized


surface, whereas 0% planarization is just the opposite (i.e. a conformal
coating which serves as a 1 : 1 mapping of the underneath surface topog-
raphy).
The degree of planarization (DOP) of a polymer coating depends
on many factors ranging from material properties of the polymer, to
coating technology, to the bake and cure scheme, to the geometry of the
underlying features. Solid content and viscosity of the polymer solution are
the most important material properties affecting the ability of planarization.
High solid content and low viscosity enhance planarization because of
12.4 THIN-FILM MATERIALS AND PROCESSES 11-705

t ....
.,,#

Poly imide

tA~ AI

Degree of Planarization= 1-
tAl
Slope =9

Figure 12·52. Degree of Planarization

the low shrinkage and ability to flow [100]. Unfortunately, these two
properties often function in contrary to each other (e.g., high solid content
increases viscosity). There is very limited information on the effect of
coating method on planarization. Studies on the degree of planarization
of spin-coated films [101] have shown that high spin speed and solvent
evaporation during spinning deteriorate the ability of planarization. Planar-
ization during spinning is dependent on the balance of the capillary,
viscous, centrifugal, and gravitational forces. Capillary and gravitational
forces tend to planarize the film, whereas centrifugal and viscous forces
try to make the film conformal. In addition to these conformal forces,
other problems are associated with spinning, particularly as feature sizes
and planarization requirements advance. First, film uniformity is not con-
stant across the surface of a spun substrate due to evaporation and turbu-
lence in the air above the substrate. Not only is thickness radially dependent
but it varies as the metal line orientation with respect to the radius changes.
Material tends to build up where lines are perpendicular to the flow.
Compounding this with the effects of shrinkage may make spin coating
cease to be a viable method if the feature sizes get too small. The effects
of bake, exposure, and curing on planarization are quite complicated and
material dependent. There is no generalized processing guidance to be
follow. The loss of planarization is a composite of two processes: drying
and curing [22]. During baking and curing, the solvent is driven off and
11-706 THIN-FILM PACKAGING

the polymer is cross-linked. At some point, there is a transition from


liquid to solid, from able to flow to unable to flow. The transition occurs
quite quickly and a critical point is associated with the loss of flow [102] .
When the polymer is liquid, surface tension tends to keep the top surface
flat. After the critical point, the solvent continues to be driven off and
the thin-film coating shrinks to a certain percentage of its original height.
Because the layer's height varies in the presence or absence of the underly-
ing topography, the total amount of height reduction varies. The effect
of shrinkage on planarization is best illustrated in Figure 12-53 by compar-
ing the initial coating surface to the final surface after curing. In principle,
a better planarization is obtained when the flow of the coated polymer is
facilitated, the time duration of flow is prolonged, and the film shrinkage
after the critical point when the solution is unable to flow is minimized.
The dependence of degree of planarization of a polymer coating on the
surface topography of the underlying substrate fall into three forms. First,
it is straightforward to evince that better planarization is obtained over
substrates with smaller step heights. Second, planarization is poorer over
large features than that over smaller features [101]. In other words, confor-
mal coating is obtained over large features, whereas at least partially
planarized coating is obtained over small features. The third substrate
topography affecting the planarization is the distribution of the features.
Isolated features are easier to planarize than densely populated features.

Poly

Metals-
Substrate-

(a) Good planarization of polymer


thin film over surface topography
after soft bake.

(b) Poor planarization after cure


due to film shrinkage.
Figure 12-53. The Effect of Shrinkage on Planarization of Polymer Thin Films
12.4 THIN-FILM MATERIALS AND PROCESSES 11-707

This difference is a result of the fact that there is a large reservation of


polymer fluid around the isolated features which flows to or from the
feature. Such an ability of flow is necessary for a good planarization, as
discussed above.
Mechanical polishing was employed by IBM in the processing of
planar multilevel copper-polyimide thin-film substrate for its ES 9000
processors as described elsewhere in this section and Section 12.7 [24,30].
This planar structure has the advantages of superior impedance control
and short current path and thereby reduced voltage drop as a result of its
stacked stud signal and power via structure. In addition, stacked vias take
less realty in the multilevel thin-film structure and enhance wiring density.
Two variations of stacked stud process were developed within IBM: the
dual-level metal (DLM) process [1] where excessive metal is removed
by mechanical polishing for planarization and another process developed
at a later time where excessive polyimide is removed by mechanical
polishing for planarization. Figure 12-54 [30] is an illustration of the
process flow of these two technologies.
There are also other ways to achieve a planarized structure in multi-
layer interconnection as discussed elsewhere in this chapter. The feasibility
of fully planarized structure by fully additive electroless plating of copper
was demonstrated by Hitachi [103]. Oki Electric Industry Co., Ltd has
developed a plasma etching technology for planarization [104]. In this
later process, a sacrifice or erodible photoresist layer was coated on to
the polyimide layer to serve as a planarization layer and was then plasma
etched away along with the excessive polyimide. The plasma was opti-
mized so that the photoresist and the polyimide are etched at the same rate.

12.4.3.4 Relative Cost Comparison of Thin-Film Processes

The prohibitively high cost ofthin-film packaging has been the major
impedance for the widespread application of thin-film package in the low-
end, cost-sensitive consumer electronics market. It is widely acknowl-
edged that the high cost of thin-film packages must be reduced significantly
for this technology to compete effectively in the future market. The goal
is to reduce the cost of thin-film MCM substrates from the current level
of $50-$100/in. 2 to around $5/in. 2, as is indicated in Table 12-16. Great
effort has been devoted and is being proposed to the research and develop-
ment of inherently low-cost thin-film materials and processes. Cost analy-
sis and modeling are important and indispensable parts of this effort.
A number of cost models have been developed and applied to deter-
mine the intrinsically lowest-cost process. The models are typically based
on unit operation analysis of the capital, material, utilities, and manpower
requirements to deposit dielectric, form vias, metallize vias, and pattern
metal conductor wiring. Two rival processing approaches, one based on
11-708 THIN-FILM PACKAGING

Sputter Cr-Cu Seed Layer, Apply


Photoresist and Define Wiring Level
Prepare ubstrale

Additive Electroplate Cu and Strip


Apply and Cure Polyimide Photoresist

Apply Photoresist and Define Vias


Apply and oft bake PSPI
l-=-==-iiii==;
Additive Electroplate Cu to form Stud •
and trip Photoresist
Define Wiring Level by P PI Lithography

Flash Etch Seed Layer and Cap


Define Vias by La er Ablation

Apply Poly imide


Sputter or Electroplate Conductor Metal

Planarize and Expose Studs by Mecbanical


Poli hing
Planarize by Mechanical Polishing

~,---~
A B
Figure 12-54. eu-PI Planar Thin-Film Processes for High-End Systems

the 6-in_ wafer conventional IC technology and the other based on large-
area processing (LAP) technology of flat panel display (FPO), are often
compared to demonstrate the impact of large-area processing on the overall
cost. Of these cost models, the one by Dow Chemicals [105] assumes a
base case of the LAP process with a spin coated dielectric polymer of
12 f.lm (cured) thickness over a 18 in. x 24 in. glass substrate, laser-
etched vias through an erodible photoresist mask, metallization of sput-
tered bottom and top Cr barrier layers, and sputtered Cu seed followed
by electroplated Cu of a total of 4 f.lm conductor thickness with subtractive
12.4 THIN-FILM MATERIALS AND PROCESSES 11-709

Table 12-16. Comparison of IC and MCM Manufacturing Requirements

High-Volume
Current IC Current MCM MCM Goal

Wafer size 6-in. 6-in. IS-in.


Unit size O.S-in. I" x I" to 2" x 2" 3" x 3"
Number of chips N!A 4-20 typical IO-S0
Linewidth <1.0 ~m 10-20 ~m 10-20 ~m
Line thinckness <1.0 ~m S-IO ~m 10-20 11m
Photoresist thickness I ~m I-S ~m S-IO ~m
Photolithography Projection/reduction Proximity! I : I projection Projection/enlarge
Die placement N!A SO ~m typical <Sl1m
Testing (# nodes) 300 2,000-4,000 >10,000
Bonding Wire Wire Area interconnect
Production throughput Millions/month SOO!month > SO,OOO!month
Test time (nodes/s) N!A I >100
Lines (in.lin. 2) 20,000 typical SOO 2,000
Performance N!A 80 Mhz 2S0-400 Mhz
Time to market S0-7S weeks 20 weeks 3 weeks
Yield 1-90% SO% >98%
Cost/in.2 $2S-S,000 $SO-$100 $S

etching for patterning. This base case involved a total of 43 steps, as


indicated in Table 12-17. The 6 in. wafer technology and two optimized
LAP processes are also listed in this table for the purpose of comparison
and process sensitivity analysis of the overall cost.
The model suggests a cost reduction by a factor of more than 5 with
the implementation of large-area processing (the base case). A further
50-60% cost reduction could be achieved with the optimized material
selection and process enhancement. Such an optimized process is featured
by laser ablation for via generation using a holographic mask or, alterna-
tively, a photosensitive dielectric polymer, barrierless Cu layers, and high-
usage, high-film retention of dielectric polymer. Figure 12-55 [l05] sum-
marizes the cost comparison between the base and optimized processes.
As illustrated in Figure 12-55, the development of very low-cost dielectric
polymers (about $50Ikg) is believed to be crucial for further cost reduction.
Further discussions on the development of low-cost dielectric polymers
are found in Section 12.10.2.
Aside from going to large-area fabrication, it was also suggested
that the optimization of 6-in. wafer processing technology alone could
bring down the cost of MCM-D substrates by as much as 70% [49].
Similar evaluations have also been performed by Amoco, Motorola,
IBM, and IBIS associates [67,75,106,107]. All of these studies favored
Table 12-17. Processes in MCM-D Substrate Fabrication

6-in. Base LAP Base SJ:!uttered SJ:!uttered SJ:!uttered HolograJ:!hic Base Photo QP.timum

Substrate clean UV/03 UV/03 UV/03 UV/03 UV/03 UV/03 UV/03 UV/03
Rinse SRD SRD SRD SRD SRD SRD SRD SRD
Dehydration Oven 1 Oven 1 Oven 1 Oven 1 Oven I Oven 1 Oven I Oven I
Sputter Cr Metallizer Metallizer Metallizer Metallizer Metallizer Metallizer
Sputter Cu Metallizer Metallizer 0.1 11 Metallizer 411 Metallizer Metallizer Metallizer Metallizer
Apply PR Track Coater Coater Coater Coater Coater Coater Coater
Bake Oven I Oven I Oven I Oven I Oven 1 Oven I Oven I
Lithography Lithography Lithography Lithography Lithography Lithography Lithography Lithography
Develop Wet Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD SRD
....
~ Flood expose Flood Flood Flood Flood Flood Flood Flood
0
Bake Oven 1 Oven I Oven I Oven I Oven I Oven I Oven I
Rinse SRD SRD SRD SRD SRD SRD SRD
Plate Cu Plate Plate Plate Plate Plate Plate Plate
Rinse SRD SRD SRD SRD SRD SRD SRD
Strip PR Wet Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD SRD
Etch Cu Wet Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD SRD
Dehydration Oven I Oven 1
Sputter Cr Metallizer Metallizer
Apply PR Coater Coater
Bake Oven I Oven 1
Lithography Lithography Lithography
Electroless Ni Plate Plate Plate Plate
Develop Wet Wet
Rinse SRD SRD SRD SRD SRD SRD
Flood expose Flood Flood
Bake Oven 1 Oven 1
Etch Cr Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD
Etch Cu Wet
Rinse SRD
Etch Cr Wet
Rinse SRD
Strip PR Wet Wet
Rinse SRD SRD
Dehydration Oven 1 Oven 1 Oven 1 Oven 1 Oven 1 Oven 1
Plasma clean UV/03 UV/03 UV/0 3 UV/03 UV/03 UV/03 UV/03 UV/03
Rinse SRD SRD SRD SRD SRD SRD SRD SRD
Dehydration Oven 1 Oven 1 Oven 1 Oven 1 Oven 1 Oven 1 Oven 1
Apply dielectric Track Coater Coater Coater Coater Coater Coater Coater
..... Cure dielectric Oven 2 Oven 2 Oven 2 Oven 2 Oven 2 Oven 2 Oven 2 Oven 2
~
Apply PR Track Coater Coater Coater Coater
Bake Oven I Oven I Oven I Oven I Oven I
Lithogrophy Lithogrophy Lithogrophy Lithogrophy Lithogrophy Lithogrophy Lithogrophy Lithogrophy
Develop Wet Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD SRD
Flood expose Flood Flood Flood Flood Flood
Bake Oven 1 Oven 1 Oven 1 Oven 1 Oven 1
Rinse SRD SRD SRD SRD SRD
Laser etch Laser etch Laser etch Laser etch Laser etch Laser etch Laser etch
Strip PR Wet Wet Wet Wet Wet Wet
Rinse SRD SRD SRD SRD SRD SRD
Descum Descum Descum Descum Descum Descum Descum Descum Descum
Rinse SRD SRD SRD SRD SRD SRD SRD SRD
Dehydration Oven 1 Oven I Oven 1 Oven 1 Oven 1 Oven 1 Oven 1
11-712 THIN-FILM PACKAGING

Base Case Cost Breakdown Optimum Process


36%- - 26%--

10% _ _

" " -19%

Ultimate Process

Figure 12-55. Cost Breakdown of the Base Case and the Optimized Cases in
Table 12-17

a migration from the conventional Ie technology to the LAP process


technology of flat panel display. However, the cost issue in thin-film
packaging is very complex. Many factors are beyond the capability of
these cost models, notably, the availability of equipment, yield associated
with a specific process technique, process robustness, and the ground
rules, to name a few. Practitioners of thin-film packaging are strongly
advised to conduct cost evaluation according to their own established
process.

12.5 YIELD/COST CONSIDERATIONS


12.5.1 Repair of Thin-Film Packaging
One of the key challenges in thin-film package fabrication is the
ability to create defect-free wiring over the entire surface area of a large
substrate. Achieving practical yields with reasonable complexity may
require development of novel defect detection and repair techniques for
each layer in a thin-film structure. This challenge has been identified [14]
as one of the principal obstacles in fabrication of sophisticated multilayer
thin-film packaging structures.
12.5 YIELD/COST CONSIDERATIONS 11-713

The best alternative to repair, however, is eliminating the need for


it, and it is more appropriate to focus on controlling the fabrication
processes and preventing defects than to expend resources in locating and
repairing them. This approach, however, minimizes the realities of the
manufacturing environments. Thin-film fabrication processes do not al-
ways reach the level of control that eliminates the need for repair, and
even in the most well-controlled processes, the impact of defects can
be significant.

12.5.2 Leverage of Repair Technologies


There are a number of benefits in having the ability to repair defective
conductors or restructure the conductor pattern in a thin-film package. A
circuit may need to be rerouted or optimized for design or performance
reasons. Yield issues, typically caused by defects occurring during the
fabrication process, can be resolved with the addition of new sections of
wiring or with the disconnection of bad wiring sections. Typically, thin-
film defects are either in the form of unwanted metallization that create
shorts or near-shorts or missing metallization that produces opens or
near-opens.
Figure 12-56 shows the potential benefit to having repair technologies
for yield enhancement. Assuming that the defect distribution is Poisson
and that all the defects are repairable, the plot shows the yield gain that

1.00
y=O.90
0.80

"~
..... 0.60

0.40

0.20

0.00-
0 1 2 3 4 S 6 7 8 9
Number of Defects Repaired

Figure 12-56. Plot Showing Potential Repair Leverage. Note net yield gain by
repairing small numbers of opens and/or shorts.
11-714 THIN-FILM PACKAGING

can result by repairing a small number of defects per substrate. The gain
is dramatic when the initial yield is low-repairing all parts with less
than three defects can raise the effective yield from 10% to almost 80%!
As the process matures and the initial yield (those parts with zero defects)
increases, the leverage for repair decreases, as shown by the curve showing
an initial yield of 90%.

12.5.3 Repair Methodologies

Thin-film packaging structures should be designed and fabricated


with repair in mind. For example, in the design of the first-level packages
used in the IBM ES 9000 system family, the nonrepairable signal redistri-
bution wiring normally buried in the multilayer ceramic has been replaced
with a repairable, top-surface thin-film wiring layer [108].
The choice of thin-film fabrication process can also dramatically
affect the nature of defects created, and biasing the process toward a
particular defect mode may be desired. For example, an additive process
such as lift-off creates many more opens than shorts. By understanding
the defect sensitivity of the fabrication technology being considered, it is
possible to restrict the defect set to those that are more easily repaired.
Because it is generally easier to delete or remove material than to add
it, shorts are "more desirable" than opens. With the appropriate repair
technologies, any defects created during fabrication can be easily fixed.
Techniques to repair defective conductors are strongly dependent
on the size of the features and the materials used. Components built with
thin-film technologies, which include almost all semiconductor and high-
performance packaging devices and where conductor dimensions are nor-
mally less than 25 11m, require a much different set of repair technologies
than components built with "thick-film" technologies, where circuit dimen-
sions typically exceed 100 11m.
Repairs of larger-dimension circuit elements, like those used on
printed-circuit boards or multilayer ceramic substrates, can use more
traditional, mechanically oriented procedures. There are a number of
methods available to repair regions of missing or damaged conductors in
thick-film circuits [109].
The demands of thin-film dimensions limit the techniques available
for repair of defective conductor patterns, and it is common to find lasers
and other directed energy sources playing an important role in restructuring
and repair of microcircuits. Localized deposition of conductive films and
local removal of thin-film material using these sources are now common-
place, especially for repair purposes, and it is these techniques that are
the focus of this section.
One obvious drawback of using a surgical approach for defect repair
is the necessity of locating and repairing each defect in sequence. One
12.5 YIELD/COST CONSIDERATIONS 11-715

alternative to this serial approach, which relies on redundant or parallel


processes, is discussed later. Another approach is found with semicustom
MCMs [110], which utilize programmable interconnections. In these
schemes, rather than replacing defective nets, defects are routed around
as part of the customization, as all wiring resources can be pretested and
the defective segments mapped and identified.

12.5.3.1 Defect Detection and Repair

Even in the most well-controlled environments, some defects can


be encountered. This is especially true in the earliest stages of hardware
build and debug, where rapid turnaround and asset utilization are critical.
Besides establishing a strong focus on contamination control to eliminate
defects at their source, a system for defect detection and repair should
be considered as an effective tool for yield management. A typical defect
detection and repair strategy relies on automatic inspection techniques
and electrical testing methodologies. Using these approaches, defective
nets can be readily identified and the defects pinpointed.
Inspection and test are complementary-focusing on slightly differ-
ent aspects of product performance. Inspection usually addresses physical
variations that could lead to reliability or performance problems and are
often associated with a simple visual inspection. As feature ground rules
shrink and circuit density increases, machine vision systems play an
increasingly important role in the inspection area. Inspections, however,
can add significant cost and processing time and, when done manually,
are labor intensive as well.
Electrical testing confirms that the assembly is electrically functional,
either by simulating the actual operating environment seen by the package,
or by checking the performance of each component or circuit path on an
individual basis.
Under normal circumstances, neither inspection nor test can identify
all the defects that are likely to occur. Many electrical defects are not
represented by obvious physical anomalies and so can escape detection.
Conversely, many reliability problems do not manifest themselves as
immediate electrical problems but are readily apparent when visually
inspected. Finding the correct balance between defect detection schemes
and product reliability can be a key driver in the success of a manufactur-
ing operation.
IBM [111] has established a representative defect detection and
repair system for the thin-film packaging structures used in the System!
390 line of computing systems. Using information provided by electrical
test and automated inspection tools, defective wiring patterns are quickly
identified and the physical coordinate and type of each defect established.
This information is transferred electronically to the repair tools, where
11-716 THIN-FILM PACKAGING

individual repair operations are performed. The process flow is shown


schematically in Figure 12-57.
To take full advantage of the defect detection system, reliable repair
techniques for the repair of both opens and shorts are required.
In general, defects created in the fabrication of multilayer thin-film
structures can be classified into two major types intralevel and interlevel
with both opens and shorts possible. Repair actions are typically estab-
lished only for intralevel defects, although the repair of interlevel defects
is possible in some circumstances [112,113].

12.5.3.2 Repair of Shorts


The use of laser cutting for removal of metal films has become
a very important processing operation in the production of electronic

Thin Film Build


Processing

Automated
Inspection
IDefects Found
I
Zero Short
Defects Repair
I

Electrical
Testing
........
IDefects Found
Zero
Defects I1 Defect
Locator
I
I Short
Repair
I
I I Open
Repair
I
J
I
Additional Thin
Film Processing

Figure 12-57. Defect Detection Process Flow


12.5 YIELD/COST CONSIDERATIONS 11-717

components [114], and this technology has been extended to the repair of
extraneous metal defects in thin-film circuitry [108,115]. Many conductors
used in thin-film packaging possess vaporization temperatures higher than
the melting temperatures of the dielectric substrate. This is especially true
for polymers, which are finding increasing utilization as the interlayer
dielectric. Selection of the optimum processing conditions for repair
requires an understanding of the practical considerations controlling the
laser removal process [116].
To guarantee clean metal removal, it is desirable to use laser spots
larger than the features being ablated, and so the dielectric is irradiated
by some portion of the incoming beam. As most polymers are only weakly
absorbing in the visible and infrared, transmitted laser radiation can create
catastrophic damage in thin-film structures when it is absorbed by subsur-
face structures [111]. Additional damage can be created by direct absorp-
tion of the laser radiation by the dielectric or through the transfer of
thermal energy from the irradiated region to the underlying substrate. To
avoid thermal damage during laser metal removal, short-duration laser
pulses such as those obtained from Q-switched solid-state or excimer
lasers are commonly used, especially when the dielectric is thermally
sensitive, as is the case for most polymers. Longer-duration pulses are
usually only used with inorganic dielectrics, such as glasses and ceramics.
Besides being stable to higher temperatures, these materials tend to have
higher thermal diffusivities than polymers and can more readily dissipate
the heat transferred during microsecond or millisecond laser pulses. Other
critical laser parameters, such as wavelength and pulse energy, are chosen
to minimize thermal or absorption damage to the irradiated areas while
providing sufficient intensity to completely sever the conductor. For exam-
ple, the ultraviolet radiation of excimer lasers is strongly absorbed in most
microelectronically useful polymers. This interaction typically results in
ablation of the polymer in the irradiated area and can produce loss of
dielectric integrity unless carefully controlled.
Figure 12-58 shows optical micrographs of a residual metal defect
in a thin-film package repaired using the excimer laser. The micrographs
show both the original defect and the repaired site. Note the clean removal
of the metal with minimal damage to the underlying polymer dielectric
layer.
Typical laser pulse-width conditions used for clean metal removal
from thin-film dielectrics are summarized in Table 12-18. Pulse energy
and wavelengths are typically application specific.
As dimensions shrink, laser removal techniques become less viable,
primarily because the beam focal spot size is limited by the laser wave-
length used, achieving a practical limitation of about 0.5 /lm. Although
lasers remain practical for the dimensions found in today's thin-film
packaging structures, the focused ion beam (FIB) has been introduced as
an alternative method for material removal. The ion beam removes material
11-718 THIN-FILM PACKAGING

Figure 12-58. Excimer Laser Short Repair on Polyimide Surface

by sputtering, ejecting atoms and molecules from the surface into a vac-
uum. The FIB eliminates many problems associated with laser removal
techniques and is becoming an increasingly common tool, especially
for VLSI processing. The FIB has submicron focusing capability, with
excellent material selectivity and depth control.
Although process intensive and therefore more costly, traditional
lithographic techniques can also be used for material removal. Resist or
polymer layers can be used to protect features during a subtractive process,
such as wet or reactive ion etching, allowing unwanted material to be re-
moved.

12.5.3.3 Repair of Opens


The method most discussed for repair of thin-film circuitry is laser
chemical vapor deposition (LCVD), including deposition from both gas-

Table 12-18. Typical Laser Metal Removal Conditions for Thin Films

Dielectric Laser Type Pulse Width

Polymers, ceramics Excimer, Nd-YAG 20-50 ns


Ceramics/glasses Nd-YAG ~s- ms
12.5 YIELD/COST CONSIDERATIONS 11-719

phase and condensed-phase precursors [117-119]. Localized laser deposi-


tion may be accomplished photochemically, using the laser to induce
photochemical reactions of deposition precursors, or thermally, using a
laser or directed energy source to locally heat a substrate surface. In a
typical gas-phase process, the substrate is immersed in a vapor environ-
ment containing organometallic molecules to be dissociated, and the sys-
tem is exposed to the laser beam. The beam can activate the molecules
in the gas phase or as an adsorbed layer through pyrolysis or photolysis,
or, often, by a combination of the two.
Although photochemically stimulated deposition reactions may pro-
vide superior resolution, the deposited films are typically contaminated
with carbon and have poor electrical properties. This method, therefore,
has limited value in direct repair metallization schemes, and the pyrolytic
process is almost exclusively used for repair activities.
The pyrolytic process is very analogous to conventional chemical
vapor deposition (CVD), where the substrate is locally heated by the laser
beam and the reactant atoms are liberated when the parent molecules
decompose by collisional excitation with the surface. For sufficient heating
to occur, the organometallic molecules must be sufficiently transparent
and the substrate or some film over it must be an efficient absorber of
the laser radiation. The pyrolytic process is shown schematically in Figure
12-59.
A wide range of precursors are available, permitting the laser deposi-
tion of useful microelectronic metals such as aluminum, copper, gold,
platinum, tungsten, chromium, titanium, cobalt, nickel, lead, tin, and palla-
dium. Successful film formation requires optimization of the nucleation
and growth conditions within the constraints imposed by the thermal
environment established by the substrate.
Successful repairs of defective conductors have been made using
laser deposition of gold films from dimethyl-Au-trifluroacetylacetonate
with an argon-ion laser operating at 514.5 nm on a number of thin-film
packaging structures [111].
A typical repair uses parameters similar to those shown in Tables
12-19 and 12-20, depending on the dielectric involved. In Table 12-19,
the initial scan conditions are used to induce deposition without damaging
the thermally sensitive polymer dielectric. This is followed by additional
scans at increasing laser powers to build the deposit to the desired thick-
ness. Actual scanning conditions are determined by the defect type and
structure being repaired.
An alternative method for achieving good-quality laser-deposited
metal films on thermally sensitive dielectrics involves switching from a
continuous-wave (CW) source to a high-repetition-rate pulsed laser. NEC
has demonstrated [120] the ability to produce high-quality gold lines from
dimethyl-Au-acetylacetonate and its trifluorinated analogue using this ap-
11-720 THIN-FILM PACKAGING

-: Laser

Figure 12-59. LCVD Process Schematic

proach. Using a rectangular-shaped, scanned beam from a high-repetition-


rate, frequency-doubled Nd-YAG or copper vapor laser (pulse widths
from 20 to 70 ns), chemically pure films up to 10 ~m thick and with
resistivities as low as 4.0 ~Q·cm have been formed on polyimide dielec-
trics, with negligible thermal damage to the polymer.

Table 12-19. Representative Laser Writing Conditions for Polymers

Pass Power (mW) Scan Rate(f.!/s) No. of Scans

5-8 2.0 2
2 40-50 2.0 2
3 150-175 10.0 4
12.5 YIELD/COST CONSIDERATIONS 11-721

Table 12·20. Representative Laser Writing Conditions for Ceramics

Power (mW) Scan Rate (l1m/s) No. of Scans

350-1000 10-50 1-4

With inorganic substrates, the risk of thennal damage is minimal


and higher-power conditions can be used to create the optimum repair
thickness directly. Representative conditions for this type of repair is
shown in Table 12-20. The resulting deposit has high chemical purity,
low resistivity, and good intimate contact to both the substrate and the
existing metal features. The films are typically granular, with some trapped
porosity. Although the quality of the deposited film is high, it can be
difficult to create a good electrical connection between the newly deposited
metal and the existing circuit. For example, the metallurgy in the existing
thin-film circuit may consist of multiple layers, where the top layer is a
barrier metal that is oxidized, fonning a protective, insulating layer. Before
a good connection can be made, this top protective layer must be removed.
One method of doing this is to use a pulsed laser to partially ablate the
metallurgy at the point of interconnection prior to depositing the repair
patch [121]. The partial ablation can be readily controlled by adjusting
the fluence of the laser beam and by varying the number of pulses.
Figure 12-60 shows how two adjacent thin-film lines can be repaired
by this process, using conditions especially developed for a typical thin
film on a ceramic structure. Modification of the process conditions allows
repairs to be made over a wider range of dielectrics, metallurgies, and
line structures. Successful repairs have been made with the LCVD process
on alumina, glass-ceramic, and a number of different polymer dielectrics.
Laser writing using solid thin films of metallo-organic precursors
are an attractive alternative to gas-phase processes. A suitable precursor
for laser writing must be capable of meeting a number of prerequisites.
They must be able to fonn homogeneous films, have high metal content,
have sufficient optical absorption, and decompose at relatively low temper-
atures with clean decomposition and volatilization of by-products.
A metallo-organic complex may meet these requirements, but most
materials lack good film-fonnation characteristics. In those cases, the metal
complex is combined with a compatible organic film fonner, typically a
polymer. A wide variety of metals complexes, including Pd, Pt, Ag, Cu,
and Au, are available and have been used to fonn metal films in a variety
of circumstances [122].
11-722 THIN-FILM PACKAGING

Figure 12-60. LCVD Repair of Open Line Defects

New technology approaches are also being established. Partially


open segments of a conductor can be selectively heated by passing a
current and inducing the deposition of an electrically conductive material
[123]. Repairs made in this way are self-aligning and self-limiting, as the
thermal driving force is reduced as the repair is built up. Partridge et al.
[124] have expanded this technique for the repair of open conductors by
creating a conductive connection across the open before initiating the
repair and have designated the technique as "self-induced repair." Al-
though typically done from a plating solution, the use of gas-phase pyroly-
sis is also possible.
Traditional bonding methods can also be extended to the smaller
dimensions found in thin-film packages by utilizing wire or solder techno-
logies, both permitting repair of defects found during the fabrication
process and allowing the repair of defects found or created in higher
levels of assembly, where the presence of active or passive components
may restrict the technology options available [125]. Figure 12-61 shows
a repair of a defective thin-film line made using a short segment of a 0.7-
mil-diameter wire.
Open repair methods can also take advantage of standard photolitho-
graphic processing techniques. The area requiring repair is first covered
with a layer of photoresist, followed by exposure of the defective area
using an appropriate optical source and development of the resist [126] .
12.5 YIELD/COST CONSIDERATIONS 11-723

Figure 12-61. Wire-Bond Repair of Open Line Defects

In a variation of this scheme, the area can be covered with a suitable


polymer film and the opening created using direct UV laser ablation [127].
Metal patches are deposited by conventional evaporative or sputtering
techniques. Standard lift-off techniques are then used to remove the unde-
veloped resist or polymer and the unwanted overlying metallization, leav-
ing the repair metallurgy behind.

12.5.4 Alternative Repair Methodologies


Another alternative to repair is defect tolerance. Instead of repairing
defective circuits, redundant or alternative wiring paths can be provided.
Although eliminating the need to physically locate and repair the actual
defects-the loss of functionality is a sufficient indicator-these methods
require the utilization of critical real estate for the spare features.
Processes that reduce defect susceptibility can also be considered
defect tolerant. Methods such as coating interlevel dielectrics in mUltiple
coats or providing dual dielectrics are examples of this. Wiring can also
be made vertically redundant, in that the effect of a defect in one layer
can be eliminated by the presence of a second layer.

12.5.4.1 Redundancy
Redundant wiring structures consisting of two independently pro-
cessed photolithographic layers have been used to reduce the defect sensi-
tivity associated with fabrication processes. This vertical redundancy can
11-724 THIN-FILM PACKAGING

effectively eliminate open defects while increasing the potential level of


shorts [108,128]_ Redundant or spare features can also provide an escape
from the low yields inherent in building larger and more complicated
devices. These approaches rely on having excess circuit components pres-
ent that can be substituted for nonworking components by testing, linking,
or opening lines, and retesting to obtain the desired overall functionality.
Although not in wide use for thin-film packaging, this has found
wide acceptance in integrated-circuit manufacturing, especially in the
fabrication of memory chips. Extra rows or column lines are built, which,
after the initial test, can be substituted for nonworking rows or columns.
This type of programmable restructuring, which normally involves laser
cutting to disconnect circuit elements and laser linking to reconnect ele-
ments, can greatly enhance yield and fault tolerance [129].

12.5.4.2 Rerouting
A key feature in high-performance packaging design is the ability
to repair the assembly and make engineering changes. Engineering change
(EC) is a process in which the chips or modules can be removed from
their carrier and replaced individually, or by which wiring can be rerouted.
Specific techniques are required for deleting a connection from a signal
terminal of a chip to the package wiring network and rerouting the terminal
to other networks or chip terminals as desired. Engineering changes are
carried out to replace defective signal lines and to make new connections
to correct design errors. In manufacturing, such a scheme can be used to
enhance yield [130]. Defective nets are isolated by laser delete and rerouted
by the addition of top-surface discrete engineering change wires [98].

12.5.5 Performance and Reliability


Repair procedures and methods must be capable of ensuring that a
repaired circuit is as reliable as it would be when perfect. Repair integrity
is normally assured through the application of functionality and/or reliabil-
ity testing.
Reliability stress tests are designed to simulate subsequent processing
or field use conditions. These can include, but are not limited to the fol-
lowing:

• Thermal cycling test, simulating the high-temperature (360°-380°C)


furnace cycles used to join pins and chips, as well as the conditions
(0°-100°C) used to simulate machine power on-off cycles.
• Temperature and humidity testing, typically at 85°C, 85% relative
humidity (rh), with and without electrical bias, to ensure no loss of
dielectric integrity due to repair processing.
12.6 RELIABILITY 11-725

• Electromigration testing under accelerated conditions, ensuring


that the new metallurgical structure created during open repair is
reliable under long-term current flow conditions.

12.6 RELIABILITY
Reliability of thin-film packages is a major concern. Some polymers
like polyimide dielectrics absorb significant amount of water, and unless
proper baking precedes the package sealing, corrosion problems discussed
in Chapter 5, "Package Reliability," are valid. In addition, the large ther-
mal-expansion mismatches between many polymers (except several low-
thermal-expansion polymers like PIQ LlOO and L120) and ceramic sub-
strates and the IC chips give rise to huge shear deformations at this
interface. Even though short term tests may not expose any real problems,
long-term usage at the elevated temperature of package operation in the
presence of absorbed moisture may present reliability concerns. It is
important, therefore, that the polymer selection include low-moisture ab-
sorption. Additionally, direct passivation of the polymer or sealing of
the package after appropriate outgassing may be necessary. The major
concern with moisture being corrosion of the copper conductor, other
means such as barrier materials may be used as well. Most of these
concerns may simply be due to lack of extensive experience with the use
of these materials. Metal-related problems are similar to the reliability
concerns of semiconductor thin films and printed-circuit boards, as dis-
cussed in Chapter 5, "Package Reliability."
Most of the thin-film structures discussed in this chapter should be
submitted to a variety of reliability tests. Table 12-21 is a typical example

Table 12-21. Reliability Tests of Thin-Film Packages

Test Measurement

1. High temperature storage • Adhesion of metal to • Some changes


150° C, 500 h polyimide with some metals
2. Thermal shock • Adhesion of metal to • Some change with
-65°-150° C, 15 cycles polyimide some metals
• Resistance • <5% change
3. Temperature, humidity, bias • Resistance • <5% change
85°C, 85% RH, 5V, l000h • Insulation Resistance • no change
4. Pressure and temperature • Adhesion
(peT) -Metal to polyimide • No change
2 Atm., 120° C, IOh -Polyimide-to-polyimide • Some change
-Polyimide-to-ceramic • Some change
11-726 THIN-FILM PACKAGING

of the reliability tests performed on thin-film packages, which summarized


the testing conditions and general observations.
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM
PACKAGING
12.7.1 IBM
In this section, multilevel thin-film technologies developed at IBM
for use in computer systems, telecommunication products, avionics, and
other high-performance applications are discussed. Thin-film wiring has
been used at IBM in five different high-performance applications ranging
from chip carriers for logic and memory devices to multichip modules
for large systems. The key elements of multilevel thin-film structures and
processes for high-performance multichip modules are (1) cofired-ceramic
substrate, (2) conformal via structure, (3) copper conductors for improved
conductivity, (4) polyimide dielectric for good mechanical properties,
high thermal stability, and low dielectric constant, (5) laser ablation or
photosensitive polyimide lithography for via definition, (6) additive elec-
troplating for wiring definition, and (7) additive electroplating for terminal-
metal definition. This combination of processes is extendible to finer
dimensions. Thin-film wiring on cofired ceramic is capable of flip chip,
TAB, and wirebond chip connections, as well as all industry-standard
card-level connections.
IBM's technology has moved from single-chip modules (SCM),
which needed thin-film wiring for redistribution of chip IIOs, to multichip
modules, with multilevel thin-films for signal wiring of highly integrated
devices. The first product was a 36-mm-square pin-grid-array package.
The next advancement in the technology was two levels of planar thin
films for signal redistribution of 121 chips on a 127-mm glass-ceramic
substrate for the ES 9000 processor. Planar processes were implemented
to satisfy system requirements demanding tighter dimensional control.
These devices were primarily bipolar. One plane pair of planar thin films
was also implemented on a 166-mm glass-ceramic substrate for redistribu-
tion and signal wiring of bipolar devices. For large CMOS devices used
in workstations, minicomputers, and servers, a nonplanar thin-film struc-
ture on either a silicon wafer or cofired ceramic was practiced. It is
interesting to note that the size of the substrate grew from 36 mm square
for a single-chip module to 166 mm square for the processor of mainframe
computers using bipolar devices, and decreased again in size for computers
based on highly integrated CMOS devices.
12.7.1.1 Thin-Film Packaging Evolution
Table 12-22 shows the evolution of multilevel thin-film technology
from the small single-chip interconnection to the conformal via intercon-
nection on large substrate.
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-727

Table 12-22. Evolution of Multilevel Thin-Film Technology at mM

Metallized ES9000 AI-PI Cu-PI


Ceramic Processor Comformal Cu-PI Stacked Conformal
Key Features Polyimide Substrate Via Process Stud Process Via Process

Wiring Chip I/O Chip I/O Signal Signal Signal


redistribution redistribution distribution distribution distribution
Structure Nonplanar Planar and Nonplanar Planar Nonplanar
redundant
Size of substrate 36mm 127 mm 125 mm 166mm 127 mm
square square square square square
Number of levels 2 2 5 7 5-6
Dielectric material PI PI PI PSPIIPI PSPIIPI
Via definition Wet etch Laser ablation RIE Laser PSPI lithography!
Laser ablation
Conductor metal Cu Cu Al Cu Cu
Wiring definition Subtractive Lift-off Subtractive PSPI lithography Additive
etching etching and electroplating electroplating
Bonding pads Subtractive Lift-off Evaporation Lift-off Additive
etching through mask electroplating
Wiring pitch 75 ~m 55 ~m 25 ~m 40~m >25 ~m
Wiring width 25 ~m 35 ~m 8 ~m 15~ >lO~m

IBM has developed and implemented multilayer thin films (MLTF)


for both high-end and cost-performance systems since the early 1980s.
Copper-polyimide and aluminum-polyimide multilayer thin films have
been implemented on silicon, alumina, and glass-ceramic substrates. The
various MLTF implementations are as follows:

• Two layers of Cu-polyimide interconnection on dry-pressed alumina


for single- and dual-chip applications, using wet etch of polyimide
and subtractive etch of Cr-Cu-Cr for wiring.
• One to two layers of planar Cu-polyimide interconnection on alumina
and glass-ceramic multichip modules for redistribution of chip I/Os
to the ceramic vias, using laser ablation for via definition, subtractive
etching of Cr-Cu-Cr for wiring definition, and lift-off for bonding-
pad definition.
• Four layers of Al-polyimide multichip interconnection on silicon
substrates, using reactive ion etching for via definition, subtractive
etching of aluminum for wiring definition, and evaporation through
a mask for bonding-pad definition.
• Four to five layers of planar Cu-polyimide interconnection on glass-
ceramic substrates, using laser ablation for via definition, photosensi-
11-728 THIN-FILM PACKAGING

tive polyimide (PSPI) lithography, and blanket electroplating for


wiring definition, and lift-off for bonding-pad definition.
• Four or five layers of nonplanar Cu-polyimide interconnection on
alumina, silicon, and glass-ceramic substrates, using laser ablation
or photosensitive polyimide lithography for via definition, additive
electroplating or subtractive etching of Cu for wiring definition, and
additive electroplating for bonding-pad definition.

The evolution of these thin-film technologies, as well as the rationale


behind the process choices and their relative merits, are discussed below.

12.7.1.2 Multilayer Thin-Film Implementation


Multilayer thin-film (MLTF) interconnection technology enhances
the electrical performance of multichip modules [13,131] providing a
variety of features such as

• Low-capacitance redistribution
• High-density wiring
• Ability to support high-liD-density chips
• Improved propagation delay
• Increased off-chip switching activity

Although thin-film interconnections provide large performance im-


provements, not all applications require these improvements and hence a
custom design is necessary for cost-effectiveness. An important element
of this design is the choice between planar and nonplanar thin-film struc-
tures [94]. Nonplanar, unfilled via structures require staggering of vias
in successive thin-film levels. Planar stud structures permit stacking of
vias in successive levels, producing a higher wiring density than nonplanar
structures for a given linewidth and pitch. The number of levels of thin
films is dependent on the wiring and electrical requirements of the package.
Thin films can be used for redistribution of chip liDs, for a single level
of wiring, or for multiple levels of wiring. An additional thin-film layer
is utilized to define the bonding pads for chip connections. Depending
on the electrical characteristics of the package, a number of thin-film
power distribution levels are needed. In this section, the different processes
and structures that were developed for different IBM systems over the
last decade are described.

12.7.1.3 Single-Layer and Multilayer Single-Chip Packages


Metallized-ceramic-polyimide (MCP) is a pin-grid-array chip carrier.
The MCP substrate is a 1.5-mm-thick perforated dry-pressed alumina
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-729

ceramic wafer with 2.5-mm holes for pins. Two layers of Cr-Cu-Cr wiring
isolated by a polyimide layer are used to interconnect the chip 1I0s to
the module pins. This is shown in Figure 9-8 in Chapter 9, "Ceramic Pack-
aging."
The wiring and the dielectric layers are defined by thin-film technol-
ogy [132]. The vias in the polyimide dielectric are defined by wet etching.
The wiring levels and the bonding pads are defined by subtractive etching
of Cr-Cu-Cr. The previous generation of this technology, known as metal-
lized ceramic (Me), had only one level of wiring defined by subtractive
etching of Cr-Cu-Cr [133]. Figure 12-62 shows the process flow for MCP.
A solder dam (Cr) is used to prevent solder bleed from the bonding
pads to the wiring traces during the flip-chip connection process. This is
achieved by using an additional lithography step after the wiring traces
are defined to selectively remove Cr from the pin and bonding-pad areas,
and to leave it on the wiring traces. Copper is used as the bonding-pad
metal for flip-chip connection and for the soldering of pins because very
few rework and reflow cycles are required during flip-chip connection of

Prepare Substrate; Define Wiring Level by


Subtractive Etching of Cr-Cu-Cr
'--_____I i Ceramic'

Apply and Partially Cure Polyimide; Apply


Photoresist and Derme Vias

Wet Etch Polyimide and Strip Photoresist

Define 2nd Level Metal by Subtractive Etching of


Cr-Cu-Cr' Define Solder Pads by Etching Cr

Connect Pins and Chips using Solder

Figure 12-62. MCP Process Flow for Redistribution of Chip I/Os


11-730 THIN-FILM PACKAGING

single-chip modules. The 110 pins are wedged (force fit) and soldered to
the wiring layer.
Metallized ceramic (MC) and MCP packages are generally used for
packaging of high-performance logic and memory devices. The ability to
fabricate two levels of wiring in MCP allows the use of the module as a
multichip module for 10w-II0 chips. Alternatively, the first level of wiring
could be used as a ground and/or power plane for high-performance de-
vices.

12.7.1.4 Thin-Film Structure and Process for Mainframe Computers


The ES 9000 processor module was developed for mainframe sys-
tems. The ES 9000 processor module was the result of advances in both
thin-film and ceramic technologies [108,134]. Some of the advance-
ments included

• 127-mm multilayer glass-ceramic/copper cofired substrates


• Multilayer copper/polyimide thin-film structure
-Projection laser ablation for via patterning
-Mechanical polishing for planarization
-Laser chemical vapor deposition for repairing thin-film opens
-Laser ablation for repairing thin-film shorts
• Buried conductors for quick engineering changes

A planarized stud process is used to fabricate the high-density thin-


film wiring structure. The stud structure is formed by using sputtered Cr-
Cu as the conducting metal, and polyimide as the dielectric material.
Mechanical polishing is used for planarization. Figure 12-63 shows the
process flow used.
The wiring level is defined by subtractive etching of Cr-Cu-Cr and
the bonding pads are defined by metal lift-off. Redundancy is incorporated
into the wiring level to enhance yield by making use of the reactive ion
etching (RIB) etch stop necessary to fabricate the bonding pads by metal
lift-off. A polyimide passivation layer is applied on top of the wiring and
bonding-pad levels. Openings for flip-chip connection and engineering
changes are formed in this passivation level by laser ablation. Figure
12-64 shows the cross section of the thin-film structure.
The design parameters for the thin-film structure in the ES 9000
processor module are 0_065 pF/mm capacitance, O.18Q/mm resistance,
and 30 mV maximum coupled noise [135].

12.7.1.4.1 Aluminum-Polyimide Multilevel Thin-Film on Silicon


Aluminum-polyimide (AI-PI) multilevel thin-film (MLTF) intercon-
nections on silicon were developed to package high-performance CMOS
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-731

Evaporat. Torml"ol Mot;lI. o"d LiftolT Ste"dl


l·reIHIn-° Substr:H~: I)e-fine ('rnuuir C011llure P:ld.s:
by Subtrllrlivo i::trhillC of Cr-Cu-Cr

Deli". I{edundanl ~ iring l..v.1 by ublracli ...


Elching of Dlankel r-Cu-Cr

1':lS iV;lIt \ Iring Lt.tl by Polyim id. '\I'I,lir"lIolI

Spuurl'" "r·Cu; "huwri"lC' nnll 1l.4:'IUO\·C' E\C'c S


and Lour Alll";on orVi ••
~1('1I11 hy .\ lerh:wic'lll PoB hing

Spuurr mankel C,·-(',, · ('r

n" ihl Llft·orr ·1.II,·il 10 1I.lin. Tcrmi,,:ol M<tllis

Figure 12-63. Planar Thin-Film Process Flow Used in the Fabrication of the
ES9000

chips in multi chip modules [136]. The thin-film structure is defined by a


confonnal via process, with aluminum as the conductor metal and polyim-
ide as the dielectric. This structure is shown in Figure 12-65.
The polyimide dielectric is patterned by reactive ion etching through
a photoresist stencil, and the wiring level is patterned by subtractive
etching of aluminum. The process flow is shown in Figure 12-66. Cu-
Au flip-chip bonding pads are defined by evaporation through a mask.
The process and the materials were chosen to be compatible with the
CMOS manufacturing line, and trade-offs were made between electrical
properties and process constraints. Aluminum was used in spite of its
high resistivity, and the use of thin dielectric layers led to a confonnal
dielectric coating and a highly nonpianar structure. The silicon substrate
offers the following advantages: (1) It has the same TCE as the VLSI
11-732 THIN-FILM PACKAGING

Figure 12·64. Cross Section of CopperlPolyimide Redistribution Layer on


Glass·Ceramic Substrate

chips and (2) it is compatible with existing semiconductor tooling and


processes. The dimensions of the smallest feature in the substrate were
8 J.lm wide and 3 J.lm thick on a 25-J.lm pitch. The 54-mm square substrate
with nine chips contained about 40 m of wiring and more than 15,000
vias, and was fabricated two-up on a 5·in. wafer (two substrates on
a wafer).

Figure 12·65. AI·PI ML TF on Silicon Substrate Prior to Dicing


12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-733

Prepare Substrate; Define Wiring Level by


Subtractive Etching of Aluminum

Silicon I

Apply Polyimide and Define Vias in Photoresist


-..

Transfer Image to Polyimide by RIE and Strip


Photoresist
~ ,r---------------------~~

...
Defme 2nd Level of Wiring by Subtractive Etching
of Aluminum
.-

Figure 12·66. AI·PI MLTF Thin-Film Process Flow

12.7.1.4.2 Stacked Stud Copper-Po/yimide


The planar stacked stud structure has the advantage of superior
electrical characteristics compared to the conformal via structure with
staggered vias. The stacked stud structure provides a shorter current path
compared to the staggered via structure, thereby reducing the voltage drop
in the power via stack.
Two different types of stacked stud processes have been used for
high-performance processor applications. The first process is the so-called
dual-level metal (DLM) process shown in Figure 12-54.
In this process, a layer of polyimide is first applied and cured,
followed by a layer of photosensitive polyimide. The wiring traces are
defined in the photosensitive polyimide layer by photolithography and
then the vias are patterned in the underlying polyimide layer by projection
laser ablation using a dielectric mask [32]. The composite structure is
11-734 THIN-FILM PACKAGING

metallized by sputtering or blanket electroplating. Mechanical planariza-


tion is used to remove the excessive metal and to define the wiring features.
A plane pair (two signal wiring layers and two ground/power layers) is
completed by repeating the above process multiple times [25]. Figure
12-13 shows the cross section of the planar stacked stud multilevel thin
films built by the DLM process on a glass-ceramic substrate.
The multilayer thin films were built on 166-mm square glass-ceramic
substrate with a measured impedance of 39Q, a line capacitance of 0.17
pF/mm, and a line delay of 6.8 ps/mm [137]. Because the structure is
planarized at the wiring level and the polyimide dielectric is applied on
a planar surface, precise control of the dielectric thickness is possible.
In the other type of planar thin-film process that was studied, the
wiring and the stud layers are formed in separate lithography and plating
steps [95]. The electroplated metal features are then overcoated with a
single polyimide coating. Planarization of the polyimide coating removes
the excessive polyimide over the metal topography and uncovers the tops
of the studs. Mechanical polishing is used for planarization. The uniformity
of polyimide removal during the planarization process determines the
uniformity of the dielectric thickness. Figure 12-67 shows the cross section
of one level of planar stacked studs built by this process.

12.7.1.4.3 Conformal Via Copper-Polyimide


Since the introduction of the ES 9000 processor substrate and the
development of aluminum-polyimide multilayer thin films, several ad-
vancements have been made in the process and structure of multilevel
thin films to reduce the cost of the process and tools and to improve the
electrical performance of the interconnection [24]. Figure 12-14 shows
the process flow for a conformal via process using polyimide as the
dielectric and copper as the conductor metal.

Figure 12-67. Thin-Film Layer Fabricated by Additive Electroplating


12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-735

Multilevel thin-film structures have been fabricated by combining


the photosensitive polyimide or laser ablation via process with the electro-
plated conductor line process. These two basic processes are repeated
multiple times to construct one or more plane pairs of thin-film wiring
as shown in Figure 12-15.
For high-performance workstations and minicomputers, the thin-
film structure consists of two signal levels, two ground and/or power
levels, and one bonding-pad level-a total of five thin-film metal layers,
on a cofired-ceramic substrate. It is arranged in an offset strip line structure
with the bonding pads on the top for interconnection to the semiconductor
chips. Before fabricating the signal and power levels, an initial layer of
large pads is needed to capture the ceramic vias. This layer is defined by
subtractive etching of Cr-Cu-Cr. When it is electrically feasible, the first
ground/power level is combined with the capture pad level. The thickness
of the conductor lines and the dielectric is tailored to meet the impedance
and resistance requirements of the product.
Photosensitive polyimide vias are patterned with a full-field projec-
tion printer. The wall profile of the vias is optimized for the metallization
process which follows. The degree of planarization (Eq. 12-11) is 35%
for a 5-~m-thick PSPI over 5-~m-thick conductor lines on a 25-~m pitch.
The topography of the structure provides a processing challenge. By a
judicious choice of materials, tools, and process conditions, a highly
manufacturable process with very high yields was developed. The conduc-
tor lines are patterned in a thick photoresist (developed in house) in
conjunction with projection printing. The linewidth control, even over
topography, is much better than ± 10% (30) for a linewidth of 10 ~m.
The vias for this nonplanar application have also been fabricated by
projection laser ablation using a dielectric mask. The laser ablation process
and the PSPI lithography process showed equivalent defect levels, and
structures fabricated by both processes passed reliability stress tests. Also,
yield was very high for both via processes. Figure 12-68 shows the top
and bottom surfaces of a 54-mm multichip module.
The plane pair contains nine layers of thin films-five metal levels
and four polyimide layers. The thin-film ground rules were as follows:

• Wiring pitch: 25 ~m
• Linewidth: 13 ~m
• Conductor thickness: 5 ~m
• Via diameter: 10 ~m
• Dielectric thickness: 5 ~m

The substrate size was 54-mm square and was built four at a time
on a 127-mm square ceramic. Each of the substrates were diced prior to
11-736 THIN-FILM PACKAGING

Figure 12-68. Cu-PI Interconnection on PGA Alumina Substrate (Top and


Bottom Views)

test. The cycle time for thin-film build, dice, and test was 24 days. The
thin-film technology is highly manufacturable and can be implemented
on a large substrate to be used as a multichip module. Alternatively, the
ceramic carrier can be diced into smaller substrates for single-chip module
(SCM) or few-chip module (FCM) implementation. It is built four-up on
a 127-mrn square cofired alumina substrate using PSPI lithography for
vias and additive electroplating for wiring. The top surface has bonding
pads for flip-chip connection and the bottom surface has brazed pins for
connection to the planar next-level packaging.
Extensive analysis and measurements verified the electrical charac-
teristics of the thin-film package. The signal levels are always arranged
in an offset strip-line structure. Various mesh designs were studied to
determine their effect on the impedance of the thin-film package. The
impedance was measured for a structure whose linewidth, line thickness,
line pitch, and dielectric thickness were 13 J.lm, 5 J.lm, 25 J.lm, and 5 J.lm,
respectively. The impedance is a function of the mesh design and the
capacitive line loading and was measured between 35 and 45n, depending
on the design. For a particular mesh design and line loading, the impedance
was within 10% of the nominal value. Thin-film structures on a 75-J.lm
pitch and > 12-J.lm dielectric thickness have also been fabricated for a
> 50n impedance.

12.7.2 Siemens: Microwiring Thin-Film Technology for


Mainframes

Microwiring technology has been developed for multichip module


applications used in the Seimens Mainframe H90 or enhanced ma-
chines [138].
The multichip module connector has 1850 pins, 1100 of which are
signal pins, 847 are power pins, and 2 are for the temperature sensors.
Up to 5200 point-to-point connections with a total length of approximately
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-737

150 m can be realized as printed wiring on the microwiring substrate.


The total wire length per substrate is 480 m, representing a line density
of 480 cmlcm2• The outer dimensions of the module are 115 mm x 115
mm x 26 mm.

12.7.2.1 Microwiring Technology


The 98-mm x 98-mm substrate is approximately 2.1 mm thick and
is composed of up to 19 conductive layers. Due to the physical design
of the module, the micro wiring substrate has to meet certain mechanical
specifications with respect to thickness tolerance, flatness, and elasticity,
as well as compressive and flexural strength. The substrate consists of
a complex of acrylic resin and polyimide layers prepared by a novel
relamination process and a fiberglass-reinforced polyimide complex pre-
pared by largely conventional processes [139].
The 17-layer complex comprises one IC connecting layer, two IC
redistribution layers, six power layers, and eight signal layers. The rein-
forcing complex comprises a core with one connector connecting layer
and one connector redistribution layer, as well as a wiring layer.
Two adjacent conductive layers are connected by a plated blind via.
Layers that are farther apart are electrically connected by a corresponding
number of such plated blind vias lined up in the z direction and staggered
with respect to one another in the x and y directions because of the laser
beam used for drilling the holes. Figure 12-69 shows schematically the
module microwiring substrate.
The signal layers contain the printed wiring as well as pads and the
via holes spaced 0.5 mm apart. In the channel between two grid lines,
there are two copper conductors with a width of 0.08 mm and a thickness
of approximately 0.03 mm. The power layers consist of a Cu layer-0.03
mm thick with a contact groove and a clearance. Figure 12-70 shows the
design of the signal and power levels.
The electrical parameters of the printed wiring are characteristic
impedance Zo = 53Q, signal delay 6.7 ps/mm, and DC resistance 13
Q·mm. The module heat sink is made of an aluminum-magnesium alloy;
it is 9.5 mm thick, has a high-grade surface, and is anodized on all
sides for electrical isolation. Its great stiffness guarantees the dimensional
stability of the module in the presence of the strong internal forces of
1500 N.

12.7.2.2 Sequential Multilayer Thin-Film Process


Figure 12-71 is a schematic representation of the principal process
steps for fabricating a microwiring substrate [140]. A 18-J.lm-thick Cu
foil is laminated with the aid of an adhesive foil onto a previously tested
and, if necessary, repaired, defect-free power or signal layer. This is
11-738 THIN-FILM PACKAGING

10

Figure 12-69. Multilayer Structure of a 17-Layer Microwiring Substrate: (1)


Spider, (2) engineering change wire, (3) programmable via, (4) pad layer, (5) redistribution
layer, (6) power layer, (7) signal layer, (8) reinforcing layer, (9) contact layer, (10) contact
spring, (11) repair wire.

followed by the photoimaging and etching of the Cu foil with 80-J.lm via
holes for the electrical interconnection of layers nand n + 1. After stripping
the photoresist, a laser beam is used to remove the exposed insulating
resin in the via holes down to layer n. A Cu layer is now deposited over
the entire surface area, including the blind holes. The top and bottom Cu

Figure 12-70. Design Layout. (a) Signal wiring; (b) power wiring.
12 .7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-739

Layer n
Res; ---~--........
U nitlal State:
Layer n (signal or ground layer) completed ,
tested and re aired _
Cu foi
65 11m Laminating (18 11m thick copper foil)

S 90 IIm-'l_.... _-r
Photoimaging

Etching
Formation of up to
S90llm
poooI----..-:.~- t Photoresist Stripping
6000 blind holes
between
layer n and layer n + 1

Laser Drilling

Metal Deposition (Cu)

Photoimaging

_ _ _ _ _ _ _ _.l-I_
M_eta
_ 1Deposition (Cu +Sn)

I
I
Photoresist Stripping

Etching
J Paltern formation on
layer n + 1

-- -
Layer
n+1 Sn Stripping L
Figure 12-71. Process Steps for Fabrication of Microwiring Substrate

layers are, in this way, electrically interconnected. The interconnecting


pattern is generated on the layer n + 1 by photoimaging, electroplating
and etching steps. A temporarily deposited thin Sn layer serves as the
etch resist. After optical and electrical testing the performance of any
repairs that may be found necessary, the described process steps are then
repeated for the next layer.
The same sequence of process steps-lamination, laser beam drilling,
photoimaging, electroplating, and etching-are likewise used for the fabri-
cation of the chip-mounting layer and the connector layer.

12.7.2.2.1 Reinforcement Substrate


The reinforcement substrate is a O.6-mm-thick double-sided printed
board of polyimide glass material. It has 1848 plated-through-holes of
11-740 THIN-FILM PACKAGING

0.3 mm diameter and nickel/gold plated pads on the connector side. All
other layers are added to the reinforcement substrate by a relamination
process. The reinforcement substrate has a form of 150 mm x 150 mm.

12.7.2.2.2 Signal, Power, and Redistribution Layers


The microwiring substrates are fabricated as single pieces in a work-
ing form of 150 mm x 150 mm. A completed, tested, and repaired layer
is laminated with a 18-J..lm-thick copper foil. The bonding material is a
glass-free polyimide layer of 25 J..lm thickness coated on both sides with
an adhesive of25J..lm. Two different bonding materials have been qualified:
acrylic adhesive and, more recently, epoxy adhesive.
Lamination is performed using hydraulic presses or, in the future,
in an isostatic vacuum gas press which is already qualified. The lamination
process and the bonding material meet the following requirements:

1. Void-free filling of the pattern, including the filling of blind holes.


2. The underlying interconnection pattern must not show up in the
surface of the applied 18-J..lm-thick Cu foil.

The desired hole pattern for 80-J..lm blind holes is transferred by


photoimaging from a chromium mask to the Cu layer by a contact exposure
process. A liquid aqueous processable positive photoresist is applied by
electrostatic spray. For the accurate alignment of mask and substrate, a
technique has been developed with which it is possible to align the mask
with the inner layer pattern with a tolerance of less than 10 J..lm.
After developing the photoresist and etching the Cu layer, the hole
pattern generated in the Cu foil is checked for completeness and the
observance of diameter tolerances by automated optical inspection using
CAD data as a reference. Missing or deformed holes can, if necessary,
be mechanically reworked.
The copper layer with the etched holes works as a mask during the
subsequent laser beam drilling. The intensive UV radiation of a KrF
excimer laser decomposes the unprotected resin by photoablation into
gases which are exhausted. The laser beam hits an area of 15 mm2 and
travels over the board in a scanning mode.
Laser beam drilling stops at the lower copper layer and assures an
absolutely clean base and nearly vertical hole walls.
A Cu-deposition step establishes a reliable electrically conducting
connection between the upper and lower layers in the blind holes drilled
by the laser beam. The Cu plating of these holes must exhibit high ductility
and good adhesion at the base of the hole so that no damage will be
caused by stresses that may build up at the elevated temperatures used
for fusing and soldering.
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-741

Metal deposition starts in a colloidal palladium catalyst with ultra-


sonic agitation and is followed by an electroless copper bath in which
about 0.5 !lm of copper is built up. Copper thickness in the blind holes
is increased by electroplating to about 10 !lm.
For photoimaging of the conductor pattern, a solvent-type photofoil
was initially used. Intensive work was done on the replacement of chlori-
nated hydrocarbon solvents, which has been prohibited by the European
Community since 1991, with aqueous type of photofoils and some progress
has been achieved.
Pattern plating of copper to a 20-!lm thickness in the blind holes
and 5 !lm of tin follows.
After alkaline etching of the copper and stripping of the tin metal
resist, an automatic optical inspection step and an electrical test on continu-
ity and isolation finishes the layer build. For the electrical tests, the inner
layers are mechanically contacted with bending needle adapters.
Line discontinuities which are found by optical and/or electrical
tests can be repaired by an electrode-gap welding process. Formed ele-
ments with all existing line configurations are available. Shorts and insuf-
ficient clearances are eliminated by a special machine. An ultrasonic hard-
metal stylus removes unwanted copper residues down to a minimum width
of 50 !lm.

12.7.2.2.3 Chip-Mounting Layer


The process and the base material for the chip-mounting layer are
identical with the inner layers. Instead of tin as a metal resist, eutectic
tin-lead is electroplated 8 !lm thick. This tin-lead deposit is fused in a
hot oil bath after etching the chip-mounting layer.
The further development of packaging systems [141] will be deci-
sively influenced by the progress achieved in microwiring substrate tech-
nology. Beginning with the present size of the substrates of 100 mm x
100 mm, the substrate area will increase by a factor of 4 or more.
In initial phase, the built-up layers will still be on one side; layers
on both sides will be implemented only in a further step. For this, a very
intensive development of this technology in relation to the quality of the
process and yield is necessary; new materials with a lower dielectric
constant have to be evaluated.
To limit the number of layers, the conductor pattern has to become
smaller. However, this can be done only to the extent permitted by the
requirements for the electrical performance of EeL (emitter-coupled
logic) systems. For CMOS application the wiring structures will shrink
dramatically. Linewidths and via holes in the range of 20 !lm are available.
The demand for fine patterns in the device layers will be decisively
influenced by the development of the outer-lead pitch of TAB devices.
It is expected that in the 1990s this pitch will shrink to 50 !lm.
11-742 THIN -FILM PACKAGING

12.7.3 Hitachi: Thin-Film Packaging for Mainframe Computers


Thin-film technology has been extensively applied in the compact
chip carrier named MCC (microcarrier for LSI chip) which presents the
possibly smallest hermetic package and is shown in Figure 12-72 [142].
Hitachi's M880 mainframe introduced a high-performance multichip mod-
ule [143] which consists of a multilayer ceramic substrate (MLC) and a
number of chips, each packaged on a MCC. The MCC has been designed
to offer the highest va interconnection density and smallest packaging
surface area by adopting full-surface flip-chip bonding and built-in termi-
nation resistors. It was generally agreed that for TAB or wirebonding
where terminals are wired out only from the fringe ofthe chips, it becomes
difficult to connect 400-600 terminals per chip. This is because the pin
pitch gets smaller than 100 Ilm for a chip size of 10 mm x 10 mm, and
it can cause interpin electrical interference, problems with reliability, and
manufacturability concerns. The design parameters of MCC are summa-
rized in Table 12-23. The MCC is based on a mullite/tungsten cofired
substrate of a size of 10 mm x 10 mm to 12 mm x 12 mm, depending
on the size of the LSI chip to be mounted. Mullite ceramics were chosen
for their low thermal expansion coefficient (3 .5 x 1O-6/0C) matching that
of silicon (3.0 x 1O-6/0C) and its low dielectric constant (5 .9) which gives

.. . . . . . . . . . . . . . . . . . . . . . . . . . _ ........................................ 0 ...................... eo . . . . . . . . . . . . . . . . . , . . . . . . . . . ... . . . . . . . . . . . . . . ..... . . . . . . . . . . . .

, AIN IIp
I ..... .......... .. . ........ . .. . . ...................... eo ........................ • .. ... • ... • .. • .. · _ ................................................... .

: !
i !,
~ ,
i i
; j

!
i
~~~L.._ _ _ _ _piiiiiiiill~
Ir-- " ' ' " - - - Thin film

::~::::;DI! t
T

Layer .
erumlc

1
ubslrale

Mullite
~-----L.--'----~-

Figure 12·72. Cross-Sectional View of MCC with LSI Chip and AIN Seal-
ing Cap
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-743

Table 12-23. Design Parameters of MCC Substrate

M880
Parameter 10 x 10 - 12 x 12 mm

Number of pads Signal: 252


Termination Resistor: 99 Total 528
Power, GND and Monitor: 177
Pad pitch MLC Side: 0.450 mm
BLM pitch LSI Side: 0.250 mm
Number of layers Thick film Conductor: 7
Thin Film Conductor: 5
Thin Film Resistor:

better electrical performance than conventional alumina. Matching of


the thermal expansion coefficients is crucial for highly reliable solder-
ball joints.
The thin-film wiring acts as redistribution layers to match the grid
pitch of the LSI chip ball-limiting metallurgy (BLM) and that of the MLC
through-hole. The thin-film conductor is aluminum, and the dielectrics are
polyimide (PI) and photosensitive polyimide (PSPI). The MCC has not
only thin multilayer circuit patterns but also incorporates thin-film termina-
tion resistors in it. The LSI packaging area is considerably reduced owing
to these built-in termination resistors compared to that of the conventional
modules with discrete resistors.
The fabrication process of the MCC has been well documented [142]
and starts with a mullite-tungsten cofired substrate. The substrate has
nickel-plated through-hole pads on the top which are arranged in a grid.
The ceramic substrates tend to have dimensional inaccuracy caused by
process variations, such as firing shrinkage. Hence, a pattern-matching
layer is first applied on the ceramic substrate to ensure electrical contacts
between the thin-film circuit and the thick-film through-hole pads which
may have location variations within the range of tolerance. The pattern-
matching layer is shaped into concentric circles around the through-hole
pad and is able to absorb locational mismatch between the pads and the
vias in the first thin-film layer.
The Al-polyimide combination was chosen as the thin-film multilayer
interconnection materials for the reason that plenty of process and reliabil-
ity data exist and there is a long history of IC fabrication utilizing these
materials in Hitachi [144,145]. At first, a wet-etchable polyimide is applied
as the first-level dielectric. The polyimide layer is thick enough to effec-
11-744 THIN-FILM PACKAGING

tively planarize the rough topography of the ceramic surface to assure


the process latitude for photolithography _The termination resistor material
is then deposited as very thin film on the polyimide by RF magnetron
sputtering_ The Cr-Si-O cermet was chosen because of its suitable sheet
resistance when deposited in a processable, submicron thickness range_
After fabrication of the thin-film resistors and formation of the first-level
vias, the first conductor layer is formed by sputtering A 1 followed by
photolithographic patterning. The subsequent layers are made by repetition
of the above processes, except that a photosensitive polyimide is used
instead of the wet-etchable polyimide. Finally, BLM's are formed by
sputtering CrINi-Cu and wet etching of the metal layer. The BLM's are
finished by nickel and gold chemical plating.
The built-in termination resistor is one of the special attributes of
the MCC. One MCC carries about 100 termination resistors. Figure 12-73
is a depiction of the structure of the termination resistors, which are
patterned in a doughnutlike shape and located around the first-level vias
in the most compact design.
Some other technological breakthroughs for the MCC fabrication
process are summarized as follows:

1. Stable and simple via formation through the thick interlevel dielec-
tric polyimide

Outer
14--- b - - - . t
Electrode

Inner
Electrode

R::rl2 lt 'n(b/a)
R: Device Re i tance
r: Sheet Re -istance

Figure 12-73. Structure of Built-in Termination Resistors


12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-745

2. Optimum metal composition and its processing to form reliable


via contact
3. Homogeneous large-area deposition of the thin resistor film
4. Stable etching processes for metals deposited on the surface of
reactive organic materials like polyimides
5. LSIs with over 500 lIO pins in the smallest area

The main purpose of the MCC is to reliably interconnect between


the LSI lIOs and the MLC lIOs by matching the lIO pad pitch of the LSI
to those of the MLC. However, the MCC offers some additional significant
functional merits as follows:

1. Highest chip-level interconnection density achieved at the time of


the implementation of this technology
2. Higher packaging density by substituting discrete resistor compo-
nents with built-in termination resistors
3. Full LSI testability and easy LSI screening
4. Easy replacement and safe handling of individual LSI chips
5. Smallest possible hermetic LSI package

12.7.4 nCHIP: Packaging with AI/Si0 2 on Silicon Substrates


nCHIP's nClOOO@/nC2000@ Silicon Circuit Board (SICB@) techno-
logies provide high-density interconnect substrates for a wide range of
applications. Wiring pitch as low as 25 !lm and the ability to route
interconnect under the die without interference from thermal vias enables
close placement of the ICs. Signal routing is accomplished on two metal
layers; the wiring density allows even interconnect-rich devices to be
routed in two layers.
nClOOO SiCBs are appropriate for most CMOSITTL (transistor-
transistor logic) designs. The circuit speeds that can be supported are
dependent on the complexity of the design, the size of the SiCB, and the
edge speeds of the lIO drivers. For designs incorporating typical CMOS
drivers, practical system clock frequencies range from 40 MHz for very
large designs to about 100 MHz for smaller designs. It is usually possible
to handle a few signals of even higher speed if care is taken in layout.
nCHIP's nC2000 SICBs are appropriate for fast BICMOS, ECL or GaAs
designs whose performance requirements exceed the nC 1000 capabilities.
Two power planes are used for power distribution. One of the planes
may be split if an additional power supply is required, Both nClOOO and
nC2000 technologies incorporate an integral decoupling capacitor between
the two planes, providing all of the decoupling capacitance required for
most designs. The integral capacitor is effective to much higher frequencies
11-746 THIN·FILM PACKAGING

than surface-mounted chip capacitors because the capacitor is distributed


throughout the SICB, the effective inductance between it and any die is
only that of the wirebonds to the chip's power supply bond pads (typically
1.2 nH divided by the total number of power supply bond wires).
In addition to providing better power-supply decoupling, the integral
capacitor can contribute to a significant reduction in module size. In
MCMs built with other technologies, large numbers of chip capacitors
can add 10-20% or more to the substrate area required, increasing module
size and cost while reducing performance.
The integral decoupling capacitor is one of two major features that
distinguish nCHIP's substrate technologies from those used by most other
thin-film substrate manufacturers. The other distinguishing feature is the
use of Si02 as the dielectric between metal layers. The reasons for the
choice of Si02 as the dielectric are more varied and perhaps less obvious
than the reasons for including an integral decoupling capacitor, so they
will be discussed at some length.

12.7.4.1 The Si0 2 Dielectric


The cross-section drawing in Figure 12-74 shows the construction
of an nClDOO Silicon Circuit Board. All of the intermetal dielectrics are
made of Si02, except for the thin dielectric forming the decoupling capaci-

Bond
Wire

Wire Bond
Pad

. - - M3 ( ignal)
A)IiIII-----iiiiiiiii-_II....-'~:::a.IIIIIIJ+--- M2( ignal)

--iiiiiiiiiijiiF::::
iO,

Ca
De;C::U:P::li:;":.g MI(Power)
MO(Ground)

Figure 12-74. Cross Section of nCIOOO Silicon Circuit Board


12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-747

tor. Si02 is routinely used as the primary dielectric in integrated circuits


(although ICs use much thinner oxides) and is efficiently integrated into
the SICB manufacturing process. The selection of the dielectric material
is critical to an MCM technology, with issues of cost, electrical and
thermal performance, density, manufacturability, and reliability all being
important. nCHIP's Si02 dielectric was chosen for a combination of
properties which make it well suited for MCM applications.
The specific characteristics of SiOz which make it attractive can
be briefly summarized as follows: high thermal conductivity, low cost,
inertness and impermeability, desirable mechanical properties, and good
consistency and control.

12.7.4.1.1 Thermal Conductivity


The thermal conductivity of nCHIP's Si02 dielectric has been mea-
sured (using specially designed thin-film test structures) at room tempera-
ture as 1.35 W1m K. This number is approximately an order of magnitude
higher than typical polymer dielectric materials (virtually all of which
fall in the range of 0.1-0.2 W/m K). For example, the most common
form of polyimide, Dupont's Kapton@, has a published room-temperature
thermal conductivity of 0.155 W/m K approximately one-ninth that of
Si02•
This difference in thermal conductivity can have a significant effect
on the design of MCMs. It permits modules to be built with no requirement
for thermal enhancements such as "thermal vias" or "thermal wells," both
of which impair wiring density and hence lead to larger modules. Thin-
film terminating resistors, a significant source of local heat (as high as
400 W/cmz in typical applications), can also be embedded in the Si02
dielectric without problems.
As an example, the thermal resistance for the complete nC2000
series dielectric stack is 0.15 cm2.oCN. Thus, a chip dissipating 20 WI
cm2 would incur only a 3°C temperature rise as a result of heat conduction
through the dielectric layers. As a result, no thermal vias are necessary
and the full area underneath the chip is available for routing. In contrast,
an MCM built with a polyimide dielectric of equal thickness would see
a 26°C temperature rise, which would normally be unacceptable. Such a
chip would have to located in a thermal well (an area where no wiring
is permitted underneath the chip), compromising the routability and hence
increasing the size of the MCM. Further, some alternative MCM techno-
logies have total dielectric thicknesses substantially greater than that used
in the nC2000 architecture; in these technologies, even chips dissipating
as little as 2 WIcm2 may require some use of thermal vias in order to
meet a tight thermal budget.
Certain polyimide-based processes use a special form of thermal
via, the "solid thermal pillar." Such thermal pillars can be quite compact
11-748 THIN-FILM PACKAGING

and highly effective, thereby minimizing the need for added substrate
area. The disadvantage of such processes, however, is higher complexity
and cost. They are generally believed to be at least twice as expensive
to manufacture as the conventional "open" or "staircased" vias, due to
the need for planarization or via filling steps at each layer.

12.7.4.1.2 Process Cost


Many factors enter into the cost of an MCM. One key factor is the
size of the module. Other things being equal, a denser MCM will allow
the fabrication of more substrates per wafer, and with higher substrate
yields. As a result, the cost of an MCM increases with area, at a faster
than linear relationship (e.g., an area increase of 20% would typically
increase cost by more than 20%). In addition, there are other system-
related costs (the package for the MCM and the PC board on which it is
typically mounted) which also scale in proportion to area.
As a result, to the extent that an MCM technology permits the
fabrication of denser modules, it will lead to a lower-cost product, both
at the module level and at the system level. Because nCHIP's Si02 dielec-
tric does not require the use of thermal vias, it generally allows the design
of denser modules.
Another important element of MCM cost is the cost of the materials
themselves. A Si02 dielectric is economical, representing approximately
10% of the total processed wafer cost (accounting for raw materials, labor,
and depreciation). In contrast, the polyimide dielectric in typical multilayer
interconnect substrates is estimated as costing 20-30% of the total pro-
cessed wafer cost, depending on the type of polyimide and other factors.

12.7.4.1.3 Inertness and Impermeability


In some applications, inertness and impermeability can be critical
in the choice of dielectrics. The moisture absorption of nCHIP's Si02 is
too small to be measured by conventional techniques. Moreover, the
diffusion constant for moisture, gases, and ionic contaminants is many
orders of magnitude less than that of polymers; this is why glass seals
are deemed hermetic, whereas polymer seals are not. These attributes are
among the reasons why Si02 is the preferred interlayer dielectric for most
integrated circuits, particularly those intended for nonhermetic packages.
(Polyimides are often used as top-level overcoats, principally for alpha-
particle protection, but in such cases they are being used as a supplement
to Si02; the polyimide is seldom used as the interlayer dielectric nor is
it the primary passivation.)
Thus far, most thin-film multichip modules have been confined to
hermetic packaging schemes, often because of concern about the reliability
of the high-density interconnect in the presence of moisture and other
contaminants. Polymer-based thin-film substrates can be highly permeable
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-749

to moisture. For example, a typical polyimide (Kapton H) will exhibit a


30% variation in dielectric constant (from 3.6 to 3.9) as the ambient
humidity is varied from 0% RH to 100% RH at room temperature, due
entirely to moisture absorption. Besides the variation in electrical proper-
ties, there is a reliability concern: one well-known MCM substrate manu-
facturer has reported that its aluminum-polyimide substrates fail even
after relatively brief exposures to the conventional "steam pot" test; they
concluded that a copper interconnect is required for nonhermetic applica-
tions. Another practitioner with a copper-polyimide substrate system deter-
mined that their interconnect required 100% cladding of the copper con-
ductors with nickel in order to reliably pass a steam pot test; this, of
course, increases the cost of the process.
The longevity of a thin-film metallization system in an organic
dielectric will depend on the precise details of the dielectric chemistry
(every polymer dielectric is different), the presence of dissolved gases
(02, CO2), and residual ionic contamination present (e.g., from a die attach
adhesive). This makes the generation of valid reliability data a complex
matter, and the resulting data cannot be generalized to other MCMs
employing similar but not identical processes. In contrast, the use of an
inert, hermetic Si02 dielectric demonstrated high reliability essentially
independent of the details of the external environment, and so these results
are of general validity.
Severe environmental tests have been performed on substrates using
the nCHIP's Si02 dielectric. These include the use of a steam pot in which
5% salt (NaCt) has been added as a corrosive agent (50% of the substrate
is immersed in the saltwater solution, and the remainder is subject to the
salt spray). Such an environment would destroy many MCMs employing
polymer dielectrics. However, because of the use of a hermetic dielectric
material (Si02) as passivation and interlayer dielectric, nCHIP's substrates
exhibit no interconnect corrosion, even after 100 h at 127°C.
Finally, the absence of any measurable moisture-absorbing tendency
simplifies the manufacturing of SiOrbased substrates, as there is no need
for bakeouts to drive out moisture, nor any requirement for "vent holes"
in the metal plates to let moisture escape during the bakeouts.

12.7.4.1.4 Mechanical Properties


Si02 films possess mechanical properties which are important in the
fabrication of multichip modules. These include a low thermal-expansion
coefficient (-0.5 ppm/0C), a high tensile strength (>1 GPa, or 150,000
psi; over six times that of polyimide), a high elastic modulus, and an
ability to be deposited under a precisely controlled compressive stress.
One of the difficulties associated with manufacturing MCMs is the
large net stresses which can be built up in the substrate, resulting in a
tendency for the substrate to bow and perhaps even delaminate. This
11-750 THIN-FILM PACKAGING

tendency is quite pronounced when using polymer dielectrics on silicon


substrates. Polymer dielectrics inherently have a residual tensile stress,
which is the result of their high thermal-expansion coefficient and high
cure temperatures. A typical polyimide will have a residual tensile stress
of over 40 MPa at room temperature (about 6,000 psi). In addition, metal
layers such as copper or aluminum will themselves have residual tensile
stresses of around 160 MPa as a result of the thermal annealing that occurs
during normal processing at 300-400°C. Thus, both the metal and the
polyimide films tend to bow the substrate upward into a concave shape.
The bow can be quite large (e.g., 10 mils of warp for a typical process after
all the films are deposited). This can pose difficulties for manufacturing,
including problems with lithography, vacuum chucking, heat transfer dur-
ing vacuum processing, and so on. It also leads to difficulties during
dicing of the wafer due to the large built-in stresses. For silicon-on-silicon
"flip-chip" modules, additional complications can result from the tendency
of the substrate to flex during thermal cycling (as both the metal and the
polyimide have high expansion coefficients relative to silicon), leading
to problems with solder fatigue.
In contrast, nCHIP's SiOrbased fabrication process produces a
stress-balanced substrate. The Si02 dielectric layer is deposited in such
a way as to be under a moderate residual compressive stress at room
temperature (about 50 MPa). This compensates for the tensile stresses in
the metals. In fact, because there is on average about three to four times
as much dielectric material as there is metal, the net stress in the wafer
is very nearly zero (i.e., the final product has essentially no tendency to
bow). (nCHIP uses a wafer flatness measurement as one of its in-process
SPC steps.) Also, the ratios of thermal expansion, elastic moduli, and
thicknesses of metal and Si02 are such that the wafer exhibits very little
flexing tendency during thermal cycling.
The strength of Si02 is an important feature when assembling wire-
bonded multichip modules. For example, during ultrasonic wedge bond-
ing, local compressive and shear stresses of 200 MPa (30,000 psi) are
routinely generated by the bond tool. Such a stress level is essential in
order to generate a high-quality weld between the wire and the bond pad.
Such stress levels can exceed the yield point of polyimide, causing it
to deform during the bonding process. Although the exact mechanisms
involving in wirebonding are poorly understood, it is known that wirebond
quality can be compromised when bonding onto thin-film metallization
atop polyimide dielectrics. Solutions to this difficulty have been found
(e.g., by attaching very thick metal pads over the bond-pad areas), but at
added cost.
The strength and adhesion of Si02 does not degrade with temperature.
This is particularly important when performing rework operations on a
multichip module (chip removal). A typical chip rework process involves
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-751

briefly heating the module up to 200°-250°C in order to remove the


chip from the substrate. Other investigators have reported an inability to
perform such rework procedures without damaging (delaminating) the
thin-film interconnect underneath the chip attach site. As a result, they
have had to switch to soft, low-Tg adhesives in order to have a viable rework
process. Such adhesives introduce compromises such as low strength,
excessive moisture absorption, high expansion, and unfavorable creep
properties.

12.7.4.1.5 Consistency and Control


Si02is a well-understood material and easy to work with in manufac-
turing. Compositionally, it is essentially identical to that used in multilevel
metal ICs as an interlayer dielectric. Thickness control is typically ±2%
across a wafer and wafer to wafer. This leads to better impedance control,
which can be important in very high-frequency applications.
In contrast, organic polymer dielectrics are relatively variable materi-
als. Their primary virtue (i.e., that they can readily be chemically engi-
neered to have specific properties for specific applications) makes it more
difficult to control a process in high volume. Cure cycles can be complex
and must be controlled precisely if the final result is to be consistent.
Film thickness can seldom be controlled to better than ±1O% through spin
or spray coating. Some of the largest electronics manufacturers have
decided to manufacture polyimide dielectrics in-house to maximize their
control over material consistency.

12.7.4.1.6 Si02 Ruggedness


As with all ceramic materials, Si02 is a brittle material (i.e., it
does not plastically deform prior to fracturing). However, when prepared
properly, this fracture strength is very high; nCHIP has determined from
extensive flexure tests that its Si02 multilayer films are stronger than that
of the silicon substrate-that is, their measured flexural strength exceeds
1010 dyn/cm2 (i.e., 1,000 MPa, or 150,000 psi). This is three times the
strength of typical alumina ceramic materials, which are used in most
high-reliability single-chip IC packages.
Some people with past experience in semiconductor processing recall
issues with Si02 cracking at thicknesses beyond 1-2 ~m. There is indeed
such an issue if the Si02 is deposited with conventional techniques that
do not control its stress state. Without a controlled compressive stress at
deposition, the resultant film will be in a biaxial tensile state and will
indeed crack at large thicknesses. nCHIP's deposition process, however,
results in a slight compressive stress on the final film, which has no
tendency to crack.
11-752 THIN-FILM PACKAGING

12.7.4.2 Key Electrical and Thermal Characteristics

• Signal layer resistance 15 mWsquare (nCl000), 6 mn/square


(nC2000)
• Intermetal (Si0 2) dielectric constant ~3.75
• Power/ground-plane resistance 30 mWsquare (nClDOO), 15 mW
square (nC2000)
• Integral decoupling capacitance ;?:50 nF/cm2
• Designed for low-inductance wirebonding-typically 1.2 nH
• SICB thermal resistance (including all standard layers)-0.17
cm2.oCIW
• Temperature coefficient of resistance (TCR) of aluminum intercon-
nect-O.4%fOC
• Excellent thermal match between silicon base, power/ground/inter-
connect layers and Si02 dielectric. Overall substrate expansion is
matched to silicon integrated circuits.

Si02 is often listed in references as having a dielectric constant of


3.9. That number is valid at low frequencies for the thermally grown
oxides commonly used in integrated circuits. nCHIPs deposited (PECVD)
oxide has a somewhat lower dielectric constant, particularly at higher fre-
quencies.

12.7.4.3 nC1000 SICS Feature Sizes

Listed in Table 12-24 are some of the key feature sizes for nClOOO
and nC2000 layouts. All are minimum dimensions.
Because of the available routing density of the M2 and M3 signal
layers, utilization of the SICB area is generally determined by module
assembly requirement rather than by routing constraints.

Table 12-24. SICB Feature Sizes

Feature nClOOO nC2000

Metal width (M2 and M3) 10 ~m 16 11m


Metal space (M2 and M3) 15 ~m 20 11m
Metal pitch (M2 and M3) 25 11m 36 11m
Via pitch 35 11m 62 11m
Via diameter 5 11m 101lffi
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-753

12.7.4.4 MCM Package Assembly


In the nCHIP assembly process, dice are mechanically and thermally
bonded to the SICB using thermally conductive epoxy. Back-side bias,
if required, is provided by contact pads underneath the die attachment
site. The SICB, in turn, is bonded to the package body or heat spreader
with another epoxy layer. These attach methods are proved to have good
thermal performance and high reliability.
The module's dice are electrically interconnected through very short
(25-mil horizontal length), low-inductance, fine-pitch wirebonds. nCHIP's
wirebonding capability employs ultrasonic wedge bonding of I-mil Al
wire. These can be placed on a 4-mil pitch and have a typical inductance
of only 1.2 nH, which is much less than in a conventional IC package
because of the very short wire and the matched pad-outs of the chip
and SICB.
nCHIP SICBs can be designed for use in a wide variety of packages
or carriers. Most common are the well-known pin-grid-array quad ftat-
pack formats, although unusual configurations can be designed to meet
special needs. Because nCHIP SICBs do not contain moisture-absorbing
organic materials, they may be used in both hermetic and nonhermetic
packaging. For thermal reasons, most packages are of a cavity-down
configuration to allow for heat-sink mounting. Low-power or conductively
cooled designs, however, can use a cavity-up configuration.
Because MCMs are typically used for high-performance (and there-
fore high-power) designs, a module's ability to dissipate a large amount
of heat is an important issue. nCHIPs SICB technology is well suited to
high-power designs, because the SICB materials both tolerate and conduct
heat well. Thermally conductive epoxies can be used to attach IC dice
to the SICB and to attach the SICB to the package or carrier. The package
material(s) may be chosen to meet the module's thermal performance
requirements (e.g., copper-tungsten instead of alumina for high-power ap-
plications ).

12.7.5 Digital: VAX 9000 Multichip Packaging


A high-performance multichip packaging device needs to provide
high speed, high density, controlled-impedance interconnections for inter-
chip signal distribution [13,146-148]. It needs to provide a high-current
DC power distribution system with small DC voltage drops and also a
stable AC power supply system under a large transient, simultaneous
current-switching condition. From the packaging point of view, a low-
thermal-resistance path is required to cool the high-power chips, and a
stable, inert environment ensures long-term reliability in a variety of
storage, transport, and use conditions.
As the advanced semiconductor devices used in high-performance
11-754 THIN-FILM PACKAGING

computers improve in performance and circuit density, corresponding


increases in chip I/O pins, switching speeds, power dissipation, and cooling
requirements also result. These requirements then drive toward higher
interconnect wiring density with low-defect-tolerant manufacturing,
thicker DC power distribution metal layers, and thinner interlayer dielec-
trics for power decoupling. Furthermore, chip attachment and assembly
technologies must be developed to handle the higher chip power density
and to maximize the total chip area as a percentage of the package area.
In the design of the VAX 9000 technology [149,150], a wafer fab
technology supplemented by a printed wiring board type of technology
is chosen to meet the above-mentioned substrate signal and power distribu-
tion requirements. The resultant technology is called HDSC (high density
signal carrier) which has been shown to be highly manufacturable and
extendible as the semiconductor chip technology advances. In chip attach-
ment, packaging and connector approaches, a back-side chip attachment,
cooling on a base plate, a two metal layer TAB tape, together with signal
and power flex circuits in a mechanical housing form the MCV (multichip
unit). This approach facilitates chip and module testing and impingement
air cooling and is field replaceable.
In this section, the HDSC and MCV designs are reviewed in some
detail and process overviews are described. The technology evolution
strategy for future high-end packaging is discussed based on some pro-
jected high-performance, high-power semiconductor device characteris-
tics. Several cases of actual multichip packaging designs are given for
meeting these chip characteristics with incremental technology improve-
ments for optimized system performance.

12.7.5.1 The HOSe Design and Process

12.7.5.1.1 The HDSe Design


A complete cross section of the HDSC is shown in Figure 12-75
where there is a total of nine Cr/Cu metal layers interfaced with poly-
imide layers as dielectric isolation. The whole structure is laminated on
a Cu/Ni base plate where chip site cutouts are made for semiconductor
devices and where cooling fins are mechanically attached for heat removal.
The bottom four metal layers of the structure is called the power core for
power distribution purposes. Each of the metal layers is 18 /lm thick,
patterned with drilling pads, breathing holes for moisture evolution during
polyimide curing and chip site cutouts, and so on. Between Ml and M2
and also between M3 and M4, a thin polyimide layer of 10 /lm is deposited
to form two 30-nF capacitors for decoupling. A thicker polyimide layer
(25 /lm) between M2 and M3 is used to planarize the surface. A top and
a bottom polyimide layer are used for electrical isolation and passiv-
ation purposes.
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-755

Figure 12·75. Cross Section of High-Density Signal Carrier (HDSC)

The top half of the structure is known as the signal core for signal
distribution. It consists of a triplate transmission line with a characteristic
impedance of 60 O. Each of the X and Y lines has a resistance of 1.0 Of
cm. The dimensions are 18 J.Lm wide and 10 J.Lm thick and the wiring
tracks are on 75-J.Lm centers. Plated copper posts are used to connect
between M2 and M3 and also M3 and M4. Ground and reference metal
layers M I and M4 are 4 J.Lm thick. The total active wiring area on the
HOSC is 10 cm x 10.5 cm.
Through-holes are drilled between signal and power cores which
are subsequently metallized for electrical connection. The same metalliza-
tion step is used to pattern the top layer pads for TAB tape attachment
and also for signal/power flex circuits. The top metal pads are made of
25-J.Lm-thick copper. The pad pitch is 100 J.Lm for TAB pads and 150 J.Lm
for signal flex pads. The interlayer polyimide has a thickness of 25 J.Lm
processed with multiple coatings.

12.7.5.1.2 The HOSC Core Processes


Power Core: The power core process starts with an aluminum wafer
coated with a 1O-J.Lm-thick poly imide layer, a thin CrfCu layer is deposited
by sputtering which is subsequently plated to 18-J.Lm-thick copper. The
top Cr cap layer is then sputtered for adhesion purposes to the subsequent
poly imide layer. The metal is then coated with photomasking followed
by a wet-etching step to pattern the metal and a dry-etching step by plasma
to clear the polyimide surface leakage. A thin layer of polyimide is
deposited on top of the metal for isolation. The process then repeats itself
three more times to complete the fabrication. In-process tests are performed
to monitor the plating and etching steps to stay within specifications.
11-756 THIN-FILM PACKAGING

Signal Core Similar to the power core, a polyimide passivation


layer is first deposited on the aluminum wafer. Then the bottom reference
plane, a 4-Jlm-thick Cr/Cu/Cr metal layer, is sputtered on the wafer surface
which is masked and etched. The top of the metal is now coated with a
25-Jlm polyimide layer by multiple spinning and baking, followed by a
final cure. Next, a Cr/Cu plating base is sputtered on the poly imide surface.
Photoresist and lithography steps are performed to define a stencil of lines
and pads. After electrolytic plating in a sulfuric-acid-based solution to
plate up the first interconnection copper metal (M2) layer of 10 Jlm, the
photoresist layer is stripped. On top of the lines, a second thicker photore-
sist layer is coated which is exposed and developed for the post masking.
The second plating step is performed to plate up the copper post (M2P)
on top of the M2 layer for a thickness of 30 Jlm. The photoresist layer
is then stripped and the plating base is etched.
For dielectric isolation, a poly imide layer (33 Jlm) is coated on top
of the posts. To accomplish planarization, several spinning and baking
steps are again taken, which is followed by the final cure. At this point,
a photoresist layer is coated on top of the polyimide, patterned, and used
as a mask for a plasma etch of the polyimide to form vias. The vias are
designed to fall on top of the posts for interlayer contact. The residue
photoresist is stripped after plasma etching. The process then repeats for
the M3, M3P, and the vias on top of M3P. The 4-Jlm M4layer is processed
the same way as Ml. Finally, a passivation polyimide layer is coated on
top of M4 and etched with plasma vias, with the same process as previously
described to complete the signal core process. The cross sections of a
completed signal core and a power core are shown in Figure 12-76.

12.7.5.1.3 The HOSe Assembly Process


Stainless-steel metal rings are attached on top of the completed
signal cores with epoxy adhesive. The front side of the signal core is
protected with an organic tape when the Al wafer is dipped into acid
and etched away completely. A copper-polyimide "decal" is now fabri-
cated which is in tension and hence is taut like a drum head. This is
because of the thermal-expansion properties of the materials chosen. The
decal is aligned on top of a matching power core and laminated onto it
with a vacuum press after an epoxy adhesive layer is attached on the top
of the power core. Again, a decal is formed as shown in Figure 12-77.
Subsequent process steps are similar to those used in manufacturing the
advanced printed wiring board.
The decal is drilled with fine (250-Jlm) holes to connect power pads
between the two cores. Sputter and wet chemical steps are taken to seed
the top surface and inside the drilled holes. Dry-film resists are used to
pattern and to form a stencil which is followed by acidic copper plating
for the PTH (plated-through-hole) metal and the top metal. After the steps
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-757

Signal Core

Power Core
Die-Site
Cutout Vee
PTH Site Vent Hole
Area
~ ~
~

.. ,

- \-
'---- Polyimide 1-'- ___ Aluminum Substrate Removed

Figure 12·76. Cross Sections of HDSC Cores

of plating, photoresist strip and plating base etch, solder is deposited on


top of the metal by screening and reftow. Laser is used to cut through
predefined polyimide lines where copper is removed to delineate die site
cutout regions. The finished decal is then laminated on a copper-nickel
base plate. The extra decal is trimmed away by laser to complete the
HDSC process.

12.7.5.2 Evolution of High-Performance Computer Packaging


As the IC technology continues on its path for a narrower linewidth,
more on-chip circuit density and speed improvements, the IC chip size,
the number of chip I/O pads, pad density, power, cooling requirements,
and switching speed will all increase. Furthermore, the use of the 6-in.
Cu-polyimide base wafer fab technology to develop the improved HDSC/
MCV for meeting the projected IC chip performance requirements will
continue. In the technology development planning, packaging of maximal
chip density and the number of chips are also assumed; obviously, the
actual circuits used on the chip and the number of chips used on the
package will have to be dependent on detailed system architectural design
and partition, circuit and packaging designs, and other considerations.
A set of projected bipolar ECL chip technology development cases
,
11-758 THIN-FILM PACKAGING

Ring

" ' - - - - - Signal Core

" ' - - - - - Power Core

Figure 12-77. Formed Decal for HDSC Assembly

BPI, BP2, and BP3 are listed next to the base case of the Motorola Mosaic
3 technology used in the VAX 9000 MCV as shown in Table 12-25.
Some of the key chip technology characteristics are shown together with
the corresponding requirements placed on the multichip units which will
be referred to as cases A, B, and C As the linewidth reduces, the circuit
speed increases significantly for each step; however, a far larger increase
in circuit density is realized such that the number of logic circuits is more
than doubled in each case and the on-chip embedded RAM is also provided.
Based on these IC chip requirements, it is clear that every aspect of
the multichip interconnect and packaging specifications must be improved
(i.e., higher DC and AC power distribution requirements, higher intercon-
nect density, smaller delay, more efficient cooling, etc). From the physical
technology of the multichip packaging point of view, this implies thicker
DC power core metal, thinner interlayer dielectrics between power core
metals, smaller interconnect line pitch for the signal core, and so forth.
Some analysis, design, and experimental work were performed to come
up with an interconnect and packaging design architecture appropriate
for meeting the chip physical and performance requirements.
For MCV power distribution, the VAX 9000 MCV power core
process architecture of area plating and wet etch of copper-chromium
followed by a thin polyirnide coating is believed extendible from the
present thickness of 18 I..Im to at least 25 11m (50% increase), the interlayer
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-759

Table 12-25. Chip Technology Development

MCU Technology MCU MCU-A MCU-B MCU-C

IC Technology MOSAIC 3 BPI BP2 BP3


Emitter Size (11m x 11m) 1.2 x 3.5 0.5 x 1.5 0.3 x 1.2 0.2 x 1.0
Metal I pitch (11m) 4.0 2.8 2.2 1.5
Unloaded gate delay (ps) 80 50 40 30
Chip type Gate array Gate array Standard cell Standard cell
Max. No. of gates/chip 10K 25K lOOK 300K
Embedded RAM bits None 16K 64K 256K
TAB
ILB (11m) 100 90 75 Flip chip
OLB 11m) 200 140 75 Flip chip
Power/chip (W) 30 45 100 150
PinS/chip
Signal 256 350 500 940
Power 104 150 200 380
Total 360 500 700 1320
Die size (cm) 0.98 1.18 1.40 1.72
Note:
I. MCU Case B assumes a power TAB to bring power to the center of the chips.
2. MCU Case C assumes flip-chip direct bump to interconnect (no TAB).

dielectric coating from the present thickness of 10 Jlm to 7 Jlm on the


thicker power core metal layer. Additional DC power distribution improve-
ments can be made by bringing in the DC power from all four sides of
the HDSC (the existing design is two sided). The total DC power in-feed
can therefore increase by another factor of 2 for the same percentage of
DC drop. For AC power distribution, the simultaneous switching require-
ments on the MCV can improve by 30% by going to a 7-Jlm polyimide
coating. Further improvements on the MCV technology requirements will
need a new power core technology architecture; that is, thicker metal
layer laminates for DC power and a separate "capacitor core" with thin
metal and thin polyimide layer built inside the signal core to provide
additional low-inductance capacitance for decoupling.
For a signal interconnect density increase, a straightforward scaling
factor of 2 for reducing both the vertical dimensions and horizontal dimen-
sions of the VAX 9000 transmission-line structure can be used. For
example, a factor of 2 scaling of the transmission-line dimensions for the
existing MCV design will result in a 9-Jlm x 5-Jlm line cross section, a
dielectric layer thickness of 12.5 Jlm, and a line pitch of 38 Jlm, thereby
realizing a line density increase of a factor of 2. But a factor of 3 scaling
increase will result in a 3x line density increase at an expense of line DC
resistance increase by a factor of 9 to 9 Wcm, which is too high. However,
11-760 THIN-FILM PACKAGING

by allowing the characteristic impedance of the line to drop to 50 n from


60 n, the use of low-dielectric-constant polyimide (from 3.5 to 2.4) and
an increase of the metal line cross-section aspect ratio (thickness: width)
from 0.5 to 0.66, the metallinewidth and thickness can both be increased
from the scaled-down version and with the corresponding adjustments of
dielectric thickness, the line DC resistance can be held to 4 Q/cm, an
acceptable value for these applications.
For die attach and assembly, it is believed that the TAB technology
can be extended by reducing both the ILB (inner lead bonder) and OLB
(outer lead bonder) pitches to 75 !lm in one to two steps. By designing
the same OLB and ILB pitch, this advanced approach is called an orthogo-
nal TAB in which a fan-out from a tight ILB to a wider OLB is not
needed and the TAB lead lengths can significantly be reduced to improve
signal integrity. Further significant density increase will require flip-chip
bump technology.
Based on the above discussions, the multichip packaging require-
ments and the trend for the physical features are summarized in Table
12-26.

12.7.5.3 Evolution of High-Performance Packaging Technology


Once the interconnect and packaging materials and process architec-
ture are determined, the physical features options can be developed and

Table 12-26. Evolution of MeU for High Performance Systems

MCV Perfonnance
Features Requirement Trend Physical Feature Trend

Signal interconnect density Higher Smaller wire pitch, more layers


Z. 60~50n Thinner dielectric, smaller wire pitch
Line resistance Can increase from Thinner dielectric and metal
1.0 n/cm to 4.0 n/cm
DC power Higher Thicker metal layer in power-core
layer, laminated power core
AC power Higher transient Thinner dielectric in power core,
separate cap core, orthogonal
T AB/wire, flip-chip bumps
Size, number and density Higher Orthogonal T AB/wire, flip-chip
of chips, chip I/O TAB, flip-chip bumps
number and density
Cooling Higher Improved die attach on baseplate,
top side cooling
Off-chip connector Higher Higher density signal flex, pin grid
array
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-761

evaluated with respect to the performance and density gains on one hand
and to the schedule and risk factors on the other hand. In addition, the
resultant physical technology features for a given design across different
areas (i.e. signal/power core, HDSC assembly and MCV assembly) have
to be consistent and compatible to a given chip set; hence, further design
trade-offs have to be made. The three MCV designs, A, B, and C for the
corresponding bipolar technologies 1, 2, and 3 specified in Table 12-25,
are discussed in the following.

12.7.5.3.1 MCU-A
To meet the requirements of bipolar technology of BPI, no change
is necessary of the HDSC and MCV materials and process technology
architecture from the base case. There is no change of equipment set
either. However, only incremental technology design rules are needed to
meet the higher-performance requirements, and the technology risks are
considered low.
The power core metal thickness is increased to 25 Ilm per layer and
the interlayer polyimide coat is reduced to 7 Ilm. For the signal core, a
0.7 scaling factor is used for the transmission linewidth, pitch, its thickness
and the corresponding dielectric thickness, and the metal post heights.
The signal line pitch is therefore 50 Ilm with a resistance of 2 Wcm. The
rest of the design and the core processes remain the same. The ILB and
OLB pitches become 90 Ilm and 140 Ilm, respectively. Signal flex circuits
stay the same, power flex circuits are attached on all four sides. BPI chips
in a 3 x 3 array remains the same. The total circuits on the MCV increase
by 7.6x with a higher speed, and the power dissipation goes up 50% to
450W.

12.7.5.3.2 MCU-8
Because of the large increase of the on-chip integration level of the
BP2 technology (lOX over the base case) and the corresponding increase of
chip I/O pad number, switching speeds, and power dissipation, significant
process changes are needed for the signal core, as shown in Figure 12-78.
A thin polyimide layer of 4 Ilm is deposited between two power/ground
layers to enhance the decoupling capabilities of the signal core. A 2x

Capacitor
Core i~::;:;==:::;;::::==:::=======:s:~~==::$

Figure 12-78. MCU-B Signal Core Design


11-762 THIN·FILM PACKAGING

signal line density increase with a line pitch of 36 Ilm over the base case
is needed. The line resistance becomes 3 Q/cm. The power core stays
basically the same as MCU-A, but the power vias and PTHs are increased
to account for the lower DC drop and the power supply inductance needed
for the maximum 900 W specification for this MCU.
For the chip attach, the TAB ILB and OLB are both 75 Ilm pitch
to bring the total chip 110 to 714. Because of the orthogonal TAB tape
design, the leads are much shorter (200 Ilm) and the tape can be simplified
to a single metal layer. For OLB bonding, single-point lead bonding is
shown to be more viable compared with gang bonding to reduce bonding
defects. In addition, epoxy die bond materials, the bond line, and the pin
fin design all need to be improved to account for the 3x power dissipation
increase. The base case chip design is still 3 x 3 and several small SRAMS
chips can still occupy a chip site.

12.7.5.3.3 MCU-C
As shown in Table 12-25, the BP3 technology further increases the
on-chip circuits by a factor of 3 over the previous case of BP2, the chip
1I0s increase by a factor of 2 and the power dissipation on chip increases
by 50%. Because of the number of chip 110 pads increases, two plane
pairs of signal interconnect layers are required, as shown in Figure 12-79.
The lower signal plane pair is the same as MCU-B for the long intercon-
nects. For the upper plane pair, the wiring pitch is reduced to 25 Ilm for
high-density localized wiring. The line resistance becomes 4.5 Q/cm, but
the more than 2x wiring density increase over the MCU-B case is expected
to support an array of 4 x 4 BP3 chips.
For chip attach, flip-chip solder bumps are used in this case. To
support 1320 1I0s on a 1.72-cm chip, 3 rows of perimeter 110 pads with
a pad size of 75 Ilm and a space of 75 Ilm are needed to meet the
requirements.
Two significant requirements of the power core now drives the
physical design changes: the total DC power increase to 2000 W and the
low thermal-expansion rate of the power core to match that of the silicon
to ensure the reliability of the solder bumps. The construction of power
core in this case is made oflaminates of individual low-thermal-expansion
ceramic substrates sheets such as a glass-ceramic or AIN substrate. Each
substrate consists of a thin bottom metal layer (4 Ilm), a thin dielectric
film (1.8 Ilm) for the capacitor, and the thick top metal layer (75 Ilm) for
power distribution. The laminates contains PTH for intersheet connection,
bottom metal pads for pin attachment, and top metal pads for signal core
decal attachment.
The signal core is attached to the power core by solder/adhesive
lamination as shown in Figure 12-79 where the solder bumps on the signal
core is reflowed and laminated onto the metal pads on the power core
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-763

Ch'Ip

- -
~

- -
I I I I
(

- - -
Interconnect
Reference Plane
Signal
Core
Interconnect

- - Reference Plane

Solder pad----../

Power
Core

Figure 12-79. Solder/Adhesive Lamination of Signal to Power Core for MCU-


C Design

substrate in a heated vacuum press. A high-temperature B-staged thermal


setting adhesive starts to activate and cross-link after the solder-reflow
temperature to bond the signal core to the power core to complete the
lamination process.
The thermal conduction path is now through the back side of the
chip and the top side of the package. Well-known approaches that have
been reported in the literature such as TCM [151], bellows, and so forth
can be used to accomplish the thermaVmechanical requirements.

12.7.5.4 Technology Summary

Under the assumption that the bipolar technology, driven by perfor-


mance and density considerations, continues to increase in circuit speed,
110 pin number, and power, the challenges for packaging/interconnect
designers and materials/process architects are to come up with a multi chip
module designs that can match these requirements. The challenges are in
signal distribution, power distribution, chip attach/rework, thermal man-
agement, and environmental protection. Although the focus throughout
the section has been on the bipolar circuit technology, similar design
11-764 THIN-FILM PACKAGING

concepts could also be applied to other IC technologies such as CMOS


and GaAs.

12.7.6 NEe: Thin-Film Packaging for Mainframes and


Supercomputers
12.7.6.1 Polyimide-Ceramic Substrate for High-Performance
System
As described in Chapter 2, "Packaging Wiring and Terminals,"
interconnection performance of the wiring substrate is primarily deter-
mined by the unit signal delay of the wiring patterns and the packaging
density. In order to minimize the unit signal delay, a reduction of electrical
capacitance of the signal lines is necessary. The electrical capacitance
depends mainly on the substrate technologies, which can be reduced
by using a low-dielectric-constant material for the insulator layers and
fabrication of fine signal lines. Increased packaging density apparently
requires a closer LSI chip mounting and higher cooling rate. When the
distance between the LSI chips becomes shorter, the interconnection wir-
ing density must be higher; therefore, either narrower signal lines and
pitches with less layers or wider signal lines and pitches with more
layers is inevitable. The former demands the implementation of thin-film
technology with low-dielectric materials and the latter indicates thick-
film technology with inorganic materials for insulator layers such as
cofired alumina-ceramic or glass-ceramic. In high-performance multichip
modules (MCM), the LSI chip power dissipation is generally high; there-
fore, a higher thermal-conductive structure is imperative. Such a high-
thermal-conductive structure is practicable by separating the heat and
wiring paths on the substrate. Usually the heat generated in the LSI chips
is removed through a thermal-conductive mechanism attached on to the
multichip module like IBM's Thermal Conduction Module [151].
Thin-film wiring technologies are featured by low electric capac-
itance and fine signal lines of high conductivity as compared to thick-
film technology. As a consequence, more LSI chips of high 110 terminals
(>400) can be interconnected onto a thin-film substrate with fewer signal
layers for faster signal transmission. As discussed previously, polyimide
material is a suitable candidate for thin-film dielectric layers due to its
excellent electrical, thermal, and mechanical properties. Alumina- or glass-
ceramic multilayer base substrates with electrical wiring in them (active
substrates) provides more wiring flexibility and the associated performance
improvements, in addition to mechanically supporting the LSI chips and
cooling mechanism. For these reasons, a package of polyimide thin-film
wiring on alumina- or glass-ceramic multilayer base substrates has been
developed [11,152]. Figure 12-80 shows a polyimide-ceramic substrate
developed by NEC for high-end mainframes and supercomputers [152].
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-765

Figure 12-80. NEC Multilayer Substrate Cross Section

The alumina- or glass-ceramic base substrates are fabricated by


cofired thick-film technology. Power/ground layers and clock distribution
layers are formed inside the base substrate because these wirings are
common among the personalized substrates for each system. The NEC
thin-film technology is unique for the combination of thick-film inorganic
base substrate and organic thin-film signal layers. The major benefits of
cofired alumina- and glass-ceramic substrates are good mechanical
strength for LSI chip mounting and attachment, compatibility with the
cooling mechanism, thermal coefficient of expansion (TCE) matching that
of silicon chips, and capability of high 110 terminal number interconnection
with high power supply. The advantages of organic thin-film layers are
fine signal lines with low electric capacitance for higher system perfor-
mance. As shown in Figure 12-80, the thin-film multilayer wiring has
eight conductive layers, of which four layers are for signal. The signal
wiring pitches are 75 I..Im, wiring width 25 !lm, and via holes in the
polyimide layer 60 !lm square.
The low dielectric constant of polyimide allows high-speed signal
11-766 THIN-FILM PACKAGING

transmission at 6 nslm and high-density interconnections of 53 lines/mm.


The size of the substrate is 225 mm x 225 mm square and 5.5 mm thick.
Photosensitive polyimide technology is used to fabricate the thin-film
multilayer substrate for the simplified patterning process relative to non-
photosensitive polyimide [153]. Up to 60,000 via holes of 60-IJm square
are formed in each polyimide layer of 20 Jlm thickness using photolitho-
graphic processes for mass production.

12.7.6.2 Wafer Scale Substrate for GaAs Packaging

Thin-film substrate is the best choice for attaining low-capacitance


signal lines because of its fine-line capability. The currently used alumina-
ceramic substrates for thin-film package, however, often have very fine
voids on the surface generated by the pulling out of alumina particles
during polishing. These voids are undesirable for fine-signal-line pattern-
ing. On the other hand, the surface of crystallized silicon or sapphire
wafers is smoother and more suitable for fine-line patterning. Apparently,
this is the main reason for their wide application in VLSI technology.
Initially, these two types of wafers were both employed for the pilot
fabrication of a thin-film structure, but it was found that silicon wafers
were not good candidates for this purpose. In the process of thin-film
layer stacking, silicon wafer of 0.5 mm thickness was broken due to the
stress developed within the thin-film structure. However, the sapphire
wafer had sufficient strength to withstand the stress produced by a thin-
film structure with up to five layers of polyimide without breaking. The
sapphire substrate was 10 cm square which was cut off from a 6-in. wafer.
Figure 12-81 shows the cross section ofthe sapphire wafer-based package
for supercomputers [154], onto which high-speed bipolar gate arrays (300
gates, 1.5 W max.), GaAs RAMs (Ta = 1.5 ns, 4 W max.), and GaAs
gate arrays (2.6K gates, 6.1 W max.) are mounted.
The two layers of thin-film signal lines of 25-Jlm width and 60-Jlm
center-to-center spacing were formed by sandwiching between ground
mesh layers. The conductor was plated copper, and insulator polyimide
layers were spin-coated. In this package, TAB-style bipolar gate arrays
were directly mounted onto the polyimide thin-film layers face down and
leaded chip carrier (LCC)-type GaAs devices were mounted by soldering
onto the copper pads on the substrate.
The sapphire wafer is attached to the alumina base substrate by
silver epoxy resin. The alumina base substrate has bonding pads on the
top periphery, 110 pads on the bottom periphery, and through-hole wirings
through the substrate to interconnect the top and bottom pads. The electri-
cal interconnections between the sapphire wafer and alumina substrate
are provided by gold ribbons using the gold-to-gold thermal-compression
bonding technique. This multichip package was 11 x 11 cm2 in size with
12.7 COMMERCIAL APPLICATIONS OF THIN-FILM PACKAGING 11-767

Polyill1;d~
Dielectric
L.y~r (20 "'"

Au Rlbbou

.pphire Wdu

Alumina Sub.... le

Figure 12·81. Cross Section of Sapphire-Wafer-Based GaAs MCM for NEC


Supercomputers

a total of 420 1I0s (228 signals and 192 power and ground) and consumes
up to 150 W. To cool this new package, the board with four of these
multichip packages mounted on it was immersed into fluorocarbon liquid.
By convecting the fluorocarbon liquid at 8 cmls flow rate, all the semicon-
ductor devices were kept below 55°C of junction temperature.

12.7.7 MicroModule Systems: Thin-Film Packaging Compatible


with IC Processing
Thin-film multi chip module packages are manufactured using stan-
dard automated semiconductor equipment in a Class 100 cleanroom [37].
Multilayer copper-polyimide interconnects were processed on 150-mm
aluminum or silicon wafers. The aluminum wafer has the advantages of
low cost and good thermal conductivity, and, at the same time, can be
employed as the power plane and one electrode of the integrated capacitor,
whereas, the silicon wafer provides an excellent thermal-expansion match
to that of Ie chips, allowing for high-density chip attach. The main features
of polyimide processing are spin coating, curing, proximity printing for
conformal mask patterning, and plasma etching for signal and thermal
via formation through the cured polyimide layers. For copper processing,
a sputtered seed layer is followed by electroplating. The metal layer is
defect sensitive and, therefore, is patterned by projection lithography to
11-768 THIN-FILM PACKAGING

increase the yield. A total of four or more metal layers can be sequentially
built up this way. A cost model for the implementation of this technology
justified the cost-effectiveness of this technology. At volume production,
substrate cost can be as low as $5/in. 2 and the total cost of a MCM-D,
consisting four chips each with a pin count of 100, is around $10-$12.
An example of the application of this technology is the fabrication of
either 2- or 4-Mbit cache SRAM memories for PCs and workstations
where space and cost are the premiums [155].

12.7.8 Other Commercial Schemes


Similar technology capabilities have been reported by other Japanese
as well as American companies. Table 12-27 lists an example of thin-

Table 12-27. Thin-Film Technology

Material Specification Remarks

Substrate
Material Alumina, glass-ceramic
Size 150 mm maximum
Conductor Metals
Conductor metal Au, Cu, Al
Middle metal Pd, Ni, Mo
Interface layer Cr, Ti, Mo, TiIW
Thickness 5 11m
Resistance Electroplating Cu, Au 5 mntsq.
Linewidth 25 11m minimum
Line spacing 40 11m minimum
Adhesion 4-6.5 kg/2 mm' (soldered) Peeled by 6-mm-dia. wire
~8 kg/1.5 mm dia (brazed) 45° Pull by 0.4-mm-dia. pin
Resistor
Material Ta-N, Ni-Cr-Si
Sheet Resistance 10-50 ntsq.
TCR ±IOO ppml°C
Others Cr-AuTiIW-Au
Dielectric
Material Polyimide
Thickness 20 11m
TCE (200-700) x 1O-1rC
Dielectric constant 3-4 50 Hz
tan Ii 0.2% 50 Hz
Volume resistivity >10'6 ntcm 20°C
Insulation resistance >106 V/cm 20°C

Source: Courtesy of NTK.


12.8 EMERGING COMMERCIAL TECHNOLOGIES 11-769

film technology being practiced by NTK on substrates up to 150 mm


using multiple levels of polyimide-copper or gold metallization.
Planar polyimide (20-40 Jlm thick in these examples) is deposited,
followed by reactive ion etching of the polyimide to form via holes with
a Si02 layer as the etching barrier. The holes are subsequently metallized
by sputtering a conformal layer of Cu or Au. Patterning of both vias and
the next conductor layer is performed with a single photolithography and
etching step. Reactive ion etching processes have been successfully used
to etch vias as small as 12 Jlm in diameter in the 20-40-Jlm-thick polyimide
films at an etch rate of 1 Jlmlmin. The sidewall angle is controlled at
65°±2°.
Another example involves the use of photosensitive polyimide and
copper plating technology to form very-high-density multilayer intercon-
nections. The outstanding accomplishment by NTT has been the develop-
ment and application of a selective electroplating and photolithographic
process to achieve fine copper conductors (3 Jlm wide) and very small
via holes (15 Jlm in diameter) through relatively thick photosensitive
polyimide [77]. This technology has been claimed to result in low-cost
fabrication of high-density multilayer interconnections because of the
short process turnaround time.

12.8 EMERGING COMMERCIAL TECHNOLOGIES IN


THIN-FILM PACKAGING
As described above, all the multichip module technologies can be
generally classified into three generic categories; those based on ceramic
(MCM-C), those based on printed wiring board (MCM-L), and those based
on thin-film deposition (MCM-D). In practice, however, combinations of
these are more common, such as thin films on cofired-ceramic substrates
with one layer for redistribution of the chip grid to that of ceramic grid
or more layers for interchip connection, as indicated in Figure 12-82.
Another combination that is receiving great interest is thin films on PWB
by sequential deposition of one or more layers, in particular by additive
processes. In addition, a number of companies, IBM and Hitachi in particu-
lar, have also practiced the other combination, lamination on ceramic
substrate. A number of other alleged low-cost processes have also been
reported, mostly based on either large-area or roll-to-roll processes which
are capable of high-throughout, high-yield polymer film deposition and
processing, as well as metallization. A few of these are described below.

12.8.1 IBM's Surface Laminar Circuitry


The best thin-film package proposed and developed to date is perhaps
the surface laminar package developed by IBM Japan in 1986 [16]. In
general, it overcomes the limitations of drilled through-holes in printed
11-770 THIN-FILM PACKAGING

r--:--:-----:-;-------.+---Si hip

+ - - - - Pb- n older

Cr-Cu- i-Au --+--....... "';""';"-4-- Fan Qui Lin.

Iud
Polylmid~
oplur. Pld
63 Layen or
Glas -Ceramic Substrate
with Copper
+--i-- VIA

Figure 12-82. Thin-Film Redistribution on Glass-Ceramic Substrate

wiring board technology, enhances wiring density to the level of conven-


tional thin-film photolithography technology and reduces the cost of the
total wiring because of large-area processing. The process is relatively
simple in that one or two layers of polymer-copper thin films, as illustrated
in Figure 12-83, are deposited sequentially on a standard printed wiring
board with or without internal wiring, using materials and processes that
are very similar to the printed wiring board fabrication. On a printed
wiring board about 600 mm x 700 mm in size, a photosensitive epoxy
(Probimer 52®) from Ciba Geigy is deposited to about 50 f..lm thickness
followed by exposure and development processes. Copper metallization is
subsequently achieved by e1ectroless seeding and electroplating. Typical
dimensions attainable to date are 100-f..lm vias and 75-f..lm linewidths.
IBM compared this technology with its other thin-film and ceramic techno-
logies, as discussed in the previous section, for a 200-MHz workstation
application and concluded that direct chip attach to this SLC board met
all the performance requirements that were previously believed to be met
only by ceramic and thin-film technologies.

12.8.2 Sheldahl's Roll-to-Roll Process


Figure 12-84 illustrates the process starting with 200 ft of polyimide
film [156]. The typically 50-f..lm-thick polyimide film has a low modulus
with expansion matching that of copper. Via generation is accomplished
in 12 in. x 12 in. sections by laser ablation on a 200-f..lm grid. A 2000-
A thin layer of copper is then sputtered on both sides as well as in the
vias, instead of the conventional electroless deposition. Using this as the
electrical contact, a 5-f..lm film of copper is subsequently deposited, and
wiring is subtractively etched. In applications where thicker copper is
- -
12.8 EMERGING COMMERCIAL TECHNOLOGIES 11-771

(a)
I Formation or internal circuit

1- - Coating or interlayer insulator


with photo-epoxy

g
. __7-J\1..-'___
- -'
Formation or via-hole by
expose &develop

Drill through hole

Catalyst seeding & masking

Electroless copper pattern plating

(b)
Item New Process Subtractive

Pattern
bape

ub Inl. ub....I •
...
-
.,
0 Pattern ± 2 - 3pm ± 10 - 20 11m
u
Thickness
...
0
Pattern
Widtb ± S - IOllm ± 20 - 40 I'm
AbUUy M. production MISS production

~-~~
-IOO/lSO
~'
~
Patterning
Ability
" ....•....... ,,' --.J

Mounting
ReliabiUry Conductor
ai:>
l. ...

Figure 12·83. Low-Cost Fine-Line Thin-Film Process on PWB


11-772 THIN-FILM PACKAGING

(c) MCM-L(SLC)
Flip-chip Assembly Process

SLC Panel MCM Chip

~
SMCIPIN Thru-Hole
~
(Burn In)
Parts Assembly

~
Chip Place +-------f

Reflow Chip Remove

Electrit Test ~
Encapsulation

Final Inspection
Figure 12-83. (Continued)

required, additive plating through photoresist is accomplished. Metalliza-


tion is critical for stress control within the film. Next, a liquid polyimide
is applied to the metallized layer and cured followed by laser ablation to
open up opennings in overlay film to expose underlying pads. Electroless
Au or Pd is done on electroless Ni to form pads for layer-to-layer connec-
tion. A number of these layer pairs are stacked, aligned, and laminated
using Z-axis conductive adhesives. A typical structure built by this tech-
nology is illustrated in Figure 12-85. Typical thin-film dimensions achiev-
able are 25-~m vias, 50-~m lines and spaces, and 125-~m pads.

12.8.3 GE's Conformal Multichip-Flex


Figure 12-86 depicts the overall process of this approach [157]. A
typical I-mil-thick flexible polyimide film with copper patterned traces
12.8 EMERGING COMMERCIAL TECHNOLOGIES 11-773

on both sides and plated vias connecting the bottom and top traces is
available from several vendors, such as the one by Sheldahl discussed
above. A dielectric layer is deposited on either or both of the two surfaces
to protect metallization or serve as a dielectric layer, which is done by
the vendors. Chips and components are precisely placed face down onto
the bottom side of the panel flex with a layer of polymeric thermoset
adhesive. Vias are laser drilled through the flex down to the chip pads
to provide chip interconnection. Next, a sputtered Ti barrierl2-6-~m Cu
provides both via and upper-level wiring metallization. Thus, a total of
three layers of interconnection is fabricated to satisfy most low-cost MCM
applications, although more layers can be fabricated onto it. The final
operation involves molding the IC and the thin-film package. Typical
thin-film dimensions are 2-mil vias and lines spaced 2 mil apart. The low
cost of this package is attributed to the large-area as well as low-cost
materials and simplified processes. The flexible film provides the addi-
tional benefit of conformality.

12.8.4 DYCOstrate Thin-Film Technology


The process developed by Hewlett-Packard GmbH, Germany [158]
was claimed to be of low cost enabled by major enhancement of plasma
technology that was previously abandoned by IBM due to its uncontrolla-
bility and high cost. The concept of DYCOstrate is simple. It starts with
a double-copper-cladded 50-~m- thick polyimide foil that is patterned by
typical processes of expose, develop, and subtractive etch to expose the
polyimide at via sites. The via holes are generated in the polyimide film
by dry plasma etching with a shortened processing time of about 20-25
min. The vias are plated for top to bottom connection and lines are
patterned and etched afterward. The final step to complete the 2D DYCO-
strate fabrication is an electroless NilAu. Additional interconnection layers
are fabricated by sequentially laminating single-copper-cladded dielectric

Via
MetaUize
Electroplate
~Expose Develop
~
Geueration Copper

~
d __ Polyimide i...t'_ Polyimide ~
Plate Cure I'" Image
i-_--I

~E' ZLink® Stack


Laminate
Final
~ Apply &AIlgn Test

Figure 12-84. Sheldahl RoU-to-RoU Process Flow


11-774 THIN-FILM PACKAGING

material to both sides of the 20 OYCOstrate. Again, the outer copper


foil is patterned to expose the dielectric polymer for plasma etching of
vias to provide interlayer connections. The sequential buildup process
described above can be repeated to obtain substrates with six or more
layers. Among many variations of this process, one allows the 20 OYCO-
strate to be laminated to a carrier by means of standard B-stage epoxy
prepreg. In this case, electrical connection between the 20 OYCOstrate
and the FR4 multilayer substrate is accomplished by mechanical drilling
and plating. Figure 12-87 illustrates various structures that can be built.
The typical wiring dimensions of this substrate are compared in Table
12-28 with standard PWB.

12.8.5 IBM's Integrated Rigid-Flex Thin-Film Package


One of the most sophisticated multilayer thin-film carrier technolog-
ies developed to date is a package developed by IBM referred to as high
perfonnance carrier (HPC) [159, I 60]. It preserves the advantages of low
cost and volume production characteristic of PWB technology while pro-
viding the high wiring densities and high 110 counts of ceramic and thin-
film technologies. HPC is also highly integrated in that carrier, flex cable,
and connector are integrated into one structure.
The basic building block of HPC technology is a five-layer core
consisting of two signal and three power/ground layers as shown in Figure
12-88. The structure has a triplate cross section for reduced coupled noise
and EMI protection. The HPC is fabricated using a composite material
consisting of amorphous silica filler particles in a polytetrafluoroethylene
(PTFE) or Teflon®matrix supplied by Rogers Coporation. Wherever neces-

CondUC-llvt:
•
P.. nlde

opper
•
onduC'lor

m Addltl,·.
~ Polylmld.

O FUm
rolylmldr

Figure 12-85. Sheldahl Substrate Cross Section


to Lase~ dorill vias in I \ I \ I 60 Laser drill vias
polYlmlde film 1 -_ _-' to chjp

2. eu metallization \ J \ J"'-...oI!

7. Sputter TilCu, plate J ",..


3. Pattern Cu "~_,, I
thick Cu, pattern

........
U'I

40 Deposit upper
layer dielectric

50 DoDd chip with 8. Add plastic substratE.. • !II


adhesive

Substrate Chip

Figure 12·86. GE's Multichip-on-Flex Process Flow


2DxC DxC
2D 4D
".~ ,..,~ . :"
r .....·.... '-i(- ,.... ~ ....
C"
1- 1:-X -;
2DC-F
4DC-F
"': ::3iC.:,:;..,:.~";",,:l.:J
-
.....
.....
Ol
)! :~'!I~,"!i!',~!~!i~!r;:~i:~:I! i ,~i 2DxC2D DxCD
2DC 4DC
.-!II ,
.~

i!I ~,i1~I!i :i l!i' :!lil'~l:~lit;I~li ~i :;'I~if i~ ~:


Figure 12-87. Overview of Various DYCOstrate Construction Alternatives
12.8 EMERGING COMMERCIAL TECHNOLOGIES 11-777

Table 12-28. Comparison of Design Features on Standard PCBs and


DYCOstrate

Feature Standard PWB DYCOstrate Today DYCOstrate Tomorrow

Via diameter 225-325 Ilm 751lm 30--50 Jlm


Pad diameter 550--525 Ilm 200--350 Ilm 100--150 Ilm
Conductor width 100--150 Ilm 75-125 Ilm 40--60 Jlm
Space width 100--150 Ilm 75-125 Ilm 40-60 Jlm
Constructions Not applicable 2D; 2D x C; 2D x C2D 4D, ... , x D; x D x C x D; +F

sary, a copper-clad invar alloy typically 1 mil thick is utilized as an


expansion compensator and reference plane to provide thermal-expansion
control of the core as is shown in Figure 12-88. The properties of the
PTFE and copper-invar-copper compensator used in fabricating the core
are listed in Table 12-29. The details of the fabrication are discussed
elsewhere [159,160]. The multilayer joining of the cores is accomplished
by means of lamination which provides both high-yield metallurgical
bonding of via lands on adjacent cores and adhesion of polymer between
the lands, as illustrated in Figure 12-89. The key attributes ofHPC technol-
ogy are listed in Table 12-30. The integration of carrier and flex cables
into a single structure is illustrated in Figure 12-90.

12.8.6 IBM's Thin-Film Transfer Technology


This technology was developed for the purpose of reducing the cost

f/.J-;. . .
of MCM-D processing. Its uniqueness lies in that the multilayer polymer-

Signal Line

-- - -
Signal Land

-- - Compensator
-
Figure 12-88. High-Performance Carrier Two-Signalfl'hree-Power Core
11-778 THIN-FILM PACKAGING

Table 12-29. Properties of Dielectric Material and Copper-Invar-Copper


Compensator for HPC

Dielectric Copper-Clad Invar

E,. 2.8
Dissipation 0.0012 at 10 Mhz
TCE 5.5 ppml°C (20-120°C)
9.4 ppml°C (20-380°C)
TCE x,y 25 ppml°C
TCEz 40 ppml°C
Elastic modulus 1.0 OPa (146 kpsi) 90 OPa (13 Mpsi)
Thickness 0.5-2 mils (12-50 11m) I mil (25 11m)
Dimensional stability 20 ppm
(Repeatability)
Water absorption 0.15%
Melt transition temperature 327°C
Decomposition temperature 400°C
Elongation 495% 1.4-2.0%
Tensile strength 550 MPa
Mass resistivity 0.60 Q·gmlm2

metal thin-film structure is built up on a reusable large-format glass


substrate and then transferred onto other product substrates of choice, like
laminates, silicon, alumina, cofired ceramic, and so forth. This process
allows the tailoring of the thin-film structure before laminating to the
product substrates. Cost reduction is achieved from several aspects. First,
the thin-film process becomes independent of the choice of product sub-
strate and a single standardized foundry would be able to support the
thin-film fabrication for many different product types. As a consequence,
the cost of retooling the production line is spared. Second, the reusable
large-format glass substrate enables the optimization of processing and
the adoption of large-area processing from the flat panel display industry
for high-throughput, high-yield and low-cost fabrication. Third, this tech-
nology allows the production of low-volume applications at the cost of
high-volume manufacturing.
The basic steps and key features of this technology are illustrated
as in Figure 12-91 [161] and summarized as follows. A polymer (polyim-
ides as in this technology) is first coated on the glass substrate serving
as the release layer. Standard multilayer thin-film interconnection structure
is subsequently fabricated. A rigid frame is attached to the surface after-
ward for the purpose of handling and testing the multilayer thin-film
aggregate after release, as well as controlling the distortion of the thin-
film pattern when it is separated from the glass substrate. Stresses origi-
12.8 EMERGING COMMERCIAL TECHNOLOGIES 11-779

Figure 12·89. High Magnification of Metallized Bond Between Adjacent Core


Vias. Courtesy of David Light, IBM.

nated from the processing and material property mismatches can result
in severe distortion of the thin-film aggregate after release. With proper
frame choice, distortion was controlled well below 0.02%. The separation
of the thin-film aggregate from the glass substrate was accomplished by
a scanning laser ablation process through the underlying glass substrate.
The laser fluence was well controlled so that the polymer release layer
is ablated without any damage to the thin-film wiring layers. The glass
substrate must be transparent to the laser radiation and the polymer release
layer must be thick enough (> 10 Ilm in this case) to avoid any significant

Table 12·30. HPC MCM·L Technology Capabilities

Typical Min./Max.

Via pitch, mil (~m) 20 (500) 9 (225)


Via diameter, mil (~m) 4 (100) 3 (75)
Linewidth, mil (~m) 2 (50) I (25)
Line spacing, mil (~m) 4 (100) 1.5 (28)
Lines/channel 2 I
Wiring capacity, in.lin.2/1ayer 100 (40) 100 (40)
(wires/cm/layer)
Layer count 6-25 36
Cross-talk, % 1.5 0.1
Time delay, ps/cm 55.8
Zo, n 50 ± 5 50-80
11-780 THIN-FILM PACKAGING

Flex Cable

Figure 12-90. Multiple Flex Cables Exiting HPC Carrier

stressing of the thin-film structure induced by the ablation process. De-


pending on the product substrate onto which the thin-film decal is lami-
nated, the process differs. For passive substrates, the lamination is simply
done with polymeric adhesives. For active substrates, surface preparation
may be necessary for both the substrate and the thin-film aggregate.
Electrical joining can be realized by any of Au!Au thermal-compression,
liquid-phase, or transient liquid-phase processes. Uniform application of
pressure was found very critical to ensure good electrical joining.

12.9 INTEGRATED PASSIVES IN THIN-FILM PACKAGE


The integration of active circuit functions using silicon integrated
circuits is a major success story of the twentieth century. This integration
12.9 INTEGRATED PASSIVES IN THIN-FILM PACKAGE 11-781

I-+-I-+-I-+-I-+-I Build thin film multi-up on glass

Separate from glass


by a dry laser process
Dice and laminate to Silicon,
diamond or AIN substrates
1r=~7 • ....4t-------I!~!!!!II~8

Dice and laminate to FR-4 epoxy or other cards

Figure 12-91. Flexible Manufacturing Process for Multichip Modules

has brought forth the development of the computer and telecommunication


industries and the infonnation age in which we now live. This success
of silicon devices merely further underscores the lack of advancement
in integration of passive components, notably resistors, capacitors, and
inductors that are the complementary components of active microelec-
tronic circuits.
The role of passive components has changed from a minor portion
of electronics system to a major occupant of package realty and important
cost driver. The number often exceed by an order of magnitude the number
of active les in many mixed-signal devices in both consumer and industrial
applications. Significant efforts have been and are being made to develop
a new class of packages with integrated inductor/capacitor (LlC) and
resistor/capacitor (RIC) circuits using advanced materials and processing
to provide cost and reliability improvements over conventional noninte-
grated components.
The primary advantage of passive integration is the reduced number
of electrical contacts or transitions. A traditional assembly has its internal
contacts between the components, the transition from the components to
the attachment materials, usually solder or epoxy, and finally the transition
11-782 THIN-FILM PACKAGING

from the attachment materials to the interconnects on the substrates. By


integration, these transitions and the associated electrical performance
losses are reduced drastically or eliminated entirely.
A second advantage is increased reliability. Failures occur mainly
at mechanical transitions or interfaces between dissimilar materials and
geometrical inhomogeneities. Stress concentration often present at these
places originated from mismatch of material properties (e.g., thermal
coefficient of expansion (TCE) and Young's modulus) and geometry
effects. Merely by reducing the number of transitions and simplifying the
geometry within the package by means of passive integration, reliability
of the package is greatly increased.
An additional advantage of integrated passive components is cost
savings. Few additional processing steps are required for passive compo-
nent integration, whereas a large number of assembly steps are eliminated
resulting in cost saving. Less materials are needed for integrated passives
than discrete passives resulting in additional cost reduction.
Higher packaging efficiency is another big incentive for passive
integration. Component size and component count are the typical drivers
for assembly size. By burying the passive components in the substrate,
the footprint constraints are removed. The same passive components can
be effectively spread "two dimensionally" within the package substrate
or the package itself in traditionally unused or waste area.

12.9.1 Current Development and Applications of


Integrated Passives
There are various approaches to implement integrated passive com-
ponents in thin-film packaging. Several materials like tantalum-silicon,
TiW, or tantalum nitride have been employed as thin-film resistors avail-
able at various resistance values. For integrated capacitors, Si3N4 , Si02,
or (Pb, La,Ba, Zr, Ti) oxides with a high permittivity have been used
for dielectrics. Nickel-iron permalloy magnetic material combined with
copper conductor were used to fabricate integrated inductors. Several
investigators have successfully implemented integrated passive compo-
nents for MCM-D fabrication which are summarized below.

12.9.1.1 Resistors
AT&T Bell Labs [162] utilized existing IC facilities for the develop-
ment and manufacturing of integrated silicon-on-silicon MCM packages
using polyimide/aluminum thin-film processes. They have fabricated Ta-
Si thin film resistors by sputter deposition and reactive ion etching (RIB)
patterning. Sheet resistance values of 8-20 Q/square have been achieved.
At W.L. Gore and Associates [163], standard IC equipment was
used to fabricate polyimide-Cu on glass-ceramic substrate MCM-D pack-
12.9 INTEGRATED PASSIVES IN THIN-FILM PACKAGE 11-783

ages with embedded resistors. Better performance and cost savings were
claimed for the reduced parasitics, increased routability and component
reduction. In a total of eight metal layers, the fourth layer from the bottom
was dedicated for thin-film resistors. TiW was chosen as a resistor material
for compatibility with the existing thin-film deposition and etching equip-
ment. The resistor was sputtered using a 15/85 wt.% TiW target. The as-
deposited sheet resistance was 2.4 Q/square with a 300-nm film thickness
and deposited resistivity of 72 m Q·cm. Curing processes of the subsequent
polyimide layers above the thin-film resistor layer would further increase
the sheet resistance to 3.2/square. The patterning of the thin-film resistors
was accomplished by the deposition and patterning of a 500-nm Si02
hard mask layer followed by TiW layer etching in hydrogen peroxide.
The shape of the thin-film resistors are determined by the 62-Q termination
resistance to the power plane. One potential concern with this technology
is the poor thermal conductivity of polyimide which limits the prompt
transfer of heat generated by the thin-film resistors.
Sputtered, high-resistance tantalum nitride has been used as a resistor
material for thin-film multichip-module application by NIT [34], NTK
[57] and Boeing [164]. In NTT's high-speed ATM switching systems,
thin layer of tantalum nitride was deposited on a cofired alumina-ceramic
substrate and patterned to form resistors. Six layers of Cu-polyimide
interconnection were built thereafter. The tantalum nitride thin-film resis-
tors were shown to be very reliable upon current supply and self-heating
for 3000 h and humidity and high-temperature tests as well. No significant
resistance change was observed. Signal reflection was virtually eliminated
with the integration of termination resistors. In the application by Boeing,
Ta2N thin film was deposited directly on the uppermost po1yimide layer.
The resistor films were patterned using RIE in a SFJ02 mixture. The
patterned resistors had a value of 50±1 Q and a temperature coefficient
of resistance (TCR) of ±100 ppm/°C. Experiments were performed to
assess the feasibility of laser trimming a resistor located on the polyimide
without causing any damage to the underlying interconnects which indi-
cated the possibility of rework on this type of resistors.
The High Density Electronics Center at University of Arkansas has
investigated into the feasibility of embedding thin-film resistors, capacitors
and inductors in flexible polyimide films for both MCM-D and MCM-L
applications [165]. NiCr, TaN, and CrSi were chosen as the resistor
materials. Standard fine line lithography was used to fabricate the devices
on a 25-flm or 50-flm-thick polyimide film. A wide range of resistance
from 50 Q to 106 Q was obtained. In addition, the same technology was
also used to fabricate thin-film capacitors of up to 22-nF capacitance and
inductors of up to 137-nH inductance. TaxOy and BaTiOx were used as the
dielectric materials, whereas a Cu line was used to form the spiral inductor.
Deutsche Aerospace (Dasa) has employed NiCr buried layer as the
11-784 THIN-FILM PACKAGING

thin-film resistors in their double-face populated MCMs for microwave


transmit/receive (TR) modules in radar application [166]. Typical sheet
resistance was 35-100 Q/square. Laser trimming was used for the defini-
tion of the resistors.

12.9.1.2 Capacitors
Several inorganic dielectric materials have been studied for inte-
grated thin-film capacitors as listed in Table 12-31 [167]. In the AT&T
Bell Labs integrated silicon-on-silicon MCM packages discussed above,
single-layer Si3N4 or dual-layer Si02/Si 3N4 were the dielectric materials
to form the thin-film integrated decoupling capacitors. The additional Si02
layer safeguards against power/ground shorts due to defects in the nitride.
In principle, the dielectric is sandwiched between the highly doped Si
substrate acting as a ground plane and the subsequent aluminum contact.
For the formation of floating capacitors, a single Si3N41ayer is sandwiched
between a Ta-Si bottom electrode and the first-level aluminum contact
(Fig. 12-92 [162]). The Si3N4 layer was deposited by LPCVD whereas
the Si02 layer was simply by thermal growth in an oxidation furnace.
Capacitance values of 33-40 nF/cm2 have been obtained. In Fujitsu's
process for integrated capacitors [168], Ba(Zr,Ti)03 (BZT) thin film was
deposited on a Pt/SiOiSi substrate by multiple cathode RF magnetron
sputtering. A dielectric constant value of 146 was achieved at a Zr content
of 1.5 mol%. Dimos et a1. [169,170] at Sandia National Labs used
(Pb,La)(Zr,Ti)03 (PLTZ) thin film as the dielectric material for integrated
capacitors. They have achieved high dielectric constant (Er ~ 900), low
dielectric loss (tano = 0.01), good leakage resistance (p>lO i3 Q·cm at
125°C), and good breakdown field strength (EB -1 MV/cm). The sol-gel
technique was employed for the fabrication of the PZT and PLZT thin
films. After spin coating of the solution, the substrate was heated to
-300°C to pyrolyze the organic species in the precursor film followed by
firing at 650°C for crystallization. Pt was sputter-deposited to form the

Table 12-31. Dielectric Materials for Thin-Film Capacitors

Dielectric Band Gap Temperature Coefficient of


Material Constant Energy {eV~ Dielectric Constant (ppm/0C)

Si02 4 IJ.O
Si3N. 9 5.1
Ta20s 25 4.0 +250
Ti02 98 3.0 -750
12.9 INTEGRATED PASSIVES IN THIN-FILM PACKAGE 11-785

Tbin-Film
um

Silicon (ground plane)

--L
JWv-o I...
Decoupling
o-i~
Floating
Resistor
Capacitor Capacitor

Figure 12-92. Passive Device Structures

bottom and top electrodes of the capacitors. The typical dielectric film
thickness was between 90 and 100 nm and thicker films may be obtained
by multiple coatings. The heat treatment should be optimized for multilayer
films to avoid cracking. nCHIP has adopted another kind of decoupling
capacitor technology for its MCM fabrication [171-173]. A 2-f..lm-thick
Al layer was deposited onto a silicon substrate as the ground plane,
followed by a 0.15-f..lm-thick anodized Al 20 3 layer as the dielectric. An-
other layer of Al was deposited to 2 f..lm thickness afterward to fonn the
power plane. Multilayer interconnection was then fabricated using CuI
Si02 processing. Figure 12-75 [171] is a schematic showing such a MCM
structure. A sheet capacitance of 50 nF/cm 2 was obtained with this integral
capacitor. Ground bounce was shown to be greatly reduced by substituting
the discrete capacitors with integral decoupling capacitor.
Recently, the Packaging Research Center at Georgia Tech demon-
strated a novel technology for the fabrication of thin-film integrated decou-
piing capacitors [174]. Several photosensitive and nonphotosensitive di-
electric polymers (e.g., polyimides) were filled with high-dielectric-
constant barium titanate or lead magnesium titanate fine powders to high
volume-fraction. These polymer/ceramic composites have the combined
advantages of high dielectric constant (up to 65) from the barium titanate
and lead magnesium titanate ceramics, and the low processing temperature
of polymers. An additional advantage of this type of composites is the
capability of patterning by photolithography due to the transparency of
the ceramic component in the UV range and wet etching as is with the
original polymers. A specific capacitance as high as 22 nF/cm2 and a loss
11-786 THIN-FILM PACKAGING

tangent of <0.032 at 100 KHz were achieved with about 60% ceramic
loading as illustrated in Figure 12-93. The dielectric constant of the
composite was found to be stable over a wide frequency range. With
higher ceramic loadings and thinner films, a higher dielectric constant of
the composite and higher specific capacitances are expected to be achieved.

12.9.1.3 Inductors
Two micromachined integrated inductors, bar type and meander type
as shown in Figure 12-94, are implemented by Georgia Tech [175] on
a silicon wafer using modified, Ie-compatible, multilevel metallization
techniques. In the case of the bar-type inductor, a 25-j..tm-thick nickel-
iron permalloy magnetic core bar is wrapped with 30-j..tm-thick multilevel
copper conductor lines. For an inductor size of 4 mm x 1 mm x 110
j..tm thickness with 33 turns of multilevel coils, a specific inductance of
approximately 30 nHlmm2 at 1 MHz is achieved. In the case of the
meander-type inductor, the arrangement of conductor wire and magnetic
core is reversed (i.e., a magnetic core is wrapped around a conductor
wire). This inductor size is 4 mm x 1 mm x 130 j..tm and consists of 30
turns of a 35-j..tm-thick nickel-iron permalloy magnetic core around a 10-
j..tm-thick sputtered aluminum conductor lines. A specific inductance of
35 nHlmm 2 is achieved at 1 MHz.

70

60
.
l~LL
I. s.~
r

....C SO iii ~..

....<U
(/J
c
~~ J
V
40 /'-
U
0

U
,
....
'i::
A iii. .
/
30
U
Ql
.$'0
Qi ~.<..~

1- -
20 / ~;$'-
(5

~
-~
~ ..
.:.$'~ ;:.0:)

10
~
I~
0
o 10 20 30 40 so 60 70

% v 0 Iu me filled

Figure 12-93. Effective Dielectric Constant of Polymer/Ceramic Composite


Lower Conductor (Cu) • J)olyi mide (30 11m) coated
,/
POJ: and cured

• S;O, d'T"d s#ffi


Seed Lay er
• Ni-Fe electroplated into
I
molds and polyimide
Via (Cu)

• Ti/CulCr deposited

•
1
Polyimide (40 11m)
~
II
-:,
.• . '.• •
•
I
Coil vias dry-etched
~ I,
COllted lind cUn'd ~I " i , '~
I II •

....co
.... pper cond uctor (Cu)

• Polyimide molds made • Polyimide coated and


by dry-etc bing aod copper Seed Layer eu red, followed b) seed
electroplated into molds layer j
•
+
Polyimide (10 /1m) coated
and cured
• Copper paled into
/
Seed Layer to be removed
~
• Second seed layer for pbotorCliist mold
plating of magnClic core
deposited and pallcrned

•
1
Field regions etcbed
to tbe surfllce, bottom
1 seed layer wet-etched

Figure 12-94. Thin-Film Integrated Inductors


11-788 THIN-FILM PACKAGING

12.9.2 Future Directions of Integrated Passives in


Thin-Film Package
As discussed above, the integration of passive components has a
profound impact on the performance, space, and cost of thin-film packages.
A variety of organic and inorganic materials and their composites have
been investigated and applied to the fabrication of integrated passives in the
form of thin films. Depending on the specific type of passive components
(resistors, capacitors, or inductors), the materials employed, and the in-
tended applications, the fabrication processes vary widely. The merits of
enhanced performance, specifically, signal integrity and clock speed, and
increased packaging efficiency with the implementation of integrated pas-
sives are straightforward and technically feasible; it is also the major
driver for the current research and development activities. Cost reduction
through the integration of passives is less studied and more complicated
and controversial. Despite some isolated applications of integrated pas-
sives and the current research activities as described in Section 12.9.1
there is still a general lack of experience and knowledge on the technology
and the widespread commercial application of integrated passives. Com-
plex materials and processing technology are the key contributors to the
cost of integrated passives. As a consequence, the majority of the current
applications of integrated passives are mostly performance driven rather
than cost driven [42], and integrated passives are mostly found in high-
end applications. Significant effort is to be made to lower the cost of
integrated passives so that this technology can compete effectively with
the vast discrete passive industry in the high-volume, low-end market.
This demands major progress in the development of low-cost materials
and processes compatible with the thin-film technologies in multichip-
module manufacturing.

12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS


AND PROCESSES
Thin-film or fine-line MCM substrate with wiring density in excess
of 400 cmlcm2 has been and continues to be very expensive, typically
$50-100/in.2 for a total of five wiring levels (two for signal, one for
power, one for ground, and one for reference plane) as shown in Table
12-32. The technology has not yet taken off in industry, except by IBM,
NEC, and Hitachi for their mainframes and supercomputers where perfor-
mance rather than cost is the priority. The reasons for this high cost
are multiple. Materials and processing costs of thin-film substrates are
unacceptably high compounded by other obstacles, such as the general lack
of industrial infrastructure, lack of known good die (KGD), nonstandard
customized design and manufacturing, long turnaround time and small
volume. The material and processing costs include expensive dielectric
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-789

Table 12-32. Ultralow-Cost Characteristics of Proposed MCM

Approximate
Today PRC Proposed Improvement Factor

Size of substrate (in.) 5x5 16 x 16 lOx


Facilities cost ($/ft2) 1000 (Class 1(0) 200 (Class 1(00) 5x
Equipment cost ($) 20-150 M 4-20M 5x
Material cost ($/kg) 1000 100 lOx
Material usage (%) 20 (spin) 80 (nonspin) 4x
Raw process time (hr) 200 100 2x
Cost ($/in2) 30-100 2-5 50X

materials, expensive cleanroom dry processes, expensive tools, and the


long processing time to fabricate MCM. Fundamental issues related to
MCM substrate materials, processes, and tools to achieve the low-cost
objective and meet the industry needs have not been solved yet and,
hence, form the basis of future directions.
The future goal of low-cost MCM substrate in the industry is about
$2-5/in. 2 down from the current $50-100/in.2 as mentioned above. The
strategy to achieve this goal appears to be exploration, development, and
prototype of large-area processing: 18 in. x 18 in. (324 in. 2) from the
current 6-in. round wafer (28 in. 2). Figure 12-95 depicts the advantage
of large-area processing: a factor of 20x improvement in terms of number
of MCMs by replacing the traditional 6-in. round wafer substrate with
I8-in. x I8-in. large substrate. Figure 12-96 [176] illustrates the cost
reduction both as a function of size (6 in. vs. 24 in.) and production
volume of MCMs. Because the cost of manufacturing, in terms of process
steps, raw process time, and so forth, is approximately the same for both
sizes, the adoption of a large-size substrate should reduce the cost by
about lOx from the current $50-1 OO/in2• Two challenges must be overcome
to meet the proposed target of $2-5/in2• For one, the above lOx improve-
ment assumes the yield of large substrate is the same as small substrate.
This is clearly not the case with current fine-line multichip-module technol-
ogy. In addition, significant cost reduction must be made to bring the
cost further down to $2-5/in. 2• The science, technology, prototype, and
manufacturing leading to the cost goal involves, therefore, low-cost novel
materials and chemical processes that are immune to airborne particles,
and low-cost novel polymer- and metal-deposition processes. The expected
improvements from all of these are listed in Table 12-32.
The low-cost substrate requires both fundamental and practical ap-
proaches to materials, processes, and tools. DARPA has taken the initiative
18"

14"
8"
6"
4"

.....
:g 4 MCMs
12 MCMs
21 MCMs

121 MCMs

235 MCMs
Today Tomorrow

Figure 12-95. The Advantage of Large-Area MCM


Cost Modeling - Generic MCM-D Fabrication Process
$50

Assumptions: C $40
6" Round ~

-Substrate I "xl" ~
In $30 6" Round
-Unyielded
--~
1e
$20
~
.c
.... 24" x24" LAP 00-
~
-..=
-Substrate 1"xl" ~ $10
~

-Unyielded =
U 24 "x24" LAP
$0
o 2,000,000 4,000,000 6,000,000
Annual Volume (in2)
Model Prediction (Generic Model)

-6" round wafer format cost approaches S30/in2 (1 "xl" -S layer)


-LAP otTers the lowest unit cost - approaching $3/in 2 (1 "xl" -S layers)
Figure 12-96. Economics of LAP
11-792 THIN-FILM PACKAGING

to develop large-area tooling in cooperation with Hughes, nChip, Micro-


Module Systems, Texas Instrument, and Boeing to alleviate the high
cost of production equipment. Georgia Institute of Technology has been
awarded a national Engineering Research Center (ERC) by NSF to concen-
trate on novel materials and processes that are compatible with these tools
in complementary mode and to demonstrate the $2-5/in.2 goal for MCMs.
Figure 12-97 illustrates one process proposed by Packaging Research
Center at Georgia Institute of Technology to achieve low cost. It involves
large-area deposition of polymer by extrusion, curtain, or meniscus coat-
ing, followed by low-cost via formation in photosensitive polymer by
large-area, full-field, and low-cost photolithographic tool. The via and
pattern metallization is accomplished by direct electroless and/or elec-
troplating of copper.
There are basically two ways to reduce the cost of polymer deposi-
tion. One is to apply conventional $1000Ikg polymer onto a large area
with high material usage such as by an extrusion coater. The other is to
introduce novel dielectric polymers that are about 1O-20x lower in cost
($50-100Ikg) than those being currently used. When low-cost dielectric
polymers are available, the cost of polymer deposition becomes relatively
less sensitive to the selection of coating technology, both the existing or
modified spinner or, for that matter, extrusion or other large-area coaters
can be used. When low-cost dielectric material is not available and cost
reduction solely relies on large-area processing, the material cost of a five
layer thin-film substrate can make up as much as 70% of the total cost,
as illustrated in Figure 12-98 [176]. From this discussion, it is obvious
that low-cost materials and large-area processes are vital for future thin-
film packaging technology. These are reviewed below.

Planar Ceramic ...), Deposit Low- Develop


~ Expose to Full ~ ~
or PWB Substrate
with Metal 1I0s
,- Cost Polymer by -y Field I or G-Line
Non-Spin Process 100-500 mJ!cm' ~
Via
Holes rY
Cure
Polymer

Electro- Sputter!
Plate Screen
Inductor Resistor

Dice into
Single!
Multi-
Chips
Adhesive Chips

Figure 12-97. PRe Process for Next-Generation Packages


12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-793

70%

60%
-
-
~
, ............... ................. ---
='.'. ~~ ........-..--..-
... 50%
-
-'. ...... ~ ...'"
' .......
CII
- ,-

i
-
---: ,.,....... ......
40°.4, - ...... .... .......... .......
-= ...... .........
...= 300/0 =
(/.l
eo..
-=
i'
..... ... ·-·-~'!iPIJJC!IIt
~

U=
'"
"
" .....- .........

---
-
"1ft. -
-
..
20% ~ •.A' _ _ Faellities
, '
10% -,'
!' Labor

0%
o 2,000,000 4,000,000 6,000,000

Annual Production Volume (inl)

Figure 12-98. MCM-D Substrate Cost Breakdown. As volume increases, the per-
centage of substrate cost contributed by equipment capital significantly decreases and the
percentage contributed by materials increases.

12.10.1 Polymeric Materials

Although both organic and inorganic materials have been used as


dielectrics in thin-film packages as discussed in Section 12.4, polymeric
materials are the most promising candidates for the future thin-film pack-
ages. The versatility of materials and properties, low material cost and
process easiness make polymers unparalleled by any other dielectric mate-
rials. As in many other fields, polymers are assuming an increasingly
important role as dielectrics in thin-film package, despite the unstable
(relative to most inorganic materials) and nonhermetic nature of polymers.
Excellent electrical and process properties and very low materials cost
are the key elements for an idea dielectric polymer, which should form
the emphases of future material development.
There are generally three types of dielectric polymers that are in
use or development today, as discussed in Chapter 11, "Polymers in
Packaging." These are polyimides in a number of chemical families such
as PMDA-ODA, BPDA-ODA, and others, Benzocyclobutene (BCB) and
epoxy. The important properties of these polymers are summarized in
Table 12-33. Polyimides have long been used for thin-film packaging as
a dielectric. A variety of derivatives have been developed by a number
of companies to fulfill the various requirements as described in Section
Table 12-33. Dielectric Polymers

Dielectric Dissipation Tensile Percent Moisture


Polymer Name Polymer Type Constant Factor TCE Ig Modulus Strength Elongation Absorption
(lO-6/°C) (0C) (GPa) (MPa) (wt.%)

Amoco Ultradel4212 Fluorinated polyimide 2.9 0.005 50 295 2.7 0.9


Amoco Ultradel 7501 Photosensitive polyimide 2.8 0.004 24 >400 3.4 3.4
Cemota IP 200 Polyphenylquinoxaline 2.7 0.0005 55 365 2.0 117 8-12 0.9
Dow Cyc10tene 3022 Benzocyc1obutene 2.7 0.0008 52 >350 2.0 85 6 0.2
Du Pont PI-2545 Standard polyimide 3.5 0.002 20 >400 1.4 105 40 2-3
Du Pont PI-2555 Standard polyimide 3.3 0.002 40 >320 2.4 135 15 2-3
-.J Du Pont PI-261OD/2611D 3 (x-y)
Low stress polyimide 2.9 0.002 >400 8.4 350 25 0.5
...'" Du Pont PI-2732/2733 Photosensitive polyimide 2.9 25 >400 6.0 192 8 1.5
Hitachi PIQ-13 Standard polyimide 3.4 0.002 45 >350 3.3 130 20 2.3
Hitachi PIQ-L100 Low stress polyimide 3.2 0.002 3 410 11 380 22 1.3
National Starch EL5010 Preimidized polyimide 3.2 <0.002 34 214 2.8 150 7 1.3
National Starch EL5512 Fluorinated preimidized 2.8 <0.002 35 225 2.8 150 6 0.8
polyimide
OCG Probimide 400 Preimidized 3.0 0.003 37 357 2.9 147 56 2.0
photosensitive
polyimide
OCG Probimide 500 Low-stress polyimide 2.9 0.003 6-7 400 11.6 444 28 0.7
Toray UR-3800 Photosensitive polyimide 3.2 0.002 40 280 3.4 140 11 1.1
Ciba Geigy LMB 7081 Photosensitive epoxy 4.1 0.025 130 3.3 <0.5
Shipley Multiposit Photosensitive epoxy 3.2 0.007 60-70 >180 2.1 6-8
XP-9500 CC (Novolac)
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-795

12.4. Although po1yimides are still the major dielectric polymer for today' s
thin-film package, some inherent concerns with polyimides remain. These
include (a) the reactivity with copper necessitating the cladding of copper
lines in most applications which complicates the processing and increases
the cost, (b) the high material cost, (c) significant moisture uptake causing
reliability concern, and (d) the relatively poor adhesion to copper which
requires the application of adhesion promoter. Benzocyclobutene (BCB)
was developed as an alternative to address the problems associated with
polyimides. It has a lower dielectric constant, lower moisture uptake,
higher film retention, better planarization, and compatibility with copper
metallization and is capable of rapid thermal curing which leads to in-
creased throughput and reduced processing cost. Adhesion promoter is, in
general, required. Nevertheless, issues related to materials cost, oxidation
resistance, and further process development are to be addressed properly
for the wide acceptance of BCB. Epoxies are the prevalent dielectric
materials used for printed wiring board fabrication and are known for
low-end, low-cost, and high-volume applications. Increased interests have
been shown to extend the applications of epoxies into thin-film packages.
Not included in Table 12-33 is a material that has not received much
recognition. Polyolefins are long being known for numerous nonelectronic
applications and are being developed for electronic applications as dielec-
trics by B.F. Goodrich. Two key fundamental properties of polyolefins,
a very low dielectric constant and a very low moisture absorption, made
them very attractive candidates as electrical-grade polymers. Primary
results on the polyolefin patented by B.P. Goodrich showed that it can
be spin or extrusion coated onto a printed wiring board and cured below
250°C. Its material cost is expected to be about 1O-20x lower than that
of BCB or polyimide and is hence about the same as epoxy. However,
it has superior electrical and thermal properties over all the three aforemen-
tioned electronic polymers. The synthesis of polyolefin is by transition-
metal catalyzed polymerization of cyclic olefin monomers derived from
the Diels-Alders adducts cyclopentadiene. Flexibility in monomer and
catalyst selection can lead to polymers with a broad range of properties
and structures.

12.10.2 Large-Area Polymer Deposition

As discussed in Section 12.4 the capability of a large-area process


(LAP) is essential for future low-cost thin-film packages. A large-area
polymer deposition technology with good film thickness uniformity, very
low defect rate, high throughput, and low material waste is one of the
critical enabling technologies for the implementation of this process. When
spin coating is applied to large-format substrate and/or square and rectan-
gular substrates, uniformity becomes a potential concern. There are several
11-796 THIN-FILM PACKAGING

fundamental reasons why uniformity is not to be expected from spin


coating on large-area, noncircular substrates. In addition, a major drawback
with spin coating is that much of the polymer solution (> 85%) is wasted
as it is thrown off of the edge of the wafer during the spin process. This
increases the fabrication cost of thin-film packaging significantly when
expensive dielectric polymer ($lOoo/kg) is used, as with most of the
current applications. It should be noted, however, that flat panel displays
of notebook computers are large (6 in. x 12 in.) rectangular substrates
and are primarily fabricated by spin coating today. The future of spin
coating is unclear despite the general interest in a coating technology
capable of large-area, noncircular substrate coating with high polymer
utilization at low cost which seems to be heading away from spin coating.
Among the many coating methods, extrusion coating is receiving
increased interest and is being investigated as a candidate for potential
applications in polymer thin-film coating over a large-format substrate.
FAS -Technologies is developing liquid extrusion coating technology and
equipment specifically for microelectronic and flat panel display (FPD)
applications. A digital electrohydraulic pumping system is used to inject
the polymer into the die. It has been shown that this extrusion system is
capable of dispensing and coating polymer fluid more precisely than the
conventional gas-driven diaphragm system. Materials of higher viscosity
(> 400 poise) may be deposited by this extruder, which means that thicker
films may be laid down in a single processing step. Uniform thicknesses
« 5% thickness variation) from 0.15 up to 100 /lm have been demonstrated
in a single pass. Coating speeds of 1-3 in./s can be routinely achieved
to ensure a high throughput. Virtually 100% material usage with this
extrusion technology would have a significant impact on the cost and
environmental issues of thin-film coating technology.
Roller coating, meniscus coating, and screen printing as discussed
previously have been used to a much less extent in thin-film packaging
and are less investigated. Careful feasibility studies are necessary to further
clarify their role in large-area coating of polymer thin films for future
thin-film packaging technology.
In comparison, spin coating is the principle method in use today to
deposit and planarize polymers because it is easy, cheap, and well estab-
lished. Nevertheless, the amount of material that is wasted as it gets
thrown off of the surface is excessive, on the order of 80%. Spinning by
nature favors circular substrates. Because of this, some of the substrate
is wasted when the round edges are cut off. Finally, the difficulty of
spinning increases with increasing substrate size, and the push is toward
larger boards to increase the number of circuits processed at one time. For
large-area processing of square or rectangular boards, extrusion appears to
be a much better method than spin coating. The material wasted with
extrusion is much less than with spin coating and even less than the other
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-797

methods listed above. The deposition process is very similar to spin


coating in that the liquid polymer solution is spread uniformly on the
substrate. Thus, aside from the rotary motion, the planarizing mechanisms
are the same. Gravitational and capillary forces work together with surface
tension to planarize the liquid. In short, the planarizing mechanisms are
the same except that the harmful effects (and material waste) of rotational
motion are absent. The problem of evaporation due to airflow over the
substrate is reduced and shrinkage is the primary factor for planarization.
However, extrusion and other candidate coating techniques, including
meniscus, roller, and screen printing, are more susceptible to slit inhomo-
geneities (i.e., misalignment, blockage, and mechanical vibration) [177].
The effects of these factors on the film quality and process robustness
must be carefully evaluated and controlled in the development and choice
of thin-film coating technology.
In conclusion, coating uniformity and waste reduction seem to indi-
cate a shift away from spin coating in low-cost, high-performance elec-
tronic packaging. Extrusion seems to be emerging as a viable and promis-
ing alternative. Intelligent use of extrusion and the implementation of
feedback control offer the possibility to continue to scale dimensions
down and reduce cost while increasing yield on larger substrates.

12.10.3 Via Formation


As is discussed in the previous sections, two rival technologies for
via formation in dielectric polymers for high-density interconnections
are commonly regarded as low cost: photosensitive polymers and laser
ablation. At this time, there is no strong conclusive evidence which favors
one technology over the other. Both are employed in actual fabrication
of MCM packages. This uncertainty is largely due to the fact that both
technologies are under constant evolution which makes any confident
prediction and judgment difficult if not impossible. However, a general
account of some of the essential attributes of the two technologies can
be made and is briefly reviewed in the following. Table 12-34 [30] com-
pares the major attributes of different via formation technologies in poly-
imides. It is noted that wet etching and reactive ion etching (RIE) are
becoming obsolete for both process complexity and cost reasons.

12.10.3.1 Laser Ablation


Both scanning laser ablation and projection laser ablation have been
successfully used for via formation in electronic packaging. Scanning
laser ablation was first adopted by Siemens in via formation in printed
wiring board for MCM-L packaging used for its Siemens 7500 H 90
mainframes [178] Hitachi and Motorola extended this technology later
on for via formation in MCM-D substrates [29,60]. A scanning laser
11-798 THIN-FILM PACKAGING

Table 12-34. Comparison of Via Process Techniques

No. of Critical Unit Choice of Aspect Ratio/ Via Wall


Via Process Steps Processes Pol~imides Limitations Profile

Wet etch 9 Polyimide apply and Most 2:1, 15 IJl11 Trapezoidal, 50°
bake, resist min. dia.
lithography, wet etch
RIE 9 Resist lithography, All 2:1, 411m max. Trapezoidal, 50°
RIE, residue removal thickness
Laser ablation 6 Laser ablation All 1:1, 10 11m min. 30°_65°
dia., trapezoidal
PSPI 7 PSPI lithography Few 1:1, 10 11m min. Barrel,500-80°
dia.

ablation tool is distinguished from projection laser ablation tool in the


utilization of a conformal mask made from patterned metal or photoresist
for imaging, and the laser beam is scanned across the substrate to ablate
the exposed polymers. Apparently, more processing steps are required for
scanning laser ablation relative to projection laser ablation; nevertheless,
simpler tool and masking technology are involved. With photoresist as
the erodible mask as mentioned in Section 12.4, a significant cost reduction
is expected due to the elimination of the metal mask deposition, patterning
and stripping steps. More effort is necessary to develop and characterize
a robust SLA via generation process using photoresist erodible masks.
Projection laser ablation has been employed by IBM for the fabrica-
tion of ES 9000 systems [31]. A via size of 75 Jlm in polyimide of 18-20
Jlm thickness was achieved with high yield (99% substrate yield with
more than 100,000 vias per substrate) and high throughput (less than 15
min cycle time). The up-to-date smallest achievable via size is 6 Jlm [31].
Uniform trapezoidal via profile with wall angles ranging from 30° to 65°
is readily obtained by altering the focus conditions. Unlike photosensitive
polymer processing, laser ablation is not limited by the via aspect ratio
and has good tolerance to the nonuniformity of polymer thickness due
to the etching-stopping effect of the underlying metal. Key processing
parameters include laser fiuence, number of pulses, and focus. A plasma
cleaning is performed after via formation to remove the carbonaceous
debris. The major drawback of this technology is the use of a quartz
dielectric mask which is expensive and is not universally available. The
issue of reworkability is also a potential concern with laser ablation. It
has been shown in Section 12.4 that large-area processing is crucial for
cost reduction. This is valid only under the assumption that the costs of
large-area tools and masks are not increased significantly. IBM has already
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-799

demonstrated the capability of laser ablation of a 300 Jlm square substrate


[31]. However, the wide acceptance of this technology in cost-sensitive
applications requires further technology simplification and tooling cost re-
duction.
There are currently many ongoing evolution in laser ablation technol-
ogy covering all the aspects from radiation source (e.g., TEA CO2 laser
and solid-state laser to lower operation cost), to tool design (e.g., Anvik
Corp's scanning system for large-area, high-throughput via drilling) [179],
to masking technology [180]. Efforts are also underway to develop a
maskless via formation technology by direct laser drilling with a focused
fine laser beam. All these are expected to have a great impact on the cost,
performance, and reliability of laser ablation technology.

12.10.3.2 Photosensitive Polymers

A variety of photosensitive versions of the conventional dielectric


polymers as mentioned in Section 12.10.1 have been formulated by adding
sensitizers, initiators, and other photoactive components for the purpose
of processing simplification, cost reduction, and improvement of through-
put and quality of via formation [30]. The basic requirements on photosen-
sitive polymers are low exposure dose, good resolution, short develop
time, and good resistance to solvent attack during develop (e.g., swelling,
dissolution, cracking, and crazing), in addition to the common require-
ments on dielectric polymers as listed in Section 12.4. Mechanical proper-
ties of photosensitive polymers are generally inferior to the corresponding
nonphotosensitive polymers with the same backbone due to the fact that
the photo components cannot be removed completely [181]. Optimization
of the overall process of via formation is more a matter of experience
and experiment, with many factors ranging from ground rule design to
material properties to processing conditions coming into play [76]. Via
sidewall profile, aspect ratio and residue at the via bottom are among the
most important issues affecting the processibility, performance, yield, and
reliability. Vias with aspect ratio of 1 : 1 have been achieved in film
thickness less than 8 Jlm, whereas for thicker films, 1 : 2 is a more realistic
figure [76]. Typical vias formed by photosensitive polymer technology
has a sidewall profile steeper in the top portion of the via with a very
shallow foot [30]. Several factors are the possible causes of residue at
the via bottom, including pattern geometry, expose energy, substrate re-
flectivity, insufficient rinse, and Cu-polymer interaction. However, with
good process control and the implementation of postdevelop treatment
for residue removal [76,182], residue usually does not pose a severe
problem. Cost reduction with the application of photosensitive polymers
is realized mainly by the reduced processing steps and low capital costs.
However, the process is less controllable in terms of via wall profile and
11-800 THIN-FILM PACKAGING

planarization among other processing qualities, which is the future focus


of photosensitive polymer material and process development. Relatively
high material cost is another shortcoming associated with photosensitive
polymers. Additionally, environmental concerns related to the treatment
and dispose of organic solutions for photosensitive polymer processing
should also be properly addressed.

12.10.4 Planarization

The planarity of a thin-film multilayer substrate has a multiple impact


on the electrical performance and reliability of thin-film package and as
well as the processability. As discussed in Section 12.4, the requirement
on planarization is becoming more compelling and challenging as the
wiring density continues to increase. An additional planarization process
step (e.g., the mechanical polishing used by IBM) would necessarily
accrue to the overall cost of thin-film packaging, while every effort is being
made to reduce the cost of thin-film packages to extend the application into
the consumer market where low cost could hardly be overstressed. Hence,
polymer deposition and/or metallization processes capable of full or partial
planarization would avoid the additional planarization steps which had
been costly and, meanwhile, technically complex. In the case of planariza-
tion by polymer thin-film deposition, both material properties and coating
techniques are important as pointed out in Section 12.4. Studies on BCB
have shown degree of planarization >85% under controlled baking and
curing conditions. In the case of planarization by metallization, the selec-
tive deposition of copper allows the filling of vias and trenches through
the dielectric polymer layer and planarizes the structure. Both electro-
and electroless copper have the capability of selective copper deposition
for the purpose of planarization. However, uniformity of electro copper
plating degrades when a large substrate is used, and fully electroless
copper plating is still under significant investigation and development
as briefly discussed in the next section. Therefore, this approach for
planarization still does not find wide application.

12.10.5 Electroless Copper Plating

Electroless copper plating has the advantages of simpler equipment


and the associated low capital cost compared to electroplating. Among
other advantages of electroless plating are the capability of large-area
uniform plating over all surfaces-regardless of size and shape, variable
plating selectivity, and 100% throwing power. It is most widely used in
the printed wiring board (PWB) industry for high-aspect-ratio plated-
through-hole (PTH) seeding and/or subsequent buildup of the desired
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-801

thickness. The cost of current electroless copper plating is, however, much
higher compared to electroplating in the PWB industry. Complex bath
chemistry, expensive chemicals and catalysts, high requirement of impu-
rity control, and slow plating rate are the major cost drivers, which are
compounded by the increasingly environmental and health concerns re-
lated to electroless copper plating.
Electroless plating was described as an autocatalytic process of de-
positing a metal in the absence of an external source of electrical current.
Electroless copper solutions generally contain a source of divalent cupric
ions, a reducing agent that is capable of reducing the cupric ions to the
metal, and a complexing agent to prevent precipitation of the cupric ions
and other additives [183]. Such a system is thermodynamically unstable
but kinetically inhibited in the absence of catalytic surfaces. Once plating
starts, the copper deposition process is generally autocatalytic. Cupric
sulfate (CUS04) is the commonly used source of copper ions, and formalde-
hyde (HCHO), and, more recently, hypophosphite and dimethylamine
borane are employed as the reducing agents. The overall reaction for
electroless copper plating with formaldehyde as the reducing agent is

Cu2+ + 2HCHO + 40H- ~ Cu + 2HCOO- + 2H20 + H2


[12-12]

This reaction is a mixed-potential reaction of the anodic oxidation of


formaldehyde and the cathodic reduction of cupric ions at the substrate.
In order for the reaction to proceed at an appreciable rate, the pH values
of the solution are generally maintained in the range of 11-13 by accelera-
tors, such as NH30H, KOH, NaOH, and LiOH [184], to lower the anodic
half-cell potential. The accelerators are designated to remove the surfacant
and then allow the adsorbed palladium catalyst on the substrate surface
to contact with the plating solution. For a bath containing KOH or NaOH,
there is a distinct tendency that the deposition rate of electroless copper
is reduced with a decrease of the pH value. However, the deposition rate
is maintained constant even at low pH in the case of LiOH [185]. Simple
cupric salts are almost insoluble at a pH above about 4; the addition
of various complexing agents I ligands, such as EDTA (ethylenediamine
tetraacetic acid), tartrate, Rochelle salt, and triethanolamine (TEA) is
usually required to prevent the precipitation of cupric salt in the alkaline
solution. TEA is found to be the most effective ligand for increasing the
electroless copper plating rate. The maximum rate is over 20 times faster
than the rate observed in the EDTA system [186]. Trace amount of
additives are present in the plating baths to provide good-quality deposits
[187]. The most commonly used additive is cyanide, which is proven to
improve ductility of the deposits Normally, the reduction of cupric ions
initiates at a catalytically active site on the substrate surface. For noncon-
11-802 THIN-FILM PACKAGING

ductive substrates, an additional catalyzation step is required to activate


the surface. The most commonly used catalytic solutions are the tin-
palladium colloids [188]. The palladium is a catalytically active element,
and the tin acts as a protective colloid. When the nonconductive substrates
are immersed in the tin-palladium colloid, the Pb and Sn particles are
adsorbed on the substrates and become the catalytic sites (Fig. 12-99).
Electroless copper deposition then initiates on these Sn-Pd catalytic parti-
cles and continues autocatalytically. In the SnCI2-PdCI2 activation treat-
ment, the sites, which are available for copper nucleation on the insulating
substrate, depend on the density of small isolated catalytic particles. Basi-
cally, the adherence of the catalyst is determined by chemisorption or
physisorption to the matrix. A new compound of seeding palladium for
electroless copper plating was developed by IBM [189]. The organometal-
lic compound, allylcyclopentadienylpalladium, was found to be easily
prepared in high yield due to the unusual volatility at room temperature.
Photochemical decomposition of allylcyclopentadienylpalladium results
in the deposition of the film that consists of a homogeneous mixture of
palladium particles in the insulating matrix. Without further activation,
the deposited film is an active catalyst for electroless copper plating to
form an uniform plated layer which exhibits good electrical properties.
The plating rate of conventional electroless copper baths is generally low
(1-5 ~mIh) [190]. Reaction kinetics are controlled by the proper selection
of additives and by the variation of bath temperatures. A high-build bath

Insulating, Catalytic
Substrate

H P
'c
/, --
M
,
H o·
~

C,
OH
+. +H ___
M ----
~
.
M-

_.....---- Conducting
H OH
e'\ - Catalytic
- -........
---- ........- - Site

0,. = H~C~
H OH M

Figure 12-99. CataIyzation of Insulating Substrate in Electroless Copper


Plating
12.10 FUTURE DIRECTIONS OF THIN-FILM MATERIALS AND PROCESSES 11-803

with a plating rate up to 23 ,.un/h has been reported through the joint
addition of organics and an increase in the bath temperature [190]. How-
ever, bath instability, nodule formation, and inferior deposit morphology
are the major problems encountered in the high-speed plating bath. A
considerable amount of work is required before optimal balance can be
achieved between high plating rate and good-quality deposits. Another
new electroless copper plating process for preparing a copper layer with
strong adhesion to glass substrate was developed by Yoshiki et al [191].
The new process incorporates a ZnO thin film on the glass substrate,
instead of etching and sensitizing the substrate surface as in a conventional
electroless plating process. The peel strength becomes much greater than
that obtained using a conventional plating method. Koyano et al. [192]
has proposed that a higher electroless copper plating rate of 7-10 Ilmlh can
be achieved in an alkaline plating solution of a copper-glycerin complex at
pH values above 12.2, with formaldehyde as the reducing agent and
bipyrizyl as an additive. The stability of the solution is greatly enhanced
by controlling the glycerin/Cu molar ratio and by intermittent filtration.
Electroless copper plating at a low pH level is desired to minimize
the attack of the caustic bath solution on the dielectric polymers. Common
results of caustic solution attack on polymers include swelling, cracking,
dissolution, and delamination at interfaces and performance degradation
of the polymer dielectric layers. In IBM, a stable bath chemistry was
developed which operates at a pH level of <9 [193]. Multidentate nitrogen
donor ligands was employed as the cupric-ion complexants in conjunction
with triethanolamine (TEA) as the buffer and dimethylamine borane as
the reducing agent to enhance the bath stability. High-quality copper
deposits was obtained on submicron structures at a deposition rate of
2-3 Ilmlh.
The operation of an electroless copper plating bath has narrower
process window compared to electroplating bath [194]. The range of
components and contaminants as well as other processing parameters,
such as temperature, must be well controlled to ensure a good deposit
quality and prolonged bath life. Impurities, especially, can cause bath
decomposition and poor quality of copper deposit and should therefore
be minimized. Basic bath controls include replenishing the bath regularly
in small amount each time, and checking the bath temperature, formalde-
hyde, caustic and copper concentrations at certain time intervals and when
a problem is suspected. Other important guidelines for maintaining an
electroless copper plating bath are redundant air agitation even during
shutdown, continuous filtration during plating, use of deionized water and
reagent-grade caustic and formaldehyde, and striping of copper plated on
the inner surface of the tank and on the racks.
In large-format MCM-O substrate manufacturing, the requirement
of uniform metallization on an area of 450 mm x 450 mm becomes critical
11-804 THIN-FILM PACKAGING

and is hardly met by electroplating of copper. This provides an unique


opportunity for electroless copper. However, technological improvement
of electroless plating in many aspects, such as higher plating rate, acid
or neutral solution plating to alleviate attack on polymer dielectric and!
or photoresist, stable and simple bath chemistry, good process control,
improved copper deposit property, and fully additive process must be
made before the implementation of this metallization process. In addition,
lowered cost of electroless copper plating is essential to survive the con-
sumer market.

12.11 ACKNOWLEDGMENT
The authors wish to acknowlegde the write-ups provided by Dr.
Chung Ho ofMMS, T. Inoue of Hitachi, T. Watari ofNEC, H.R. Krauter
of Siemens, and Stan Drobac of nCHIP.

12.12 REFERENCES
1. M. E. Jones, W. C. Holton, and R. Stratton. "Semiconductors: The Key to Computa-
tional Plenty," Proc. IEEE, 72: pp. 1380-1409, 1982.
2. M. Terasawa, S. Minami, and J. Rubin. "A Comparison of Thin Film, Thick Film,
and Co-Fired High Density Ceramic Multilayer with the Combined Technology: T
& T HDCM (Thin Film and Thick Film High Density Ceramic Module," Int. J.
Hybrid Microelectron, 6(1): pp. 607-615, 1983.
3. N. Naclerio. "ARPA Strategy and Programs in Electronic Packaging," ARPA WWW
homepage (http://ETO.sysplan.comlETOIEI-packaging/present), 1996.
4. B. S. Landman and R. L. Russo. "On a Pin vs. Block Relationship for Partitions of
Logic Graphs," IEEE Trans. Computers, EC-20: pp. 1469-1479, 1971.
5. T. Chiba. "Impact of the LSI on High-Speed Computer Packaging," IEEE Trans.
Computers, EC-27: pp. 319-325, 1978.
6. C. T. Goddard. "The Role of Hybrids in LSI Systems," IEEE Trans. Components
Hybrids Manu! Technol., CHMT-2: pp. 367-71, 1979.
7. T. S. Steele. "Terminal and Cooling Requirements for VLSI Packages," IEEE Trans.
Components Hybrids Manu! Technol., CHMT-4: pp. 187-191, 1981.
8. D. Balderes and M. L. White. "Package Effects on CPU Performance of Large
Commercial Processors," Proc. 35th ECC, pp. 351-355, 1985.
9. R. T. Evans. "Interconnection and Packaging of IBM's Large Processors," Proc.
34th ECC, pp. 374-378, 1984.
10. A. J. Blodgett. "Microelectronic Packaging," Scientific American, 249(1): pp. 86-
96, 1983.
11. T. Watari and H. Murano. "Packaging Technology for the NEC SX Supercomputer,"
Proc. 35th ECC, pp. 192-198, 1985.
12. E. E. Davidson. "The Electrical Design Methodology for the Package Used in the
IBM 3090 Computer," IEEE Proc., Wesconl85 Electronic Show and Convention, p.
I, 1985.
13. C. W. Ho, D. A. Chance, C. H. Bajorek, and R. E. Acosta. ''The Thin-Film Module
as a High Performance Semiconductor Packaging," IBM J. Res. Devel., 26(3): pp.
286-296, 1982.
14. C. W. Ho. High-Performance Computer Packaging and the Thin Film Multichip,
VLSI Electronics Microstructure Sciences, pp. 103-143, Academic Press, New
York,1982.
12.12 REFERENCES 11-805

15. C. A. Neugebauer. "Comparison of VLSI Packaging Approaches to Wafer Scale


Integration," Proc. IEEE Int. Conf. Computer Design, pp. U5-l20, 1984.
16. Y. Tsukuda, S. Tsuchida, and Y. Mashimoto. "Surface Laminar Circuitry Packaging,"
Proc. 42nd ECTC, pp. 22-27, 1992.
17. S. Ray, D. Berger, G. Czornyi, A. Kumer, and R. Tummala. "Dual-level Metal
(DLM) Method for Fabricating Thin Film Wiring Structures," Proc. 43rd ECTC,
pp. 538-543, 1993.
18. S. Poon, J. T. Pan, T. C. Wang, and B. Nelson. "High Density Multilevel Copper!
Polyimide Interconnects," NEPCON West, pp. 426-427, 1989.
19. N. Iwasaki and S. Yamaguchi. "A Pillar-Shaped Via Structure in a Cu-Polyimide
Multilayer Substrate," IEEE Trans. Components Hybrids Manuf. Technol., 13(2):
pp. 440-443, 1990.
20. J. F. McDonald, J. A. Steakl, C. A. Neugebanner, R. O. Carlson, and A. S. Gergun-
dahl. "Multilevel Interconnections for Wafer Scale Integration," J. Vacuum Sci.
Technol., 4(6): pp. 3127-3136, 1986.
21. H. J. Levinstein, C. J. Bartlett, and W. J. Bertram. "Multichip Packaging Technology
for VLSI-Based System," IEEE Computer Soc. Proc., March 1987.
22. P. Garrou. "Polymer Dielectrics for Multichip Module Packaging," Proc. IEEE,
80(12): pp. 1942-1954, 1992.
23. W. Lautenberger, B. Auman, J. Butera, D. Goff, C. Lazaridis, H. Seiler, and
1. Labadie. "Polyimides for Large Area Multichip Modules," Proc. ICEMCM '95,
pp. 501-506, 1995.
24. E. Perfecto, K. Prasad, P. Wilkens, G. White, R. Tummala, C. Prasad, and T.
Redmond. "Multi-level Thin Film Packaging Technology at IBM," Proc. ICEMM
'93, pp. 474-482, 1993.
25. R. Tummala. "Multichip Packaging in IBM-Past, Present and Future," Proc.
ICEMM '93, pp. I-II, 1993.
26. G. Adema, M. Berry, and I. Turlik. "Dielectric Materials for Use in Thin Film
Multichip Modules," Electronic Packaging and Production, 32(2): pp. 72-76, 1992.
27. R. Tummala, H. Potts, and S. Ahmed. "Packaging Technology for IBM's Latest
Mainframe Computers (S!390IES9000)," Proc. 41st ECTC, pp. 682-684, 1991.
28. T. Tessier and P. Garrou. "Overview ofMCM Technologies: MCM-D," Proc. ISHM
'92, pp. 235-247, 1992.
29. K. Miyazaki, A. Takahashi, O. Miura, R. Watanbe, Y. Satsu, T. Miwa, andJ. Katagiri.
"Fabrication of Thin Film Multilayer Substrate Using Copper Clad Polyimide Sheets,"
Proc. 43rd ECTC, pp. 306-310, 1993.
30. K. Prasad and E. Perfecto. "Multilevel Thin Film Packaging: Applications and
Processes for High Performance Systems," IEEE Trans. Components Packaging
Manuf. Technol.-Part B: Advanced Packaging, 17(1): pp. 38-49, 1994.
31. S. Patel, T. Redmond, C. Tessler, D. Tudryn, and D. Pulaski. "Laser Via Ablation
Technology for MCMs Thin Film Packaging-Past, Present, and Future at IBM
Microelectronics," Proc. ISHM '94, pp. 31-41, 1994.
32. P. B. Chinoy. "Processing and Electrical Characterization of Multilayer Metallization
for Microwave Applications," Proc. ICEMCM '95, pp. 203-208, 1995.
33. R. 1. Jensen. "Polyimides as Interlayer," in Polymers/or High Technology-Electron-
ics and Photonics, ed. by M. 1. Bowden and R. S. Turner, ACS Symposium Series
346, American Chemical Society, Washington DC, 1987.
34. S. Sasaki, T. Kon, and T. Ohsaki. "New Multi-Chip Module Using a Copper Polyimide
Multi-Layer Substrate," Proc. 39th ECC, pp. 629-635, 1989.
35. T. Shimoto, K. Matsui, and K. Utsumi. "CulPhotosensitive-BCB Thin-Film Multi-
layer Technology for High-Performance Multichip Module," Proc. MCM '94, pp.
115-120, 1994.
11-806 THIN-FILM PACKAGING

36. D. Darrow and S. Vilmer-Bagen. "Low-Cost Patterned Metallization Technique


for High Density Multilayer Interconnect Applications," Proc. 43rd ECTC, pp.
544-549, 1993.
37. C. W. Ho and U. A. Deshpande. "A Low Cost Multichip Module-D Technology,"
Proc. [CEMM '93, pp. 483-488, 1993.
38. A. Strandjord, Y. Ida, P. Garrou, B. Rogers, S. Cummings, and S. Kisting. "MCM-
D Fabrication with Photosensitive Benzocyclobutene: (Processing, solder Bumping,
System Assembly, and Testing)," Proc. [SHM '95, pp. 402-417, 1995.
39. T. Dudderar, Y. Degani, J. Spadafora, K. L. Tai, and R. Frye. "AT&T JlSurface
Mount Assembly: A New Technology for the Large Volume Fabrication of Cost
Effective Flip-Chip MCMs," Proc. MCM '94, pp. 266-272, 1994.
40. W. Radlik, K. Plehnert, M. Zellner, A. Achen, R. Heistand II, D. Castillo, and R.
Urscheler. "MCM-D Technology for a Communication Application," Proc. MCM
'94, pp. 402-409, 1994.
41. V. Rao, D. Hutchins, J. Reagan, D. Scheid, R. Streif, T. Syrstad, C. Eggerding, and
T. Redmond. "Manufacturing of Advanced Cu-PI Multi Chip Modules," Proc. 43rd
ECTC, pp. 920-934, 1993.
42. M. Chopra and R. Sechler. "Comparison of MCMs for a High Performance Worksta-
tion'" Proc. [CEMM '93, pp. 457-467, 1993.
43. G. B. Leung and S. A. Sands. "A Thin Films on MLC Application," Proc. 41st
ECTC, pp. 10-13, 1991.
44. S. Numata, S. Oohara, K. Fujisaki, 1. Imaizumi, and N. Kinjo. "Low Thermal
Expansion Polyimide," J. Appl. Polymer Sci., 31(101): pp. 81-86, 1986.
45. S. Numata, T. Miwa, Y. Misawa, D. Makino, J. Imaizumi, and N. Kinjo. "Low
Thermal Expansion Polyimides and Their Applications," Mater. Res. Soc. Proc., p.
I, 1987.
46. See, for example, Japanese Patent 61-111598, 1987.
47. R. Tummala, "Glass Composition for Glass-Metal Package," U. S. Patents
3,640,738, 1971.
48. T. Matsuzaki, K. Sato, T. Suzuki, and M. Terashima. "High Density Multilayer
Wiring Substrate Using Copper Thick Film and Thin Film," IEEE Trans. Components
Hybrids Manuj Technol., CHMT-8: pp. 484-491, 1985.
49. R. H. Heistand II, D. C. Frye, D. Burdeaux, 1. Carr, and P. Garrou. "Economic
Evaluation of Deposited Dielectric MCM Manufacturing Costs," Proc. 2nd [EPS/
[SHM MCM Conj, pp. 441-450, 1993.
50. P. Ricker!, 1. Stephanie, and P. Jr. Slota. "Evaluation of Photosensitive Polyimides
for Packaging Applications", Proc. 37th ECC, pp. 220-225, 1987.
51. 1. H. Lai, Polymers for Electronic Applications, CRC Press, Boca Raton, FL, 1989.
52. M. Berry, T. Tessier, I. Turlik, G. Adema, D. Burdeaux, J. Carr, and P. Garrou
"Benzocyclobutene as a Dielectric for Multichip Module Fabrication," Proc. 40th
ECTC, pp. 746-750, 1990.
53. D. H. Klockow. "Expanding Hybrid Circuits Through Thick and Thin," AT & T Bell
Laboratories Record, pp. 25-29, March, 1984.
54. A. Strandjord, R. Heistand, J. Bremmer, P. Garrou, and T. Tessier. "A Photosensitive
BCB on Laminate Technology (MCM-LD)," Proc. 44th ECTC, pp. 374-385,1994.
55. H. F. Bok. "Process Enhancements in Meniscus Coating for Flat-Panel Displays,"
SID 94 Digest, pp. 940-942, 1994.
56. L. J. Laursen and P. Garrou. "Consortium for Intelligent Large Area Processing
(CILAP)-Program Review," Proc. [CEMCM '95, pp. 253-258,1995.
57. R. Kambe, M. Kuroda, R. Imai, and Y. Kimura. "Copper Polyimide Multilayer
Substrate for High Speed Signal Transmission," Proc. 41st ECTC, pp. 14-16, 1991.
12.12 REFERENCES 11-807

58. E. Fogarassy and S. Lazare. Laser Ablation of Electronic Materials: Basic Mecha-
nisms and Applications, pp. 39-53, Elsevier Science Publishers, Amsterdam, 1992.
59. E. Fogarassy and S. Lazare. Laser Ablation of Electronic Materials: Basic Mecha-
nisms and Applications, pp. 239-253, Elsevier Science Publishers, Amsterdam, 1992.
60. T. G. Tessier, W. F. Hoffman, and 1. W. Stafford. "Via Processing Options for MCM-
D Fabrication: Excimer Laser Ablation vs Reactive In Etching," Proc. 41st ECTC,
pp. 827-834, 1991.
61. T. F. Redmond, 1. R. Lankard, J. G. Balz, G. R. Proto, and T. A. Wassick. ''The
Application of Laser Process Technology to Thin Film Packaging," Proc. 41st ECTC,
pp. 1066-1071, 1991.
62. R. Srinivasan. "Interactions of Polymer Surfaces with Ultraviolet Laser Pulses,"
Photochemistry and Polymeric Systems, ed. by J. M.Kelly, C. B .McArdle, and M. J .de
F.Maunder, pp. 46-53, Royal Society of Chemistry, London, 1992.
63. S. Lazare, H. Hiraoka, A Cros, and R. Gustiniani. "Ultra-Violet Laser Photoablation
of Thermostable Polymers: Polyimides, Polyphenylquinoxaline and Teflon AF," in
Polyimides and Other High-Temperature Polymers, ed. by M. 1.M. Abadie and B.
Sillion, Elsevier Science Publishers, Amsterdam, pp. 395-406, 1991.
64. T. G. Tessier and G. Chandler. "Compatibility of Common MCM-D Dielectrics with
Scanning Laser Ablation Via Processes," Proc. 42nd ECTC, pp. 763-765, 1992.
65. F. Bachmann. "Excimer Laser Drill for Multilayer Printed Circuit Boards: From
Advanced Development to Factory Floor," MRS Bull., pp. 49-54, 1989.
66. T. Shimoto, K. Matsui, M. Kimura, and K. Utsumi. "High Density Multilayer
Substrate Using BCB Dielectric," Proc. IMC, pp. 325-326, 1992.
67. T. G. Tessier and E. G. Myszka. "Approaches to Cost Reducing MCM-D Substrate
Fabrication," Proc. 43rd ECTC, pp. 570-578, 1993.
68. G. E. Wolbold, C. L. Tessler, and D. J. Tudryn. "Characterization, Set-Up and Control
of a Manufacturing Laser Ablation Tool and Process," in Excimer Lasers: Applica-
tions, Beam Delivery Systems, and Laser Design, SPIE Proc. No. 1835, pp. 62-69,
SPIE,1992.
69. J. H. Brannon and J. R. Lankard. "Patterning of Polyimide Films with Ultraviolet
Light," US Patent No. 4,508,749. 1985.
70. T. G. Tessier, G. M. Adema, S. M. Bobbio, and I. Turlik. "Low Temperature Etch
Masks for High Rate Magnetron RIE of Polyimide Dielectrics in Thin Film Packaging
Applications," Proc. 3rd Int. SAMPE Electronics Conj., pp. 85-87, 1989.
71. M. J. Rutter. "Via Formation in Thick Polyimide Layers for Silicon Hybrid Multichip
Modules," in Microelectronic Packaging Technology: Materials and Processes, Proc.
2nd ASM Inter. Electron. Mater. Processing Congress, ed. by W. T. Shieh, ASM
Press, Metals Park, OH, 1989.
72. J. Paraszczak, J. Cataldo, E. Galligan, W. Graham, R. McGouey, S. Nunes, R. Serino,
D. Shih, E. Babich, A. Deutsch, and G. Kopcsay. "Fabrication and Performance
Studies of Multilayer Polymer/Metal Interconnect Structures for Packaging Applica-
tions," Proc. 41st ECTC, pp. 362-369, 1991.
73. D. Y. Shih, H. Yeh, C. Narayan, J. Lewis, W. Graham, S. Nunes, 1. Paraszczak, R.
McGouey, E. Galligan, J. Cataldo, R. Serino, E. Perfecto, C.-A. Chang, A. Deutsch,
L. Rothman, and J. Ritsko, "Factors Affecting the Interconnection Resistance &
Yield in the Fabrication of Multilayer Polyimide/Metal Thin Film Structure," Proc.
42nd ECTC, pp. 1002-1014, 1992.
74. R. R. Tummala and E. J. Rymaszewski. Microelectronics Packaging Handbook, von
Nostrand Reinhold, New York, 1989.
75. H. J. Neuhaus. "An MCM-D Substrate Fabrication Model: Comparing Dry Etch,
Wet Etch and Photosensitive Dielectrics," Proc. 2nd IEPSIISHM MCM Conj., pp.
46-50, 1993.
11-808 THIN-FILM PACKAGING

76. E. D. Perfecto, C. Osborn, and D. G. Berger. "Factors that Influence Photosensitive


Polyirnide Lithography Performance," Proc. 2nd IEPSIISHM MCM ConJ., pp. 40-
45, 1993.
77. K. Moriya, T. Ohsaki, and K. Katsura. "High Density Multilayer Interconnection
with Photosensitive Polyirnide Dielectric and Electroplating," Proc. 34th ECC, pp.
82-83, 1984.
78. H. Takasago, M. Takada, K. Adachi, K. Yamada, Y. Onishi, and Y. Morihiro. "Fine-
Line Multilayer Hybrids with Wet Processed Conductors and Thick Film Resistors,"
Proc. 34th ECC, pp. 324-326, 1984.
79. O. Ahimada, K. Ito, T. Miyagi, S. Kimijima, and T. Sudo. "Electrical Properties of
a Multilayer Thin Film Substrate for Multichip Packages," Proc. Japan IEMT Symp.,
pp. 123-124, 1989.
80. E. W. Rutter, E. S. Moyer, R. F. Harris, D. C. Frye, V. L. St. Jeor, and F. L. Oaks.
"A Photodefinable Benzocyciobutene Resin for Thin Film Microelectronics Applica-
tions," 1st Int. MCM Conf., pp. 394-396, 1992.
81. P. Mukerji, T. G. Tessier, G. Demet, and S. Dasgupta. "MCM Via Generation by
Wet Processing," ISHM Polymer and Ceramic Low K Dielectrics: Advances in
Processing and Characterization Workshop, 1992.
82. H. 1. Neuhaus. "A High Resolution, Anisotropic Wet Patterning Processing Technol-
ogy for MCM Production," 1st Int. MCM Conf., pp. 256-263, 1992.
83. T. Rucker, V. Murali, and R. Shukla. "A Wet Etch Polyirnide Process for Multichip
Modules," 1st Int. MCM Conf., pp. 251-254, 1992.
84. C. E. Diener and J. R. Susko. "Etching of Partially Cured Polyirnides," Proc. 1st
Technical Conf. Polyimides, pp. 353-364, 1992.
85. T. Inoue, H. Matsuyama, E. Matsuzaki, Y. Narizuka, M. Tanaka, and T. Takenaka.
"Micro Carrier for LSI Chip Used in the HITAC M-880 Processor Group," Proc.
41st ECTC, pp. 704-706, 1991.
86. A. Kimura, T. Tsujimura, K. Saitoh, and Y. Kohno. "Fabrication and Characteristics
of Silicon Carrier Substrates for Silicon on Silicon Packaging," 1st Int. MCM Conf.,
p. 23, 1992.
87. Y. H. Kim, J. Kim, G. F. Walker, C. Feger, and S. P. Kowalczyk. "Adhesion and
Interface Investigation of Polyimide on Metals," J. Adhesion Sci. Technol., 2(2): pp.
95-98, 1988.
88. G. M. Adema, L. T. Hwang, G. A. Rinne, and I. Turlik. "Passivation Schemes for
CopperlPolymer Thin Film Interconnections Used in Multichip Modules," Proc.
42nd ECTC, pp. 776-782, 1992.
89. G. Czornyj, K. Chen, G. Prada-Silva, A. Arnold, H. Souletis, S. Kim, M. Ree, W.
Volksen, D. Dawson, and R. DiPietro. "Polyimide for Thin Film Redistribution on
Glass-Ceramic/Copper Multilevel Substrates," Proc. 42nd ECTC, pp. 682-687, 1992.
90. S. P. Kowalczyk, Y.-H. Kim, G. F. Walker and 1. Kim. "Polyimide on Copper: The
Role of Solvent in the Formation of Copper Precipitates," Appl. Phys. Lett., 52(5):
pp. 375-376,
91. R. Hiestand, R. DeVellis, P. Garrou, D. Burdeaux, T. Stokich, P. Townsend, T.
Mania, and L. Bratton. "Cyciobutene 3022 (BCB) for Non-Hermetic Packaging,"
Proc. ISHM '92 ConJ., pp. 584-590, 1992.
92. T. G. Tessier, I. Turlik, G. M. Adema, D. Sivan, E. K. Yung, and M. J. Berry.
"Process Considerations in Fabricating Thin Film Multichip Modules," Proc. 1989
Int. Electronic Packaging Symp., pp. 248-270, 1989.
93. W. J. Bertram. "High Density, Large Scale Interconnection for Improved VLSI
System Performance," Proc. 19871EDM, pp. 100-103, 1987.
94. E. D. Perfecto, K. Prasad, C. Osborn, S. Swaminathan, G. White, and C. Prasad.
"Comparison of Planar and Non-Planar Thin Film Processes," Proc. 43rd ECTC,
(Supplement to the regular proceedings), 1993.
12.12 REFERENCES 11-809

95. J. T. Pan, S. Poon, and B. Nelson. "A Planar Approach to High Density Copper-
Polyimide Interconnect Fabrication," Proc. 8th [EPS, pp. 174-189, 1988.
96. T. J. Buck. "Substrates for High Density Packaging," Proc. 1990 NEPCON West,
p. 650, 1990.
97. S. K. Ray, K. Beckham, and R. Mater. "Flip Chip Interconnection Technology for
Advanced Thennal Conduction Modules," Proc. 41st ECTC, pp. 772-778, 1991.
98. S. K. Ray, K. Seshan, and M. Interrante. "Engineering Change (EC) Technology for
Thin Film Metallurgy on Polyimide Films," Proc. 40th ECTC, pp. 395-400, 1990.
99. L. B. Rothman. "Properties of Thin Polyimide Films," J. Electrochem. Soc.: Solid-
State Sci. Technol., 127(10): pp. 2216-2220, 1980.
100. C. C. Chao and W. V. Wang. "Planarization Enhancement of Polyimides by Dynamic
Curing and the Effect of Multiple Coating," Proc.1st Technical Con! Polyimides,
pp. 783-793, 1982.
101. L. E. Stillwagon and R. G. Larson. "Topographic Substrate Leveling During Spin
Coating," Proc. Symp. Patterning Sci. Technol., 90(1): ed. by R. Gleason, J. Hefferon,
and L. White, The Electrochemical Soc., Pennington, NJ, pp. 230-238, 1989.
102. D. R. Day, D. Ridley, J. Mario, and S. D. Senturia. "Polyimide Planarization in
Integrated Circuits," Proc. 1st Technical Con! Polyimides, pp. 767-780, 1982.
103. H. Akahoshi, M. Kawamoto, T. Itabashi, O. Miura, A. Takahashi, S. Kobayashi, M.
Miyazaki, T. Mutho, M. Wajima, and T. Ishimaru. "Fine Line Circuit Manufacturing
Technology with Electroless Copper Plating," Proc. 44th ECTC, pp. 367-373, 1994.
104. Y. Kasuya, Y. Takahashi, Y. Uno, Y. Iguchi, and T. Kanamori. "PlanarizationProcess
of Copper-Polyimide Thin Film Multilayer Substrate," Proc. 14th [EEElCHMT Japan
Int. Electronic Manufacturing Techno/. Symp., pp. 13-17, 1993.
105. D. Frye, M. Skinner, R. Heistand II, P. Garrou, and T. Tessier. "Cost Implications
of Large Area MCM Processing", Proc. MCM '94, pp. 69-80, 1994.
106. G. White, E. Perfecto, T. DeMercurio, D. McHerron, T. Redmond, and M. Norcott,
"Large Fonnat Fabrication-A Practical Approach to Low Cost MCM-D", Proc.
MCM '94, pp. 86-93, 1994.
107. L. H. Ng, "Economic Impact of Processing Technologies on Thin Film MCMs,"
Proc. 42nd ECTC, pp. 1042-1045, 1992.
108. T. F. Redmond, C. Prasad, and G. A. Walker. "Polyimide-CopperThin Film Redistri-
bution on Glass Ceramic/Copper Multilevel Substrates," Proc. 41st ECTC, pp. 689-
693, 1991.
109. A. H. Landzberg. Microelectronics Manufacturing Diagnostics Handbook, van Nos-
trand Reinhold, New York, 1992.
110. G. Messner, I. Turlik, J. W. Balde, and P. E. Garrou. Thin Film Multichip Modules,
International Society for Hybrid Microelectronics, Reston, V A, 1992.
111. T. A. Wassick. "Repair of Thin Film Wiring with Laser-Assisted Processes," Proc.
42nd ECTC, pp. 759-762, 1992.
112. L. Economikos and D. W. Onnond. "Thin Film Repair Process for Interlevel Electri-
cal Connectors," IBM Tech. Disclosure Bull., No.2, pp. 17-19, 1991.
113. G. P. Flayter, C. H. Perry, and S. K. Ray. "Repair Technique for Inter-Level Short,"
IBM Tech. Disclosure Bull., 26(1): pp. 242-243, 1983.
114. M. L. Cohen, R. A. Unger, and J. F. Milkovsky. "Laser Machining of Thin Films
and Integrated Circuits," Bell Sys. Tech. J., 47: pp. 385-407, 1968.
115. R. L. Waters and G. N. Ravich. "Circuit Surgery Using Xenon and YAG Lasers,"
Int. Symp. Testing Failure Analysis, pp. 86-91, 1982.
116. G. R. Levinson and V. I. Smilga. "Laser Processing of Thin Films (Review)," Sovi.
J. Quantum Electron., 6(8): pp. 885-897, 1976.
117. T. H. Baum and P. B. Comita. "Laser-Induced Chemical Vapor Deposition of Metals
for Microelectronic Technologies," Thin Solid Films, 218(1-2): pp. 80-94, 1992.
11-810 THIN-FILM PACKAGING

118. H. G. Muller, C. T. Galanakis, S. C. Sommerfeldt, T. J. Hirsch, and R. F. Miracky.


"Laser Process for Personalization and Repair of Multi-chip Modules," in Lasers in
Microelectronics Manufacturing, ed. by B. Braren, SPIE Proc. No.1598, pp. 132-140,
SPIE, 1991.
119. A. Reisman, D. Temple, and I. Turlik. "Metal-Organic Chemical Vapor Deposition
for Repairing Broken Lines in Microelectronic Packages," European Patent Applica-
tion No. WO 92/08246, 1992.
120. Y. Morishige and S. Kishida. "Highly Conductive Gold Direct Writing by Laser
Induced CVD," Spring Meeting Japan Soc. App/. Phy., Vol. 2, No. 564, 1991.
121. T. H. Baum, P. B. Comita, and R. L. Jackson. "Process for Interconnecting Thin-
Film Electrical Circuits," US Patent No. 4,880,959, 1989.
122. M. E. Gross. "Laser Direct-Write Metallization in Thin Metallo-organic Films,"
Chemtronics, 4: pp. 197-201, 1989.
123. R. L. Melcher and N. S. Shiren. "Method for Selective Plating and Etching Using
Resistive Heating," IBM Tech. Disclosure Bull., 6(81): pp. 246-247, 1981.
124. J. Partridge, B. Hussey, C. Chen, and A. Gupta. "Repair of Circuits by Laser
Seeding and Constriction-Induced Plating," IEEE Trans. Components Hybrids Manu!
Technol., 15(2): pp. 252-257, 1992.
125. E. F. Handford, J. M. Harvilchuck, M. J. Interrante, R. A. Jackson, R. N. Master,
S. K. Ray, W. E. Sablinski, and T. A. Was sick. "Method and Structure for Repairing
Electrical Lines," US Patent No. 5,153,408, 1992.
126. J. C. Logue, W. J. Kleinfelder, P. Lowy, J. R. Moulic, and W. Wu. "Techniques for
Improving Engineering Productivity of VLSI Devices," IBM J. Res. Devel., 25(3):
pp. 107-115, 1981.
127. D. L. Klein, P. A. Leary-Renick, and R. Srinivasan. "Ablative Photodecomposition
Process for Repair of Line Shorts," IBM Tech. Disclosure Bull., 2(84): pp. 4669-
4671, 1984.
128. T. L. Michalka, W. Lukaszek, and J. D. Meindl. "A Redundant Metal-Polyimide Thin
Film Interconnection Process for Wafer Scale Dimensions," IEEE Trans. Semicond.
Manu!, 3(4): pp. 158-167, 1990.
129. J. I. Raffel. "Laser Linking for Defect Avoidance and Customization," in Proc. SPIE
Lasers in Microlithography, ed. by D. J. Ehrlich, J. Y. Tsao, and J. S. Batchelder,
Vol. 774, pp. 93-100, SPIE, 1987.
130. W. G. Burger and C. W. Weigel. "Multilayer Ceramics Manufacturing," IBM J. Res.
Devel., 27(1): pp. 11-19, 1983.
131. M. Swaminathan, E. Perfecto, and K. Prasad. "Thin Film Multichip Module Technol-
ogy at IBM: Design and Electrical Performance," IEPS Con! and Exhibit., pp.
780-787, 1993.
132. M. E. Williams. "Production of MCP Chip Carriers," Proc. 40th ECTC, pp. 408-
411, 1990.
133. D. J. Bendz, R. W. Gedney, and J. Rasile. "CostlPerformance Single-Chip Module,"
IBM J. Res. Deve/., 26(3): pp. 278-285, 1982.
134. R. R. Tummala, J. U. Knickerbocker, S. H. Knickerbocker, L. W. Herron, R. W.
Nufer, R. N. Master, M. O. Neisser, B. M. Kellner, C. H. Perry, J. N. Humenik, T.
F. Redmond, "High-Performance Glass-Ceramic/Copper Multilayer Substrate with
Thin-Film Redistribution.," IBM J. Res. Devel., 36(5): pp. 889-904, 1992.
135. E. E. Davidson, P. H. Harding, G. A. Katopis, M. G. Nealon, and L. L. Wu. "The
Design of ES 9000 Module," Proc. 41st ECTC, pp. 50-54, 1991.
136. M. G. Bregman, A. Kimura, T. Matsui, H. Nishida, K. Nishiyama, H. Ohkuma, A.
Tanaka, C. Kovac, and D. McQueeney. "A Thin Film Multichip Module for Worksta-
tion Applications," Proc. 42nd ECTC, pp. 968-972, 1992.
12.12 REFERENCES 11-811

137. R. Tummala and B. Clark. "Multichip Packaging Technologies in IBM for Desktop
to Mainframe Computers," Proc. 42nd ECTC, pp. 1-9, 1992.
138. H. Wessely, W. Turk, K. H. Schmidt, and G. Nagel. "Computer Packaging," Siemens
Res. Devel., Rep Bd 17: pp. 234-244, 1988.
139. H. Brosamle, B. Brabetz, V. V. Ehrenstein, and F. Bachmann. "Technology for a
Microwiring Substrate," Siemens Res. Devel., Rep Bd 17(5): pp. 249-253, 1988.
140. H. Brosamle. "High Density Multichip Module Based on PWB Technology," 2nd
Inter. Symp. on Printed Circuit: Future European Trends and Printed Circuit Tech-
nology, 1991.
141. H. Wessely, O. Fritz, P. Klimke, W. Koschnick, and K. H. Schmidt. "Electronic
Packaging in the 1990's: The Perspective from Europe," IEEE Trans. Components
Hybrids Manu! Technol., 14(2): pp. 272-284, 1991.
142. T. Inoue, H. Matsuyama, E. Matsuzaki, Y. Narizuka, M. Ishino, T. Takenaka, and
M. Tanaka. "Microcarrier for LSI Chip Used in the Hitachi M-880 Processor Group,"
IEEE Trans. Components Hybrids Manu! Technol., 15(1): pp. 7-14, 1992.
143. F. Kobayashi, Y. Watanabe, M. Yamamoto, A. Anzai, A. Takahashi, T. Daikoku,
and T. Fujita. "Hardware Technology for Hitachi M-880 Processor Group," Proc.
41st ECTC, pp. 693-703, 1991.
144. K. Mukai, A. Saiki, K. Yamanaka, S. Harada, and S. Shoji. "Planar Multi-level
Interconnection Technology Employing a Polyimide," IEEE Trans. Solid-State Cir-
cuits, SC-13: pp. 462-467, 1978.
145. T. Nishida, K. Mukai, T. Inaba, T. Kato, I. Tezuka, andN. Horie. "Moisture Resistance
of Polyimide Multilevel Interconnect LSI's," Proc. IEEE-IRPS, pp. 148-152, 1985.
146. A. J. Blodgett and D. R. Barbour. "Thermal Conduction Module, a High Performance
Multilayer Ceramic Package," IBM J. Res. Devel., 26(1): pp. 30-36, 1982.
147. F. C. Chong, C. W. Ho, K. Liu, and S. Westbrook. "High Density Multichip Memory
Package," WESCONI85 Professional Program Session Record, Session 7, 1985.
148. H. B. Bakoglu. Circuits, Interconnections and Packaging for VLSI, Addison-Wesley,
Reading, MA, 1989.
149. P. Dunbeck, R. Dischler, J. McElroy, and F. Swiatowiec. "HDSC and MCU Design
and Manufacture," Digital Tech. J., 2(4): pp. 99-108, 1990.
150. D. Marshall and J. McElroy. "VAX 9000 Packaging-The Multi Chip Unit," 35th
COMPCON, pp. 54-57, 1990.
151. P. Hardin, G. Melvin, and M. Nealon. "The ES!9000 Glass Ceramic Thermal Conduc-
tion Module Design for Manufacturability," IEEElCHMT '91 IEMT Symp., pp.
351-355, 1991.
152. A. Dohya, T. Watari, and H. Nishimori. "Packaging Technology for the NEC SX-3!
SX-X Supercomputer," Proc. 40th ECTC, pp. 525-533, 1990.
153. T. Watari, Private Communication.
154. K. Umezawa, M. Kimura, H. Nishimori, T. Mizuno, K. Kimbara, and R. Nakazaki.
"A High-Performance GaAs Multichip Package for Supercomputers," NEC Res.
Devel., 33(1): pp. 32--40, 1992.
155. S. Mok. "Volume Implementation of MCM-D Based Cache SRAM Products for
Workstation and PC Applications," Proc. MCM '94, pp. 320-325, 1994.
156. G. Genge!. "A Process for the Manufacturing of Cost Competitive MCM Substrates",
Proc. MCM '94, pp. 182-187, 1994.
157. R. Fillion, R. Wojnarowski, B. Gorowitz, W. Daum, and H. Cole. "Conformal
Multichip-on-Flex (MCM-F) Technology", Proc. ICEMCM '95, pp. 52-58, 1995.
158. M. Moser and T. G. Tessier. "Higher Density PCBs for Enhanced SMT and Bare
Chip Assembly Applications," Proc. ICEMCM '95, pp. 543-552, 1995.
159. D. N. Light, J. S. Kresge, and C. R. Davis, "Integrated Flex: Rigid-Flex Capability
in a High Performance MCM," Proc. MCM '94, pp. 430-442, 1994.
11-812 THIN-FILM PACKAGING

160. H. L. Heck, J. T. Kolias, and J. S. Kresge. "High Performance Carrier Technology,"


Proc. 19931EPS Conj., pp. 771-779, 1993.
161. C. Narayan, S. Purushothaman, F. Doany, and A. Deutsch. "Thin Film Transfer
Process for MCM Interconnect Wiring," Proc. MCM '94, pp. 105-114, 1994.
162. M. Y. Lau, K. L. Tai, R. C. Frye, M. Saito, and D. D. Bacon. "A Versatile, IC
Process Compatible MCM-D for High Performance and Low Cost Applications,"
Proc.ICEMM '93, pp. 107-112, 1993.
163. D. Scheid, "Advanced MCM-D with Embedded Resistors," Proc. MCM '94, pp.
273-278, 1994.
164. J. M. Cech, A. Burnett, and L. Knapp. "Pre-Imidized Photoimageable Polyimide as
a Dielectric for High Density Multichip Modules," Polymer Sci. Eng., 32(21): p.
1647, 1992.
165. T. Lenihan, L. Schaper, Y. Shi, G. Morcan, and J. Parkerson. "Embedded Thin Film
Resistors, Capacitors and Inductors in Flexible Polyimide Films," Proc. 46th ECTC,
pp. 119-124, 1996.
166. M. Oppermann, E. Feurer, and B. HoI!. "Development and Realization of a Double-
face Populated Multichip Module in Thin Film Technology for High Frequency
Application," Proc. MCM '94, pp. 279-284, 1994.
167. R. Kambe, R. Imai, T. Takada, M. Arakawa, and M. Kuroda. "MCM Substrate with
High Capacitance," Proc. MCM '94, pp. 136-141, 1994.
168. M. Tsukada, J. Cross, M. Nishizawa, K. Kurihara, and N. Kamehara. "Preparation
and Characterization of Dielectric Thin Films for Decoupling Capacitors," Proc.
lCEMCM '95,pp. 347-352,1995.
169. D. Dimos, S. Lockwood, R. Schwartz, and M. Rodgers. "Thin-Film Decoupling
Capacitors for Multi-Chip Modules," Proc. 44th ECTC, pp. 894-899, 1994.
170. T. Garino, D. Dimos, and S. Lockwood. "Integration of Thin Film Decoupling
Capacitors," Proc. ISHM '94., pp. 179-184, 1994.
171. T. Takken and D. Tuckerman. "Integral Decoupling Capacitance Reduces Multichip
Module Ground Bounce," Proc. 43rd ECTC, pp. 79-84, 1993.
172. D. Tuckerman, D. Benson, H. Moore, J. Homer, and J. Gibbons. "A High-Perfor-
mance Second-Generation SPARC MCM," Proc. MCM '94, pp. 314-319, 1994.
173. L. Matthew, N. Brathwaite, K. Flatow, H. Whittmore, and L. Bauer. "Production
Update: Report on Volume MCM Implementation," Proc. lCEMCM '95, pp. 229-
234, 1995.
174. P. Chahal, R. Tummala, M. Allen, and M. Swaminathan. "A Novel Integrated
Decoupling Capacitor for MCM-L Technology," Proc. 46th ECTC, pp. 125-132,
1996.
175. C. H. Abn, Y. J. Kim, and M. G. Allen. "A Fully Integrated Planar Toroidal Inductor
with a Micromachined Nickel-Iron Magnetic Bar," IEEE Trans. Components Packag-
ing Manuf. Technol., Part A, 17(3): pp. 463-469, 1994.
176. W. Baker. "Low Cost Thin Film MCMs," ARPA WWWhomepage (http://eto.sysplan.
cornlETOIEI-Packaging/presentation), 1995.
177. M. Parodi, W. Batchelder, J. McKibben, and P. Haaland. "Developments and Trends
in Polymer Coating Technologies for Flat-Panel-Display Manufacturing," SID 94
Digest, pp. 933-935, 1994.
178. F. Bachmann. "Excimer Lasers in a Fabrication Line for a Highly Integrated Printed
Circuit Board," Chemtronics, 4(3): pp. 149-152, 1989.
179. K. Jain, J. Hoffman, T. Dunn, W.Folster, and D. Pancha!. "Large-Area, High-
Throughput, High-Resolution Patterning and Via-Drilling System," Proc. lCEMCM
'95, pp. 327-335, 1995.
180. R. Patel, W. Advocate, and S. Mukkavilli. "Projection Laser Ablation Mask Alterna-
tives," Proc. lCEMCM '95, pp. 320-326, 1995.
12.12 REFERENCES 11-813

181. A. E. Nader. "Photodefinable Polyimides Designed for Use as Multilayer Dielectrics


for Multi-Chip Modules", Proc. MCM '92, pp. 439-413, 1992.
182. B. D. Kotzias, A. Nader, D. Murray III, and W. Lautenberger. "Application of a
Photodefinable Polyimide Process for Manufacturing Multi-Chip Modules," Proc.
ICEMM, '93, pp. 46-50,1993.
183. C. H. Ting, M. Paunovic, P. L. Pai, and G. Chiu. "Selective Electroless Metal Deposi-
tion for Integrated Circuit Fabrication," J. Electrochem. Soc., 136(2): pp. 462-
466,1989.
184. S. M. Ho, T. H. Wang, H. L. Chen, K. M. Lian, and A. Hung. "Metallization of
Polyimide Film by Wet Process," J. Appl. Polymer Sci., 51(8): pp. 1373-1380, 1994.
185. M. Matsuoka, 1. Murai, and C. Iwakura. "Kinetics of Electroless Copper Plating and
Mechanical Properties of Deposits," J. Appl. Polymer Sci., 139(9): pp. 2466-
2470,1992.
186. K. Kondo, J. Ishikawa, O. Takenada, and T. Matsubara. "Acceleration of Electroless
Copper Deposition in the Presence of Excess Triethanolamine," J. Electrochem. Soc.,
138(12): pp. 3629-3633, 1991.
187. T. N. Vorob'eva, V. A. Rukhlya, V. V. Sviridov, and E. V. Gert. "Selective Deposi-
tion of Copper on a Polyimide Film from a Chemical Metallization Bath," J. Appl.
Chem. USSR, 59(3): pp. 508-513, 1986.
188. S. Nakahara and Y. Okinaka. "Microstructure and Mechanical Properties of Electro-
less Copper Deposits," Annu. Rev. Mater. Sci., Vol. 21, pp.93-129, 1991.
189. R. R. Thomas and J. M. Park. "Vapor Phase Deposition of Palladium for Electroless
Copper Plating," J. Electrochem. Soc., 136(6): pp. 1661-1666, 1989.
190. E. K. Yung and L. T. Romankiw. "Plating of Copper into Through-Holes and Vias,"
J. Electrochem. Soc., 136(1): pp. 203-215, 1989.
191. H. Yoshiki, V. Alexandruk, K. Hashimoto, and A. Fujishima. "Electroless Copper
Plating Using ZnO Thin Film Coated on a Glass Substrate," J. Electrochem. Soc.,
141(5): pp. L56-L58, 1994.
192. H. Koyano, M. Kato, and H. Takenouchi. "Electroless Copper Plating from Copper-
Glycerin Complex Solution," J. Electrochem. Soc., 139(11): pp. 3112-3116, 1992.
193. R. Jagannathan and M. Krishnan. "Electroless Plating of Copper at a Low pH Level,"
IBM J. Res. Devel., 37(2): pp. 117-123, 1993.
194. N. V. Mandich and G. A. Krulik. "Fundamentals of Electroless Copper Bath Opera-
tion for Printed Circuit Boards," Metal Finishing, 91(1): pp. 33-41, 1993.
13

PACKAGE ELECTRICAL
TESTING
MADHAVAN SWAMINATHAN-Georg;a Tech
ABHIJIT CHATTERJEE-Georgia Tech
FRANK CRNIC-IBM
BRUCE C. KIM-Tufts University
KOPPOLU SASIDHAR-Georgia Tech

13.1 INTRODUCTION

Electronic packages provide a means for interconnecting, powering,


cooling, and protecting integrated-circuit (IC) chips. This is made possible
through interconnections that provide a connection between active devices,
such as integrated-circuit chips or to other discrete components, mounted
on the package. Because semiconductor chips are expensive, a testing
scheme is necessary to ensure the integrity and performance of all the
package interconnection paths. This is a two-step process: final testing
of the interconnection paths prior to attachment of the active devices (dies
or chips) and subsequent testing of the assembled package. In this chapter,
an integrated circuit in its barest form (wafer after dicing) is called a die
and when it's ready to be packaged is called a chip.
13.1 INTRODUCTION 11-815

The testing of interconnections prior to die attachment is called


"substrate test" and it is used to guarantee a defect-free substrate. This
is a screening process to ensure that the interconnections do not contain
defects produced during fabrication and ensures that known-good-dies
(KGD) (tested dies that are good) are not assembled onto defective sub-
strates, thus minimizing the wastage of expensive dies. Functional testing
is a follow-on to substrate testing and represents the testing of the package
after die attachment to assure a defect-free system. This ensures that no
defects were introduced into the KGD during the assembly operation and
that all the terminals of the chips are connected, in the manner desired.
Pretested dies, also called KGD, facilitate the functional testing process
as opposed to dies that are not pretested, which add additional steps to
the testing process.
Both test categories may include repair processes and engineering
change (EC) schemes to minimize the wastage of fabricated substrates
due to process defects and system design modifications. A typical test
flow is shown in Figure 13-1 which consists of a substrate test to guarantee
a known-good substrate, assembly of known good die after bum-in to the
substrate, and a final functional test prior to shipment. The flow in Figure
13-1 is similar for a wide range of packaging technologies including single

MCM substrate

Pass

Pass

known good MCM

Figure 13-1. MCM Test Flow


11-816 PACKAGE ELECTRICAL TESTING

chip and multi chip packages. The philosophy and economics of testing
have been discussed in detail by Knowles [1] and Davis [2] and serves
to provide important feedback to the upstream processes and protects the
efficiency of downstream operations. The advantages of electrical testing
in packaging are summarized in Table l3-1. Because all packages need
to be tested prior to shipment to meet quality, reliability, and functionality
goals, electrical testing represents an integral part of manufacturing. Al-
though testing may not add function to the package or product, it adds
considerable value and hence is easily justified in a manufacturing envi-
ronment.
Whereas the electrical testing (substrate and functional) definitely
adds value, it could represent a major added cost component in the develop-
ment and manufacturing of electronic components and systems, with
estimates ranging up to 50% of the product cost. This is because manufac-
turers spend considerable time and effort in assuring the reliability (see
Chapter 5, "Package Reliability") and quality of their products, through
developmental efforts in design (see Chapter 3, "Package Electrical De-
sign"), test generation, and test application. The testing method or methods
selected by the manufacturer may result in expensive test facilities with
associated complexity, which are added to the product cost. As a conse-
quence of this, it is widely agreed upon that test issues be considered up
front during system design and not as a postdesign effort, so as to select
the optimum test method that meets both the cost and the performance
goals of the system under development. This is especially true for multichip

Table 13-1. Packaging Operations Flow and Substrate Test Advantages

Benefits Provided by Substrate Test

Operation
Early check of design practices
Substrate design
Design changes verified quicldy
Substrate materials Materials problems discovered and diagnosed with less product
preparation at risk
Shorter cycle of learning for yield improvements
Substrate build Effective, timely defect diagnostics
More accurate build forecasting

Substrate electrical test


Chip attach (or package to Avoid component loss due to defective substrates
package bonding)
System assembly Excess capacity not required to process modules with substrate
defects
System testlbum-in Fewer errors, analysis simplified
System use Product reliability may be improved by substrate specific tests
13.2 SUBSTRATE TEST 11-817

modules, which are inherently complex and require the use of detailed
test procedures to ensure product quality.
This chapter provides details on substrate test and functional test.
Both pass/fail testing and performance benchmarking have been explained
as part of functional test.

13.2 SUBSTRATE TEST


Consider a multilayered package as discussed in Chapters 7 ("Micro-
electronics Packaging-An Overview") and 12 ("Thin-Film Packaging").
A simplified cross section of a package has been reproduced in Figure
13-2, consisting of multiple layers of metallization separated by dielectric
layers. The multiple metal layers may be necessary due to the density
of interconnections required, line pitch allowed by the technology, and
electrical performance desired. The metallization layers consist of planes
and wiring, the configuration of which is dictated by the electrical specifi-
cations. As an example, deposited multi-chip module (MCM-D) substrates
may consist of orthogonal (X-Y) wiring layers sandwiched between volt-
age (VDD) and ground (GND) planes, providing an impedance controlled
environment for the interconnections, as shown in Figure 13-2. Other
technologies such as ceramic multi-chip module (MCM-C), laminated
multi-chip module (MCM-L), MCM-D/C, and so forth may support similar
cross sections, with increased layer counts. For example, MCM-C sub-
strates may consist of VDD-X-Y-GND layers, repeated, with 60-120
layers stacked on top of each other. As the layers are fabricated, processes
may induce defects on the wiring layers, through the presence or absence
of conductive material along the interconnection length or through the

Mi ro-Probe
ub Irate
Testing
Terminal urrace
Metallurgy

Ground
Plane

y- ignal
Plane
Optical
x- ignaJ Inspection
Plane

Figure 13-2. Hierarchy of MCM Substrate Test


11-818 PACKAGE ELECTRICAL TESTING

thinning of dielectric layers. This results in opens on interconnections,


shorts between interconnections, via opens or shorts between interconnec-
tions and planes, all of which are detrimental to the proper functioning
of the product.
In certain other cases, physical imperfections during processing may
not render an interconnection functionally open or short but may degrade
to an open or shorted condition (near-open and near-short) at a later date.
These classes of defects are of particular concern, as they can result in
unexpected failures in further processing or during customer usage. Near-
opens and near-shorts are classified as latent defects, examples of which
are shown in Figure 13-3. The use of very narrow metal lines and polymer
insulators (as in thin-film substrates) can increase sensitivity to latent
open failures under thermal, mechanical, or bias stress during use. Ionic
residues or extraneous metal from photolithographic processing can result
in current leakage paths leading to latent shorts.
To ensure the functionality of the product, defects such as opens,
shorts, latent opens, and latent shorts have to be detected, diagnosed, and
repaired. During processing, defects are detected through visual and optical
inspection, where each layer is inspected individually and defective inter-
connections are either repaired or the substrate is rejected for unrepairable
defects. Although every layer is inspected for opens or shorts, temperature
and process stressing of subsequent layers may induce defects on intercon-
nections, which therefore need to be inspected prior to die attachment.
This obviously is not possible either through visual or optical inspection
of the individual layers, because the wiring is embedded in the substrate,
as in Figure 13-2. Hence, testing is required through the top and bottom
layer pads (terminal metallurgy pads) after all the layers are fabricated.
This test is called the substrate test and is used to detect opens, shorts,
latent opens, and latent shorts on interconnections buried in the substrate.
Although opens and shorts testing can be used for part sorting,
substrates supporting high-performance applications may require paramet-
ric measurements to ensure that the interconnections meet a set of design

i /\ i

I V I ~
~
(a)
~ (b)

Figure 13-3. Latent Defects due to Near-Opens (a) and Near-Shorts (b)
13.2 SUBSTRATE TEST 11-819

criteria such as insulation resistance, conductor resistance, and line capaci-


tance. This can be used to achieve less manufacturing variance between
substrates to guarantee success for high-speed applications. Parametric
measurements to obtain data for process control, in addition to opens and
shorts testing, is also part of the substrate test.
Because the ultimate goal of testing is to assure defect-free substrates
and lower manufacturing costs, optical inspection can be used to supple-
ment the substrate test. Table 13-2 provides a comparison of optical
inspection and substrate test with emphasis on technique, efficiency, diag-
nosis capability, and application. In manufacturing simpler substrate de-
signs, substrate test is preferred due to its functional and highly efficient
defect detection capabilities. However, an effective synergism can be
developed between test and inspection for defect detection in substrates.
As an example, substrate layers can be optically inspected for defects
prior to assembly. Once the substrate is fabricated and tested, defects can
be physically analyzed. In this manner, inspection inefficiencies can be
understood, corrected and improved. Hence a combination of test and
inspection can be a powerful tool for understanding defects and improving
the yield of substrates.
Sometimes the material set and smaller feature sizes involved in
complex substrates (e.g., thin film) may imply a set of unique test require-
ments that are not typically practiced in testing substrates. For example,
controls introduced or intensified for the thin-film substrate test may
include measures to minimize particulate contaminants, metallic debris,
organic residue, and tool or handling-related substrate damage. For the

Table 13-2. Comparison of Electrical Test and Optical Inspection for Ceramic
Substrate Defect Detection

Test Inspection

Nature of Technique
Functional measure of conductor continuity Optical observation for presence or absence of
and insulator integrity conductive material in wiring or insulation
channels respectively

Efficiency
Near 100% defect detection Variable efficiency
Defect Signature Available for Defect Analysis
Net identification Precise x/y location of defects

Application
Must be fired substrate assembly Any subassembly with exposed conductors and
sufficient conductor/insulator optical contrast
11-820 PACKAGE ELECTRICAL TESTING

most complex and advanced substrate designs, the electrical test plays a
key role in yield management (see Chapter 5, "Package Reliability") for
the build process. Tests may be inserted at interim points during substrate
build in order to reject defective parts early in the process. The proper
selection of an in-process test can be used to significantly reduce the cost
of substrates as well as decrease the cycles of yield improvement for new
products. Dense substrates used in multi-chip modules (MCMs) necessitate
a more sophisticated approach to the substrate test than the simpler single-
chip module (SCM) or printed circuit board (PCB) technologies. Hence,
although MCMs offer advantages to system designers through tighter
integration and offer higher performance, they may require more complex
substrate test techniques. Table 13-3 is a summary of the advantages of
MCMs and the resulting challenges faced by test engineers.
Numerous techniques are currently being pursued for the substrate
test such as capacitance, resistance, combination of resistance/capacitance,
electron beam, latent open testing and time domain network analysis
(TDNA). In addition, other techniques are at the research and development
(R&D) phase. All these methods either require contact probing of the top
or bottom surface pads, or use contactless probing as in electron beam
testing. The key elements that differentiate test methods are the equipment
cost, test time, throughput, and defect resolution capabilities. For example,
although capacitance test equipment cost is small, the method allows only
for large-defect detection. Hence, capacitance testing has to be augmented
by other test techniques (such as resistance) for high defect resolution.
This, in turn, may increase equipment cost and impact throughput. How-
ever, stand-alone capacitance testing is cheap and easy to use. On the other
hand, though stand-alone resistance testing has high defect resolution, it
has low throughput, increased test time, and a resulting high test cost. To
avoid substrate damage through contact probing, vacuum-based electron
beam test systems may be used which are significantly slower and more
expensive for volume production testing than the more commonly used

Table 13-3. Advantage of MCM Technology and Implied Test Challenges

MCM Advantage Test Challenge

Smaller via grid Smaller probe and tighter positioning accuracy required
More 110 terminals Premium on test/product contact integrity
Greater circuit density Higher test speeds required for reasonable throughput
Level of integration Significant influence on end-product reliability,
undetected defects (escapes) more costly
Higher value package False test errors (overkill) must be minimized
13.3 SUBSTRATE TEST METHODS 11-821

capacitance or resistance testing. The TDNA method [3] may be used for
testing substrates that require a high defect resolution, but this requires
testing at high frequencies using expensive probes and equipment.
Although TDNA has been proposed as a substrate test, it has been
commonly used for high frequency characterization of substrate intercon-
nections. A method invented by IBM called the Electrical Module Test
[4] may be used to test substrates to detect latent opens on interconnections,
if this is a primary defect causing failures. Because each technique has
its own implementation method and diagnostic procedures, test methods
need to be compared based on the system application and throughput
desired. As wiring densities increase, the applicability and practicality of
test systems may vary greatly; hence, the appropriate test method should
be selected with care. The following sections discuss the various test
methods being used in industry and methods being used in industry and
methods still in the R&D phase.

13.3 SUBSTRATE TEST METHODS


13.3.1 Capacitance Testing
Capacitance testing is the most widely practiced test method. It is
based on the measurement of the capacitance of a net with respect to a
common ground plane, where a net is defined as an interconnection
between two or more terminals. It relies on the recognition that a net
can be broken down into individual elements, with the total capacitance
equaling the sum total of the individual capacitances of the elements.
Based on this scheme, the measured net capacitance is smaller than the
expected value for an open defect and larger than the expected value for
a short defect, with the defects occurring on or between interconnections,
respectively. For shorts between interconnections and ground plane, the
measured net capacitance is zero. Parametric testing using capacitance
measurements may be used to supplement opens/shorts testing to ensure
that the nets meet the required specifications for high-speed communi-
cation.
The basic concept behind capacitance testing is best illustrated by
Marshall et al. [5] as shown in Figure 13-4. Figure 13-4a shows a defect-
free net (with two terminals) that has a capacitance C associated with it.
The measured capacitance at each end of the net therefore agrees with
the expected value C. Figure 13-4b is the same net which is defective with
one open defect, resulting in two unconnected elements. The measured
capacitance from each end (CI and C2) for the defective net is therefore
less than the expected capacitance C, with the sum total equaling the
expected value C. Similarly, Figure 13-4c is the same net with one short
defect between two adjacent nets, resulting in a four-terminal net. The
measured net capacitance at each terminal is therefore greater than the
11-822 PACKAGE ELECTRICAL TESTING

1;o~
I I
CI C2
-: "'='

Cmea ured= Cmea. ured < C Cmcasured >


(a) (b) (c)

Figure 13-4. Capacitance Measurements of (a) a Defect-Free Net, (b) a Net


with an Open Defect, and (c) Two Nets Shorted Together

expected value C and equals the sum total of the individual net capacitances
assuming the short circuit has negligible capacitance associated with it.
Hence, based on a set of expected capacitances, interconnections can be
diagnosed as being defect free , using the above method.
Because probing of the net terminals and ground plane are required,
capacitance testing can be performed with a single two-point probe ar-
rangement or with a movable single-point probe and a probe connected
to the ground plane. As testing is typically done in the frequency range
1 kHz-1O MHz, close proximity of the two points of the probe is unneces-
sary. Also, because either the near or far end (not both) of the net requires
probing, probe movements are greatly simplified. Marshall et al. [5],
Economikos et al. [6], Hamel et al. [7] and Wedwick [8] provide further
details on capacitance testing of substrates.
Because capacitance testing is based on expected data which are
generated through design and modeling, a careful choice of the minimum
and maximum pass/fail limits is key to the successful use of this mode
of testing. The pass/fail limits can be explained using Wedwick's [8]
defect scenarios shown in Figure 13-5, which are tested using capacitance
testing. In the figure, the fail limit is a 100% or greater increase in
capacitance at the measurement point for a short, and a 50% or smaller
decrease in capacitance for an open. Although a single terminal requires
probing at a given time, both ends ofthe net require probing to completely
test that net for opens or shorts. As an example, although the defect is
undetectable in Figure 13-5b because the change in capacitance is <100%,
probing of terminal B enables short detection due to >100% increase in
the measured capacitance, as compared to the expected capacitance of
net B. During testing, process variations may introduce an additional 15%
13.3 SUBSTRATE TEST METHODS 11-823

A -----22:-
F - - - - -..
F
B ________________ Short ~_
I Cexpected = 2pF
Crneasured =4pF
% Change = + 100%
Diagnosis = short
(a) Short between two conductors of equal length.

I
2pF

A--~~--------------­
Short
Cexpected =2pF
Crneasured =2.5pF
% Change = +25%
B
0.5pF Diagnosis =Undetectable

(b) Capacitance Measurement on a shorted net.

A _ _I..p.F___ __...._____ B
IpF
Cexpected = 2pF
Crneasured = 1pF
(c) Open circuit at Center % Change =-50%
Diagnosis = open

iiiiIiIi________
A __....1.5pF 0.5pF
-B Cexpected =2pF
Crneasured = 1.5pF
% Change = -25%
(d) Open circuit at Far end Diagnosis =undetectable

Figure 13-5. Capacitance Measurement at Terminal A

change in expected capacitance values and has to be accounted for during


defect detection.
The most difficult defects to detect using capacitance testing is an
open circuit that divides the net exactly in two, or a short circuit between
two nets of similar lengths. Undetectable multiple defects could also occur
as a result of an even number of defects which are equally divided between
opens and shorts [5]. However, the occurrence of these defect scenarios
is small. Usually, the maximum and minimum fail limits are set no
further than ±50% from the expected value to allow for process variations,
expected value inaccuracies, and instrumentation tolerances. This margin
typically applies for opens testing because an open at the far end causes
only a small decrease in measured capacitance at the near end, whereas the
measured capacitance at the far end is much larger than the expected value.

13.3.1.1 Test Implementation


The sequence of operations for the successful implementation of
capacitance testing of substrates has been discussed by Marshall et al.
[5] and shown in Figure 13-6. The data file consists of expected capacitance
values of the various nets in the substrate. The data file (also called the
11-824 PACKAGE ELECTRICAL TESTING

Figure 13-6. Capacitance Test Implementation

test file) can be generated using measured capacitance values from the
first few substrates produced. A more popular and robust technique is to
use a CAD-based approach that allows for the test file generation before
the first substrates are produced. The popularity of the latter technique is
because physical design data are used to extract the capacitances and,
hence, enables the AC screening of substrates through a correlation of
line capacitance to line length. However, the major difficulty in extracting
accurate capacitances from the physical layout is the inability to account
for complex interactions between multiple conductors and planes. A sim-
plistic approach is generally used for capacitance extraction which is
based on the superposition of capacitance contributions from individual
elements. As an example, a net may be composed of pads, vias, and lines
whose individual capacitances are extracted using numerical simulation
tools. The expected capacitance value of the net is the sum total of the
individual capacitance contributions. This method obviously does not
include the interactions between elements. However, the method is compu-
tationally fast, requiring little time for test file generation once the physical
layout is complete. Commercial software tools and layout interfaces are
available that could be used to automate the process from the physical
design to test file generation.
13.3 SUBSTRATE TEST METHODS 11-825

Companies such as Micro Module Systems [5] and IBM [7] have
implemented a net property extractor that scans the computer-aided design
(CAD) file to extract data such as line lengths with associated geometries
(width, thickness, distance from ground plane), via information, pad de-
tails, and number of crossovers with other nets. Using two-dimensional
(2D) and three-dimensional (3D) parameter extractors, the individual ca-
pacitances are extracted and summed to compute the total capacitance of
a net. The linear predictor used by Hamel et al. [7] for computing the
total net capacitance CT is based on:

Cr = ~ CLi X L; + ~ C x ~ + ~ Cv X Nk + ~ C x N
p c m [13-1]
; j k m

where C Li is the z"lh line capacitance per unit length, 1.;. the z"lh line length,
Cp the pad capacitance, ~ the number of pads, Cv the via capacitance, Nk
the number of vias, Cc the cross over capacitance and Nm the number of
crossovers. A typical test file generated from the CAD design file is listed
in Figure 13-7 providing details on the sample number, net name, net
length, expected capacitance, impedance and capacitance per unit length.
The test file which is a list of expected capacitance values can then be
compared to measured data on a net by net basis to create an error listing
file that provides details on the defective nets. A typical measurement
system is shown in Figure 13-8 consisting of a moving probe fixture and
a capacitance meter operated at 1 MHz. In the figure, the shorting block
is used to provide the ground contact for the probe by shorting the VDD
and GND I/O's at the bottom surface.

13.3.1.2 Limitations
Capacitance testing, though easy to implement, is limited in its ability
to resolve opens and shorts due to the wide test window required (±50%).
Since defect detection is subject to process variability and is a function
of the frequency used, the test frequency can be varied to improve defect
resolution with a trade-off in test time. The typical capacitance test set-
up has little capability for providing current or voltage stressing to the
product under test. However, capacitance testing is ideally suited for
flexible in-line product screening and low volume developmental sub-
strate testing.

13.3.2 Resistance and Continuity Testing


Resistive shorts detection and continuity measurement can be com-
bined to provide good resolution and precise definitions of short and open
defects. This method is based on resistance measurements between the
extreme ends of the net using two probe heads. If the measured resistance
11-826 PACKAGE ELECTRICAL TESTING

MAXIMUM 0.1870684257
MINIMUM 0.1328073643
AVERAGE 0.1471803629
STD. DEV 0.006501338716
RANGE 0.0542610614
NO. OBS 1042

SAMPLE NO. NETNAME LENGTH EXPECTED ZO PF/MM


CAP.
1 1A3AAAB40 253.30 40.40 51.572 0.154
2 1A3AAAB54 220.00 38.30 48.000 0.165
3 1A3AAAA41 193.50 34.50 47.154 0.168
4 1A3AAAB57 188.55 34.00 46.666 0.170
5 1A3AAAA24 182.25 33.20 46.263 0.171
6 1A3AAAA44 178.65 31.90 47.321 0.167
7 1A3AAAB35 171. 45 32.40 44.667 0.177
8 1A3AAAB51 169.60 26.80 54.163 0.146
9 1A3AAAB38 166.50 29.30 48.383 0.164
10 10CAAAAC2 163.80 26.60 52.736 0.150
11 1E1AAAA03 162.00 26.50 52.369 0.151
12 1A3AAAA86 160.15 26.90 50.939 0.155
13 1A3AAAA45 159.75 29.90 45.348 0.175
14 1A3AAAE71 158.80 24.70 55.405 0.143
15 lA3AAAE72 158.80 24.90 54.921 0.144
16 1A3AAAF71 158.40 24.70 55.266 0.143
17 lA3AAAF72 158.40 24.90 54.783 0.145
18 1ElAAAB09 158.40 25.60 53.158 0.149
19 lA3AAAE05 157.90 24.60 55.335 0.143
20 1A3AAAB48 156.60 25.40 53.003 0.144

Figure 13-7. Capacitance Test File

Capacitance
Meter

Movable
Probe
GndProbe
(fixed) Substrate

Bottom Surface
..cb.....:r::I:::r:::r::r:::O::I::I:::r::r::CO:::I::EE::;- Pads

Shorting Block

Figure 13-8. Capacitance Measurement System


13.3 SUBSTRATE TEST METHODS 11-827

is within the expected range, the net is assumed to be defect-free. However,


any deviation from the expected value represents a defect with the magni-
tude of the resistance providing details on the defect size and type. The
open and short test setup is shown in Figure 13-9. It is based on a four-
point probe measurement with probes contacting the terminals of the nets.
For the opens test (Figure 13-9a), a known current I is passed through
the outer two probes, the potential V is thereby developed and measured
across the inner two probes. The ratio of V/I provides the resistance of
the net, which can then be compared to expected values. For shorts testing
(Figure 13-9c), a potential difference V is applied between two nets, and
the current I is measured. The ratio V/I provides details on the dielectric
leakage. For good dielectrics, a typical pass limit for dielectric leakage
is> lOOMQ. The equivalent circuits for opens and shorts testing are shown
in Figures 13-9b and 13-9d, respectively.
The resistance test is ideally suited for opens testing due to the
increased probings that are required for shorts testing. This is best ex-
plained using a substrate containing N nets, each consisting of n terminals
that require testing. Terminals represent pads accessible from the top or
bottom surface connected to the same net. Assuming two probe heads
(each with two probes) are used for resistance measurement, one at each
terminal of the net, all the terminals require probing at least once for
opens testing, for a total of (n-l)N tests. However, for shorts testing,

(a) (b) Deleclion:

tl
Rn (Voltage
Mea uremenl)

nel
(cl (d)

I Mea uremenl
(Currenl
Measuremenl)

Figure 13-9. Resistance and Continuity Testing. (a) Opens testing; (b) equivalent
circuit for opens testing; (c) shorts testing; (d) equivalent circuit for shorts testing. Vc =
constant voltage source (current limited); Ie = constant current source (voltage limited);
Rn = net under test.
11-828 PACKAGE ELECTRICAL TESTING

probing on the N-I remaining nets is required for each net tested, requiring
a total ofN(N-I)/2 tests assuming no duplication in measurements. Hence
the number of tests required is proportional to N for opens testing and
proportional to N2 for shorts testing. For shorts testing, each test requires
significant time due to the N(N-I)/2 relative mechanical probe movements,
between the two probe heads. Test times can however be reduced by
processing the design data from the physical layout, to identify adjacent
nets for shorts testing at the expense of increased risk of fault escapes.
Resistance testing's advantage is that it measures opens and shorts
directly and can detect low-resistance opens and high-resistance shorts.
The large number of probe movements for shorts testing heavily favors
the use of a matrix, or cluster probe (also called a "bed of nails" probe)
for complex products. This approach employs an array of test probes
placed in contact with the substrate and switched, thus greatly minimizing
the mechanical movements required of a tester. A cluster probe coupled
with mechanical relays or solid-state switching can be used to deliver
current and voltage stress to the product in an extremely efficient manner.
This method, however, requires a significant outlay for fixturing but pro-
vides the best quality and fastest test available. For example, products up
to 50 mm square can be tested in less than 20 s, using the cluster probe
method. An alternative method is to use "flying probes" for point-to-
point measurements, but this method is extremely slow for all but very
simple substrates.
Resistance testing is implemented in a manner similar to capacitance
testing. A test file is generated from the physical layout of the substrate,
providing a listing of the various nets and the associated resistances for
continuity testing. The expected resistance values are then matched against
measured data. A similar approach is used for shorts testing, where the
high resistance between nets is compared to measured data. One source
of error in resistance testing is the contact resistance associated with the
probes, which may have a significant effect on the measured data, espe-
cially for high-density interconnections as in thin film substrates. This
error can be minimized by using a four-point probe arrangement (Figure
13-9), as opposed to a two-point probe arrangement connected to a resis-
tance meter.
To eliminate lengthy resistance test time and improve the quality
of test, capacitance and resistance testing may be combined. A capacitance
tester can be initially used to measure the capacitance of every net in the
substrate. The faulty nets can then be verified using resistance testing by
applying a stimulus either within the net for an opens test, or between
nets for a shorts test, as discussed by Woodard [9].
Resistance testing is ideally suited for thin-film technology, where
leakage paths are prone to fail under voltage bias, temperature, or me-
chanical cycling. The most effective method for detecting a leakage path
13.3 SUBSTRATE TEST METHODS 11-829

is through a sensitive resistance test between conductors, as in Figure


13-9, and is practiced by IBM on thin-film substrates. This test operates
at a low current to limit the energy injected into the circuit under test
and allows defect detection without alteration or arcing. The use of higher
test voltages allows for more accurate leakage testing and may permit
wetting of poor electrical contacts (contact resistance). High voltages also
permit the breakdown of oxides or very thin air gaps that could mask a
latent short.

13.3.3 Electron Beam Testing


Noncontact testing has shown promise and offers the advantage
of eliminating mechanical probing on the substrate. Electron beam test
technology is an attractive alternative for high-throughput, layout-indepen-
dent, noncontact, nondestructive testing of unpopulated substrates. Be-
cause electrons can be positioned at any location on the substrate using
computer control, this mode of testing provides high flexibility with respect
to layout changes and, hence, can be used to test different products.
Electron beams have no mechanical impact and, thus, do not destroy the
pad surface or crack fragile substrates. Very fast beam deflection and
charge storage allow for high test speed and thus high throughput.

13.3.3.1 Test Implementation


A focused electron beam can be positioned on various locations of
the substrate using an electromagnetic deflection system. The beam can
be switched on and off by a blanking system in order to form "charge"
and "read" pulses. A voltage contrast results as a consequence of switching
the beam for a short time duration, which can then be detected and used
for discrimination between charged and uncharged pads. Ross et al. [10]
describe a means for testing substrates using a voltage-contrast electron
beam, as shown in Figure 13-10. Consider a net consisting of three
terminals that are accessible from the top surface, as shown in Figure 13-
lOa. The net is intact between terminals 1 and 2 and contains an open
between terminals 2 and 3. In the figure, the net is charged by positioning
the electron beam at terminal 1 of the net. The beam is then directed to
terminal 2 followed by terminal 3 and pulsed for a short duration at each
terminal to read the charge, as shown in Figure 13-lOb. The voltage-
sensitive detector indicates that terminal 2 is charged and terminal 3
uncharged, indicating an open between terminals 2 and 3. As a final test
sequence, all the charge on the substrate is erased by an electron flood
gun for subsequent test algorithm implementation, as shown in Figure
13-lOc. A similar sequence is used for shorts testing with the difference
that charge on the remaining nets are read to detect voltage contrasts as
a consequence of charging. To allow for pass/faillimits, charge and read
11-830 PACKAGE ELECTRICAL TESTING

3
Read Charge on Terminals 2 & 3
Read Voltage for Shorts Test

J~:N~ ~:~
Charged

TT=:TI
L-_G_o_od_ _O..:,p_en_ _---.J Substrate Open Substrate

(a) (b)

Discharge Net using Flood Gun

1 2 3

TT=:TI
I.-_G_o_od_ _O..:,p_en_ _....I Substrate

(c)

Figure 13-10. Voltage Contrast Electron Beam Test Method. (a) Charge net; (b)
charge detected on terminal 2 and none on terminal 3; (c) discharge net for next test.

threshold voltage levels are used in the tester. Nets are charged by using
the charging current to bring up their voltage levels to a charge threshold
which is above the read threshold. For this purpose, the beam is switched
on to form a charge pulse that is typically longer than the read pulse.
Brunner et aI. [11] discuss the application of electron beam testing
to MCM substrates as in Figure 13-11, which represents a group of
substrate nets consisting of multiple terminals which are accessible through
pads from the top or bottom surface of the substrate. As an example, net
N consists of two terminals while net N+ 1 consists of three terminals.
To test the nets using an electron beam tester, the following sequence
is used:
1. One terminal of net N is charged.
2. All other terminals of the same net are read. They are expected to
be charged and any voltage contrasts indicating otherwise represents
an open.
3. Terminals on net N+ 1 are next read. They are expected to be un-
charged and any voltage contrast suggesting otherwise represents a
short to a previously charged net.
4. Net N+ 1 is next charged.
5. Step 2 is repeated on all the terminals of net N+ 1 to detect opens.
6. Step 3 is repeated on all the terminals of net N+2.
7. Steps 4, 5 and 6 are repeated on nets N+2,N+3 ... until the test
sequence is complete.
13.3 SUBSTRATE TEST METHODS 11-831

Energy
charged

uncharged

BOllom

t t t t t t t t t t t t
Electrons

Figure 13·11. Electron Beam Testing

The above test sequence obviously detects and locates opens. How-
ever, this sequence is limited for shorts testing, because although it detects
shorts, it does not provide information on the nets that are shorted to each
other. Short location is done through a separate test cycle. For this purpose,
the shorted net is charged and all other nets read with a pulse to detect
the nets involved in the short. The second test cycle is short, limited to
defective nets and requires little time. The main test sequence requires
the access of each terminal twice one each for the detection of opens and
shorts. Hence, the test time increases in proportion to the number of
terminals and is dependent on the ability of the nets to hold charge during
the test duration.
The electron beam can also be used for testing connections to the
rear side of the substrate by a separate flood beam which charges all
bottom pads simultaneously. Charged pads on the top side belonging to
nets connected to bottom pads indicate correct connection. Uncharged
pads on the top side belonging to these nets represent opens. Shorts
between top to bottom nets show up during the main test sequence from
the top side.
11-832 PACKAGE ELECTRICAL TESTING

13.3.3.2 Limitations
The size of defects detectable using electron beam testing is important
in judging the compatibility of the test method with the fabrication process
being used. As an example, consider two nets shorted to each other with
a resulting shorting resistance R, as shown in Figure 13-12 and discussed
in detail by Brunner et al. [11]. The voltage level measured at pad 2 as
a result of charging pad 1 is a function of the resistance R and the total
capacitance of the shorted net, resulting in a time constant Re, assuming
that the charged voltage has reached the appropriate measurable level.
Using the electron beam method, the voltage is measurable if the time
constant is larger than the time required to read the charge on pad 2.
Hence, the resolution for shorts detection depends on the capacitance of
the shorted nets, the resistance of the short and the time required for
reading the charge. The shorted resistance detectable ranges from less
than lQ to lOOMQ. Electron beam testing is therefore ideally suited for
the detection of dielectric leakage between nets. For opens testing, Figure
13-12 can be modified to include the resistance R as part of a single net.
Typical values of open resistances that can be detected are in the range
lO-lOOMQ. Lower resistances are considered connected; hence, the
method is not suited for the detection of low-resistance opens.
For substrate stressing, electron beam testing is plagued by severe
current-carrying limitations. The test system is more expensive for volume
production testing than the more commonly used resistance or capacitance
test systems. Passives integrated into the substrate, such as resistors and
capacitors, affect defect detection. As an example, a net terminated by
an integrated resistor (of small value) to ground cannot be charged by
the electron beam to retain charge for the appropriate time (due to the
resulting small time constant) and, hence, shorts between nets cannot be
detected, as explained by Brunner et al. [11]. However, opens can be
detected on these terminated nets if the defect results in the disconnection
of the termination. Integrated capacitors to ground limit the resolution of
defects due to their influence on the charge time. Like other test methods,

Figure 13-12. Electron Beam Defect Resolution


13.3 SUBSTRATE TEST METHODS 11-833

electron beam testing does not provide details on the precise defect location
or the cause of the defect.

13.3.4 Electrical Module Test

As wiring densities increase with a consequent decrease in line width


and spacing (as in thin film substrates with line widths of 25Jlm on a
pitch of 75Jlm), defects such as latent opens have a greater chance to
occur. Latent opens are near-opens that transform into complete opens in
the field, producing failures. Detection of latent opens such as cracks,
notches, via line connections, and so forth are usually based on optical
inspection or stress testing. However, optical testing can be applied only
to visible areas and stress testing such as thermal cycling or mechanical
fatigue tests are time-consuming. For expensive substrates, the lack of a
latent open detection technique makes the package vulnerable to failure
during subsequent assembly processes. Testing for latent opens as a result
of bum-in stressing [12] on thin-film substrates has been addressed by
IBM through the formerly proprietary and currently published electrical
module test (EMT) or latent opens test [4]. Bum-in stressing involves the
application of heat cycles and/or electrical bias to the substrate in order
to reduce failure to detect latent opens. Properly chosen bum-in conditions
can help to weed out unreliable substrates. However, stress conditions
must be carefully developed so that defective circuits can be forced to
fail during bum-in without significantly weakening others. Bum-in should
also not introduce failure modes inconsistent with field conditions. Burn-
in testing is often performed with integrated-circuit chips mounted on the
substrate. This, however, could result in expensive perfectly good chips
being lost, or requiring rework due to substrate failures. Bum-in or other
steps taken at points well removed from substrate build can result in an
extended feedback time for latent defect understanding. Once corrective
action is implemented, initial parts must be measured at bum-in conditions
to verify the solution. Thus, the cycle time to reach a measurement point
(from solution to bum-in condition) is a key parameter in determining
reliability improvement rates.
The latent open testing adopted by IBM [4] is based on the application
of alternating and direct currents through metal interconnections to detect
latent electrical opens such as line narrowing, notches, nicks, cracks,
weak connections and interface contaminations. Due to the nonlinear
relationship between the voltage across and current through the intercon-
nection, distorted signals are generated based on the nature of the defect.
The phase of the distorted signal from defective interconnections are
compared to the reference phase from defect-free interconnections for
fault diagnosis. Application of this technique can improve product reliabil-
ity and reduce manufacturing cost through early defect detection.
11-834 PACKAGE ELECTRICAL TESTING

The method is based on non-linear conduction characteristics of


metal conductors due to the temperature dependence of the DC resistance
and its thermal properties. Halperin et al. [4] explains the method, using
Figure 13-13 to show the I-V characteristics of an ideal conductor which
is a straight line. As a positively biased sinusoidal current passes through
the conductor, the corresponding voltage is also a sinusoid as shown in
Figure 13-13. When the conductor has a nonlinear characteristic caused
by a resistance change due to an increase in current-induced temperature,
a distorted sinusoidal voltage is generated with a corresponding shift in
the I-V curve, as shown in Figure 13-13. The amount of distortion depends
on the size of the defect. Hence, phase-sensitive detection techniques can
be used to compare phases between defective and defect-free interconnec-
tions for pass/fail testing. For high defect resolution, the method requires
a 90 degree phase shift between good and defective conductors. This can
be achieved by selecting an operating frequency range that can generate
adequate phase difference and amplitude [4]. The phase detector output
[4] as a function of operating frequency is shown in Figure 13-14 for
similar-size notch defects on copper thicknesses of 3511m, 311m and 511m
used in epoxy-glass, polyimide, and alumina substrates respectively, show-
ing the effect of the operating frequency on three different technologies.
As an example, an operating frequency of 10KHz is desired for polyimide
substrates, which maximizes the detector output signal.

13.3.4.1 Test Implementation


The low frequency tester implemented by Halperin et al. [4] is ideally
suited for printed circuit-board applications and uses a 1KHz drive current.

Voutput
v

time

I I I
1"'1
ITI
I

Figure 13-13. Nonlinear Characteristics of Metal Conductors


13.3 SUBSTRATE TEST METHODS 11-835

0.00035

..
-=
~
ri.i
:I
So
0.00030

:I
0
100

~ 0.00025
~
GI
~

0.00020 0
10 10' UP
Frequency

Figure 13-14. Operating Frequency vs. Detector Output for Latent Opens
Testing

The method is implemented using two probes, one at each end of the
interconnect. DC and superimposed AC current flow through the intercon-
nect via the drive probe and return probe. The sensed voltage passes
through receiver circuits which consist of filters for unwanted frequency
rejection. Since a sinusoidal input signal oscillating at frequency f is used,
the sensed signal consists of the fundamental frequency f and higher
frequency harmonic (2f, 3f, 4f.. ... ) components. A linear amplifier ampli-
fies the second harmonic (2f) of the sensed signal which passes through
a phase detector. To allow for phase discrimination, the fault signal is
phase-sensitive demodulated and converted to a direct current voltage
that is amplified and read on a meter. A calibration procedure is initially
used to generate the reference phase of good interconnections.
A higher operating frequency (1 MHz) is used for testing thinner
conductors as depicted in Figure 13-14 for alumina substrates. The use
of the second harmonic for detection in the low frequency regime cannot
be applied at higher frequencies due to signal distortion of the AC source
produced by commercially available oscillators. This is overcome using
the sum of two sinusoids, with the difference frequency if2 - I,) used for
defect detection.
Using EMT, defects that can be detected are a function of the defect
resistance and the total resistance of the line. The low frequency tester
is capable of 3 mn defect resolution for a 5 n line and is capable of
11-836 PACKAGE ELECTRICAL TESTING

testing long defects (125 J,lm). The high-frequency tester has a 10 mQ


resolution capability for a 20 Q line resistance and is capable of detecting
20 J,lm long defects.

13.3.4.2 Limitation
Latent open testing requires probing on both ends of the interconnec-
tion and, hence, requires two probe heads, one for the drive signal and
the other for the return signal. Contact resistance and I-V non-linearity
between the probe and sample must be low to minimize distorted signal
generation that can be confused with defect signals. Typically gold plated,
palladium alloy or beryllium copper probes are used to minimize contact
resistance and I-V non-linearity. EMT is limited due to the need for
expensive equipment and is not readily applicable for detecting latent
short defects (e.g., line flaring).

13.4 COMPARISON OF TEST METHODS

Table 13-4 provides a qualitative and quantitative comparison among


test techniques practiced in the industry. Although TDNA has been used
for high-frequency characterization of interconnects and is not a viable
test method, it has been included in the table. The number of probe heads
depends on whether one end or both ends of the interconnect require
probing for a two-terminal net. This is related to probe movements with
the complexity arising due to the necessity for the two probe heads to
be in synchronization. The complexity also manifests itself through the
requirement for expensive test equipment. The test time required for
implementing each test method is based on the number of probings required
and assumes that the set-up time is similar and small for all the methods.
In the table N is the number of nets with n terminals and Tn is the time
required per test and assumes simple probes such as flying probes. A test
time of 5 tests/s has been assumed in all the cases, although test times
of 50-200 tests/s are possible with the current testers. The method used
dictates the dwell time on the pads for the probes, assuming similar probes
are used (flying probes versus cluster probes).
For example, high-speed capacitance testers with dwell times <10 ms
are available today, whereas the dwell times for EMT testers are closer
to 50 ms. Test times can be further reduced using the bed-of-nails tester
discussed by Crnic [13] and Woodard [9]. A qualitative comparison is
provided in Table 13-4 for the total test time assuming high wiring densi-
ties, with further details available in Woodard [9]. The opens and shorts
resolution in the table provides a measure of the nature of the defects
that are detectable by the various test techniques. A small value for
opens and large value for shorts represent good resolution. For example,
Table 13-4. Comparison of Test Techniques

Capacitance and
Details CaQacitance Resistance Resistance Electron Beam Latent ()pen TDNA

Frequency 1-IOMHz DC 1-10 MHz I kHz-I MHz 30-70 GHz


and DC
Probe head 2 2 None 2 112
00
Co> Probe movement Simple Complex Simple Complex Complex Complex
.....
Probe time per test (T.) 200 ms 200 ms 200 ms 15 ms charge per net 200 ms Large
(T.) + 10 ms test
Total test time nNT. N(N-I)TJ2 + nNT. 2nNT.+NT. (n-l)NT. Large
medium (n-I)NT. medium small large
large
Opens resolution I MO IOMO I MO 10-100 MO 3-10 mO Small
Shorts resolution I MO 300 MO 300 MO In-IOO MO Not applicable Large
Equipment cost Small Small Small Large Large Large
11-838 PACKAGE ELECTRICAL TESTING

capacitance testing is ideal for opens testing but has poor resolution for
shorts testing.
However, electron beam testing is ideally suited for shorts testing.
Among the methods shown in Table 13-4, only EMT has the capability
of detecting latent opens, with none of the methods providing a capability
for detecting latent shorts.

13.5 CONTACTS AND PROBES

The application of the various test methods to products requires the


developement of contact techniques using high-density, reliable probes,
capable of minimal substrate damage, and incorporation of the probes
into testers for automation. For example, ceramic substrate products manu-
factured by IBM in mid 1990's contain over 7500 usable nets. These
designs require probing of over 70,000 top and 3,500 bottom surface
features. Due to the presence of interconnections between chips in the
substrate and interconnections exiting the substrate through I/O's, the
probing scheme must be capable of providing contact to both the top
surface pads and bottom-surface I/O's. The top-surface contacts are as
small as 90 /lm in diameter on centerline spacings of 250 /lm. Due to the
fine pad dimensions and the number of pads requiring probing, test probes
may have to contact many points, simultaneously. Commercially available
flying probes (-1996) are capable of contacting ultrafine-pitch geometries
down to 50 /lm pads on a 50 /lm pitch.
IBM has been very successful in applying resistance and continuity
test methods through its contact technology for its cofired multilayer
ceramic substrates. IBM [17] invented and developed the buckling beam
test probe based on the principle of a buckling column. The test probe
delivers relatively uniform contact force over a wide range of displace-
ments. This allows for consistent electrical contacts to be made across
the surface of the substrate despite Z-dimensional variations. The buckling
beam probes used for ceramic substrate testing are round and made from
Paliney 7 alloy (includes Ag, Cu, Sn, Au, Pt). The test probe has a built-
in offset in the probe guide as shown in Figure 13-15, such that all beams
buckle in the same direction under the load. The space transformer design
makes replacement of a damaged probe beam possible.
For ceramic-substrate testing, approximately 35 g of force is applied.
The contact force applied is a function of the unsupported beam length
and the cross section of the beam. In addition to the contact force for testing
IBM's ceramic substrates, tHe probe tips have to contact the terminal pads
which are distorted from their design locations due to dimensional changes
that occur during the sintering process. This shifts the top-surface features
from the nominal grid location. Hence, the pads must be mapped for
testing. For current products, IBM has chosen to map the substrate on a
13.5 CONTACTS AND PROBES 11-839

(a)
Direction of Space transformer
motion
Conductor

1 Top guide
Middle guide --.,I-----'T""""~~~
Spherical-end of probe beam

Holes in middle guide is offset


from holes in top and bottom guide

Bottom guide
I(' Probe tip
~Substrate test pad

(b) (c)

1 1
Figure 13-15. Buckling Beam Cross Section. (a) Before contact; (b) at contact; (c)
after contact.

separate tool to avoid burdening expensive testers. For future products,


alignment requirements have mandated optical mapping and registration
capability on each tester. The mapper uses a fixed product stage and a
camera mounted on precision X and Y tables. The camera is set up with
special lighting to create sufficient contrast between the ceramic and via.
Feature recognition is used to define via centers by determining the metal
area center of mass. This method is superior to the edge detection method
as shown in Figure 13-16, which detects a shifted feature center. The
center-of-mass technique is less sensitive to the confusing optical effects
of the Culglass composite via used by IBM than via perimeter or edge
detection. The substrate is held for mapping and tested in a four point
locating device. The alignment device allows determination of the sub-
strate's mechanical center. Future testers will incorporate top surface
referencing and achieve improved alignment capability with some cost
in set-up time. Each mapper includes a benchmark fixture with precision
markings for camera pixel site calibration. The benchmark is also used
11-840 PACKAGE ElECTRICAL TESTING

Edge detection
(Object in a box)

Detected Actual feature center

Area centroid
(Center of mass calculation)

Figure 13-16. Comparison of Feature Recognition by Edge Detection and


Area Centroid

as a fixed point for XY distance checking. Glass standards are used to


calibrate the mapper and tester. A single feature is mapped with the
standard located in the four point fixture and then re-mapped after the
standard is rotated 180 degrees. This establishes the machine center. The
mapper then collects coordinates for a feature near each of the 100 chip
sites on the standard. These coordinates are saved for use as correction
factors and must repeat within 3 /lm. Buckling beam test probes with
over 700 130 /lm diameter contact beams in a 100 rom2 area of a single
chipsite are currently in use by IBM. High-density ceramic substrate test
probes with over 1000 beams are under development at IBM.
For ceramic substrates, electrical testing is performed prior to 110
pin brazing using an evaporated Cu pad 1.4 rom in diameter for bottom
side contact. Contact to the bottom-side 110 pads is made with a "bed of
nails" contactor with an array of spring probes as shown in Figure 13-
17. The contact pins are a tempered Be-Cu alloy plugged into a Hypertac
[18] socket. The 0.9 mm diameter pins have a 0.1 mm contact tip spaced
on a 2.50 mm grid, with nearest-neighbor interstices 1.25 mm apart. The
I/O contactor has a changeable plate to allow for either 2772 or 3526 I/
O's. Testers with "universally switchable" I/O contactors are also under
development which does not require a regular I/O grid and can handle
over 6000 bottom side features. The 110 contactor provides a contact
force of approximately 30 g per pin. The cumulative upward contact force
of the bed of nails 110 contactor is balanced by the vacuum hold-down
mechanism of the tester. Testers with a hold-down clamping device for
very high 110 substrates are also under development.
The test probe used by IBM for testing thin-film substrates is a
13.5 CONTACTS AND PROBES 11-841

~ Buckling beam

Top surface pad

__ ~~I!!!!!,,!! Ir---,"!!!!~~l-J Substrate


Bottom surface pad

Bottom surface pin

Figure 13-17. Cluster Probe Tester

refinement of the technology used for ceramic substrates. Due to thin-


film wiring between contact pads on the top surface (for redistribution,
repair and engineering changes), the probe beam is smaller as compared
to ceramic substrates. For the thin-film substrates, a test probe of 80 Ilm
in diameter is used. Smaller probe beams are being developed for future
products. The thinner probe wire necessitates use of a controlled-direction
multiple-buckling beam probe assembly, which was invented and devel-
oped by IBM. This assembly, called the sandwich probe, is necessary to
achieve sufficient Z dimensional travel in the probe without overstressing
the Be-Cu probe beams. A key feature of this device is the probe guide,
shown in Figure 13-18. The guide facilitates multiple-buckling action that
occurs during probe contact and governs the beam spacing and alignment.
A special photolithographic reactive ion etch (RIE) process is used to
produce the oblong holes in the guides which control the undulating
buckling segments of the sandwich probe. Another unique feature of the
sandwich probe is that the probe wire is continuous from the point of
contact with the substrate to the connecting plug at the switching matrix
of the tester. This feature minimizes resistive loss by reducing contact
interfaces and allows for maximum sensitivity test.
To efficiently administer opens and shorts test (OST) for IBM's
large thin-film substrates, the full cluster tester was developed. The tester
utilizes an array of sandwich probes to contact up to 121 chip sites at a
single time. This configuration results in over 26,000 simultaneous top
and bottom contacts and a test time of about 80 s. The complete test
procedure including load/unload and a re-test after 180 degree substrate
rotation takes about 5 min. Approximately 27 miles of probe wire are
used to test the substrate.
11-842 PACKAGE ELECTRICAL TESTING

(a) (b)

Figure 13-18. Sandwich Probe Guide. (a) Beam deflection before touching a substrate;
(b) beam defection after touching substrate.

The full cluster tester includes two new test features as compared
to existing testers. The first of these is a 50 rnA current pulse used to wet
contacts during opens testing. This pulse is of short duration (up to 0.25 s)
so that circuitry is not harmed. The current pulsing feature has been very
successful in reducing false open errors. The second is break-downlburn-
out detection. This circuitry discerns the transient current spike which
occurs at the onset of a dielectric breakdown or arc. The test stimulus is
immediately shut down and the incident can be reported, facilitating an
investigation. The breakdown detection feature is very useful in high
voltage testing of thin-film substrates. Surface residues or contaminants
resulting from thin film processing can allow an arc to be struck between
metal features. Continued application of the test stimulus can result in
burning or "charring" of the polyimide. By interrupting current flow,
breakdown detection minimizes the damage that can occur.
A special tester was developed by IBM for latent opens (same as
EMT) screening of thin- film interconnections. This tester operates in a
serial fashion with two probes placed at the ends of a circuit path for
testing and moved to the next circuit when the test is complete. X and
Y axis movements are accomplished via commercially available tables.
A Z-axis actuator as shown in Figure 13-19 was developed to provide
straight line motion with high acceleration and a limited (1 mm) stroke.
13.5 CONTACTS AND PROBES 11-843

Probe flex
cable
--l"-7t~~~~;ts=-- Connector block

Stator race Top stator and


pins (2)
Ball cage -----.0£-011
Armaature race ---fI!:.~..Jl

Insulator:Shm"--r-if~~~~~ Stator coil

Armaturre

Probe shield Bottom stator

Ball (4)

Probe SpItB_

Figure 13-19. Miniature z-Axis Actuator

This device has continuous position control, ability to maintain a steady-


state force and controllable impact velocity. The Z actuator operates
similar to an electronic stepping motor. The probe is mounted to the
armature with a set of balls rolling in V-grooves acting as separators and
bearings between the stator and armature. Position sensing is accomplished
via an infrared light emitting diode (LED) which differentially illuminates
two photo-transistors through an aperture mask which is mounted to the
moving armature. With a 2 ms up/down stroke, this device can perform
approximately 4 tests/s with the commercial tables, but is designed to
operate with the Hummingbird mini-positioner device being developed
for use with future IBM testers. The Z actuator mounted on the miniposi-
tioner is capable of a 5 mm X-Y probe movement with a total move time
of 8.0 ms. Allowing approximately 10 ms for measurement, 40-50 EMT
tests/second can be achieved within a 12 mm workplace. A tester utilizing
two probe heads and an XYZ positioning system is shown in Figure 13-
20. The two point probe tester uses "confidence fixtures," containing
examples of conductors whose cross-sections have been reduced. The
confidence fixtures are tested before and after each substrate in order to
11-844 PACKAGE elECTRICAL TESTING

~-+-~--x·y s....

81_

Adjooot. .nt Block (21

Cb""fLlSk..... _ b l y

Figure 13-20. IBM Two-Point Probe Tester

guarantee the validity of the test. The two point probe tester uses a top
surface reference to register the substrate and can accept maps generated
for other IBM substrate testers. Top-to-bottom substrate nets are latent
opens tested by evaporating a shorting plate on the top surface and probing
from the bottom side. Four point testers with top and bottom probing
capability are also in use.
In addition to the requirement for probing fine geometries, the advent
of thin-film substrates has lead to new challenges for eliminating the
interference between the substrate and the tester. An additional objective
is to avoid alteration of either the test output or the substrate due to
particulate contaminants, metallic debris, organic residues and tool-
handling related substrate damage. Due to the sensitivity of the probes
and testers, particulate contamination can affect the test sequence and
results. For example, non-conductive debris of almost any size can prevent
contact between the test probe and the substrate surface feature. The
results of the test could therefore erroneously indicate an open circuit.
Similarly, debris from the probe can also be transferred onto contact
surfaces and may render bonding or joining to these surfaces problematic.
Furthermore, large, hard particulates can result in the mechanical alteration
or destruction of substrate features or even the test probes themselves and
could create a shorting path if they adhere to the substrate surface. Organic
residues such as stains can result in dielectric leakage paths and could
also inhibit further processing. Hence particulate control is necessary and
13.5 CONTACTS AND PROBES 11-845

is facilitated greatly by the use of a clean room processing area with


appropriate garments and procedures. Substrate cleaning is also recom-
mended, both before and after test, to allow for particulates and residues
to be removed. To minimize the generation of metallic particles during
test, probing of metal surfaces with varying hardness should be avoided.
As scratching of pads during probing is sometimes unavoidable, debris
and particulate interference should be considered in diagnosing failures
that may occur during the test process. Post-test process control inspection
of the substrate surface may therefore be necessary and prove to be helpful
in the timely detection and correction of substrate damage problems.
One of the most challenging problems related to thin film surface
damage is the control of motion in the XIY, and Z axes, due to fine
geometries. Figure 13-21 illustrates the problem, wherein the test probe

Z drive 1
(a) Z sense 0
XY drive 0
XY sense 1

(b) c:::/e Z drive 0


Zsense 1
XY drive 0
XY sense 1

(c) o XY Sensor Z drive 0


Z sense 1
XY drive 1
XY sense 0

(d) Zdrive 0
Z sense I
XY drive 0
XY sense 1

(e)
Correct probe placement

(0 Probe drag during lift-off


XY drive withou[ Z up

(g) Probe drag during touchdown


Z drive With XY not ready

Figure 13-21. Probe Control in X/Y and Z Axis. The above sequences are schematic
depictions of test probe or probe cluster movements from one test point to another.
Segments (a)-(d) show alignment of the XIY and Z markers (at right angles on the probe
body) with their expected sensor position as indicated by a lighted lamp. Boolean states
of sense and drive for the axes involved are shown at right. Segments (e)-(g) are graphs
of XIY and Z positions realized during movement from pad A to pad B.
11-846 PACKAGE ELECTRICAL TESTING

must be moved from Pad A to Pad B without disturbing the intermediate


features or insulation that may lie between the two pads. The intermediate
features could represent wiring for redistribution or repair. They are on
the same metallization level as the terminal pads. The example shows a
single probe and short probe movement but the principle is applicable to
clustered probes and long distance probe travel as well. Figures 13-2lf
and 13-21g show the resulting pad damage due to uncoordinated probe
movement in the X,Y and Z directions. The motions described in Figure
13-21 are relative, where either the probe or the substrate may be raised
or lowered because both could result in similar damages. To avoid damage,
separate sense controls for XfY, and Z axes, with suitable hardware and
software for sense and drive feedback loop, may be necessary, particularly
for high-speed test sequences. The feedback loop could, however, have
potential downfalls due to false sense or drive signals due to circuit
noise, interference between drive and sense signals, consumption of Z
dimensional process window by mechanical tolerances (probe to probe,
probe plane to product plane skew, table inaccuracy or wear, unexpected
product deviation, sensor mounting), oscillation about sense points, and
inability to instantly interrupt/reverse drive sequences or timing inconsis-
tencies.
As testing represents a significant portion of the substrate cost, a
method of decreasing manufacturing cost is through the use of in-process
testing. As shown in Figure 13-22, based on the yield, few substrates
require a final test because parts can be rejected early in the manufacturing
phase. In-process testing can be achieved either through visual or opti-
cal inspection.
13.6 OPENS AND SHORTS TEST IMPLEMENTATION
Test, contact, and probing methods discussed in the earlier sections
have been applied here to test substrates. The example used is the ceramic
substrate, although the test flow is applicable to other substrates as well.
For complex substrates (such as ceramic), the top-surface features
are contacted in two separate tests as shown in Figures 13-23 and 13-24.
The two-step method minimizes the complexity of the test probe design
as all contact features need not be included in each test head. The first
operation is a single-site top-surface probe head making contact with the
pads within a chip site which tests circuitry within the chip site and to
I/O pins. Two test heads can be operated simultaneously to reduce time,
each testing half of the substrate. This test represents an opens/shorts test,
also called the single-site common opens/shorts test (COST) which checks
voltage planes, redistribution nets, and I/O pins. Neither planes, redistribu-
tion, nor I/O nets are repairable. Thus, the relatively inexpensive COST
operation can be used to reject substrates with unrepairable defects prior
to further processing (e.g., application of thin film to the substrate).
13.6 OPENS AND SHORTS TEST IMPLEMENTATION 11-847

I-- Shipped~j.- Rejecled --+l


FinalleSI
TFVPTI

TFVPTI

Ceramic ub Lrale Ie I

Ceramic voltage plane lell

D Electrically good _ Rejecled

o Defeclive D Repaired

Figure 13-22. Reducing Substrate Manufacturing Cost with In-Process Test.


a) Yield advantage from realizing rework/repair opportunities not available later in process;
b) labor, material, and capital costs avoided by rejecting defective parts early in process.

The second operation to a complete substrate test is personalized


opens/shorts test (POST) which is usually performed with two multi-site
test probe heads, as in Figure 13-24. The POST operation consists of one
probe head (A head) that is maintained stationary on a chip-site cluster
and a second (B) head that is stepped around to all other chip site clusters.
The A head then moves to the second chip-site, the B head steps to all
but the first two sites, and so on until all chip sites are tested to all
others for continuity and shorts. In actual practice, tester movements are
optimized to minimize wear of mechanical components. During any test
sequence, the probe heads remain stationary on the substrate while the
test engine is switched electronically from point to point. The field-effect
transistor (FET)-based solid-state switching hardware can develop test
speeds of up to 1O,000/s with the capability of delivering up to 250 V
and 20 rnA.
The tester is numerically controlled (NC) using design data. The
design data give expected values of opens or shorts to all substrate top
and bottom contact features and thus allows complete comparison of
11-848 PACKAGE ELECTRICAL TESTING

To open and shon test scanner

Head
(In first position)

Figure 13·23. Single.site Contact Scheme

product to design expectation. As an example, a 1OO-chip ceramic substrate


fabricated by IBM requires approximately 5 MB of data for testing. NC
data can be specified by test engineers and programmers to coincide with
the single-chip site contact used for COST or multi-site testing used for
POST. NC data can be applied for test through rules-driven tool control
software. The software can be adapted for new products of different sizes,
shapes, and electrical specifications.

To open and hon test scanner

Figure 13·24. Site·to·Site Personality Opens/Shorts Test


13.6 OPENS AND SHORTS TEST IMPLEMENTATION 11-849

Figure 13-25 is a simple example of how opens/shorts testing is


performed on substrates. In the figure, both two and three-point nets are
shown with all nodes or terminals ordered from top to bottom for shorts
testing. In the figure, net 7-8 is shorted to net 10-11 and net 12-13 has
an open. Nodes are tested for shorts starting from the bottom and only
checked to lower-order points. For example, in Figure 13-25, net 1-2 is
tested first for continuity and no shorts testing is done with any other
nets. In the next step, net 3-6-9 is tested for opens, and tested for shorts
using net 1-2. The sequence is continued until all the nets are tested. As
each net node is tested, lower-order nets are electronically coupled, allow-
ing "bulk" examination of the net under test to all lower- order points.
If a short is detected, the coupled nets are systematically isolated so that
the individual nets can be unambiguously identified as shorted together.
Opens in multipoint nets are detected from the ends of the net and isolated
to a particular subnet. As an example, in Figure 13-25, the open occurs
between points 12 and 13. The open is initially detected between nodes
12 and 14 of the multidrop net, through a continuity test. To obtain the
location of the open, new test data is generated on a real time basis and
the open is isolated by testing nodes 12 and 13.
To increase test accuracy, error analysis programs may be used for
duplicate error checking. For example, the part can be rotated 180 degrees
and reprobed for continuity checking. If opens detected on the initial test
are not found after rotation, they can be assumed to be false and dropped.
This technique has been used very successfully in testing for reducing
false open errors, but retesting must be managed carefully to prevent any
fault escapes. The rotation test can also be used for adding redundant
shorts testing of single-node nets. Reduction in false errors can be enhanced
through the use of a verification step following the initial test. This
operation is performed on equipment consisting of two top-side and two
bottom-side probes or two top-side probes and a bottom-side shorting
plate. The probes are fitted with Kelvin tips as shown in Figure 13-26
that allows pad contacts to be verified. Errors called by the prime testers
are subject to retest by the verification tool and are declared invalid if
not found. Information on verification actions and results are stored to
monitor tester efficiency and verification problems. At IBM, aggressive
actions to reduce defect counts such as duplicate probing have resulted
in very low levels of test escapes «0.025% for high-volume programs).
For testing, mapping of the substrate is necessary to align the probes,
based on coordinates of the terminal pads. For IBM's ceramic substrates
(see Chapter 9, "Ceramic Packaging"), this is done through a capacitor
pad on a chip site near the center of the substrate. The mapping head
then moves one-chip site diagonally toward the upper left comer of the
substrate and locates the next capacitor and proceeds until the upper left
chip site is reached. The centermost via of that site is then located by the
11-850 PACKAGE ELECTRICAL TESTING

Example:
114 *2 :;r,int nets
CD 114
13 Net 1-2 tested for
13 1- 4-5 7-810-11 continuit~ no lower

i
*3 p'!:!int nets 3-6-9, nets for sorts testing

i
12-13-14 1:z!11 Tl2
1:z!11 Tl2 *7-8 shorted
to 10-11 -" 8 9
-" 8 9 *O~n in segment ............. 5 6
............. 5 6 12-13
1______ 2 3
l:;!2 3

@ 114 Q) 114 Bulk shorts test:


13 13
a) 3,,;,6-,9 tested for a) 4-5 continuity
test;

y~~2
COntlnUl~;
b~ 3-6- tested for 1:Z:11 T12 b) 4-5 shorts tested
sorts to 1-2 to 1-2 3-6-9
............. 5
1 2
6
3
~9
1
5
Z
6
3
cO~fcl~ by ele,<tronic
SWI chmg matrix

@ l!j ~ 114
13 a) 10-11 continuity

{Ei
a) 7-8 continuity test;
test' b) 10-11 shorts tested

~12
11Tl2 b) "1-8 shorts tested to 1-2,3-6-9,4-5,
7 8 to 1-2,3-6-9,
9 4-5 8 9 7-8 short detected
456 456
1 2 3 1 2 3

1 14
13 Short isolation:
a) 10-11 short
114
13
a) 10-11 short
tested to 4-5;

~
1O
~
11Tl2 tested to 1-2, Ol BTl2 b) 10-11 short
7 8 9 3-6-9; A 7. 9 tested to 7-8
short isolated
5 6 b) 1O-11 short 4 5! 6
1 2 3 tested to 4-5, 1______ 2 .3
7-8, short
detected

@~c
O~n detection 14 0Pttn subnet
and isolation: 13 testmg:
A 1
a) 12-13-14 tested a) 12 short tested
1~
i 12 B for continuity, open to 1-2 3-6-9 4-5
detected' 7-8 Ib-11;' ,
7;.Le8 9 b) If- (1 tested for b) 13-14 short
............. 5 6 continuity, open tested to 1-2,3-6-9,
1______ 2 3 isolated; 4-5,7-8,1O-11
c) 13-14 tested for
continuity

Figure 13-25. Accurate Open/Shorts Testing with Minimal Cycle Time


13.6 OPENS AND SHORTS TEST IMPLEMENTATION 11-851

Probe footprint on pad

ViewAA Dual contacts permit probe-pad contact


verification and 4-point resistance
measurement

Figure 13-26. Verification Probe

mapper and its coordinates are recorded. The mapper then moves one
nominal chip site distance down and continues to map the location of the
centermost via of each chip site in a serpentine pattern. The result of the
mapping operation is an array of coordinates locating the center of each
chip site on the substrate. These coordinates are used by the tester to
place the test probes on the substrate terminal pads. The mapping operation
is shown in Figure 13-27 where the coordinates of six chip sites in the
center row are used to calculate a best-fit overall surface pattern rotation
angle. The tester then rotates the substrate by the estimated angle to
achieve parallelism between the cluster of probes and the surface metal-
lurgy pattern. An alternate way is to use the glass alignment standard to
calibrate the testers, as in Figure 13-28. The standard is located in a four-

Machine center determination

•
(Xo.Yo)

* True machine center


calculated from 0, 180 * Coordinates from
measurements each chip-site collected
and stored
* Chip-site coordinates
used to calculate overall
-+ substrate "theta"

Figure 13-27. Mapper Calibration and Substrate Mapping


11-852 PACKAGE ELECTRICAL TESTING

0000000
0000000
0
000 [Test
; ] 000
0
00 probe 00
0000000
0000000
(a) (b)

f®®l
(c)
~
(d)
Figure 13-28. Test Probe Alignment Using the Glass Standard. a) Croon-glass
alignment standard; b) alignment of standard showing optical cavity; c) probe/target
alignment before adjustment; d) proper alignment.

point fixture and a fiber-optic bundle is inserted into a tunnel drilled in


the body of the standard (Fig. 13-28b). The optics is aimed upward such
that the pattern on the standard surface is in view (Fig. 13-28c). The test
probe is brought into contact with the glass standard and alignment of
probe to pad on the surface is observed. The tester X and Y tables are
moved until alignment between the probe and the standard is optimized
(Fig. 13-28d). This calibration is performed on a chip site at the center
of the standard which can then be repeated at the top and bottom of the
standard to check the alignment of the probes.
For opens/shorts testing, test parameters are required to enable pass/
fail sorting and for parametric measurements. For IBM's glass-ceramic
substrates, to enable shorts testing, a potential of 100 V is applied to the
point under test. Up to 1 rnA of current is injected and lower-order points
tested for leakage currents (as in Fig. 13-25), which allow the detection
of resistive shorting paths of up to 10 MQ. Similarly, for opens testing,
a 20-rnA constant-current source is used with a voltage of 100 V to detect
open circuitry. To enable pass/fail sorting, an open is defined as a product
resistance of 35 Q or greater between points under test. Both opens and
shorts testing with the above voltage and current levels stress the circuitry
significantly to increase confidence in product reliability. For parametric
measurements, a small sample of substrate nets are measured weekly
for conformance to limits on resistance and capacitance. Based on the
acceptable capacitance for IBM's glass-ceramic nets, capacitance mea-
surements must guarantee an average value between 1.15 and 1.35 pF/
cm. Resistance measurements are taken using the four-point Kelvin probes
13.6 OPENS AND SHORTS TEST IMPLEMENTATION 11-853

and must demonstrate an average value between 0.15 and 0.30 Ulcm.
Any deviation from these ranges constitutes an unacceptable substrate.
The logistics and control of the substrate-test process is a key factor
in the quality and efficiency of the operation. The sequence and completion
of operations and flow of data must therefore be carefully managed.
The progression of test operations is controlled by a computer-based
technology routing or tech route which contains a preset sequence of
operations, as in Figure 13-29, controlled by electronic log-on and log-
off functions. A job must log on and log off from the first operation
before it can be logged on to the second operation, and so on. Several
rework options exist for substrate repairs and corrective actions. Product
flow through the network operations is computer controlled. Following
network completion, the substrate is returned to the primary tech route
for continued processing.
Figure 13-30 depicts the flow of data required to successfully perform
the desired testing and store test results. Substrate testing requires map
data for probe placement and design data for test sequencing. The map
data are generated and passed to a storage node. When the substrate part
number is released to the manufacturing line for build, numerical control
(NC) data are created for build and test. The test data are electronically
delivered to a storage node at this time. When the substrate is logged
onto a tester, the map data and NC data are automatically retrieved for

Figure 13-29. Ceramic Substrate Test Operations


11-854 PACKAGE ELECTRICAL TESTING

I• Storage and Control Systems

•• •• • ••
. .
Job •
• Job • • •• •
Data. Data • • Defects. Job •

••
Data:
•• Map Map
• • Repairs

•·• ..
·
• Defects.

Map ...... Test


•
••
•
•• ·••
•• •••
...
Alto • Defects
Job
Data •
•
• •• •• Ship
Defects.
•• •• ••
··
Map . .

•• ..
4 Verify
••
. •• ...

. . . . Dataflow •• ••
-+ Substrate flow
....
Repair

Figure 13-30. Data Flow for Ceramic-Substrate Electrical Test

the particular job and part number. The map data for a particular job
number remain in the storage node while the substrate is active in the
manufacturing line.
The map data are archived if it is rejected or shipped. Design NC
data for frequently used part numbers are stored in the tester. Defect data
generated by the tester are sent to a storage node called the serial history
file (SHF). Test data for any test pass are stored in SHF. Defect data are
passed to the verification tool if errors were detected. Verification results
are then recorded in SHF. Completion of all data transmissions are auto-
matically verified. Most substrate signal nets terminate at engineering
change (EC) pads and can be repaired if found defective. Defective nets
identified during testing can be deleted through a laser cut operation.
These nets are eventually rewired on the substrate surface. Net deletion
data are stored and used to modify design data for testing such that the
deletion is an expected result of the test. All repaired nets are rewired
prior to shipment.

13.7 FUTURE ISSUES AND CHALLENGES IN


SUBSTRATE TESTING
Small, tightly spaced terminal features and high circuit counts (as
in thin-film substrates) place a premium on the test contact technology
13.8 FUNCTIONAL TESTS 11-855

and on tester speed. Accurate high speed testers are therefore necessary,
which require advanced contactor, positioner, and switching technology.
To automate advanced testers, carefully conceived tool control software
is required for operation. Smaller and more fragile conductors and the
increasingly more complex chemical processing required to define them
will continue to place significant emphasis on reliability assurance testing.
Fine-feature dimensions will create very stringent debris and contamina-
tion control requirements for the test processes and equipment. Thus,
product damage and test accuracy will remain concerns as substrate inte-
gration and its value increases.
Waste or rework of highly functional integrated-circuit chips cannot
be tolerated due to the undetected substrate defects. Similarly, capable
and valuable substrates cannot be reworked or rejected due to false test
errors. High-density substrates with high-reliability specifications require
a test scheme capable of detecting latent opens and latent shorts defects.
Because manufacturing costs are to be kept low, low-cost test methods
which are capable of high defect resolution are necessary. Testing becomes
extremely complicated for substrates containing embedded passives such
as resistors, inductors, and capacitors. Advancement in tester technology
may be necessary because the testing of passives may impose new chal-
lenges, such as inaccessibility from the top surface to the pads connected
to passives. Control of manufacturing costs for highly integrated substrates
can mean additional requirements for in-process testing. Consistent with
the need for a new low-cost substrate test technique capable oflatent opens
and shorts detection, universities such as Georgia Tech, in partnership with
industry, are pursuing new methods for substrate testing.

13.8 FUNCTIONAL TESTS


The functional tests follow the substrate test, after the mounting of
the semiconductor dies (or chips) onto the MCM substrate, as depicted
in the test flow in Figure 13-1. As mentioned earlier, the primary objectives
of functional test are as follows:

1. To ensure that no defects were introduced into the known-good-die


during the MCM assembly operation in which the bare dies are
attached to the tested substrate
2. To ensure that the bare dies are properly bonded to the substrate
interconnects and that all the terminals of the attached bare dies are
connected to each other in the manner desired
3. To ensure that the MCM performs at speed according to the product
specifications (this is accomplished through design for testability
(DFf) methodologies, wherein I/O's of the die contained in an MCM
are accessed from some MCM I/O's, for test purposes).
11-856 PACKAGE ELECTRICAL TESTING

Boundary can Cell Boundary Scan PaLh

TOI
--r-~-----.,-------------~ ~
TM
--+---.
Application Logic

TOO

Figure 13-31. Chip Architecture for IEEE 1149.1

This eliminates the need for a bed-of-nails testing procedure to make


contact with all the MCM VO's, as a test access bus may be used to
access all the die VO's, through a test access port. The resulting standard
(designed by a consortium of industry partners), called the JTAG 1149.1
standard (JTAG stands for Joint Test Action Group), although initially
developed to solve the problem of testing printed-circuit boards, has since
evolved and forms the basis for functional test ofMCMs. The chip architec-
ture conforming to the above standard is shown in Fig. 13-31. For functional
testing, it is important to differentiate MCMs with and without pretested
dies (or chips), as the former simplifies the testing process.
Because MCMs contain multiple chips, each one requires a specific
set of functional test vectors for excitation, to determine whether defects
were introduced into the chip during the MCM assembly process. Because
of the large number of test vectors required for testing all the chips, the
common practice is to include a built-in self-test (BIST) capability in the
chips, which allows for the generation of test stimuli and comparison of
the respective responses with reference "signatures". Built in self test is
the capability of a circuit (chip, board, MCM or system) to test itself.
The BIST procedure may be activated through the JTAG test port using
special test commands. The necessary BIST functions may be performed
at-speed thereby validating objective (3) of the functional test. If desired,
13.8 FUNCTIONAL TESTS 11-857

additional functional test vectors may be applied to the MCM through


the JTAG interface. Once the individual dies are tested, the remaining
task is that of testing the MCM interconnections through the test access
bus. With both tests being successful, the test engineer can be assured
with high confidence that primary objectives (1) and (2) of the functional
test, have been met. In subsequent sections, the boundary scan test standard
that provides test access to all the die on an MCM from an external test
port is explained, followed by the implementation of BIST through this
standard. Some functional test strategies have also been discussed based
on the above concepts.

13.8.1 Boundary Scan Test


Boundary scan is a general testability strategy that allows controlla-
bility and observability of the pins of a semiconductor IC embedded in
a printed-circuit board through a serial test port. The method is equally
amenable for accessing I/O's of chips mounted on an MCM. The IEEE/
ANSI standard 1149.1-1990 [19] is an industry-wide accepted set of
guidelines to implement boundary-scan. The primary components of this
standard include the design and use of a test bus which resides on a board,
the protocol associated with this bus, elements of a bus master which
controls the bus, input/output (I/O) ports that tie a chip to the bus, and
some control logic that resides on each chip on the board, to interface
the test bus ports to the DFr hardware residing on the application portion
of every chip. In addition, the IEEE 1149.1 standard also requires the use
of a boundary-scan register to exist on every chip. Boundary scan facilitates
isolation and testing of the chips either via the test bus or by built-in-
self-test hardware. Testing of interconnections for opens and shorts is
also possible through the boundary-scan test hardware.
The boundary-scan bus consists of four ports, a test clock (TCK),
a test mode select (TMS), the test data in (TDI) line, and the test data
out (IDO) line, as shown in Fig. 13-32. Test instructions and test data

IDI --Ir----.!' - - - - . . - - - '


TMS, TCK --Ir-----+--------I

IDO .-+---,
.-------S._-,

Figure 13-32. Boundary Scan Including an MCM Design


11-858 PACKAGE ELECTRICAL TESTING

are sent to each chip on the board over the IDI line. Test results and
status information are sent from a chip over the IDO line to whatever
device is driving the test bus. This information is transmitted serially.
The sequence of operations is controlled by a bus master, which can be
either automatic test equipment (ATE) or a component that interfaces to
a higher level test bus that is part of a hierarchical test and maintenance
system. Control of the test-bus circuitry is primarily carried out by the
test access port (TAP), which responds to the state transitions on the
TMS line.
The test and associated logic operates as follows.

1. An instruction is sent over the TDI line into the instruction register.
2. The selected test circuitry is configured to respond to the instruction.
In some cases this may involve sending more data over the TDI line
into a register selected by the instruction.
3. The test instruction is executed. Test results are shifted out of selected
registers and transmitted over the TDO line to the bus master. New
data are shifted into registers using the TDI line while results are
shifted out and transmitted over the TDO line.

As an example of boundary-scan test implementation, Figure l3-32


shows a board containing four chips which support IEEE 1149.1. The
boundary-scan cells are connected in a single scan path, where the TDO
of one chip is tied to the TDI of the next chip in the boundary-scan chain,
except for the initial TDI and TDO ports which are tied to distinct terminals
of the board. Some of the interconnections between chip pads are also
shown in the figure. Using this configuration, various tests may be carried
out, such as (1) interconnect test for opens/shorts, (2) snapshot observation
of normal system data, and (3) testing of each chip.
The test procedure for multichip modules is similar to that for printed
wiring boards. The bare chips on an MCM are analogous to ICs on a
printed wiring board from a test perspective. Hence, the boundary-scan
test hardware and test procedures are, in general, applicable to MCMs.
There are several issues related with the design and use of the IEEE
1149.1 test bus standard:

(a) The physical structure of the test bus and the rules that govern how
the bus is interfaced to the individual chips
(b) The data transfer and control protocol associated with the use of
the bus
(c) The on-chip test bus circuitry associated with a chip that allows the
chip to communicate with the bus.
13.8 FUNCTIONAL TESTS 11-859

The latter includes a set of boundary-scan registers and the test


access port (TAP) controller which is a finite state machine that decodes
the state of the bus. Figure 13-31 shows the general form of a chip that
supports IEEE 1149.1, which may form part of an MCM or printed-circuit
board. The application logic represents the normal chip design prior to
the inclusion of logic required to support IEEE 1149.1. This circuitry
may include DFT or BIST hardware. If so, the scan paths are connected
via the test bus circuitry to the chips's scan-in and scan-out ports. This
is illustrated by the connection from the test data input line TDI to Sin
and Sout of the test data output line, TDO. The normal 110 terminals of
the application logic are connected through boundary scan cells to the
chips 110 pads. The test bus circuitry is referred to as the bus slave and
consists of the boundary scan registers, a I-bit bypass register, an instruc-
tion register, several miscellaneous registers, and the TAP controller.

13.8.2 Built-In Self-Test


Built-in self-test (BIST) is the capability of a circuit (chip, board,
MCM, or system) to test itself. This section deals primarily with structural
off-line BIST. Thus, linear feedback shift registers (LFSRs) are used
extensively for BIST purposes. BIST architectures consist of several
key elements:
(a) Test-pattern generators (TPG)
(b) Output-response analyzers (ORA)
(c) Circuit under test (CUT)
(d) Distribution system (DIST) for transmitting data from TPGs to CUTs
and from CUTs to ORAs
(e) BIST controller for controlling the BIST circuitry and CUT during
self-test

Often all or parts of the controller are off-chip. The distribution


system consists primarily of direct interconnections (wires), buses, multi-
plexers, and scan paths. The general form of a centralized BIST architec-
ture is shown in Figure 13-33. Here several CUTs share TPG and ORA
circuitry. This leads to reduced overhead but increased test time. During
testing, the BIST controller may carry out one or more of the follow-
ing functions:

1. Single-step the CUTs through some test sequence


2. Inhibit system clocks and control test clocks
3. Communicate with other test controllers, possibly using test buses
4. Control the operation of self-test, including seeding of registers,
keeping track of the number of shift commands required in a scan
11-860 PACKAGE ELECTRICAL TESTING

BIST

I I
I I I
--------~---------~-----~

BIST controller

Figure 13·33. Generic Form of BIST Architecture

operation and keeping track of the number of test patterns that have
been processed.

Various levels of chip-level BIST have been used during the last
decade as explained in Refs. 20-23. The BIST schemes that are needed
for MCM test are those that provide very high fault coverage. Most of
these schemes run at normal operating speed and typically achieve the
required fault coverage.

13.8.3 Functional Testing of MCMs


The testing of MCMs is a difficult task due to both the complexity
of MCMs and the fact that no established infrastructure for MCM testing
exists [24]. MCM testing combines the complexities of chip and printed-
circuit board testing [25-28]. An MCM test which meets product quality
requirements has to ensure that all die are properly connected, that they are
functionally correct, and that the MCM as a device meets its performance
specifications. This requires an interconnect test, a full functional test,
and a performance test. Due to very limited accessibility of MCM circuit
components (access to other than MCM primary input/output pins may
be nonexistent), it is impossible to test the interconnects and the chips
with conventional test techniques. Therefore, it is essential to incorporate
testability features during functional design of the chips that are to be
used in MCMs.
Interconnect test detects mechanical assembly defects and does not
identify functional failures. On the other hand, a full-functional at-speed
test of the die detects defective dies. The at-speed test is usually not as
important with regard to printed-circuit board testing as opposed to testing
MCMs, because the packaged chips used in printed circuit boards are
13.8 FUNCTIONAL TESTS 11-861

fully tested. Due to die-handling and chip-bonding processes during MCM


assembly, there is a chance of damaging the dies. Hence, thorough testing
is required after mounting the dies on the substrate. In addition, there
are thermal stresses which are present while the module is powered.
This can also result in damage to the individual dies. Hence, assembled
MCMs must be subject to the same kind of rigorous testing as the
individual chips.
If failures are detected during testing, an MCM may enter a repair
cycle. The repair cycle may sometimes include cap removal if the repair
is done after final test and rework. The cost-effectiveness of the repair
depends on the MCM technology adopted. In general, the test procedure
must include a diagnostic procedure in order to identify the defective
element in the MCM.
To enable very high fault coverage, a structured testability approach
for testing MCMs is desired and is explained in the next few sections.

13.8.4 A Structured Testability Approach

The testability approach proposed by Zorian [29] is based on imple-


menting chip-level BIST and module-level boundary-scan in the MCM
design. Both BIST and boundary-scan, as explained earlier, are design-
for-testability techniques; that is, they are meant to improve the controlla-
bility and observability of circuit performance to make it easily testable.
The use of both techniques necessitates early considerations, because they
impact the chip design process.
The most popular testing approach executes BIST through the bound-
ary-scan TAP [30]. The incorporation of BIST and boundary-scan in a
chip and techniques to obtain very high fault coverages at the chip and
the module level test are covered in the following discussion. This section
also addresses the benefits of having BIST and boundary-scan in bare
dies and discuss their utilization for bare die and MCM assembly testing.
Functional at-speed test of a chip becomes possible due to the incorporation
of BIST in its design. In fact, the two main problems of wafer testing,
namely obtaining very high fault coverage functional test and running
it at system speed for performance test, are greatly simplified due to
the use of BIST. The execution of this operation (i.e. running BIST)
is autonomous. It only needs the application of boundary-scan standard's
RUNBIST instruction [19, 24] through the boundary-scan TAP to start
the BIST execution, and upon completion, the BIST response needs
to be scanned out for evaluation. This is also done through the
boundary-scan TAP [30].
The use of BIST and boundary-scan impacts reliability testing of
chips (i.e., dies). During burn-in, BIST can be kept active on the die under
test, which needs only the boundary-scan TAP interface. This allows
11-862 PACKAGE ELECTRICAL TESTING

continuous monitoring of the response status for additional in-process


information. The above test processes result in known-good-die (KGO)
that leverages the assembly of competitively strong MCMs.

13.8.5 Assembled-MCM Test and Diagnosis Procedure


To perform a typical test of MCMs with BIST and boundary-scan
the following tests need to be conducted:

(a) Integrity and identity check


(b) Interconnect (substrate) test
(c) Functional at-speed chip test
(d) MCM performance test
(e) MCM parametric test

A brief explanation of each of the operations follows.

13.8.5.1 Integrity and Identity Check


This is to verify the integrity of boundary-scan circuitry prior to its
use in subsequent tests. The boundary-scan standard provides a certain
methodology for performing this test, yielding relevant diagnostic infor-
mation regarding the location of the failure if any [11]. Figure 13-32
shows an MCM design composed of four ICs. Each IC contains its
corresponding boundary-scan facilities. The scan chain which connects
all four ICs is represented by a line starting from TOI (test data input)
and ending with TOO (test data output). The remaining boundary scan
signals are distributed to all ICs in parallel. Following the integrity test,
a test is performed to check the identity of each die. The boundary-scan
standard allows permanent storage of chip-level ID codes. Each MCM
die has its own ID code, which is read through the boundary-scan TAP
and is compared with a reference value. This test detects the existence
of an incorrect chip or a chip that is not oriented properly on the substrate.
This is especially important, as many of the chips on a MCM may have
the same footprint, causing problems with visual inspection methods
for reliability.

13.8.5.2 Substrate Test


This test checks whether dies are properly bonded to the substrate
interconnections. The existence of a boundary-scan chain in the module,
consisting of the boundary-scan registers of each die and the connections
between these registers through the substrate, creates a virtual electronic
bed of nails built into the module. This bed of nails built into the module
is independent of the module's density. Boundary-scan latches provide
13.8 FUNCTIONAL TESTS 11-863

the stimulus and can monitor logic values across all interconnections.
This permits testing for opens or shorts between inputs/outputs of two
dies and between inputs/outputs of a die and the inputs/outputs of the
MCM. To minimize test application time and optimize detection of inter-
connect defects, special test algorithms are used. The test processes are
automated and also provide diagnostic information identifying the faulty
net and the type of fault. This information is very useful for the repair
process. Even though the substrate is usually unrepairable, the die to
substrate attachment can be reworked or the die can be replaced and
rebonded to the substrate.

13.8.5.3 Functional At-Speed Test


It is necessary to run the BIST operations on each die after the latter
are bonded to the substrate in order to detect defects that may be introduced
into the chip during the die attach process. In order to perform this test,
access is needed to the TAP of each die and that is possible through the
boundary scan chain of the MCM. Figure 13-34 shows an MCM design
with BIST in its individual chips. As an example, the test pattern generator
(TPG) and output data evaluator (ODE) of ICI (integrated circuit chip
#1) are illustrated.
The use of BIST also provides diagnostic information. The analysis
or automatic comparison of the BIST responses can identify which chip
has failed. If some of the chips in an MCM are not designed with BIST
capabilities, such as RAM chips, after MCM assembly other measures
are taken in order to test them.

13.8.5.4 Performance Test


The MCM performance test is an at-speed test on the assembled
MCM. It verifies that the finished module meets all the performance

Primary I nput
~

ITPGII CUT I IODE I ITPOII CUT I IODE I


Primary Output
IC 1 IC2

ITPGII CUT I lODE I


IC4
ITPol I CUT I El
IC3
Figure 13-34. An Example of Bare Die BIST in an MCM Design
11-864 PACKAGE ELECTRICAL TESTING

requirements. Further, it tests propagation delay times, including chip


delays and substrate routing delays. This test requires automatic test
equipment with high-pin-count and high-speed capabilities. The boundary-
scan SAMPLE mode is used to obtain diagnostic information to locate
any performance-related problem. In this mode, the state of boundary
scan cells are captured at fixed intervals of time (snapshots) and scanned
out through the boundary-scan chain.

13.8.5.5 Parametric Test


The boundary-scan chain is used for the input/output parametric test,
by allowing proper input/output initialization to predefined values [31].
The above five test operations can be applied during the MCM
assembly process, and/or after the assembly completion. It is preferable
to have in process testing in order to identify the faulty element(s), and
perform repair before the final MCM is capped.
The repair cycle includes faulty element isolation and rework. The
diagnostic information, obtained during each one of the five testing opera-
tions, is utilized to isolate the faulty element. For instance, the BIST
signature identifies the faulty chip, and the Boundary-scan-based intercon-
nect test identifies the failed net. The rework involves the replacement of
the faulty element, and this is followed by a retest.

13.9 FUTURE ISSUES AND CHALLENGES IN


FUNCTIONAL TESTING
The rapid increase in the complexity of VLSI systems has made the
task of test, diagnosis and repair more costly and time consuming. This
has necessitated the incorporation of unique design for testability features
into system designs. Such features include, for example, hardware for
built-in self-test and boundary scan. With the advent of multi-chip modules
and other methods of integrating high-density electronic modules, such
as 3-D chip and module stacking, the testing problem has been stretched
to its limits. Future systems will include ultra-fast processors with giga-
bytes of static and dynamic memory and multiple co-processors for graph-
ics, video, audio and other utilities. Clearly, methodologies for breaking
the testing problem into smaller tractable parts will evolve. Solutions to
these smaller problems will then be amalgamated to provide a testability
solution for the complete electronic system. We foresee a greater emphasis
on intelligent built-in self-test mechanisms in the future. These built-in
self-test mechanisms will be tailored to specific classes of hardware and
will need to provide high levels of fault coverage (almost 100%) in order
for overall system "yields" to be high. Also, these mechanisms will need
to accommodate non-conventional fault models such as those for speed-
testing of digital circuits. As circuit dimensions become smaller, clock
13.10 RECENT AND FUTURE DEVELOPMENTS 11-865

speeds are increased and supply voltages are lowered, issues relating to
signal integrity will become an integral part of functional test. Moreover,
with increased design of mixed-signal circuitry, standards for boundary
scan based test and built-in self-test of mixed-signal systems will evolve.
Eventually, schemes for hierarchical boundary-scan based testing of
mixed-signal systems will need to be developed and issues such as parti-
tioning of scan chains for ease of testing and reliability of design for
testability hardware will draw increased attention.

13.10 RECENT AND FUTURE DEVELOPMENTS


13.10.1 Substrate Testing
To overcome the difficulties and deficiencies with the present tech-
niques, a number of test methodologies are under development at compa-
nies (IBM, lSI) and universities (e.g., Georgia Institute of Technology).
At IBM and lSI, two new test methods, namely, time domain opens/
shorts (TDOS) and high voltage ramp monitor (HVRM) are being used in
production applications for advanced ceramic and laminate based MCMs
[109]. Both methods suggest a significant improvement in test speed and
fault detection capabilities as compared to TDNA and resistance testing. A
third test methodology [l09], network sensitive pulse generation (NSPG) is
under development that is designed to identify opens and shorts up to 10
mn resistance, requiring a short dwell time. Trial runs are currently in
progress using NSPG on flex and ceramic substrates.
The RF-resonator method being developed at Georgia Institute of
Technology is a useful method for latent opens and shorts testing of
interconnections. The method is based on the use of a single-probe head
in series with a high Q resonator, as discussed by Kim, Swaminathan and
Chatterjee [14-16]. The resonator modulates the AC response of the
interconnections, producing a change in the frequency response, which
changes as a function of the defects. Both the magnitude and phase
responses are shown in Fig 13-35a and b, respectively. Interconnect defects
change the magnitude response, causing a reduced magnitude as compared
to the defect free response, in the vicinity of the resonant frequency.
Similarly, the phase response in Fig 13-35b varies with open or short
defect, allowing for fault identification. Based on look-up tables created
through calibration measurements, the modulated AC response is com-
pared to stored data (for defect-free interconnections), resulting in a differ-
ential voltage that provides details on the defects. The look-up table is
created through analytical expressions using CAD design data and is
similar to the test file generated for capacitance testing. In this method,
the magnitude and phase responses (as a function of frequency) are com-
bined to diagnose the defect such as its position, size and so forth. The
method is based on the movement of the resonating frequency of the
11-866 PACKAGE ELECTRICAL TESTING

20.0 r-~--~----r--~-----,
(a)
15.0
4>
'0
2
. ~ 10.0
:E'"
(I) defect-free
5.0 (2) defecti ve
pole frequency

Frequency (MHz)

- defect-free ( 1)
- - - opens defect (2) I (b)
,...... 50.0 hons defoct (3) I
'"
~ I

e
4>

0.0 t'======--~-.1
4>
'"
'"
..c
Il-
-50.0

(2)
- I OO. OJ~--""'*"',...........---:;~-~.,,-t
7 705 710 715
Frequency (M Hz)

Figure 13-35. Frequency and phase responses of a defect. (a) Frequency response
of open (IOOllm long) for 2 em long lossy line; (b) phase response of opens (Iill) and
shorts (4pF) for 2em long line.

resonator due to the presence of a defect, resulting in a change in the


magnitude and phase responses. Hence, the resolution of the method is
a function of the quality factor (Q) of the resonator, and for the method
to be viable, a high Q resonator is desired.
To further explain the method, consider a 2 cm line fabricated on
glass using gold metallization, modulated by using a 711 MHz resonator.
Though lower or higher frequency resonators can be used, 711 MHz was
chosen since it was commercially available, had an adequate Q (360) and
resonated at a frequency which was far removed from the standing wave
resonances of interconnections in the range 1-4 cm, which are typical
interconnect lengths in MCMs. Table 13-5 shows the movement of the
resonating frequency as a function of the position and nature (opens or
shorts) of the defect. Assuming a fixed frequency (e.g., a 3-dB point) is
13.11 REFERENCES 11-867

Table 13-5. Resonating Frequency versus Defect for a 2-cm Interconnect

Type of Defect Frequency

Fault-free 667 MHz


Open (0.5 cm from far end) 681 MHz
Open (1.0 cm from far end) 691 MHz
Open (1.5cm from far end) 701 MHz
Short 639 MHz

used for the measurement, the shift in frequency in Table 13-5 produces
a corresponding change in the magnitude and phase responses at the given
frequency. This measurement is then used for pass/fail testing of the
interconnections. Because passives integrated into the substrate produce
shifts in the resonating frequency of the resonator, defects on passives
connected to interconnections can be detected by probing from the top-
surface metallurgy pads. Due to the use of a single probe head and
low test frequency, probe movement, test time, and equipment costs are
expected to be low. This method combines the simplicity of capacitance
testing with the resolution of latent opens testing.

13.10.2 Functional Testing


In the past, many boundary scan techniques have been proposed for
interconnect tests of printed circuit boards and multi-chip modules. A
generalized optimal test vector generation approach for interconnect test
of boards has earlier been proposed. Further, new techniques for self test
of systems and modules have been developed. In the recent past, new
deterministic built-in self-test techniques that address specific classes of
circuits like RAMs, ROMs, PLAs, adders, multipliers, and so forth, have
been developed. Recently, there has been work in automated insertion of
built-in self-test hardware into specialized classes of integrated circuit
designs. There is ongoing work in the area of improving fault coverage
of built-in self-test techniques and in finding optimal combinations of
built-in self-test and boundary scan hardware for faster, cheaper and better
test of complex systems that have hundreds of Ies.

13.11 REFERENCES
I. T. Knowles. Automatic Testing-Systems and Applications, McGraw-Hill, New
York, 1976.
2. B. Davis. The Economics of Automatic Testing, McGraw-Hill, New York, 1982.
3. A. Deutsch et al. "Defect Detection using 70 GHz TDR," IBM Research Report RC
16699, April 1991.
11-868 PACKAGE ELECTRICAL TESTING

4. A. Halperin, T. H. DiStefano, and S. Chiang. "Latent Open Testing of Electronic


Packaging," IEEE Multi-Chip Module Conference, MCMC-94. 1994.
5. J. Marshall et al. "CAD-Based Net Capacitance Testing of Unpopulated MCM
Substrate," IEEE Trans. Components packaging manuf. techno!. B. CPMT-17 (1),
1994.
6. L. Economikos, T. Morrison, and F. Crnic. "Electrical Test of Multichip Substrates,"
IEEE Trans. Components packaging manuf. techno!' B, CPMT-17 (1), 1994.
7. H. Hamel, S. Kadakia, and H. Bhatia. "Capacitance Test Technique for the MCM
of the 90s," Proceedings of the International Electronic Packaging Conference. pp.
855-872, 1993.
8. R. W. Wedwick. "Continuity Testing by Capacitance," Circuits Manuf.. pp. 60-
61. 1974.
9. O. C. Woodard. "High Density Interconnect Verification of Unpopulated Multichip
Modules." Proceedings of the Eleventh IEEElCHMT International Electronics Manu-
facturing Technology (IEMT) Symposium. pp. 434-439, 1991.
10. A. W. Ross. R. R. Goruganthu. and O. C. Woodard. "High Density Interconnect
Verification Using Voltage Contrast Electron Beam," Proceedings of the Eleventh
IEEElCHMT International Electronics Manufacturing Technology (IEMT) Sympo-
sium. pp. 270-274, 1991.
11. M. Brunner, R. Schmid, R. Schmitt, M. Sturn, and O. Gessner. "Electron-Beam
MCM Testing and Probing," IEEE Trans. Components packaging manuf. techno!.
B, CPMT-17 (1), 1994.
12. F. Jensen and N. E. Petersen. Burn-In: An Engineering Approach and Analysis of
Burn-In Procedures. Wiley, New York, 1982.
13. F. Crnic. "Electrical Test of Multi-Chip Substrates," Proceedings of International
Conference and Exhibition on Multi-chip Modules (ICEMM). pp. 422-428, 1993.
14. B. Kim, A. Chatterjee, M. Swaminathan, and D. Schimmel. "A Novel Low-Cost
Approach to MCM Interconnect Test," IEEE International Test Conference. 1995.
15. B. Kim, A. Chatterjee, and M. Swaminathan. "A New Technique for Testing MCM
Substrate," IEEElISHM MCM Test II: Advanced Technology Workshop. 1995.
16. B. Kim, M. Swaminathan, and A. Chatterjee. "A Novel MCM Interconnect Test
Technique Based on Resonator Principles and Transmission Line Theory," Proceed-
ings of IEEE 4th Topical Meeting on Electrical Performance of Electronic Packag-
ing. 1995.
17. R. Bove, U.S. Patent Number 3,806,801.
18. E. Dalrymple. "Very Low Insertion Force Connectors for High Mating Cycle Applica-
tions", Connection Techno!.. December 1986.
19. IEEE Standard Test Access Port and Boundary-Scan Architecture. IEEE Std. 1149.1-
1990, IEEE Standards Office, NJ, 1990.
20. V. D. Agarwal, C. J. Lin, P. W. Rutkowski, S. Wu, and Y. Zorian. "Built-in Self-
Test for Digital Integrated Circuits," AT&T Tech J., pp. 30-39, 1994.
21. J. A. Jorgenson and R. J. Wagner. "Analyzing the Design-for-Test Techniques in a
Multiple Substrate MCM," Proceedings of 12th IEEE VLSI Test Symposium. pp.
360-365, 1994.
22. T. Storey. "A Test Methodology for VLSI Chips on Silicon," Proc. ITC. pp. 359-
368, 1993.
23. M. M. Pradhan, E. O'Brien, S. L. Lam, and J. Beausang. "Circular BIST with Partial
Scan," Proc. IEEE International Test Conf.. pp. 719-729, 1988.
24. C. M. Maunder and R. E. Tulloss. The test access port and boundary-scan architec-
ture. IEEE Computer Society Press, New York, 1990.
25. D. A. Doane and P. D. Franzon. MultiChip Module Technologies and Alternative
The Basics, pp. 615-658, Van Nostrand Reinhold, New York, 1993.
13.11 REFERENCES 11-869

26. A. Flint. "Testing Multichip Modules," IEEE Spectrum, pp. 59-62, 1994.
27. P. R. Mukund and J. F. McDonald. "MCM: The High-Perfonnance Electronic Packag-
ing Technology," IEEE Computer Mag. (Special Issue), 1993.
28. Y. Zorian. "A Universal Testability Strategy for Multichip Modules Based on BIST
and Boundary-Scan," Proc. IEEE International Conference on Computer Design,
pp. 59-66, 1992.
29. Y. Zorian. "A structured Approach to Macrocell Testing Using Built-in Self-Test,"
Proc. IEEE Custom Integrated Circuits Conj., pp. 28.3.1-28.3.4, 1990.
30. C. W. Yau and N. Jarwala. "The Boundary-Scan Master: Target Applications and
Functional Requirements," Proc. IEEE International Test Conference, pp. 311-
315, 1990.
31. K. E. Posse. "A Design-for-Testability Architecture for Multichip Modules," IEEE
International Test Conference, pp. 113-121, 1991.
32. Y. Zorian. "A distributed BIST control scheme for complex VLSI devices," Proceed-
ings of 11th IEEE VLSI Test Symposium, pp. 4-9, 1993.
33. C. J. Lin, Y. Zorian, and S. Bhawmik. "PSBlST: A Partial-Scan Based Built-in Self-
Test Scheme," Proc. IEEE International Test Conference, pp. 507-516, 1993.
34. M. Nicolaidis. "An Efficient Built-in Self-test Scheme for Functional Test of Embed-
ded RAMs," Proc. 15th International Symp. Fault-Tolerant Comput., pp. 118-
123, 1985.
35. Y. Zorian and A. Ivanov. "An Effective BlST Scheme for ROMs," IEEE Trans.
Computing, C- 41(5), pp. 646-653, 1992.
36. A. 1. Van de Goor and Y. Zorian. "An Effective BIST Scheme for Ring-Address
FIFOs," Proc. IEEE International Test Conference, pp. 96-101, 1994.
37. S.D. Golladay, N.A. Wagner, J.R. Rudert, and R.N. Schmidt. "Electron-Beam Tech-
nology for Open/Short Testing of Multi-chip Substrates," IBM Jo. Res. Devel., 34(21
3): pp. 250-259, 1990.
38. A. Hopper et al. "A Feasibility Study for the Fabrication of Planar Silicon Multichip
Modules Using Electron Beam Lithography for Precise Location and Interconnection
of Chips," IEEE Trans. Components Hybrids Manu! Techno!., CHMT-15(l): pp.
97-102, 1992.
39. V. Murali, T. Rucker, J. Fu, and R. Shukla. "Yield and Reliability Concerns in
Polyimide Based Multi-Chip Modules," 1992 IEEE Multichip Module Conference,
pp. 98-10 1, 1992.
40. A. Landzberg. Microelectronics Manufacturing Diagnostics Handbook, Van Nos-
trand Reinhold, New York, 1993.
41. S. DeFoster, M. Mancini, and J. Zalesinski. "Automatic Testing of Metallized Ce-
ramic-Polyimide Substrates," IEEE Trans. Components Hybrids Manu! Techno!',
CHMT-13(4): 1990.
42. M. Sriram and S.M. Kang. Physical Designfor Multichip Modules, Kluwer Academic
Publishers, Boston, 1994.
43. John H. Lau. CHIP ON BOARD: Technologiesfor Multichip Modules, Van Nostrand
Reinhold, New York, 1994.
44. T. Storey et al. "A Test Methodology to Support an AS EM MCM Foundry," IEEE
International Test Conference, pp. 426-435, 1994.
45. A. Flint. "Test Strategies for a Family of Complex MCMs," IEEE International Test
Conference, pp. 436-445, 1994.
46. J. M. Jong, B. Janko, and V. Tripathi. "Equivalent Circuit Modeling of Interconnects
from Time-Domain Measurements," IEEE Trans. Components Packaging Manu!
Techno!., CPMT-16(1): 1993.
47. S. F. Gong et al. "Investigation of High-Speed Pulse Transmission in MCM-D,"
IEEE Trans. Components Hybrids Manu! Techno!., CHMT-18(1): 1993.
11-870 PACKAGE ELECTRICAL TESTING

48. G. White et al. "Large Fonnat Fabrication-A Practical Approach to Low Cost MCM-
D," IEEE Trans. Components Packaging Manuf. Technol., CPMT-16(1): 1995.
49. C. L. Ratzlaff and L. T. Pillage "RICE: Rapid Interconnect Circuit Evaluation Using
AWE," IEEE Trans. Computer-Aided Design Integrated Circuits Sys., CADIC-
13(6): 1994.
50. J. Peeters, E. Beyne, and G. Brandli. "A Broad Band Loss Model for MCM Intercon-
nections," IEEE Multi-Chip Module Conference, 1993.
51. H. Liao and W. Dai. "Wave Spreading Evaluation of Interconnection Systems,"
IEEE Multi-Chip Module Conference, 1993.
52. L. Prokopchak. "Development of a Solution for Achieving Known-Good-Die," IEEE
International Test Conference, pp. 15-21, 1994.
53. Y. Eo and W. R. Eisenstadt, "High-Speed VLSI Interconnect Modeling Based on
S-Parameter Measurements," IEEE Trans. Components Packaging Manuf. Technol.,
CPMT-16(5): 1993.
54. X. Zhang and K. K. Mei. "Time-Domain Finite Difference Approach to the Calcula-
tion of the Frequency-Dependent Characteristics of Microstrip Discontinuities," IEEE
Trans. Microwave Theory Tech., MTT-36(l2): pp. 1775-1787, 1988.
55. S. Voranantakul "Crosstalk Analysis for High-Speed Pulse Propagation in Lossy
Electrical Interconnections," IEEE Trans. Components Packaging Manuf. Technol.,
CPMT-16(1): 1993.
56. T. Ueno and Y. Kondoh. "Membrane Probe Technology for MCM Known-Good-
Die," IEEE International Test Conference, pp. 22-29, 1994.
57. J. R. Brews. "Transmission Line Models for Lossy Waveguide Interconnections in
VLSI," IEEE Trans. Electron Devices, ED-33(9): pp. 1356-1365, 1986.
58. K. W. Goossen and R. B. Hammond, "Modeling of Picosecond Pulse Propagation in
Microstrip Interconnections on Integrated Circuits," IEEE Trans. Microwave Theory
Tech., MIT-37(3): pp. 469-478, 1989.
59. R. W. Jackson and D. M. Pozar. "Full-Wave Analysis of Microstrip Open-End and
Gap Discontinuities," IEEE Trans. Microwave Theory Tech.," MTT-33(l0): pp.
1036-1042, 1985.
60. L. Rosze!. "MCM Foundry Test Methodology and Implementation," IEEE Interna-
tional Test Conference, pp. 369-372, 1993.
61. T. Storey. "A Test Methodology for VLSI Chips on Silicon," IEEE International
Test Conference, pp. 359-368, 1993.
62. M. Melton and F. Brglez. "Automatic pattern Generation for Diagnosis of Wiring
Interconnect Faults," IEEE International Test Conference, pp. 389-398, 1992.
63. S. Yao et a!. "A Multi-Probe Approach for MCM Substrate Testing," IEEE Trans.
Computer-Aided Design Integrated Circuits Sys., CADICS-13( 1): pp. 1l0-121, 1994.
64. F. W. Angelotti "Modeling for Structured System Interconnect Test," IEEE Interna-
tional Test Conference, pp. 127-133, 1994.
65. N. Jarwala and C. W. Yau. "A New Framework for Analyzing Test Generation and
Diagnosis Algorithms for Wiring Interconnects," IEEE International Test Confer-
ence, pp. 63-70, 1989.
66. D. Y. Shih et al. "Factors Affecting the Interconnection Resistance and Yield in
Multilayer Polyimide/Copper Structures," IEEE Trans. Components Packaging Ma-
nuf Technol., CPMT-16(l): 1993.
67. S. Y. Kim et a!. "An Efficient Methodology for Extraction and Simulation of Trans-
mission Lines for Application Specific Electronic Modules," IEEE Multi-Chip Module
Conference, 1993.
68. W. Blood and W. Yip. "Electrical Analysis of a Thin Film Multichip Module
Substrate," IEEE Multi-Chip Module Conference, 1992.
13.11 REFERENCES 11-871

69. R. Kambe. "MCM Substrate with High Capacitance," IEEE Trans. Components
Packaging Manu! Technol., CPMT-18(1): 1995.
70. T. S. Chu and T. Itoh. "Generalized Scattering Matrix Method for Analysis of
Cascaded and Offset Microstrip Step Discontinuities," IEEE Trans. Microwave The-
ory Tech., MTT-37(2): pp.280-284, 1986.
71. W. Menzel and I. Wolff. "A Method for Calculating the Frequency-Dependent
Properties of Microstrip Discontinuities," IEEE Trans. Microwave Theory Tech.,
MTT-37(2): pp. 107-112, 1977.
72. A. Gopinath and C. Gupta. "Capacitance Parameters of Discontinuities in Microstrip
lines," IEEE Trans. Microwave Theory Tech., MTT-37(1O): pp. 831-836, 1978.
73. P. B. Katehi and N. G. Alexopoulos. "Frequency-Dependent Characteristics of Mi-
crostrip Discontinuities in Millimeter-Wave Integrated Circuits," IEEE Trans. Micro-
wave Theory Tech., MTT-37(l0): pp. 1029-1035, 1985.
74. A. Iqbal, M. Swarninathan, M. Nealon, and A. Orner. "Design Trade-offs Among
MCM-C, MCM-D and MCM-D/C Technologies," IEEE Trans. Components Packag-
ing Manu! Technol., CPMT-17(1): 1994.
75. 1. M. Jong, Bozidar Janko, and Vijai Tripathi. "Equivalent Circuit Modeling of
Interconnects from Time-Domain Measurements," IEEE Trans. Components Hybrids
Manu! Technol., CHMT-16(l): 1993.
76. N. Nagi, A. Chatterjee, J. A. Abraham. "Fault Simulation of Linear Analog Circuit,"
1. Electron. Testing: Theory Applic., JETTA-4: pp. 345-360, 1993.
77. N. Nagi, A. Chatterjee, A. Balivada, J. A. Abraham. "Fault-based Automatic Test
Generator for Linear Analog Circuits," IEEE International Conference on Computer
Aided Design, 1993.
78. N. Nagi, A. Chatterjee, and J. A. Abraham. "MIXER: Mixed-Signal Fault Simulator,"
Proc. International Conference on Computer Design, 1993.
79. A. Deutsch, G. Arjavalingam, and G. V. Kopcsay. "Characterization of Resistive
Transmission Lines by Short-Pulse Propagation," IEEE Trans. Microwave Guided
Wave Lett., MGWL-2(1), 1992.
80. W. H. Kautz. "Testing for Faults in Wiring Networks," IEEE Trans. Computers,
C-23(4): pp. 358-363, 1974.
81. R. R. Tummala and E. J. Rymaszewski. Microelectronics Packaging Handbook,
Van Nostrand Reinhold, New York, 1989.
82. A. Mehta et al. "SuperSPARC mu1tichip module," IEEE Multi-Chip Module Confer-
ence, pp. 19-28, 1993.
83. A. J. Blodgett and D. R. Barbour. "Thermal conduction module: A high Performance
Ceramic Package," IBM 1. Res. Devel., 26: pp. 30-36, 1982.
84. J. U. Knickerbocker et al. "IBM Systeml390 Air Cooled Alumina Thermal Conduc-
tion Module," IBM 1. Res. Devel., 35: pp. 330-341, 1991.
85. Georgia Institute of Technology, "NSF Proposal," Packaging Research Center, At-
lanta, GA, 1994.
86. J. Conti. "Case for Moving Probe Testing of Bare Boards," Surface Mount Technol.,
5(4): p. 56, 1991.
87. T. Ninomiya and Y. Nakagawa. "Automated Pattern Inspection for Unbaked Layers
of Multi-Layer Ceramic Substrates", IEEE International Workshop on Industrial
Applications of Machine Vision and Machine Intelligence, pp. 346-351, 1987.
88. H. Oka, M. Ando, J. Serizawa, K. Fujihara and H. Tate. "Automatic Optical Pattern
Inspection System Using 3D Profile Measurement", Printed Circuit World Conven-
tion, 1990.
89. K. Smith and G. Rinne. "MCMs: Approaches for Testing and Troubleshooting",
High Speed Digital Symposium and Exhibition, 1992.
11-872 PACKAGE ELECTRICAL TESTING

90. R. Allison. "Bare Board Applications for Moving Probe Test Systems", Proceedings
of the Technical Program-National Electronic Packaging and Production Confer-
ence, pp. 792-808, 1992.
91. M. Brunner, D. Winkler and B. Lischke. "Crucial Parameters in Electron Beam
Short/Open Testing," in Microcircuit Engineering 84, ed. by A. Heuberger and H.
Benking, Academic Press, London, pp. 399-410, 1985.
92. D. L. Millard, K. R. Umstadter and R. C. Block. "Noncontact Testing of Circuits
Via a Laser-Induced Plasma Electrical Pathway," IEEE Design and Test ofComput-
ers, 1992.
93. T. Sakata and K. Numata. "Prober for Highly Integrated Multichip Modules", IEEE!
CHMT International Electronics Manufacturing Technology Test Symposium, pp.
429-433, 1992.
94. P. Slade. "Resistor Trimming", IEEE Trans. Components Hybrids Manuf. Technol.,
CHMT-14(l): pp. 2-18, 1991.
95. D. Ballew and L. M. Streb. "Board-Level Boundary-Scan Regaining Observability
with an Additional IC," Proceedings IEEE International Test Conference, pp. 182-
189, 1989.
96. 1. Bond. "Test dominates MCM assembly," Test & Meas. World, pp. 59-64, 1992.
97. M. Brunner, R. Schmid, R. Schmitt, M. Sturm, and O. Gessner. "Electron-beam
MCM Testing and Probing," IEEE Trans. Components Packaging Manuf. Technol.,
CPMT-17(1): pp. 57-62, 1994.
98. R. A. Fillion, R. J. Wojnarowski, and W. Daum. "Bare Chip Test Techniques for
Multichip Modules," Proceedings 40th EIAlIEEE Electronic Components Technol-
ogy Conference, p. 554, 1990.
99. A. E. Gattiker, W. Maly, and M. E. Thomas "Are There Any Altematies to Known-
Good-Die," Proceedings of MCM Conference, pp. 102-107, 1994.
100. S. C. Hilla, "Boundary-Scan Testing for Multichip Modules," Proceedings IEEE
International Test Conference, pp. 224-231, 1992.
101. D. C. Keezer. "Bare Die Testing and MCM Probing Techniques," Proceedings IEEE
Multi-Chip Module Conference (MCMC), pp. 20-23, 1992.
102. M. Lubaszewski, M. Marzouki, and M. H. Touati. "A Pragmatic Test and Diagnosis
Methodology for Partially Testable MCMs," Proceedings IEEE MCM Conference,
pp. 108-113, 1994.
103. R. H. Parker. "Bare Die Test," Proceedings IEEE Multi-Chip Module Conference
(MCMC), pp. 24-27, 1992.
104. P. Raghavachari. "Circuit Pack BIST for System to Factory-the MCERT Chip,"
Proceedings International Test Conference, p. 641, 1991.
105. R. Roebuck, et al. "Known good die: a practical solution," ICEMM Proc., pp.
177-182, 1993.
106. L. Roszel. "MCM foundary test methodology and implementation," Proceedings
IEEE International Test Conference, pp. 369-372, 1993.
107. H. N. Scholz, R. E. Tulloss, C. W. Yau, and W. Wach. "ASIC Implementations of
Boundary-Scan and BIST," 8th International Custom Microelectronics Conference,
pp. 43.0-43.9, 1988.
108. B. Vasquez, D. Van Overloop, and S. Lindsey. "Known-Good-Die Technologies on
the Horizon," Proceedings of IEEE VLSI Test Symposium, pp. 356-359, 1994.
109. G. Mezack. "Personal Communication," Integrated Solutions Inc., Tewksbury, MA.
14
PACKAGE SEALING AND
ENCAPSULATION

C.P. WONG-Georgia Tech


D.B. CLEGG-Motorola Inc.
ANANDA H. KUMAR-18M
K. OTSUKA-Hitachi
BERHAN OZMAT-TI

14.1 INTRODUCTION

The packaging is defined as interconnecting, powering, cooling, and


protecting integrated circuits. This chapter addresses the protection of
semiconductor devices by means of sealing and encapsulation processes
after the device packaging is completed. In addition to providing protection
of devices, sealing and encapsulation must also protect the package and
its package wiring so as to provide the required reliability of the packaged
devices. Figure 14-1 illustrates the basic concepts of sealing and encapsu-
lation.
An advanced ultralarge-scale integration (ULSI) device is a very
complex and delicate, three-dimensional structure. It consists of millions
of components in a single integrated-circuit (IC) chip. These components
11-874 PACKAGE SEALING AND ENCAPSULATION

Metal/Ceramic
Seal ~I f? CnlP 5\
I~~~~c-e~ra~m~i-c~--~I
If- Seal

Figure 14-1. Basic Concepts of Sealing and Encapsulation

are densely packaged in a multilayer structure with different metalized


conductor lines separated with dielectric insulating layers. The rapid
growth of the number of components per chip, the rapid decrease of
device dimensions, the steady increase in Ie chip size, and the ever-
increasing input/output (I/O) interconnects have imposed stringent require-
ments not only on the Ie physical design and fabrication but also on the
Ie encapsulation and package sealing. The increase of integration in ULSI
technology has resulted in the miniaturization of the device size which
has reduced the propagation delay due to higher-density packaging and
interconnection. As a result, a modem advanced device operates at a faster
speed (> 100 MHz), consumes less power, and, consequently, dissipates
less heat during operation. In addition, ULSI technology has increased
the reliability of the microelectronic devices due to the elimination of the
poor interconnections and has decreased the cost per function of the
devices which has had a profound impact on the modem electronics
industry. However, due to the high-density packaging of these devices,
the power consumption per package is also increased, so is the physical
dimension of the chip size and multichip modules (MCM). To package,
encapsulate, and seal these high-power, large IC devices and the MCM
is becoming a challenge [1,2].
The various materials used to fabricate integrated circuits (ICs) and
their related interconnect are generally not capable of surviving their end-
use operating, or field, environment without some level of additional
protection. Metals susceptible to corrosion, dissimilar metals forming
galvanic cells, structures that are stress or heat sensitive, and so on, can
all lead to performance degradation or failure in the field without proper
protection. Field environments can vary greatly for anyone device, de-
pending on the product in which it is incorporated or the customer's
application of the product. These environmental stresses are not only
limited to the thermomechanical stresses associated with device operation
but also to the pollutants present in the operating environment.
A particular product may be appropriate for many uses, even a
mobile one, so that a single product may see several environments in its
lifetime. When factors related to the device are coupled in, such as the
cost of manufacture, production volume, expected use environments or
markets, expected volume of product going to the various markets, and
so forth, packaging issues and decisions become very complex. These
14.1 INTRODUCTION 11-875

complexities require the packaging engineer to make critical decisions


on how to most cost-effectively package the device while obtaining accept-
able reliability in the field. Packaging selection can be seen as a compro-
mise between cost and performance, where performance includes reliabil-
ity issues as well as other design considerations. Two earlier chapters of
this book have discussed specific packaging methods: Chapter 9, "Ceramic
Packaging," usually involving hermetic enclosures; and Chapter 10, "Plas-
tic Packaging." This chapter introduces more general concepts and options
for the environmental protection of integrated circuits using inorganic and
organic materials, as well as combinations of the two, to obtain acceptable
field reliability.
In the early days of single transistors, hermetically sealed metal
cases protected sensitive chips from environmental stresses. Moisture-
related problems prompted the military and aerospace industries to require
hermetic packages in order to achieve high reliability and long life. Her-
metic packages utilize a sealed environment that is impervious to gases
and moisture to protect the devices. Final sealing is accomplished with
caps or lids using glass or metal seals. Moisture content within the package
is standardly limited to a volumetric maximum of 5000 parts per million
which was chosen to eliminate the possibility of condensation occurring
inside the enclosure. For extremely sensitive devices, however, moisture
limits are lowered even further, increasing both the time and cost to
manufacture and treat these hermetically sealed packages.
Hermetic packaging proved to be such a robust method of obtaining
a long field life that little has changed in hermetic packaging technology
in recent years. After a slow development and initial reliability problems,
hermetically packaged devices gave way to lower-cost nonhermetic plas-
tic-molded packaging, especially in electronics destined for commercial
and industrial markets. Today, however, nonhermetic does not imply
nonreliable. Plastic packages are not hermetic, yet demonstrate admirable
reliability and now account for approximately 90-95% of all device pack-
aging.
Many obstacles had to be overcome with respect to understanding
materials and processes associated with plastic packaging before their use
was widespread. Organic materials are not hermetic and allow moisture
to penetrate and be absorbed. Different organic material families have
greatly varying material properties related to their moisture permeation
rates and moisture-absorption percentages. In plastic packaging's infancy,
corrosion was found to be the primary cause of failure due to poor adhesion
and high levels of mobile ionic contaminants in the materials. The presence
of moisture along with mobile ions were identified as the major contribu-
tors to corrosion, among other failure mechanisms. Improvements in plas-
tic packaging materials and processes have all but removed corrosion as
a source of failure and have lead to reliability that approaches that of
11-876 PACKAGE SEALING AND ENCAPSULATION

hermetic packages. Now, stress-related failures from increasing die sizes


present the largest obstacle to be solved by the plastic packaging engineer.
These major improvements in plastic packaging were derived from
the effective use of large volumes of environmental test data that assisted
engineers in understanding factors governing the principal failure modes
associated with protection through organic materials. Accelerated environ-
mental stress tests on packaged devices were the primary source of these
data. In some cases, combinations of accelerated stress, along with corro-
sive ambients, were used to predict reliability behavior under severe use
conditions and were further extended as a means of assessing the effects
of impurities on device lifetimes. Accelerated stress tests such as the
85°C/85% rh test (85/85) became useful tools for screening and assessing
the coupling between the processes and material systems. In the beginning
of plastic packaging, 1000 h of 85/85 was not only difficult to obtain
with existing materials and processes but was considered sufficient for
many commercial markets. It is now quite common for plastic packages
to exceed 10,000 h of 85/85 without failure.
More severe environmental tests have been devised to shorten test
and evaluation times, decreasing the cost of the final product. The Highly
Accelerated Stress Test (HAST) is the most prominent example and ex-
pands upon the 85/85 approach. HAST utilizes a pressurized chamber to
allow the temperature to be elevated while maintaining humidity control,
to produce the additional activation energy for more accelerated testing.
Quite a bit is still unknown with regard to accelerated testing and its
correlation to field life. This chapter will consider these issues and their
relationship in the environmental protection of integrated circuits.
In both hermetic and nonhermetic packages, it is assumed that one
level of protection precedes the sealing or encapsulation, as discussed in
this chapter. This first line of defense is what is customarily termed
the "passivation layer" and is typically applied at the back end of the
semiconductor device circuit fabrication process. The passivation layer
covers the entire integrated circuit with the exception of the bond pads,
which remain exposed to facilitate signal and power interconnect to the
device. The most prevalent circuit passivations consist of chemically
vapor-deposited phosphosilicate glass, or silicon nitride, or photodefinable
polyimides that are dispensed in spinning process, then exposed, etched,
and cured. The passivation layer on the devices, although very important
in providing device reliability, is considered a given input to any sealing
and encapsulation packaging processes and an art within itself. Therefore,
passivations will not be discussed in detail in this chapter.
14.2 HERMETIC VERSUS NONHERMETIC PACKAGING
There is a natural division of definitions between the two most
common packaging approaches for integrated circuits: hermetic packaging
14.2 HERMETIC VERSUS NONHERMETIC PACKAGING 11-877

which presents a sealed, impermeable package to the operating environ-


ment, and what we term nonhermetic, which is permeable. The hermetic
package usually consists of a cavity package made of ceramic, metal,
glass, and so forth, that are, for the most part, resistant to permeation by
fluids and gases. On the other hand, nonhermetic packages are generally
associated with organic coatings or seals such as epoxies, silicones, poly-
imides, and so forth, which are not capable of preventing penetration of
fluids and gases. There are also new technologies that are designated
"hermetic-equivalent" packages that use novel approaches to obtain her-
meticity through depositions of metal films to seal the electronics package.
Both hermetic and nonhermetic packages are able to house single devices
or multichip systems.
Considering the issues, the packaging engineer's objective is to
evaluate the criticality of the application, the types of devices and assembly
methods that will be contained in the package, and the operational environ-
ment in which the package will be used, and to determine which type of
packaging approach provides the best trade-off among cost, reliability,
and performance. In general, the most critical applications requiring the
highest, long-term reliability will utilize hermetic packaging, but today
the choice is not always so clear-cut, as there are many factors that must
be weighed, including cost, size, and weight. With the great strides that
have been made in packaging with polymers in the last 10 years, more
and more applications are open to its use. As different as the two packaging
approaches are, so are the screening and qualification testing used to
evaluate their performance. Hot, wet tests such as 85/85 or HAST do not
have much effect on a properly designed and assembled hermetic package.
The opposite is true with nonhermetic packages, as organic materials tend
to break down, or depolymerize and hydrolyze in these environments.
The remainder of this chapter provides an overview of both packaging
methods, as well as some of their disadvantages, advantages, and technical
details of their design and production-related processes.

14.2.1 Hermetic Packaging

Condensed moisture on a device's surface during operation has


historically been shown to lead to the principle causes of failure in the
field. The extremely small geometries involved in integrated circuits,
differing galvanic potentials between metal structures, and the presence
of high electric fields all make the device susceptible to interactions
with moisture. One packaging method designed to prevent performance
degradation due to moisture's deleterious effects is the hermetic package.
Ceramic or metal is used to form an enclosure to isolate the electronic
devices from the ambient operating environment. By eliminating condensi-
ble moisture from the cavity of the package during the sealing process
11-878 PACKAGE SEALING AND ENCAPSULATION

and preventing the ingress and egress of moisture at the package perimeter
during its operating life, excellent long-term reliability can be achieved.
The word hermetic is defined as completely sealed by fusion, solder, and
so on, so as to keep air or gas from getting in or out; in other words,
airtight. In hermetic packaging practice, such seals are nonexistent. Small
gas molecules will enter the package over time through diffusion and
permeation. Eventually, these gases will reach equilibrium within the
cavity of the package. In light of this permeation, long life can still be
obtained in the field because of the extremely slow nature of this activity.
Accelerated tests and leak testing for screening and qualification are
specified in the military standard MIL-STD-883 and represent the founda-
tion and industry-accepted approach to testing for reliability in her-
metic packages.

14.2.2 Nonhermetic Packaging

The widespread availability of nonhermetic packaging followed that


of hermetic packaging by many years due to problems associated with
the initial polymer materials used for encapsulation. Both the traditional
plastic-molded package and polymer-encapsulated circuits fell into this
category. The inability of these early polymers to sufficiently retard the
deleterious effects of moisture once it reached the delicate surfaces of
the circuit, interconnect, and substrate led to poor performance both in
testing and in the field. Inadequate adhesion, contaminants within the
material itself, incompatible thermal expansion, and stress-related prob-
lems, in addition to a relatively immature knowledge of filler technology,
all combined to prevent plastic packaging's immediate acceptance. With
significant efforts in the areas of resins, fillers, material formulations, and
process development work, polymer packaging finally began to make its
presence felt in the early 1970s, although they had been used in some
form as early as the late 1950s. During this time, significant progress was
made in improving the quality of the glass passivation layer that is depos-
ited on the active area of the device as a first line of defense against
moisture-related problems. The combination of these technological ad-
vances provided the essential increase in reliability that was needed for
polymer packaging to begin seeing widespread use.
During plastic packaging's infancy period, many failure modes were
identified and material concerns addressed that resulted in different poly-
mers and polymer formulations that more closely matched the require-
ments of the packaging technology and its applications. This has led to
recent demonstrations of plastic packages that approach the reliability of
hermetic packages in many applications and environments. Certainly,
plastic-molded packages are the most dominant packaging method in use
today for commercial- and industrial-grade electronics. It is estimated that
14.2 HERMETIC VERSUS NONHERMETIC PACKAGING 11-879

over 90% of all integrated circuits are marketed in this form. Nonhermetic
packaging encompasses not only the aforementioned plastic-molded pack-
age but also plastic and ceramic cavity packages that are sealed with a
polymer, and the more recent chip-on-board (COB) and multichip module
(MCM) approaches to packaging when a hermetic seal is not used. Most
of the high-performance MCMs being produced, MCM-C (ceramic) and
MCM-D (deposited), are currently packaged in hermetic packages, but the
lower-cost MCM-L (laminated) which is a wirebonded or tape automated
bonding (TAB) device on a printed wiring board, tends to be encapsulated
with organic polymers in an overcoat fashion.
Much remains to be understood in the use of polymers for protection
of semiconductor devices. Generalities of the required material properties
are known, and how to avoid the failures associated with them, but specific
degradation processes and activities at the material-surface interface re-
main the object of many research dollars. As mentioned earlier, corrosion
was the primary cause of failure in early polymer packaging. New formula-
tions with better filler technology resulted in materials that do not impart
stress-related failures on devices and their associated interconnect, until
the recent emergence of large dies, are attempted. In the newer technologies
of COB and MCMs, corrosion is once again a primary cause of failure
with the existing encapsulation materials, especially with wirebonded
devices as in most plastic-molded packages where the thin aluminum
bond pad on the device is exposed to the coating interface.
Adhesion between metallic-organic interfaces is facilitated by a
combination of mechanical interlocking and chemical and physical bond-
ing. If the density and strength of the molecular bonds between polymer
and substrate is not high enough to prevent an aqueous phase from forming,
and the other ingredients are available, corrosion is likely to occur. Differ-
ent families of materials have different ways of attaining corrosion protec-
tion. For example, although typical epoxies do not have the density of
bonds at the interface relative to silicones, moisture permeation through
epoxies is much, much slower, thereby limiting available moisture at the
interface. Although epoxies in general have a much higher strength of
adhesion than silicones, they also have a much higher modulus which
can lead to degradation of the material interface through mismatches in
thermal expansion when thermal excursions are experienced. Corrosion
protection and adhesion properties are closely linked, and long-term relia-
bility requires long-term adhesion.
This polymer matrix density perspective carries over into the bulk
material as well. Although it is certainly desirable to have a polymer
material that is completely avoid of ionic species to begin with, it is also
highly desirable for the material to be able to limit the migratory capabili-
ties of ionic species from the ambient environment.
Moisture can also have deleterious effects on the long-term adhesion
11-880 PACKAGE SEALING AND ENCAPSULATION

between organic materials and the substrate being protected. Moisture


acts as a debonding agent through one or a combination of the following
mechanisms: (I) attack of the metal surface by moisture can form a weak,
hydrated oxide surface, (2) moisture-assisted chemical bond breakdown,
and (3) moisture-related degradation or depolymerization. The natural
protectant of metals, surface oxides, can actually be detrimental to the
durability of the polymers adhesion in wet environments. Both metals
oxides are relatively polar. Water (H20) will preferentially adsorb onto
the oxide surface and create a weak boundary layer at the metal-polymer
interface that can lead to adhesion problems. In wet environments, long-
term corrosion prevention properties of a polymer can be adversely ef-
fected by this weakening interface region.
Accelerated testing is usually the means by which nonhermetic pack-
aging is assessed during screening and qualification in the manufacturing
process, with hot wet environmental tests the most common. Temperature
cycling is the most common thermomechanical environmental test. Tem-
perature cycling does not test the corrosion-resistant properties of the
package and polymer system, but it tests the ability of the assembly to
endure the stresses imparted by the various materials that make up the
device, interconnect, and polymer packaging. An accurate correlation of
screening and qualification testing to field life has yet to be accomplished,
but individual manufacturers have historical, and most often proprietary,
data that prove that the technology achieved acceptable reliability in
specific markets and end-use environments. As more and more is under-
stood in relation to the fundamental principles surrounding the use of
polymers in packaging, its domination is expected to increase, and more
difficult markets, such as the military, will open to it.

14.3 IC FAILURE MECHANISMS


The failure mechanisms associated with moisture-induced failures
are discussed in greater detail in Chapter 5, "Package Reliability." Only
a brief description is provided here.
Moisture is the principal cause of nearly all failures in nonhermetic
packages. Even in those cases where the presence of a second contaminant
is essential to cause the failure, such failures would not normally occur
in the absence of moisture. The three common types of problems caused
by moisture are the following:
1. Charge Separation in Metal-Oxide Semiconductor (MOS) De-
vices: Here moisture provides mobility to surface charges on insulators,
sometimes leading to the formation of parasitic gates. The failure will
occur as soon as the field of the parasitic gate reaches a specific value.
This type of failure is not permanent. Gross failures are noted at moisture
levels exceeding 2%, whereas below this level, failures depend on the
14.3 IC FAILURE MECHANISMS 11-881

presence of ionic contaminants such as sodium. [3] Recently, a passivation


layer has been added, covering the IC devices and reducing the charge-
migration problem.
2. Corrosion of Metallization: A galvanic cell will be formed in the
presence of moisture and a galvanic couple. Ionic contaminants can dis-
solve in condensed moisture, greatly increasing the rate of corrosion. The
corrosion of AI-Au and AI-Ag wirebonds in epoxy-bonded microcircuits
are two examples of such corrosion [4]. In these cases, small amounts of
chlorine were also found on the corroded surfaces. Ultrasonically bonded
wires were found to be more susceptible to such corrosion than thermo-
compression bonded wires, presumably because of the effect of cold-
working on the galvanic potential.
An example of direct chemical corrosion is the corrosion of alumi-
num stripes when moisture and chloride contamination are present, accord-
ing to the following reactions:

6HCI + 2 Al ~ 2AICl3 + 3H2


AICl3 + 3HOH ~ AI(OH)3 + 3HCI
2AI(OH)3 ~ aging ~ Ah03 + 3H20

In the absence of the chloride ion, the aluminum anode stripes readily
become passivated with the oxide, but when present, these ions rapidly
accumulate at the anode surface, promoting the formation of a nonpassivat-
ing oxide [5]. In one instance, it was noted that in the absence of a voltage
gradient, failure of aluminum stripes occurred after 2000 h, whereas when
a voltage difference of 25V DC between conductors 0.5 mm apart was
present, failures took place after only 50 h[6].
Another important cause of aluminum corrosion is its dissolution in
orthophosphoric acid formed by the reaction of moisture with excess
phosphoros contained in phosphosilicate glass passivating layers on semi-
conductor devices [7]. The leachability of the phosphoros increases with
the P205 content, the annealing temperature, and time [8]. In plastic
packages, the minimum phosphorus level that is compatible with the
desired device characteristics has to be used.
3. Electrolytic Conduction: The presence of condensed water and
a continuous water path or a very thick absorption layer due to high
moisture ambient between two electrically biased stripes of the same or
different metals causes an electrolytic current to flow from the anode to
the cathode. The anodic metal dissolves in the electrolyte and redeposits
at the cathode. The deposited metal forms a projection of the cathode
toward the anode, increasing the local field strength at this site, thereby
attracting more of the metal to redeposit from the electrolyte. This escalat-
ing effect on the field strength gives rise to the characteristic dendritic
11-882 PACKAGE SEALING AND ENCAPSULATION

(treelike) growth of the metal deposit on the cathode, eventually leading


to the disappearance of the anode metal or, more likely, to the formation
of a short, causing device failure. The electrolytic corrosion of gold
conductor lines in the presence of moisture and chlorine is shown in
Figure 14-2 [9]. A definite threshold moisture level of 1.5% was noted
for the reaction described, with no failures occurring below this level
[9,10] . Other moisture-induced failures are also found to be characterized
by minimum threshold levels, some of which are listed in Table 14-1
[11]. Although the mechanisms involved in these electrolytic failures are
well understood, the estimation of failure rate for a given field strength
and geometry is not trivial.

The failure mechanisms just considered require the presence of con-


densed water on the device surface at operating temperatures. The tempera-
ture at which water vapor of a given partial pressure condenses is the
dew point. From the nomograph in Figure 14-3 it will be seen that the
moisture content within the package atmosphere must be greater than
6000 parts per million by volume (ppmv) (dew point at O°C) for any

(- -J

(+1

Figure 14-2. Migrated-Gold Resistive Shorts in Gold Metallized Microcircuit


(From Ref. 9, ©IEEE.)
14.3 IC FAILURE MECHANISMS 11-883

Table 14·1. Device Failure and Water-Vapor Concentration

Water Vapor Concentration (ppm)'


Failure-Free
Failure Mode Demonstrated Failures Upper Limits

Nichrome disappearance 5,000-10,000 500


Aluminum disappearance 50,000-250,000 1,000
Gold migration 15,000-150,000 1,000
MOS inversion 5,000-20,000 200

'Measured at package ambient temperature of 100°C.


Source: From Ref. 11, ©IEEE.

liquid water to be capable of forming. The actual moisture level that can
be tolerated will be lower because of the role that oxide surfaces and
cracks and crevices play in causing moisture condensation. For general
purposes, the acceptable total water content cited in MIL-STD-883 is
5000 ppmv. The surface resistivity on package surfaces, which is the
rate-determining property, is directly proportional to the thickness of the
absorbed or condensed water film. This, in tum, is related to the moisture
level in the package by the following equation [12]:

Number of monolayers of water L = 3.98 X 10-2 CV / S [14-1]

where
C = concentration of water vapor in ppmv
S = internal surface area
V = internal volume in cm3

For a transistor outline (TO-8) package, the equation calculates that


a 6000-ppmv water-vapor level could yield 30 monolayers of water on
the package walls if all the water vapor condensed.
Although the connection between the presence of moisture and de-
vice failures is beyond any doubt, it has not been possible to establish a
direct relationship, other than a statistical one between the level of moisture
contamination and device life. However, in a clean surface condition, the
number of absorption layers correlates with the leakage current between
the electrodes [13]. It is surmised that this may be due to the critical role
played by other factors that are not easily characterized, on corrosion
rates, such as the presence of ionic contaminants. Statistical models have
11-884 PACKAGE SEALING AND ENCAPSULATION

10,000 1,000
8,000
8,000
4,000 500
3,000
2,000 200

1,000 100
800
800
50
400
300
200 20
.,
100
80
80
10
•!
~
Q.

40
30
20 ..•
.r
c
..,..
0 .r
-50 .5
5.0 ~
E
C
2.0
II.
-60
!:::J
.,
C 0
0
JII -70 Do

..•. !
0
..
10.0 -80 'Iii 1.0
8.0 Q -80 0
8.0 :::J
-90 Do
4.0
(I -70 !:::~ 0.5
'0
3.0 :I
-100
-110 -80
•
Q
2.0 0.2
-120
1.0 -130 -90 0.1
0.8
0.8
0.05
0.4
0.3
0.2 0.02

0.1 0.01

Figure 14-3. Nomograph for Dew Points and ppmv as a Function of Pressure

related failure rate to the temperature and relative humidity in several


ways, such as failure rate (1) as proportional to exponential of temperature
and relative humidity [14], (2) as proportional to log vapor pressure of
water [15], and (3) as proportional to square of the relative humidity
[16]. The electrical conductivity on the device surface shows similar
relationships to the relative humidity, reflecting the role of ionic mobility
in the corrosion reactions.

14.3.1 Sources of Moisture


The three contributors to the moisture within a package are (1) the
sealing ambient, (2) the absorbed and dissolved water from the sealing
14.3 IC FAILURE MECHANISMS 11-885

materials, lid, and the substrate, released during the sealing process, and
(3) the leakage of external moisture through the seal itself. The sealing
ambient has to be very dry, with moisture levels at the sealing zone of
the furnaces being kept below 100 ppmv. Hot cap-type sealers use nitrogen
gas formed from a liquid-N2 source which will have moisture levels of
10 ppmv or less.
The problem of moisture evolution from sealing glasses will be
discussed later. This is the principal reason for failures in ceramic dual-
in-line packages (Cerdips). Even when sealing by soldering and welding
techniques, it is important to remove absorbed moisture from all the
materials involved by baking the parts to a suitable temperature. Typically,
a 200°C, 24-h vacuum baking at a vacuum level of a few torrs should
be effective in removing all moisture from packaging materials. Table
14-2 gives the moisture levels associated with several bakeout times and
temperatures [3]. Once baked, the package components should be stored
under dry conditions such as in N2 boxes to prevent reabsorption of
moisture. It has been noted that the very bakeout operation can often
change the moisture-adsorption characteristics of the surfaces for the
worse. Thus, although gold-coated surfaces do not absorb water to any
significant extent, the bakeout time and temperature could bring out transi-
tion-metal atoms from below. The presence of oxygen in the ambient
promotes such out-diffusion of transition metals which have a much higher
affinity for water molecules. Hence, the control of oxygen levels in bakeout
ovens is very important [17].
Another source of moisture within the package is the epoxy die
attach material. It has been shown that when silver-epoxy is heated to
about 300°C, the unreacted monomers decompose to form water as a
reaction product [18]. The moisture levels within brazed packages contain-
ing the silver-epoxy die attach compound were found to be over 10,000
ppmv. A preseal bake at 300°C in N2 significantly reduced the moisture
level (2000 ppmv), whereas a change over to weld-sealing, following a

Table 14-2. Moisture Content Versus Bakeout Time and Temperature

Temperature
Time 150°C 200°C 250°C
@ {pp!!!l {pp!!!l {pp!!!l

15,000 14,000 10,000


10 5,000 1,000 50
100 500 200 10

Source: From Ref. 3, ©IEEE.


11-886 PACKAGE SEALING AND ENCAPSULATION

preseal bake, lowered the moisture levels within the packages to below
300 ppmv. Significantly, the welded packages when heated to 300°C
reached moisture levels of 3000 ppmv. Desiccants are sometimes intro-
duced into the package to absorb the moisture within the package. A
Si-Au-Sn alloy, used for die attach, is found to be a powerful desiccant
because the silicon in the alloy reacts with water vapor [18].
Once a package is sealed with these precautions to minimize sealed-
in-moisture, the only source of moisture left is the leakage through the
seal. A hermetic seal implies a seal that will indefinitely prevent the entry
of moisture and other contaminants into the sealed cavity. In practice,
such seals are nonexistent. Smaller gas molecules will enter the cavity
over a period of time, by diffusion or permeation, and ultimately reach
equilibrium within the package. Hence, an operational definition of her-
meticity is needed. Such a definition is provided for hermetic seals of
microelectronic packages in MIL-STD-883, in terms of helium leak rates
when the package is tested in a prescribed manner. The acceptable leak
rates for military hybrid packages per MIL-STD-883 are given in Table
14-3.
It is suggested that in the absence of an explicit relationship between
device life and moisture level, the setting of the these maximum allowable
leak rates reflect more the limits of package-sealing and leak-detection
methods than any fundamental consideration [20]. For example, for a 1-
cm3 package having a leak rate of 5 x 10-6 atm-cm3/s, the rate of water
buildup will be 10-1 ppm/s or 8.6 x 103 ppm per day, indicating that the
package will be in equilibrium with the atmosphere within a matter of
days. Similar conclusions were arrived at by several investigators [20-22].
Figure 14-4 shows the measured moisture penetration rates for various
helium leak rates. A helium leak rate of 10- 10 atm-cm3/s is said to assure
adequate protection against moisture permeation. It is also noted, however,

Table 14-3. Allowable Hermeticty Leak Rates (Method 1014.4 MIL-STD-883)

Bomb Condition
Pkg. Volume Exposure Time Dwell Reject Limit
(cm3) ~ ill ill (atm·cm 3/s He'l

<0.40 60±2 2+0.2,-0 5xlO-8


~.40 60±2 2+0.2,-0 2xlO-7
~.40 30±2 4+0.4,-0 Ix 10-7

'Note: The leak ratios can be expressed in International Units (SI units through the approximate
relation 1 atm·cm3/s = 0.1 Pa m3/s)
14.4 PACKAGE SEALING AND ENCAPSULATION 11-887

108~----~------~-----r------'------'~

1 y--- S 2.210-2 atm 2.710-1


atm
100d 107r------r------r------r----~~----_R~

t 10d---108r-----1------+----~~~--~~--~~

•E
~ 1 d---105r------r----~~~--~~~~~~--~~

1 h---

1017
=
Initial Volume V 0.02 cm3 Number of Water Molecules -+
I I I
- 25"C/70% r.h. 102 103 1()4 ppm 105
- - 90"C/40% r. h. Water Content -+

Figure 14-4. Water-Vapor Penetration. The penetration of water vapor into packages
for different He leak rates. (From Ref. 20, ©IEEE.)

that the helium rates measured under test conditions could be several
orders of magnitude larger than the rate of moisture ingress under use
conditions. This is because of the absence of viscous flow contribution
to moisture movement in the leak. path in the latter case and because of
the tendency of moisture to absorb on the walls of the leak. path [21].
Actually, most hermetic packages probably have much lower leak. rates
than the specifications allow.

14.4 PACKAGE SEALING AND ENCAPSULATION

The purpose of sealing and encapsulation is to protect the electronic


components from adverse environment and thereby increase their long-
term reliability. High-performance and low-cost materials for this purpose
are the major driving forces in the 1990s.
11-888 PACKAGE SEALING AND ENCAPSULATION

14.4.1 Package Sealing and Encapsulations Material


Requirements

Electrooxidation (corrosion) and metal migration are attributed to


the presence of moisture. Generally speaking, electrocorrosion occurs
when enough moisture diffuses through the encapsulant to form a continu-
ous water path on the device in the presence of mobile ions and under
an electrical bias. Thus, a major function of an encapsulant is to serve
as an effective moisture barrier. Most organic materials are quite permeable
to moisture. Figure 14-5 shows the permeability of various materials.
Pure crystals and metals are the best materials as moisture barriers. Glass
(silicon dioxide) is an excellent moisture barrier but is slightly inferior
to pure crystals and metals. Organic polymers, such as fluorocarbons,
epoxies, and silicones are a few orders of magnitude more permeable to
moisture than glass. Obviously, gases are the most permeable to moisture
of all materials shown in Figure 14-5. In general, for each particular
material, the moisture diffusion rate is proportional to the water-vapor
partial pressure and inversely proportional to the material thickness. This
is accurate when moisture diffusion rates are in steady-state permeation.
However, moisture transient penetration rates (perhaps more important

Permeability (g/cm-s-torr)
10-6 10-8 10-10 10-12 10-14
-1 10m

..
1.0cm
it
...
II)
II)
E 1.0 mm
II)
II)
II)
C
.:Jt.
.~ 0.1 mm
.s::.
!:.
C)
0
..J
-5

1.0,..m
Min Hr Day Mo Yr 10100
Yr Yr

Figure 14-5. Effectiveness of Sealant Materials. The time for moisture to permeate
various sealant materials in one defined geometry. (From Ref. 20, ©IEEE.)
14.4 PACKAGE SEALING AND ENCAPSULATION 11-889

because they determine the time it takes for moisture to break through)
are inversely proportional to the of material thickness.
Mobile ions, such as sodium or potassium, tend to migrate to the
p-n junction of the IC device where they acquire an electron, and deposit
as the corresponding metal on the p-n junction. This consequently destroys
the device. Furthermore, mobile ions will also support leakage currents
between biased device features, which degrade device performance and
ultimately destroy the devices by electrochemical processes such as metal
conductor dissolution. For example, chloride and fluoride ions, even in
trace amounts (at the ppm level), could cause the dissolution of aluminum
metallization of complementary metal-oxide semiconductor (CMOS) de-
vices. Unfortunately, CMOS is likely to be the trend of the very-large-
scale integration (VLSI) technology and sodium chloride is a common
contaminant. The protection of these devices from the effects of these
mobile ions is an absolute requirement. The use of an ultrahigh-purity
encapsulant to encapsulate the passivated IC is the answer to some of
these mobile ion contaminant problems.
Ultraviolet-visible (UV - VIS) light radiation can cause damage to
light-sensitive optoelectronic devices. UV -VIS protection can be achieved
by choosing an opaque encapsulant. However, impurities in an encapsu-
lant, such as low levels of uranium in the ceramic or plastic packages
can cause appreciable alpha-particle radiation. Cosmic radiation in the
atmosphere can also be a source of alpha-particle radiation. The alpha
radiation can generate a temporary "soft error" in operating dynamic
random access memory (DRAM) devices. This type of alpha-particle
radiation has become a major concern, especially in high-density memory
devices. Good encapsulants must have alpha-radiation levels less than
0.001 alpha particles/cm2/h and be opaque in order to protect devices
from UV- VIS radiation. Because the alpha particle is a weak radiation,
a few micrometers thickness of encapsulant usually will prevent this
radiation damage of the DRAM devices.
Hostile environments, such as extreme cycling temperature (values
from -65°C to + 150°C in MIL-STD-883), high relative humidity (85%-
100%), shock and vibration, and high-temperature operating bias are part
of the real-life device operation. It is critical for the device to survive
these environmental influences. In addition, encapsulants must also have
suitable mechanical, electrical, and physical properties (such as minimal
stress and acceptable thermal expansion coefficient, etc.) which are com-
patible with IC devices. In addition to the above low requirements of low
moisture permeability, excellent mobile ion barrier, good UV-VIS and
alpha-particle protection, and excellent mechanical, electrical, and physi-
cal properties, the encapsulant must have a low dielectric constant to
reduce the device propagation delay and excellent thermal conductivity
to dissipate heat generated by IC devices. Furthermore, the encapsulant
11-890 PACKAGE SEALING AND ENCAPSULATION

must be high-purity material with low-ionic contaminants. Because encap-


sulation is often the final process step, and some of the devices are
expensive, particularly in the high-density multichip modules (MCM),
the ability to facilitate repair processes during production and service is
desirable. With the proper choice of encapsulant and process, the encapsu-
lation can enhance the reliablity of the fragile IC device, improve its
mechanical and physical properties, and improve manufacturing yields.

14.5 TYPES OF HERMETIC PACKAGES


No material is truly hermetic to moisture. The permeability to mois-
ture of glasses, ceramics, and metals, however, is very low and is orders
of magnitude lower than for any plastic material. Figure 14-5 illustrates
the time scales for moisture penetration for these materials. Although
polymeric-sealed devices can be designed to pass even fine leak: tests,
moisture will move through the seal in hours [23]. Hence, the only true
hermetic packages are those made of metals, ceramics, and glasses. There
are two basic types of hermetic packages in wide use today:

• Metal packages
• Ceramic packages

The common feature of these packages is the use of a lid or a cap


to seal in the semiconductor device mounted on a suitable substrate. The
leads entering the package also need to be hermetically sealed. In metal
packages, the individual leads are sealed into the metal platform or
"header" by separate glass seals. In ceramic packages, the leads are formed
by thick-film-screen and fire techniques and are commonly embedded in
the ceramic itself. Cerdip packages are unique in that the same sealing
member seals in both the cap and the leads. To satisfy the requirement
of hermeticity, cap seals in all these packages are formed by fusing a
suitable glass or metal. In brazing and soldering, a low-melting braze or
solder alloy forms the seal. In welding, the cap itself is partially melted
to effect the seal. These different types of seals are described in greater
detail later.

14.5.1 Metal Packages


Metal hermetic packages are used extensively in military applications
because of their enviable record of reliability under the harshest use
conditions. There are four types of metal package: round header (Fig. 14-
6a), platform (Fig. 14-6b), flat pack (Fig. 14-6c), and monolithic (Fig.
14-6d). The round header or the T-O packages have been in use since
the earliest days of the transistor and are still employed to package discrete
14.5 TYPES OF HERMETIC PACKAGES 11-891

(b)

or
Flatpack
Monolithic

Platform

Figure 14·6. Common Types of Metal Package

devices and smaller hybrid circuits. The platform packages provides up


to 16 pinouts, whereas the more versatile flat pack is a low-profile package
used for LSI and VLSI with up to 200 leads. The monolithic package is
a rugged package constructed of a single-piece platform providing up to
88 leads. Nearly 80% of all metal packages are welded, with the remaining
being soldered.

14.5.2 Ceramic Packages


A description of several types of ceramic packages is given in Chapter
7 ("Microelectronics Packaging-An Overview"). The brief descriptions
of some of these packages here is meant for the purpose of illustrating
different methods of package sealing.

14.5.2.1 Ceramic-Dual-in-Line Package


These were once the most widely used hermetic packages because
of the ease of their fabrication and the low cost (Fig. 14-7). In these
11-892 PACKAGE SEALING AND ENCAPSULATION

Kovar

Figure 14-7. Sealing of the Cerdip Package. (From Ref. 24.)

packages, a Kovar® lead frame is first attached to a glazed alumina


platform by temporarily softening the glass. After die attach and wirebond-
ing, the glazed ceramic cap is sealed to a subassembly in a conveyor
furnace or in the so-called hot cap sealers. The glasses used for sealing
Cerdips are high-lead-content vitreous or devitrifying glasses with sealing
temperatures of about 400cC.

14.5.2.2 Hard Glass Package


Another type of hermetic package is the so-called hard glass package
(Figure 14-8), in which a Kovar lead frame and a Kovar seal ring are first
attached to an alumina or beryllia platform using high-melting borosilicate
glass at temperatures around 11 OocC. After plating the attached lead frame
and the seal ring with suitable metals, a metal lid is soldered or brazed

Solder Kovar Lid

Glass

Kovar
Die Kovar or
Alumina

Figure 14-8. Loaded GlassIHard Glass. Typical hard glass package, which is com-
posed of the glass body through which Kovar leads penetrate. (From Ref. 24.)
14.5 TYPES OF HERMETIC PACKAGES 11-893

to the seal ring. Instead of using a Kovar seal frame, an alumina seal
frame is also often used. In this case, a glazed ceramic lid such as those
used in Cerdip sealing can be used to seal the package. Figure 14-9
compares the construction of this kind of hard glass package with that of
the more economical Cerdip.

14.5.2.3 Other Ceramic Packages


There are several other types of ceramic packages, including the
side-brazed dual-in-line package (DIP), bottom- or top-brazed chip carri-
ers, pin-grid arrays, and other multilayer ceramic packages. The DIP,
which has been the most widely used package in the last two decades,
is now losing ground to the chip carriers and to the pin-grid arrays,
including multichip, multilayer substrates. The chip carrier offers higher
lead counts than the DIP and is especially designed for economical manu-
facturing techniques such as tape automated bonding (TAB) and surface-
mounting technology (SMT).
When even higher pin counts are required, pin-grid arrays are needed.
The largest pin-grid arrays also tend to be multilayer substrates with
several layers of cofired circuit lines. In most of these packages, a metal
or ceramic cap lid is brazed or soldered to the nickel and gold-plated

Cross Section
0- Ceramic Cap
c.r.mlc~1 +-solder Glass Seal
o-ceramlc Side Wall
Solder Glass.
Seal Glass Seal
<.,q~Hard
......

Solder Glass.
Seal +- Hard Glass Seal
Metallized Area
Ceramlc~ ~ceramic Base
Base ~

(a) (b)

Figure 14-9. Alternative Packages. (a) Single seal; (b) dual seal. (From Ref. 25,
reproduced with permission from the Society of Glass Technology, England.)
11-894 PACKAGE SEALING AND ENCAPSULATION

refractory metal seal band on the substrate. Welding, glass sealing, and
metal gasket sealing are other methods used to attach hermetic caps
to these substrates. Figure 14-10 illustrates the construction of some
ceramic packages.

14.6 TYPES OF HERMETIC SEALS


14.6.1 Fused Metal Seals
Metal hermetic packages with enclosed volumes of a tenth of a
milliliter or more are commonly welded, soldered, or brazed. Ceramic
packages can also be sealed by glass sealing. To facilitate the sealing of
the ceramic substrates by soldering or welding, a metal seal band should
be provided on the substrate surface. In hard glass packages described
earlier, a seal frame made of Kovar or alloy 42 is first attached to the
substrate by using a borosilicate glass. In ceramic packages, the seal band
is formed by thick-film, cosintered molybdenum or tungsten metallurgy.
The seal frame is then suitably plated, and a metal lid is attached by
soldering or welding. The large throughputs, high yields, and reliability
associated with the welding technique is spurring a change from glass
sealing to welding for ceramic packages [24]. The major considerations
in selecting a sealing method are the availability of equipment and the
cost of the hybrid circuit. Welding is more economical because of its
speed, high process yields (approaching 100%), and reproducibility. Solder
or braze sealing is commonly employed when it is required to allow for
the ability to remove the cap and reseal the lid. Of these, the most popular
method for hermetic sealing is welding.

14.6.1.1 Soldering
The solders used for hermetic sealing are selected based on the
required temperature hierarchy for the processes that precede and follow
the sealing operation, the desired minimum seal strength, and cost. The
lid seal has to remain intact, for example, during the soldering of chip
carriers to printed-circuit boards. In this case, the solder for the seal should
have a higher melting temperature than the solder for direct mounting.
Where cap rework is required for pin-grid array packages, the sealing
solder should have a considerably lower melting point than the melting
point of the solder or braze used to attach the pins to the substrate.
Although straight tin-lead solders are widely used for hermetic sealing,
alloying additions such as indium and silver are sometimes added to
improve the strength or fatigue resistance. Use of bismuth-tin alloys for
sealing has been suggested [26]. It is found that the property of these
alloys to slightly expand on solidification (0.0005-0.0007 cm/cm) helps
to minimize the shrinkage voids in such seals [27].
14.6 TYPES OF HERMETIC SEALS 11-895

Metal
Metallization
(or Alumina) Lid
(8)

Kovar

Brazed
Leads

Metal
(or Alumina)
Lid
(b) Solder
(or Glass)
Brazed
Leads
Metallization

Kovar Multilayered
Alumina

(c)

Metallization

Die Alumina

Figure 14-10. Sealing of Other Ceramic Package Types. (a) Side braze ceramic
package; (b) chip carrier; (c) chip carrier (SLAM). (From Ref. 24.)
11-896 PACKAGE SEALING AND ENCAPSULATION

Soldering is less preferable to brazing with a eutectic gold-tin alloy


because of its lower strength (less than one-half the strength of gold-tin),
greater susceptibility to embrittlement due to intermetallic formation, and
the need to employ fluxes during sealing in most cases. The lower solder
strength translates into lower fatigue resistance, and, in fact, solder joints
in power devices have been found to fail during power cycling [27]. This,
however, is not a problem when both the cap and the substrate are made
of the same material. The softer the solder, the less likely it is to fatigue.
The brittleness of solder seals has sometimes caused seal failure
even during helium-bomb leak testing [27]. These failures were found to
be due to the formation of the AU3SI4 intermetallic through its reaction
with the plated surfaces of the cap and the substrate and its segregation
during solder spreading. It has been observed that at 240°C, a 60 : 40
Sn-Pb alloy can dissolve 10 Ilm of gold in I s [29]. The ductility of this
solder with about 8%-10% dissolved gold is practically zero. Although
pure gold coatings provide an excellent surface for soldering, the coating
thickness has to be kept to a minimum and under close control. Solderabil-
ity of bright and hard golds are strongly dependent on the alloying additions
and organic brighteners used [30]. Some of these additives release gases
during soldering, giving rise to gross porosity in the seal. Another reason
for solder nonwetting or dewetting is the formation of a gold-tin interme-
tallic compound ahead of the spreading front. A nickel-alloyed (15%-20%
Ni) gold surface, although providing good wettability, avoids the formation
of excessive intermetallics because of the slow reaction rate with
nickel [31].
The ease with which tin-lead solders oxidize makes it necessary to
handle them carefully and store in nitrogen. During sealing, a flux will
also be necessary if the solder pastes or preforms are directly used in the
operation. Flux entrapment in the die cavity is unavoidable in soldering.
Trapped flux can cause corrosion and lead to the deposition of tin on the
electrically active areas of the device [37]. Flux usage during solder sealing
can be avoided, at some cost, by resorting to the use of previously reflowed
solder. Here the solder cream or the preform is first reflowed in nitrogen
on to the seal area of the substrate [37] or of the lid [33], using a flux.
The flux and, if necessary, the surface of the reflowed solder are cleaned,
and a second reflow operation in· nitrogen is carried out to smooth and
flatten the surface of the reflowed solder. In the actual sealing operation,
the lid and the substrate are assembled under a clip or a spring-loaded
fixture and reflowed in an inert-gas ambient without using a flux. The
absence of the flux increases the criticality of other process parameters
such as the metallurgical and surface cleanliness of the sealing surfaces,
the spring loading, the furnace profile, and the ambient. The spring load
is necessary to force down the cap in order to assist the spreading of the
solder. Too high a pressure leads to solder balling and bridging [31].
14.6 TYPES OF HERMETIC SEALS 11-897

These tiny solder spheres with oxidized surfaces can cause intermittent
shorting inside the package. The furnace ambient in the sealing zone
should have no more than 100 ppm oxygen and 50 ppm water vapor.
Solder sealing is commonly performed in a conveyor furnace. A
typical furnace cycle consists of a fast preheat period (3-5 min), minimum
time (3-5 min) above the liquid's temperature, a peak temperature of
40-80°C above the melting temperature, and a fast cool-down after solidi-
fication (Fig. 14-11). The proper amount of solder, conveniently specified
as solder thickness, should slightly exceed the total camber tolerances for
the substrate and the lid. If thickness is lower, gross leaks can occur; too
much solder leads to solder balling.
Among the soldering techniques that avoid the overheating of the
device being sealed are seam soldering, which employs the same equip-
ment as seam welding, and platen or hot cap soldering [30]. The latter
offers high yields and is generally a low-cost process. In this method, a
platen with a controlled heating profile is brought into contact with the
lid while the lid, the preform, and the substrate are held in a water-cooled
fixture. The operation is performed in a controlled-atmosphere chamber.
The chamber pressure is raised slightly, when the solder melts, to counter-
act the increased pressure within the cavity. The solder fillet, commonly
formed in furnace soldering, is not formed in this method of sealing, due
to this positive pressure.
Although soldered packages are easily delidded, the rework is seldom

350
~~eak345°C 9
300 I I\..
280 C 8
/ \.
,
v

250 /I- 4 Min .J\ 7


0'
~
.,...
.. 200 / ~
6

II \
::::II 5
e!
CD
a. 150 4
.,
E
Aj \
"
t- 3
100 EntranCe Exit and

50
~eheat Cool Down
. 2

V
o
1 2 3 4 5 6 7 8 9 10
Time (Minutes)

Figure 14-11. Temperatures Versus Time Profile for Reftow Sealing. (From
Ref. 33, reprinted with the permission of Solid State Technology.)
11-898 PACKAGE SEALING AND ENCAPSULATION

carried out more than once. This is because of increased gold leaching,
dewetting, and the need to use higher and higher temperatures for each
successive seal because of solder alloying characteristics [34]. Softer
solders allow for more reworks. Gross leakers can often be salvaged by
reflowing at a higher temperature.
Solder sealing of large-cavity packages presents special challenges
because of the large volume of heated gas inside the package, which tends
to form blow holes in the molten solder or draw solder inside as solder
balls. One of the largest solder-sealed packages (64 x 64 mm) is the
multichip ceramic substrate used in the IBM 4341 System [31].

14.6.1.2 Brazing
Brazing with a eutectic (80 : 20) Au-Sn alloy is used in place of
soldering when the need exists for a stronger, more corrosion-resistant
seal and where the use of flux has to be avoided. The braze is usually
used as a thin, narrow preform tack-welded to a gold-plated Kovar lid.
The metal seal band on the substrate is also gold plated for good wettability
and corrosion protection. In furnace sealing, the typical reflow time is
2-4 min above the eutectic temperature of 280°C with a peak temperature
of about 350°C. Other methods of sealing used for soldering can also be
used with advance for braze sealing.
The very steep slope of the liquidus curve in the Au-Sn system on
the gold-rich side of the eutectic has given rise to many a hermeticity
failure. An increase in just 3-5% in gold content above the eutectic
composition raises the liquidus temperature from 280°C to 450°C. This
increase can easily and rapidly occur by the dissolution of the gold from
the plated surfaces. It is suggested that the problem can be avoided by
using a brazing alloy slightly richer in Sn than the eutectic, such as a
78 : 22 alloy [36]. This entails a slightly higher sealing temperature, but
in this case, the dissolution of plated gold would actually lower the liquidus
and facilitate wetting of the sealing surfaces.
The factors that influence the formation of a reliable seal during
brazing are summarized in Table 14-4. These include the flatness of
substrate and lid, furnace temperature profile, and atmosphere control
[35]. Furnace sealing with AuSn has been reported to provide good yields
with combined fine and gross leak losses of 2% or less [33].

14.6.1.3 Welding
The most popular method for sealing high-reliability packages, such
as those used in military applications, is welding. One survey indicated
that about 80% of military packages are welded [34]. Despite the higher
initial cost for equipment, welding is popular because of the high yields
and a good history of reliability. In welding, high-current pulses produce
14.6 TYPES OF HERMETIC SEALS 11-899

Table 14·4. Factors Affecting Brazing

Impact
Parameter Too Much Too Little Magnitude

Metallurgical
Lid and package Gold leaching raises Difficulty wetting Moderate
gold plating melting temperature
Preform volume Unnecessary wetting of Difficulty obtaining High
surrounding areas complete seal
Packagellid Cannot be too clean, but Solder wetting Moderate
cleanliness too aggressive cleaning inhibited
methods can produce
corrosion
Package thermal Poor wetting caused by None High
history prior to base metal diffusion
lidding

Mechanical
Lid and package Cannot be too flat Thicker solder preform High
flatness and/or use spring
clips
Package lid and Longer furnace preheat None Moderate
fixture mass and/or longer furnace
dwell
Lid compliance Easy deformation from Difficult to Moderate
external forces compensate for lack
of flatness
Preform attach Tacked or bonded; Chance of preform Moderate
means easy handling misalignment

Furnace
Belt speed Packagellid mass cannot Silicon component Moderate
reach melting damage. Excess
temperature solder flow.
Peak Solder craws up on lid Proper solder flow not High
temperature Silicon component achieved
damage
Fillet appearance Hermeticity probable Hermeticity Moderate
questionable
Atmosphere Poor wetting None High
Moisture and
oxygen content
Nitrogen flow rate Wastes nitrogen Poor wetting Moderate

Source: From Ref. 35.


11-900 PACKAGE SEALING AND ENCAPSULATION

local heating between 10000 and 1500°C, fusing the lid or the plating
thereon to the package. The local heating prevents damage to internal com-
ponents.
In parallel seam welding (Fig. 14-12), also called series welding,
the package and the lid are passed under a pair of small tapered copper
electrode wheels. The transformer produces a series of energy pulses,
which are conducted from one electrode across the package lid to the
other electrode. The generation of heat at the wheel-lid interface leads
to the formation of either a weld or braze type of seal, the nickel or gold
plating acting as the brazing medium [37].
In the opposed electrode welding (Fig. 14-13), the package to be
sealed is moved under a pair of larger electrode wheels. The power ply,
usually a capacitor discharge, produces a series of welding pulses, which
pass from the electrode wheel across the lid-package sidewall interface
and return via the workbench, that generate heat at the electrode-lid
interface and the lid-package sidewall interface.
Welding can accommodate greater deviations from flatness for the
package and the lid than soldering or brazing. The higher temperatures
at the welding interface can volatilize most contaminants so that cleanliness
is not as critical factor as it is in soldering and brazing.
Other less common methods for package welding are electron beam
(e-beam) welding and laser welding. Laser welding oflarge metal packages
is attractive to resistance welding because of its high speed, very limited
heat input to sensitive areas, ability to handle unconventional seal geome-
tries, and noncontract nature. Continuous-wave (CW) lasers have been
found to cause solidification cracking in the weld, whereas the cracking

AC Welding Transformer

Copper Electrode

Figure 14-12. Parallel Seam or Welding


14.6 TYPES OF HERMETIC SEALS 11-901

...------+ Welding Transformer (D.C. Capacitor Discharge)

Copper Electrode

Figure 14-13. Opposed Electrode Welding

tendency was significantly lower with pulsed lasers that fonn a series of
overlapping welds. The cracking tendency also depends on the nature of
the metal plating in the weld area, with electroless nickel-plated surfaces
prone to most cracking than electroplated nickel [37]. Gold plating tended
to cause gross leakers and excessive weld splatter according to one report
[38], whereas another report found gold plating helped the fonnation of
a sound weld [37]. The availability of reliable, versatile laser equipment
should make laser welding more popular in the future. Both CO2 and
Nd-YAG lasers are used in laser welding of hennetic packages.
Delidding of welded packages is most commonly carried out by end
milling or by precision sawing [34]. Before resealing by welding, the seal
flange has to be reground. Although delidding of a welded package is
quite common during development and the prototype-production stage,
it is usually not used for production packages because of the prohibition
on the use of delidded and resealed packages by military specifications.

14.6.1.4 Glass Sealing

Glasses have been used in semiconductor package sealing since the


beginning of the transistor age. The first use of glasses was in device
passivation, and, to this day, glasses continue to be used for this important
purpose, which serves as the last line of defense for a semiconductor
device against external moisture and other contaminants. In this section
we only consider the use of glasses in hennetic package sealing.
11-902 PACKAGE SEALING AND ENCAPSULATION

Glass sealing is encountered in a wide range of package types, from


the earliest T-0 headers used to package the first transistors to the latest
ceramic packages. In the former, glass is used to form hermetic glass-to-
metal seals for the package leads fed through holes in a metal plate or
header. In the latter, the glass is used to form a sandwich seal between
a ceramic cap or lid and the ceramic substrate on which the devices
are mounted. In Cerdip packages, the same glass seal performs both
these functions.
The versatility of glasses in package-sealing applications resides in
their chemical inertness, oxidation resistance, good electrical insulating
characteristics, impermeability to moisture and other gases, and the wide
choice of thermal characteristics that can be obtained. The principal short-
comings of glasses-their low strength and brittleness-make it impera-
tive to pay particular attention to seal design, choice of sealing glass, and,
to some extent, the sealing process used. An excellent review of stresses
in different glass-seal configurations has been given by Varsheneya [39]
from which the following discussion is drawn.
Glass seals fail for one or more of the following reasons:

1. Poor adhesion between glass and the surfaces to be sealed


2. High stresses in the glass leading to its fracture
3. Fracture due to applied stresses in handling or use
4. Failure in the metal or ceramic parts being sealed

Poor adhesion in glass sealing, although a common problem with


glass-to-metal seals, is seldom a problem in glass-to-ceramic seals. There
is extensive literature on the mechanisms of glass-to-metal adhesion and
the methods for promoting the same for specific metals. For good adhesion,
the glass should, in its molten state, exhibit good wetting to the metal
surface, which, in tum, promotes a strong chemical bond, mechanical
interlocking, or both. Wetting of liquid to a solid surface is governed by
the drive to minimize the interfacial tensions in the system. The equilibrium
condition for the wetting is given by the classic equation

s'Yg = sYl + l'Yg cos e [14-2]

where
'Y = interfacial tensions
e = contact angle
I, s, g = liquid, solid, and gas, respectively.
14.6 TYPES OF HERMETIC SEALS 11-903

The lower the contact angle, the better the wetting. The chemical
basis for good wetting is explained by the so-called oxide-saturation theory
of Pask and co-workers [40,41].
Stresses in a seal are the result of the differential thermal contraction
between the sealing components from the temperature of sealing down
to room temperature. This stress can be calculated by

[14-3]

where
Aixy = function of the elastic properties of the glass and metal and the
seal configuration
a g, am = thermal contraction coefficients of the glass and the metal, re-
spectively
Ta = set point of the glass
Tr = room temperature.
The set point is generally taken as the temperature 5°C above the strain
point (log 11 = 14.5, where 11 is the viscosity in poises) and 15°C below
the annealing point (log 11 = 13.0). It is observed that for vitreous glasses,
the set point corresponds to the temperature at which log 11 = 13.7. It is
recommended that !la = (ag - am) should be below 500 ppm for good
seal integrity.
The most common glass-to-metal seal in microelectronic packages
is the so-called bead seal used to seal the feedthrough leads into metal
headers. The glass seal also acts to insulate the leads from the header.
These seals are of two types-matched or compression seals [42].
In matched seals, the thermal coefficients of expansion of the glass
and the metal involved are similar. Sealing is dependent on the formation
of a chemical bond between the glass and the metal. To ensure the latter,
the metal surface is provided with an adherent oxide layer. The optimum
methods for providing such an oxide layer on Kovar alloy, widely used
in metal packages, is described in the literature [42-44].
The compression seal does not involve the formation of a chemical
bond between the glass and the metal. In this case, the glass is chosen
such that its expansion coefficient is lower than that of the metal. When
the assembly is sealed at the melting point of the glass and cooled, the
metal shrinks more from the sealing temperature and holds the glass
tightly in compression. Although this type of seal is generally hermetic
and stronger than the matched seal, it is not as thermally reliable as
the latter.
11-904 PACKAGE SEALING AND ENCAPSULATION

The normal glass-to-Kovar seal forms a meniscus on the leads, which


signifies good wetting. However, too long a meniscus is known to cause
a problem because it can be easily broken off, thus causing hermeticity
failures or the exposure of unplated Kovar lead areas, which then readily
corrode [24]. A polymeric coating applied to the base of the leads has
been shown to be effective in preventing the damage to the meniscus by
providing mechanical port to the leads in bending [46].
The resistance of the bead seals to fracture upon sudden temperature
variations is tested by a standard thermal-shock test prescribed in the
MIL-STD-883. The test involves alternatively immersing the packages
in a cold fluid and a hot fluid 15 times, allowing time in each for the
package temperature to equilibrate. It has been reported [3] that although
the package may be hermetic before and after the thermal shocks, it may,
sometimes, lose the hermeticity during the test itself. The reason for this,
if true, is obscure. The strength of matched bead seals has been determined
to be about 7500 psi [46].
The sandwich seal is the commonly encountered glass-to-ceramic
seal in microelectronic packages. These include the low-temperature glass
seals in Cerdip packages and the hard glass seals of alumina seal frames
to alumina or beryllia substrates with borosilicate glasses.
Most of the soft glass seals are made from lead-zinc-borate glasses.
They generally permit sealing below 420°C. These glasses are character-
ized by a low softening point and low viscosity at the sealing temperature,
as well as by a fast -setting viscosity curve. Other required attributes are low
water content, good chemical durability, and thermal expansion closely
matched to that of the ceramic. Some of these requirements are hard to
come by in low-temperature sealing glasses. The thermal expansivity of
lead-zinc-borate glasses range between 80 x 1O-3/ o C and 100 x 1O-7/ o C,
compared to a value of 65 x 1O-7/ o C for alumina and (50-55) x 1O-7/ o C
for Fe-Co-Ni alloys commonly used in packaging. As such, the stresses
in these seals would be extremely high. Hence, these glasses are modified
by devitrification and through the addition of certain low-expansivity
fillers such as fused silica and beta-eucryptite [25]. Addition of titania to
glass composition has been shown to promote the devitrification of the
glass as well as the formation of lead titanate crystals having anomalously
low expansivity and thereby lowering the overall expansivity of the seal
[48]. Typically, the thermal expansivity of the seal glass ranges from 70
x 1O-7/ o C to 80 x 1O-7/ 0 C.
It has been noted that in the high PbO region of lead-zinc-borate
system, the tendency to devitrify increases rapidly between 5% and 20%
ZnO [25]. Devitrification in these glasses is said to occur by homogeneous
nucleation despite the availability of enormous surface area for surface
nucleation in the powdered glass. Increases in strength and decreases
in thermal expansivity generally accompany devitrification, both factors
improving the seal integrity.
14.6 TYPES OF HERMETIC SEALS 11-905

A major problem in sealing Cerdip packages is the evolution of


moisture from the sealing glasses. Water is physically and chemically
absorbed on glass surfaces, included in internal pores, and structurally
combined with the glass network. The seriousness of this problem is
illustrated by the calculation that the evolution of only 0.15 Ilg of water
into the cavity will provide 5000 ppmv of water vapor to 0.04 cm3 volume
of a 16-lead Cerdip and that this amount of moisture is routinely found
in package sealing glasses [49]. Figure 14-14 shows the moisture evolution
behavior of certain sealing glasses studied by the moisture evolution
analysis (MEA) technique [47]. Three distinct regimes of moisture evolu-
tion are noted. At around 210°C, the physically absorbed moisture is
evolved. The peak around 370°C, near the softening temperature of the
glass, is attributed to the evolution of moisture from the internal pores
in the sealing glass. At still higher temperatures, in the sealing regime,
the chemically combined water is expelled. In an actual sealing operation,
some of this moisture gets released into the sealed cavity. The authors
were able to show a direct correlation between the moisture measured in
the sealing glasses by the MEA technique and the moisture levels actually
measured in the hermetic cavities sealed using the same glasses measured
by a residual gas analysis (RGA) technique, thereby establishing a monitor-
ing technique for moisture levels in Cerdip. It has been shown that the

MEA Ambient: 02
Glass: XS1175·M1

Region I Region II Region III

Q.
E
~

o 5 10 15 20 25 30 35 40 45 50
Time (Minutes)

Figure 14-14. Moisture Evolution from Sealing Glasses. (From Ref. 47, reprinted
with permission from Journal of Electronic Materials, a publication of the Metallurgical
Society, Warrendale, PA.)
11-906 PACKAGE SEALING AND ENCAPSULATION

microporosity in the sealing glasses plays a major role in determining the


extent of moisture release into the package even when the glass is used
in the preglazed condition [49]. Furthermore, a prebaking in vacuum of
the glazed lid at above the softening temperature made a significant
difference in the amount of moisture actually entrapped in the package
during sealing (4500-6000 ppm for nonprebaked lids to 200-900 ppmv
for prebaked lids). Heating to below the softening point during baking
was found to be not as effective because the internal porosity still retains
the remaining reabsorbed moisture.
Although devitrifying solder glasses offer a better seal integrity, they
are also associated with greater degree of moisture within the sealed
package compared to packages sealed with vitreous seals. In one study
[50], about half of the packages sealed with a devitrifying glass were
found to contain over 6000 ppmv water vapor. Of those packages sealed
with a nondevitrifying glass containing an aluminum silicate additive,
over 90% had enclosed moisture level of less than 500 ppmv, with all of
the packages having moisture level below 2000 ppmv. The principal
reason for this difference is that crystallized glass has much lower solubility
for water than in its glassy state, so that almost all of the chemically
combined water gets expelled, some of it into the package cavity, when
devitrification occurs during sealing.

14.6.1.5 Glass-Sealing Methods

Furnace sealing is the most common method for glass sealing. The
key factors to be controlled are the furnace ambient and the temperature
profile. Heating is provided by conventional heaters as well as by infrared
(lR) heaters. The lids and the ceramic are usually preglazed with sealing
glass A well-controlled glazing process is one that yields good seal dimen-
sions on the ceramic, and good wetting and adhesion to the ceramic and
produces little or no prenucleation in the case of devitrifying glasses.
Prenucleation causes premature crystallization during the subsequent seal
step, thereby reducing the flow, causing insufficient wetting as well as
porosity. Differential thermal analysis (DTA) has been found to be a
useful tool in determining the occurrence of prenucleation during the
glazing process [52]. It is noted that although a lODC lowering of the
crystallization peak temperature from its normal position, as measured
by DTA, may be tolerable, a 20 DC drop would cause definite yield prob-
lems, and a 30-40DC drop could reduce yields by up to 50%. Prenucleation
can be pressed by heating rapidly, consistent with the need to remove the
organics, during glazing. The same is true during sealing. A heating rate
of 75-125 DC/min is recommended. For sealing, a lO-20-min soak at the
peak temperature is required, after which the package is cooled slowly
at about 40 D C/min.
14.6 TYPES OF HERMETIC SEALS 11-907

Generally, seals are 250-400 11m thick and about 1000 11m wide.
A typical furnace profile for glazing and sealing of a semiconductor
package sealing glass is shown in Figure 14-15.
In cases where exposure to the relatively high glass-sealing tempera-
tures cannot be tolerated, the hot-cap sealing method has been employed.
Here, the lid is heated to a temperature above the melting point of the
glass while the substrate is kept at a lower, safer temperature for the
device. Even though the substrate is also preheated to avoid a large
temperature differential across the interface through sealing glass, the
process entails a relatively fast cooling of the seal and, consequently,
larger stresses in it. This has sometimes led to seal fracture [51].
A so-called graded viscosity seal is aimed at reducing the danger
of moisture entrapment during the sealing process [51] (Fig. 14-16). Here
the lead frame is first sealed to the substrate using a devitrified glass. The
package is then heated in nitrogen and transferred into a hot-cap sealer
in a dry box. The sealing surface of the substrate slopes outward toward
its edge so as to favor the outgassing of moisture from vitreous sealing
glass away from the package interior.
Another novel sealing method uses a beam of focused infrared light
to heat the seal area of a preglazed lid and substrate assembly [51].
The glasses used are specially formulated to absorb infrared radiation.
Although the seal area gets heated to about 400°C, very little heating of
the interior occurs.
Furnaces and other equipment designed for use in hermetic sealing
are described in the literature [51,53,54].

14.6.2 Gasket Seals


One clear trend in packaging for large mainframe systems is to
use multilayer substrates capable of accommodating many semiconductor
chips. In such packages, the sealing cover is generally a part of the
package-cooling system. The cover encloses a large volume and will

500
U Sealing Profile ... _____ _
• 400
!
.al! , "",
300
" -----~
&200 I
I

/ Glazing Profile
~
t-
100 ,--'
10 20 30 40
Time (Minutes)

Figure 14-15. A Typical Sealing/Glazing Profile for Cerdips. (From Ref. 51.)
11-908 PACKAGE SEALING AND ENCAPSULATION

1S:s:~;:S::S:ri~~~~~!~iE~~~:: Package Lead


....- - Ceramic Base
~~~~~~~~~

Die Attach Surface

Die

• sssssss~
;.=
. . . .0.=:=
.:.:::::: ............... :::::.:.: .
1·············1
::::::::::::: ......... ::::::.:.:.:.
S'SSSSS •

Figure 14-16. Graded Viscosity Seal Package. GVSP features elimination of outgas-
sing within the package. (From Ref. 51.)

have a large seal perimeter. To allow for repair or replacement of the


semiconductor chips during assembly, test, or service, the cover should
be easily demountable and resealable. The seal should also be capable of
accommodating the large difference in the thermal expansivities between
the copper or aluminum covers, chosen for their mechanical and thermal
properties as well as for the ease of their manufacture, and the ceramic
substrate. These considerations clearly rule out fused metal or glass seals.
Gasket seals are ideal for this purpose, but traditional rubber gaskets will
not provide for a hermetic seal due to the high permeability of these
materials to moisture and other gases. However, such gaskets made of a
malleable metal are hermetic if sufficiently high pressures are applied and
maintained over the seal area. Copper gaskets are commonly used in
ultrahigh-vacuum systems, whereas C rings of stainless steel are used in
systems operating at high pressures and temperatures. To improve the
seal reliability, these rings are covered with a soft metal such as lead
or aluminum.
In gasket seals, the possible leak paths are (1) gas permeation through
the gasket material itself and (2) the leaks at the contact surfaces. The
first consideration eliminates rubber and polymeric materials for gaskets.
14.6 TYPES OF HERMETIC SEALS 11-909

Gaskets made of metal are suitable because of the very low permeabilities
of metals to gases. The sealing load required for a given maximum
acceptable leak rate should be sufficient to elastically and plastically
deform the metal to provide a close conformal fit to the microtopography
of the surfaces. This load depends on the roughness of the sealing surfaces,
the elastic modulii, the yield strength, and the work-hardening characteris-
tics of the gasket material, the friction between the mating surfaces, and
the geometry of the gasket. The leak rate of gases through the interfaces
of a gasket seal is given by the following equation [55].

where
oa = the applied average normal sealing stress
0y = the yield stress of the gasket metal
A, b = constants dependent on the nature of the gas, the geometry of the
gasket, and the temperature
C = the leak conductance
r = mean roughness of the sealing surfaces

It is reported that the leak rate through interfaces of metal gaskets decreases
very sharply as the average stress over the seal area reaches a critical
value lying between 1.5 and 2.5 times the yield strength of the softer
metal [55]. The highest stress ratio was required for radially ground
surfaces (1 Ilm rms), whereas the lowest sealing stress was required for
circumferentially machined surfaces (0.70"y, 2.5 !lm rms). Among the
various cross sections of the gaskets, the circular cross section requires
the highest compression ratio, whereas the square cross section require
the lowest compression ratio.
The specific elastic and plastic properties of the gasket material must
require sealing load levels that are compatible with the strength, toughness,
and the static fatigue limit of the ceramic. To keep the sealing stress at
safe levels, softer metals like lead or indium could be used as gasket
materials. By themselves, these materials will creep under load and relax
the sealing stress below that needed to assure effective sealing. The
optimum gasket must have both a low-yield stress at the contact surfaces
and a large elastic compliance. This unique combination of properties is
achieved using a composite gasket. Typical seal loads for metal seal ring
with a tubular diameter of 30 Ilm and a wall thickness of 1.5 Ilm is about
30 kg/cm [56].
A gasket seal, such as described above, is utilized in a complex,
multilayer package used in IBM's 3081 computer systems [57]. The
multilayer ceramic package consists of over 30 metallized layers and
/1-910 PACKAGE SEALING AND ENCAPSULATION

Hat---v

/.r-r-74-- Seal Ring


Baseplate
Substrate

Kovar
Frame

Figure 14-17. Metal CoRing Seal to Brazed Kovar Frame

carries over 100 high-performance chips. Each chip is contacted by a


spring-loaded metal piston embedded in a water-cooled housing or hat
that is filled with helium. This arrangement is able to dissipate up to 600
W. In the early versions of this module, the hat assembly was clamped
to a Kovar seal frame brazed to the cofired molybdenum seal ring on the
bottom surface of the substrate (Fig. 14-17). In a latter version, the hat
assembly was directly clamped to the machined ceramic surface of the
substrate (Fig. 14-18). In both cases an Inconel Coring was used.

14.7 TESTING OF HERMETIC PACKAGES


It is implicit that when a package is tested to be hermetic, no other
tests are needed to assure its reliability. Hence, 100% leak testing is the
principal test for hermetic packages. However, it is not uncommon to
subject hermetic packages to temperature-humidity tests as well as to
certain thermal and mechanical stress tests aimed at determining seal

Hat ---004

/ .,,-~.-- Seal Ring


Baseplate Substrate

Figure 14-18. Metal CoRing Seal Directly to Substrate Surface


14.7 TESTING OF HERMETIC PACKAGES 11-911

integrity. It is also often required to determine contaminant levels within


the package as part of a quality-control measure or of a failure-analysis
effort. Only the methods used for leak testing and for moisture-level
measurement within sealed packages are briefly reviewed here. The reader
is referred to MIL-STD-883B and other relevant specifications for a more
detailed description of the leak-testing methods and of thermal and me-
chanical characterization of seals.

14.7.1 Leak Testing

The practice of leak testing of hermetic packages has been described


in a recent article [58]. The National Bureau of Standards (NBS) Special
Publication 400-9, entitled Hermeticity Testing for Integrated Circuits,
reviews current problems in leak testing [59].
The standard leak rate (L) is defined as that quantity of dry air at
25°C, measured in atmospheric cubic centimeters per second (atm cc/s),
flowing through a leak path or paths where the high-side pressure equals
1 atm (760 mm Hg absolute) and the low-side pressure is 1 torr (1 mm
Hg absolute). Helium leak rates, Rl> measured under the specified test
conditions, converted into the rates that would have been measured under
the above standard pressure gradient, are termed the equivalent standard
leak rates.
The spectrum of leak rates of interest is broadly divided into two
regimes. Leaks in the range 10-1_10-4 are termed gross leaks, whereas
those in the range 10-5_10- 12 are termed fine leaks. Leak rates of 10-4-10-5
are not easily characterized solely by methods used for either fine or gross
leak detection.

14.7.1.1 Gross Leak Testing

Gross leaks are tested by using a fluorocarbon liquid. In the bubble


technique, the package to be tested is first subjected to a vacuum-evacua-
tion step C~5 torrs, I-h dwell) to facilitate easy entry of the test liquid if
a leak exists. Following the evacuation soak, the package is immersed in
3M Electronic Product Division's Fluorocarbon liquid (FC-84) without
breaking the vacuum. It is left to soak in it for 30 min. The "filled"
packages are then immersed in another fluorocarbon liquid (FC-40) having
a higher boiling point than FC-84 and maintained at a temperature above
the boiling point of FC-84. Any FC-84 that had leaked into the package
during fill phase will now expand in volume and appear as bubbles
observable through a magnifying window in the FC-40 liquid container.
An alternative method, called the negative ion detection (NID)
method, uses a gas analyzer specially tuned to detect FC-84 molecules
from packages filled with the latter in a manner similar to that used in
11-912 PACKAGE SEALING AND ENCAPSULATION

the bubble method. The "filled" packages are heated in a metal chamber
maintained at 125°e and attached to the gas analyzer.

14.7.1.2 Fine Leak Testing


The common method of testing for fine leaks uses helium gas as
the testing medium. First, the package is subjected to helium gas under
pressure in a pressure chamber or "bomb." Typically, the pressure used
is 60 psig. and the dwell time is 2 h. After the pressurization step, the
packages are introduced, either singly or in small batches, into a vacuum
chamber attached to a helium mass spectrometer. If any helium evolution
is detected during the evacuation step, the whole batch is rejected.
Another method used for fine leak testing also serves to test for
gross leaks. Here, the package is subjected to pressurization in dry nitrogen
gas into which is mixed about 1% Kr-85 radioactive tracer gas. After the
pressurization, the package enters a scintillation counter, which detects
the gamma emission that accompanies the beta decay of Kr-85. Because
the gamma rays can easily penetrate the package walls, the scintillation
count directly yields a measure of the Kr-85 concentration within the
package.
It has been pointed out that the helium leak rates measured in this
test could be orders of magnitude larger than helium leak rates that would
occur under use conditions-l atm pressure in air [21]. This difference
is due to the predominant contribution of viscous flow of the gas during
bombing and subsequent testing in vacuum and to the absence of viscous
flow under use conditions. Further, the penetration of moisture into the
package could be significantly slower than for He because of the tendency
of the former to absorb on the walls of the leak path.

14.7.2 Moisture Measurement and Monitoring


As discussed previously, the actual moisture content within the pack-
age is made up of contributions from moisture released from the sealing
materials and the package itself, the sealing ambient, and the moisture
leaked in through the seal. Thus, it is necessary to measure the actual
moisture level within the package on a sampling basis, to make sure that
the sealing is being performed under the desired conditions. The moisture
content within the package is also of interest for investigating the causes
of failure of hermetically sealed devices.
Measurement of moisture level within a package is most commonly
carried out by residual gas analyzers (RGA), which use a mass spectrome-
ter to detect and measure water vapor. (See, for example, MIL-STD-883
Method 1018-2.) This technique gives a quantitative measure of the total
moisture within the package. The package is broken or punctured in a
heated vacuum chamber to release the moisture, which is then led into
14.7 TESTING OF HERMETIC PACKAGES 11-913

the mass spectrometer. Because of the extremely small amount of moisture


that is being measured, great care is needed in setting up the mass spectrom-
eter, in breaking the package, and in the interpretation of data [3,60]. A very
highly sensitive method for measuring moisture content from individual
packages using plasma chromatography has also been described [61].
Advantages of the mass spectrometer technique include its widespread
use and acceptance, its great sensitivity, its quantitative nature, and the
ability to simultaneously measure other volatile constituents of the package
besides water. Disadvantages of the RGA method are that it is a costly,
time-consuming, and destructive method and requires well-trained person-
nel to conduct the tests.
A reliable method for determining moisture-penetration rates in-
volves circulating dry He through the package placed in a controlled
humidity chamber. The moisture picked up by the circulating gas is
continuously monitored using a P205 cell [20]. As discussed previously,
moisture levels in Cerdips can be monitored by determining the moisture
content of sealing glasses measured by a moisture evolution analyzer [47].
A nondestructive method for continuously monitoring moisture lev-
els within sealed cavities utilizes several types of in situ monitors. In the
commercially available humidity sensors or humistors, chemically altered
polymers or oxides are used to absorb the moisture in their ambient,
which, in tum, causes changes in their surface or bulk electrical characteris-
tics. These humistors, once calibrated, can be sealed into the package to
continuously monitor the moisture level within the package.
One type of moisture sensor that is best suited for use in microelec-
tronic packages consists of a porous aluminum oxide layer sandwiched
between two electrodes-an aluminum base and a porous gold electrode.
The porous oxide layer is grown by the anodization of aluminum electrode.
The thin, evaporated gold electrode is formed on the porous oxide layer
and is permeable to moisture. As the moisture content in the package
changes, it is reversibly absorbed and des orbed by the porous oxide,
thereby modifying the cell's impedance. Because the entire body of the
porous aluminum oxide participates in absorbing the moisture, these sen-
sors are said to have "volume effect" and are very sensitive. This type
of humistors have a measurement range from 1 to 10,000 ppm moisture
level and a sensitivity of about 1 ppm (Fig. 14-19). Earlier versions of
this moisture sensor were made on aluminum foil tabs. Newer, more
rugged sensors of this type constructed on silicon wafers using thin-film
techniques offer improved reliability and enable further miniaturization
down to a 15 x 15 J.lm size. They are also shown to be stable up to 500°C
and exhibit little or no thermal cycling or hysteresis effects. The humistor
readings have been shown to correlate well with the industry standard
method of mass spectrometric measurements [62,63].
Another type of moisture sensor used in the microelectronics is the
11-914 PACKAGE SEALING AND ENCAPSULATION

Permeable Gold

AI

Co =Capacitance of Oxide Layer


R1 = Pore-Side Resistance
C2 = Pore Base Capacitance
=
Ro Resistance of AI2 0 3
R2 = Pore-Base Resistance

Figure 14-19. Al20 3 Moisture Sensor. (a) Cross section; (b) equivalent circuit. (From
Ref. 62, ©IEEE.)

dew-point sensor. The operating principle of this sensor is the occurrence


of a large decrease in resistance between closely spaced electrodes at the
dew point. In principle, the sensor is used to measure the dew point within
the package as long as the condensing phase is liquid. The sensor is
formed as interdigitated thick- or thin-film electrodes on a chip surface
(Fig. 14-20). As the package is slowly cooled, the resistance between the
electrodes is monitored. A sudden drop in resistance signals the moisture
condensation and, hence, the attainment of dew point. Unlike in the case
of the Alz0 3 sensor, only the surface of the sensor absorbs moisture and
causes the measured change in the resistance. For this reason, its detection
limit and sensitivity are not as good as for the former.
In both of the above cases, the sensors have to be specially fabricated
and enclosed in the package. It has been shown that adjacent metallization
tracks on the silicon device itself can be used to measure the dew point
[64]. This method makes use of the fact that the stray capacitance between
the conductors increases at the moment the chip surface reaches the dew
point. In the absence of condensed moisture (Fig. 14-21a), the stray
capacitance builds up partly through the chip and partly through the
surrounding air. When moisture condenses (Fig. 14-21b), a conducting
water film forms on the insulator and the capacitance increases. The actual
change that occurs is of the order of a picofarad (pF) and can be measured
by a stable capacitance bridge.
The measurements of dew point made using this method have been
made shown to agree with those made by the surface-conductivity method.
In another version of this type of sensor, the need for cooling the package
to the dew point is avoided. Here, the frequency dependence of the
14.7 TESTING OF HERMETIC PACKAGES 11-915

Pads

I
J
I 11-----_ _ _ _--,
J
I 1
1

I 1
1

I 1
1

Figure 14-20. Interdigitated Dew-Point Sensor. (From Ref. 62, ©IEEE.)

capacitance between the two metallization tracks is measured at one


temperature and is determined over a range of frequencies. For a given
moisture level, a sudden change in resistance occurs at a unique frequency.
Once calibrated, the sensor can be used to monitor changes in the moisture
level within the package [65].
The difficulties involved in leak: testing, on the one hand, and the
need to determine the actual level of moisture within the package, on the
other, have made it desirable to incorporate moisture sensors within the

Cp
,----I~--
•I

Dry Condition After Condensation

Figure 14-21. IC Chip Cross Section. (From Ref. 64.)


11-916 PACKAGE SEALING AND ENCAPSULATION

hermetic cavity of the package. The availability of miniature sensors of


high reliability and accuracy, such as those described here, makes it
attractive to use them routinely, at least in large packages.

14.8 RELIABILITY TESTING


To ensure the long-term reliability of the packaging sealing and
encapsulation, reliability testing is an integrated part of the manufacturing
process. Effects of moisture, mobile ions, and so forth, are directly affect-
ing the reliability and accelerated testings, such as the conventional 85°C/
85% rh, and the highly accelerated stress test (HAST) are critical to
predict this manufacturing process. These will be discussed as follows.

14.8.1 Correlation of 85°C/85% rh to HAST


Reliability testing results are used to give a degree of confidence of
a device's field reliability to the electronics merchant and the customer.
The only way this degree of confidence can be given is if there is some
correlation between the accelerated test time-to-failure (TTF) date that
has been obtained from qualification testing, and the expected field TTF
rates. This can be accomplished by formulating reliability models that
relate accelerated environmental tests to actual field life. The objective
for any reliability model is to be able to estimate device performance
accurately outside the range of experimental observations. This can be
done in either of two ways: relate the accelerated test TTF results of a
specific device to that of the same device at different accelerated test or
field conditions; or through comprehensive, fundamental modeling of the
physical activities at the molecular level leading to degradation and failure.
The latter technique is commonly termed the "physics-of-failure" ap-
proach. The former of the two approaches incorporates statistically sig-
nificant data obtained from well-planned and consistently executed envi-
ronmental test evaluations to form a probabilistic model allowing
extrapolation to other use conditions. Although the physics-of-failure
approach is the most desirable because of its fundamental nature, models
based on these theories are still at a research level in industry and are
just beginning to be investigated in-depth. This section summarizes the
evaluation of two existing predictive models for the time-to-failure of
nonhermetic packages subjected to constant temperature and relative hu-
midity. The major conclusions of that analysis are summarized here.
Two of the leading candidate models, one proposed by Peck and
the other by Denson and Brosius, are essentially empirical in their formula-
tion. To assess the relative consistency of the two models, their functional
forms are considered and compared relative to a given set of data. Again,
the drive behind these two models and others like them is the need to
extrapolate beyond the range of available observations. The main applica-
14.8 RELIABILITY TESTING 11-917

tion appears to be for the estimation of acceleration factors from laboratory


test conditions, in which temperature and humidity are much higher than
operational, or field, conditions. If the acceleration factors are reasonably
accurate, tremendous savings in time and testing costs can be realized.
If the extrapolation holds into the range of operating conditions, then the
model could have great impact for long-term reliability estimations and
the cost of reliability testing.
The underlying assumptions in the Peck model are that component
TIFs are log-normal random variables and that their variances are equal
and constant over a wide range of temperatures (0C) and relative humidities
(% rh). With these assumptions, an estimated sample median TIF, tm
eC, % rh) was obtained, Peck's model for this sample median is

[14-4]

where
k = Boltzmann's constant, 8.615xlO-5 eV/K
A = a scaling constant
n, Ea = model parameters.

No doubt, much of the success of the model is due to the fact that a
power law for relative humidity and an Arrhenius law for temperature
seems quite reasonable.
Because 85°C/85% rh (85/85) is an industry standard test condition,
Peck defined his observed ratio, Ro, between different tests conditions to be

[14-5]

Ro was compared to a calculated ratio, Reo derived from tm (OC, % rh) in


a similar manner to Ro. Specifically, Rc is found to be

[14-6]

If the graph of Ro versus Rc is linear with slope I and intercept 0,


then the model is an excellent representation of the experimental data.
From a linear least-squares analysis, Peck has estimated the parameters
to be n = 3.0 and Ea =0.90 eV.
The data used by Peck for his model formulation are shown in Table
14-5. A rather obvious result is that the estimates for the model parameters
are dependent on the data. Thus, if the failure mechanisms, materials,
11-918 PACKAGE SEALING AND ENCAPSULATION

Table 14·5. Published Data of Median TTF for Epoxy.Molded Packages of


Unspecified Origin, used by Hallberg and Peck for Their Mathematical Model
Formulated to Calculate Accelerated Factors Between Two Accelerated
Environmental Tests: 85°C/85% rh and HAST

Median Median Median


Life Life Life
°C % rh @ °C % rh @ °C % rh @

85 85 3,060 85 85 6,000 150 81 12.4


110 85 500 138 85 359 130 81 85
130 85 90 138 75 450 85 85 4,300
160 85 12 138 50 622 120 85 420
85 85 1,050 127 85 407 130 85 130
111 85 151 85 85 900 140 85 60
120 85 90 110 90 165 85 85 12,000
130 85 50 121 90 104 110 85 2,700
85 85 1,600 131 90 52 127 85 504
111 85 170 85 85 3,360 138 85 300
130 85 66 121 90 239 85 85 3,400
85 85 5,400 131 90 87 110 85 340
110 85 440 85 85 5,300 127 85 130
127 85 160 110 90 700 198 85 54
130 85 220 121 90 240 127 85 140
138 85 70 131 90 85 138 65 270
85 85 1,700 85 85 2,400 127 75 177
130 85 430 85 65 440 85 85 15,000
85 85 7,500 60 85 2,000 145 85 195
130 85 220 111 75 50 85 85 180
85 85 8,000 120 75 30 60 85 1,300
110 85 1,280 130 75 18 110 85 50
130 85 320 111 100 26 125 85 18
85 85 380 120 100 15 85 85 1,000
130 68 66 130 100 8 110 85 180
85 85 245 85 85 210 130 85 50
110 85 57 110 90 30 85 85 6,425
130 85 23 60 90 1,800 110 90 725
85 85 8,300 85 85 15 85 85 210
110 85 2,000 85 20 1,600 60 85 1,300
130 85 300 20 85 7,000 111 85 54
85 85 730 20 58 1,500 121 85 290
110 85 112 85 85 3,920 85 85 5,000
130 85 70 138.5 85 80 110 85 500
85 85 2,500 157.3 85 28 127 85 170
125 85 225 85 85 370 138 85 80
150 85 62 85 70 660 85 85 5,345
85 85 1,440 85 60 1,160 151 85 40
130 81 32 70 81 1,940 133 85 170
150 81 15 55 81 7,390 109 85 1,300
150 50 115 85 85 2,860
140 65 42 115 81 446
14.8 RELIABILITY TESTING 11-919

processes, test methods, and so on differ substantially from those utilized


and contained in Peck's data, then a much more extensive set of failure
data for these systems must be obtained or the data must be regenerated
and models developed for each specific package type. It is expected that
the devices and materials and processes are not uniform throughout this
data set. As with all statistical models, the more uniform and complete
the set of data is, the better the model will be.
Another issue is fundamentally at the heart of the role of acceleration
factors. Peck has chosen 85°C/85% rh as the reference condition in order
to extrapolate to other accelerated test conditions. However, for normal
operating conditions, the extrapolation must be from the accelerated refer-
ence condition to the decelerated conditions of the operating environment.
For the latter, Peck's model does not fit as well, based on statistical criteria
[5]. A final issue which may be somewhat theoretical is the choice of a
statistical criterion for parameter estimation. Because a linear model is
assumed for Ro versus Re, the criterion for a best fit could be based on
mean square error, best slope, or best intercept. Each yields slightly
different results. All things considered, Peck's model has utility for first-
order approximations. Nevertheless, caution should be exercised for relia-
bility estimation when accuracy and high levels of confidence are required.
The Denson and Brusius model was developed empirically in a
similar fashion to Peck's model. The equation for the median life tm (OC,
% rh) for Denson and Brusius model is

[14-7]

where A is a scaling constant, and N and Ea are model parameters. Note


that the temperature dependence in Equation (14-7) is identical to that in
Equation [14-4], i.e., both are an Arrhenius law for temperature. However,
the relative humidity is considerably different. In Equation (14-7), % rh
has an exponential dependence, but it is polynomial in Equation 14-4.
The analogous equation to Equation 14-6 for the Denson and Brusius
model is

R~C(OC, %rh) = exp{ (~a)[(OC + 273tl - (358 Ktl]} [14-8]


exp[N( %rh- I - 85-1)]

and the estimated parameters are N = 2.96 and Ea = 0.30 eV. These
estimates were obtained from data other than that used by Peck and are
not readily available for further evaluation.
Figure 14-22 is a graph of the equivalent life relative to 1000 hat
11-920 PACKAGE SEALING AND ENCAPSULATION

100

...
Ul
~
'2
0
:I: ~
0
0
~
.... 10
0
~
iii ~
~
...
Q)

.2
0-

'(ij
LL Itnp
~ rOiled
::f Olding C Plasti
'"
Q)
0tn c
> 'Pound <0.1
~ Estimate
"5
E Based on
~
() Zero
Failures

0.1
1972 1974 1976 1978 1980 1982 1984
Years

Figure 14·22. Correlation of the Observed Ratio (Ro) of Elevated Tempera·


ture-Humidity Test TTFs to 85°C/85% rh TTFs for Plastic Packages Using
the Data in Table 14·5 Versus Calculated TTF (Re) using Hallberg and
Peck's Derived Mathematical Model. (From Ref. 71).

85 D C/85% rh as a function of temperature for fixed relative humidity;


that is, a graph of Equations 14-4 and 14-7 for A = 1000 h. The first
observation is that there are five curves for Peck's model and only one
for the Denson and Brusius model. This is directly attributable to the
difference in the functional forms for the relative humidity in the two
models. Assume that the range for % rh is (50, 100). Thus, 50% rh will
maximize the effect of relative humidity over the given range, but the
contribution of the % rh term to Equation 14-7 is only a factor of 1.025,
which is graphically indistinguishable. For practical purposes, the Denson
and Brusius model is independent of relative humidity; that is, it is simply
a function of temperature.
Another striking observation is that the Denson and Brusius model
and Peck's model yield similar results for only a very small interval of
temperatures around 100D e. In fact, as the temperature drops toward
room temperature, the difference in the models is about three orders of
magnitude. Obviously, this would yield drastically different predictions
in life. This effect is, no doubt, attributable to a value of Ea = 0.30 eV
14.9 RECENT ADVANCES IN SEALING AND ENCAPSULATION 11-921

in the Denson and Brosius model versus Ea = 0.90 eV in Peck's model.


It should be noted that when Peck's data are used to find the parameters
in the Denson and Brosius model, the values are N = 133.24 and Ea =
0.86 eV. These values seem more reasonable. There should be some effect
of relative humidity, and because both models have the identical Arrhenius
factor, the parameter should be close.
Thus, there are two issues which must be addressed if either of these
models are to be used for accurate reliability estimations. One issue is
the sensitivity of the model to the data. Would Peck's model be similar
to the Denson and Brosius model if the Denson and Brosius data were
fit to Peck's model, or is there a major difference between the two? Are
the components used in generating Peck's data sufficiently similar to the
components and coatings of interest so as to be comparable? How much
data are necessary to make a selection for the model? The other issue is
the model development. Are there fundamental reasons for selecting one
model over the other? What physical principles are applicable for either
model? Until some of these questions are resolved, care must be taken
when making reliability estimates based on these models.
Both models are primarily empirical and apparently have a great
deal of application. However, the lack of sharpness in the estimates of
the parameters diminishes their usefulness for estimating failure times
beyond the range of observations. Until further information is available,
they should be used only for approximate reliability estimations, not for
reliability estimations when accuracy and high levels of confidence are
required. Data accumulation and probability modeling efforts must con-
tinue before we can obtain accurate estimations outside the range of
observations, especially from accelerated test conditions back to field
conditions. Efforts toward understanding the physics-of-failure approach
may provide the fundamental knowledge necessary for the applicable use
of reliability models.

14.9 RECENT ADVANCES IN SEALING AND ENCAPSULATION


Although technology advances in hermetic packaging have not been
a necessity and, therefore, have been fairly limited in recent years, the
technology of using polymers for encapsulating bare devices has pro-
gressed from a "black art" to science. The greatest advancements have
been made in the materials used for encapsulation. They are cleaner with
respect to ionic contamination and possess far better physical and electrical
properties. Although plastic-molded packages have been around for some
time and their materials are fairly well developed, more demanding and
novel packaging approaches such as chip-on-board (COB) and multichip
modules (MCMs) require new areas of understanding and entirely different
material requirements.
11-922 PACKAGE SEALING AND ENCAPSULATION

For overcoat applications such as COB and tape automated bonding


(TAB), polymer manufacturers are actively pursuing materials that have
the specific properties required of these applications. Coating materials
for COB and TAB have very different requirements from plastic-molding
compounds, not only because of the vastly different processing used for
encapsulation but also from the standpoint of physical properties and
performance requirements. In general, a partial "wish list" for COB or
TAB applications would entail a material that has the following properties:

1. Is easy to dispense with existing robot-syringe equipment


2. Is compatible with the remaining assembly processes
3. Does not require heating of the material during dispensing
4. Possesses an acceptable combination of viscosity and thixotropic
properties for the application
5. Has sufficient adhesion to the constituents that are to be protected
6. Has a consistent batch-to-batch ionic content of less than 5 parts
per million each of Cl-, K+, and Na+
7. Requires a short duration cure with a maximum temperature of 150°C
8. Has a Tg of 150°C or greater
9. Possesses a residual stress after curing of less than 5 MPa
10. Does not cause stress related failures of the assembly during the
temperature cycling required for qualification
11. Provides sufficient mechanical protection to the device and intercon-
nect so additional packaging is not required
12. Is able to stand up to hot-wet-electrically biased environmental tests
13. Maintains acceptable electrical properties in hot and humid environ-
ments
14. Exhibits less than 1% linear shrinkage during cure
15. Is nonvolatile and nontoxic
16. Has at least an 8-h pot life and at least a 6-month shelf life
17. Is comparatively low in cost

Granted, many of these properties are application-specific and, there-


fore, require materials that are also application-specific, but a single-
dispense material does not exist today that fits the above criteria for
commercial versions of COB, and certainly not for the more advanced
MCMs.
With the numerous materials on the market, each application requires
different considerations when it comes to decisions on how best to provide
environmental protection. Vast ranges of physical properties are available
14.9 RECENT ADVANCES IN SEALING AND ENCAPSULATION 11-923

to the packaging engineer. Silicones, in general, possess extremely high


coefficients of thermal expansion (TCEs) and an extremely low moduli,
whereas epoxies are the opposite in that they have fairly low TCEs and
relatively high moduli. For example, a well-known silicone gel on the
market has a TCE of well over 325-800 ppm/DC and a modulus of less
than 0.001 GPa, and one of the better epoxies for COB has a TCE of
less than 22 ppm/DC and a modulus of 10.5 GPa. Also, silicone gels are
very clean with respect to Cl-, K+, and Na+. This specific silicone gel is
reported to have 2 ppm of each. Epoxies, by the nature of their manufacture,
are much higher in ionic content and this epoxy is one of the best with
20 ppm of each. The two materials in this example are at opposite ends
of the spectrum with respect to their physical properties, yet both can be
used to coat a wirebonded COB device and provide thermomechanical
stress resistance and corrosion resistance. The silicone gel needs additional
packaging in most cases because it does not provide mechanical protection
with its extremely low modulus, and the epoxy will not survive nearly
as well as the silicone gel in corrosion tests such as biased 85°C/85% rh
or biased HAST due to its ionic content and adhesion characteristics
[66-76]. Continuing improvements in epoxy resins have resulted in epox-
ies that have stronger polymer matrices and are better able to withstand
the difficult-to-pass hot, wet environments common in accelerated testing.
These improved epoxies would appear to be more apt to provide a higher
degree of reliability in testing and the field, but they are at a developmental
stage today.
New materials have been formulated in an attempt to incorporate
the best of both worlds. There are recent resins that combine the physical
properties of two materials to produce a composite resin like the silicone-
modified epoxies and polyimide siloxanes. If you combine a silicone and
an epoxy into one formulation, you obtain a mix of their respective physical
properties that may in some cases provide better overall performance. The
same is true for the polyimide siloxanes. Material dispense characteristics
have been improved to facilitate more consistent dispensing and are formu-
lated with a specific application in mind. TAB is beginning to see more
use in commercial electronic assemblies, certainly in consumer-destined
electronics coming out of Japan, and normally incorporate an encapsula-
tion coating on the surface of the die and inner lead bond area. TAB
encapsulation materials are required to have different properties than those
for COB in that smaller particles and higher flow are needed to fully
encapsulate around the inner lead bonds. Whereas, for COB, overcoat
materials need to have viscosity and thixotropic properties that help hold
the desired shape of the coating over the device and interconnect
through cure.
Cleaning of the assembly prior to the coating application has been
identified as a critical process step in obtaining long-term reliability. As
11-924 PACKAGE SEALING AND ENCAPSULATION

CFCs are phased out, new cleaning surfactants and processes will be
available to the packaging engineer. Cleaning must remove not only ionic
species but organic contaminants as well. What removes ionic contami-
nants, such as a wet cleaning process, does not always remove organic
impurities. Surfactant cleaners may be good at eliminating ionic species,
but they can leave a residue themselves that may require another cleaning
process to remove. When cleaning assemblies such as COB and MCMs,
consideration must be given to the effects of the cleaning process on all
of the constituents that make up the package. Evaluation methods for
determining the effectiveness of a cleaning process on all of the surfaces
present is still developing. Some commercial equipment exists that qualita-
tively measures what contaminants remain that are dissolvable and remov-
able from a device or assembly by gauging the change of resistivity in a
rinsing fluid, usually an alcohol. Other means can be used to evaluate
cleanliness on specific surfaces by water contact angle measurements. In
any case, cleaning is not something to be overlooked or taken lightly.
Silicon dioxides, nitrides, and various polymers can be deposited
via chemical vapor deposition. Some research is going on in this area,
and a few manufacturers of chemical vapor-deposition equipment are
attempting to specialize in producing coatings for assemblies such as
COB, as either a first-layer coating to be followed by an overcoat, or as
the only coating if housed in a nonhermetic plastic or metal box. The
deposition processes, resulting material characteristics, and associated
equipment are still in an early stage of development and appear to be
immature at present, although not altogether unpromising. Because the
deposition requires a vacuum, coating would have to be performed in
batches. These coatings may only be feasible for high-end electronics if
they are to be cost-effective.
With all of the aforementioned advances in polymer materials tech-
nology, a rapid increase in the area of reliability performance has been
realized. With this has come the increased difficulty of designing reliability
tests that provide the necessary acceleration factors for reasonable test
times in the manufacturing environment. Electrically biased 85°C/85%
rh (85/85) has been the standard test in the plastic-molded package industry
for many years. With new molding compounds, 85/85 tests can last several
thousand hours before failures begin to appear. These long test times are
not conducive to the fast pace of to day's electronics design and production
floor. The Highly Accelerated Stress Test, better known as HAST, is
quickly becoming an accepted alternative to 85/85. HAST utilizes a pres-
surized chamber to allow the temperature to be elevated while maintaining
humidity control, to produce the additional activation energy for more
accelerated testing. Electrical bias in 85/85 or HAST provides additional
activation energy so that corrosion failures and mechanisms can be identi-
fied in a shorter time. Alternative environmental tests such as Battelle's
14.9 RECENT ADVANCES IN SEALING AND ENCAPSULATION 11-925

Flowing Mixed Gas test (FMG), introduce parts-per-million or parts-per-


billion quantities of contaminants commonly found in the atmosphere to
simulate corrosive processes found in some electronics applications. FMG
was designed to test connector technologies and has not yet been used
with any regularity in the testing of nonhermetic packages due to the
relatively long test times required. Correlation of 85/85, or HAST to field
life is still an issue of considerable discussion, as well as the acceleration
factors between 85/85 and HAST. A great deal of ongoing work in the
correlation and testing area is needed to be able to accurately estimate
reliability in the field from models based on these accelerated tests.
Other work has been done where environmental testing is performed
in a sequential series of steps. Temperature cycling has historically been
used in testing plastic packages as a single-step test to evaluate the thermal-
mechanical stresses arising from mismatches of TCEs. The 85/85, HAST,
or similar corrosion/polymer degradation tests are also usually run as
separate evaluations. Recent studies have been performed using tempera-
ture cycling and salt atmosphere as preconditioning tests for 85/85 or
HAST. The temperature cycling will induce cracks and delaminations of
the polymer matrix and the polymer-to-substrate interface. Next in the
sequence, salt atmosphere adds an ionic contaminant (NaCI) to the newly
created cracks and voids in the package. Finally, placing the parts in either
biased 85/85 or biased HAST to accelerate the corrosion mechanism that
is now in place with moisture, NaCI, and electrical potentials. Even with
a test sequence this severe, the biased HAST portion of the test sequence
has run in excess of 1000 h without significant performance degradation
of the wirebonded and coated test chip. With the desire to run ultraclean
85/85 and HAST chambers for the purposes of repeatability and control,
among other things, a sequence such as this does have its dangers and un-
knowns.
On the device side of the test spectrum, the fabrication of test
chips that are designed to amplify the failure mechanisms that have been
observed by the polymer packaging industry has been greatly advanced
in recent years. Sandia National Labs, for one, has taken the sophistication
of test chips to new levels. Sandia's family of assembly test chips, desig-
nated ATCs, have incorporated assorted detectors and sensors that are of
primary interest to those in the polymer encapsulation industry. The ATC
family now has five versions in production. The ATCOI is the first-
generation test chip designed for studying aluminum corrosion. The
ATCO 1 incorporates eight different triple tracks, three corrosion ladders,
and aluminum-sheet resistance, in addition to other corrosion detection
structures. The ATC04 is the most recent and advanced test chip in the
family, with passivated and unpassivated triple tracks in both single- and
double-level metal, some triple tracks have underlying polysilicon heaters
for heat dissipation studies, diode thermometers, and 48 addressable pie-
11-926 PACKAGE SEALING AND ENCAPSULATION

zoresistive stress sensors. Both the ATCOI and ATC04 have features sizes
down to 1.25 Jlmlmrn line and space widths. These chips have been made
available to industry in an effort to provide an industrywide standard test
chip family for comparative studies and data correlation.
The unquestioned superior reliability of hermetic packages will re-
main a factor in their continued use for military and critical civilian
applications despite their higher cost, size, and weight. The predicted
increases in device densities will influence failure susceptibilities in several
ways. The increased device density will mean increased power densities;
hence, generally higher operating temperatures, larger voltage gradients,
and less conductor material means there is less material loss required for
opens or shorts. Also, the packing of more devices on the same chip, in
concert with physically increasing the sizes of chips, greatly increases
the probability of chip-level failure. The cumulative effect of these factors
serves to increase the susceptibility of failure by an order of magnitude
for each doubling in circuit density. Certainly, the group of technologies
that form COB and MCM assemblies will continue to evolve, and with
it so will polymer encapsulation. These considerations, in combination
with the continuing advances in materials and processes for polymer
packaging, will constantly redefine the limits of use for nonhermetic versus
hermetic packaging.

14.10 FUTURE DEVELOPMENTS


The rapid development of IC technology has created a critical need
for packaging sealing and encapsulation using advanced polymeric materi-
als for encapsulants, interlayer dielectrics, and passivation layers. Recent
advances in high-performance polymeric materials, such as improved
silicone elastomers, ultrasoft silicone gels, low-stress epoxies, low ther-
mal-expansion coefficient and photodefinable polyimides and cyclobu-
tenes have begun to provide polymers to meet reliability requirements of
VLSI technology. But to date they have not met all of the requirements,
particularly those of defense industry needs. It is becoming increasing
clear that this will happen during the next decade or two. The advances
in high-density multichip module packaging will undoubtedly require
nonhermetic packaging for the ever-increasing large modules. However,
the demands for improved properties of materials in the areas of low
dielectric constant, high breakdown voltage strength, high sheet resistance,
and less dielectric change with humidity will continue to require the
development of high-performance polymeric materials. In addition, given
the low-cost requirements of all electronic packages, all these requirements
need to be accomplished at low cost. Their application for on-chip intercon-
nections, and wafer-scale integration structures with very fast interconnect-
ing networks as well as for high-performance packaging will become
14.11 REFERENCES 11-927

apparent. It is a challenge that the collaborative efforts between polymer


chemists, material scientists device engineers, and packaging engineers
will face in the near future.

14.11 REFERENCES
1. C. P. Wong (ed.). Polymers for Electronic and Photonic Applications, Academic
Press, San Diego, CA, 1993, and references therein.
2. C. P. Wong. "Overview ofIC Device Encapsulants as Device Packaging," J. Electron.
Packaging, 3: p. 97, 1989.
3. R. W. Thomas. "Moisture, Myths and Microcircuits," IEEE Trans. Parts, Hybrids
Packaging, PHP-12(3): pp. 167-171, 1976.
4. J. L. Jellison. "Susceptibility of Micro Welds in Hybrids to Corrosion. 'Degrada-
tion,' " 13th Annual Proceedings, Reliability Physics Symposium, pp. 70-79, 1975.
5. H. Koelmans. "Metallization Corrosion in Silicon Devices by Moisture-Induced
Electrolysis," 12th Annual Reliability Physics Symposium, pp. 168-171, 1974.
6. F. R. Neighbor and B. R. White. "Factors Governing Aluminum Interconnection
Corrosion in Plastic Encapsulated Microelectronic Devices," Microelectron. Reliab.,
16: pp. 161-164, 1977.
7. W. M. Paulson and R. P. Kirk. "The Effects of Phosphorous-Doped Passivating Glass
on the Corrosion of Aluminum Metallization," 12th Annual Proceedings, Reliability
Physics Symposium, pp. 172-179, 1974.
8. G. DiGiacomo. "Phosphorous Migration Kinetics from PSG to Glass Passivation
Surface," 22nd Annual Proceedings, Reliability Physics Symposium, pp. 223-228,
1984.
9. A. Shumka and R. R. Piety. "Migrated-Gold Resistive Shorts in Microcircuits," 13th
Annual Proceedings, Reliability Physics Symposium, pp. 93-98, 1975.
10. A. Christou. "Moisture Diffusion Through Hybrid Circuit Encapsulants," Electron.
Packaging Production, 19(4): pp. 91-99, 1979.
II. R. W. Thomas. "Moisture in Microcircuits," in Reliability Technology for Cardiac
Pacemakers, ed. by H. A. Schafft, NBS Special Publication, 400-9, U.S. GPO,
Washington, DC, 1976.
12. M. Byrnes, J. L. Carter, J. Sergent, and D. King. "Considerations in the Hermetic
Packaging of Hybrid Microcircuits," Solid State Technol., 27(8): pp. 68-77, 1984.
13. B. Van, S.1. Meilinle, G. W. Warren, and P. Wymablatt, "Water Absorption and
Surface Conductivity Measurement on a-Aluminum Substrates," IEEE Trans. Com-
ponents Hybrids Manu! Technol., CHMT-IO, No.(2): pp. 247-251, 1987.
14. B. Reich. "Acceleration Factors for Plastic Encapsulated Semiconductor Devices and
Their Relation to Field Performance," Microelectron. Reliab., 14(1): pp. 63-66, 1975.
15. F. N. Sinnadurai. "Accelerated Aging of Semiconductor Devices in Environments
of High Vapor Pressure of Water," Microelectron. Reliab., 13(1): pp. 23-27, 1974.
16. R. A. Lawson. "Accelerated Testing of Plastic Encapsulated Semiconductor Compo-
nents," 12th Annual Proceedings, Reliability Physics Symposium, pp. 243-249, 1974.
17. W. E. Swartz, J. H. Linn, J. M. Ammons, M. Kovac, and K. Wilson. "The Adsorption
of Water on Metallic Surfaces," 21st Annual Proceedings, Reliability Physics Sympo-
sium, pp. 52-59, 1983.
18. D. R. Carley, R. W. Nearhoff, and R. Denning. "Moisture Control in Hermetic
Leadless Chip Carriers with Silver-Epoxy Die-Attach Adhesive," RCA Rev., 34(2):
pp. 278-290, 1984.
19. J. G. Davy. "Calculation of Leak Rates of Hermetic Packages," IEEE Trans. Parts,
Hybrids Packaging, PHP-ll(3): pp. 177-189, 1975.
11-928 PACKAGE SEALING AND ENCAPSULATION

20. D. Stroehle. "On the Penetration of Water Vapor into Packages with Cavities and
on Maximum Allowable Leak Rates," 15th Annual Proceedings, Reliability Physics
Symposium, pp. 101-106, 1977.
21. R. E. Suloff. "A Study of Leak Rate Versus Reliability of Hybrid Packages," Interna-
tional Microelectronics Symposium, pp. 121-124, 1978.
22. A. DerMarderosian and V. Ginot. "Water Vapor Penetration into Enclosures with
Various Air Leak Rates," 16th Annual Proceedings, Reliability Physics Symposium,
pp. 179-183, 1978.
23. R. K. Traeger. "Hermeticity of Polymetric Lid Sealants," Proceedings 25th Electron-
ics Components Conference, pp. 361-367, 1976.
24. D. Nixen. "Package Reliability as Affected by Materials and Processes," Semicond.
Int., 5(4): pp. 175-182, 1982.
25. D.W.A. Forbes. "Solder Glass Seals in Semiconductor Packages," Glass Technol.,
8(2): pp. 32-42, 1972.
26. H. H. Manko. Solders and Soldering, McGraw-Hill, New York, 1964.
27. K. S. Dogra. "A Bismuth-Tin Alloy for Hermetic Seals," Proceedings IEPS 4th
International Conference, pp. 631-636, 1984.
28. M. D. Zimmer. "Reliability and Thermal Impedance Studies in Soft Soldered Power
Transistors," IEEE Trans. Electron. Devices, ED-23(8): pp. 843-850, 1976.
29. C. J. Thawaites. "Some Aspects of Soldering Gold Surfaces," Electroplat. Metal
Finishing, 29(9): pp. 21-26, 1973.
30. G. M. Stoll. "Hermetic Sealing of Thermally Sensitive Circuits with Emphasis on
Thermal Conductivity Sealing," Proceedings International Microelectronics Sympo-
sium, pp. 520-526, 1983.
31. J. M. Morabito. "Recent Advances in Solder Bond Technology," Thin Solid Films,
72: pp. 433-442, 1980.
32. J. Brady and M. Courtney. "Hermetic Tin!Lead Solder Sealing for the Air-Cooled
IBM 4381 Module," Paper presented at the IEEE, ICCD Conj. at Portchester, NY,
November 1983.
33. W.M.S. Yang. "Reflow Solder Ceramic Lids for Hermetic Packages," Solid State
Technol., 27(12): pp. 137-143, 1984.
34. F. W. Luce. "Delidding and Resealing of Microelectronic Packages," MICOM Report
MM & T Project, R793438, 1982.
35. G. Bourdelais and E. F. Hill. "Hermeticity Considerations for VISI Packaging,"
Proceedings 6th Annual International Electronics Packaging Conference, pp. 18-
24,1986.
36. D. D. Zimmerman. "A New Gold-Tin Alloy Composition for Hermetic Package:
Sealing and Attachment of Hybrid Parts," Solid State Technol, 15( 1): pp. 44-46, 1972.
37. N. R. Stockham and C. J. Dawes. "Resistance Seam and Laser Welding Hybrid
Circuits," Hybrid Circuits, 1(9): pp. 509-519, 1983.
38. S. Norrman and P. A. Torstensson. "Hermetic Sealing of Kovar Hybrid Packages
by Laser Welding," Hybrid Circuits, 1(8): pp. 21-23, 1985.
39. A. K. Varshneya. "Stress in Glass-to-Metal Seals," in Treatise Materials Science
and Technology, ed. byM. TomazawaandR. H.Doremus, vol. 22,pp. 241-306,1982.
40. J. A. Pask and R. M. Fulrath. "Fundamentals of Glass-to-MetaI Bonding; VIII. Nature
of Wetting and Adherence," J. Am. Ceram. Soc., 45(12): pp. 592-596, 1962.
41. M. P. Borom and J. A. Pasko "Role of Adherance Oxides in the Development of
Chemical Bonding at Glass-to-Metal Interfaces," J. Am. Ceram. Soc., 49(1): pp.
1-6, 1966.
42. R. G. Buckley. "Understanding Glass-to-MetaI Seals," Electron. Packaging Produc-
tion, 18(10): pp. 74-79, 1978.
14.11 REFERENCES 11-929

43. A. E. Yaniv, D. Katz, J. E. Klein, and J. Sharon. "A New Technique of Surface
Preparation for Bonding Glass to Kovar," Glass Technol., 22(5): pp. 231-235, 1981.
44. W. F. Yext, B. 1. Shook, W. S. Katzenberger, and R. C. Michalek. "Improving Glass-
to-Metal Sealing Through Furnace Atmosphere Composition Control," IEEE Trans.
Components Hybrids Manuf. Technol. CHMT-6(4): pp. 455-459, 1983.
45. M. B. Miller. "Stress Absorbing Coatings for Glass-to-Metal Seals in Microelectronic
Packages," SAMPE, pp. 19-21, 1983.
46. K. Kokini and R. W. Perkins. "Estimating the Strength of Annular Glass-to-Metal
Seals in Microelectronic Packages: An Experimental Study," IEEE Trans. Compo-
nents Hybrids Manuf. Technol., CHMT-7(3): pp. 276-279, 1984.
47. R. Shukla and N. Mencinger. "Moisture Evolution Analysis of Sealing Glasses used
in Integrated Circuit Packages," Electron. Mater., 14(4): pp. 461-471, 1985.
48. A. H. Kumar and R. R. Tummala. "Titania-Doped Lead-Zinc-Borate," Bull. Am.
Ceram. Soc., 57(8): pp. 738-739, 1977.
49. R. W. Vasofsky and R. K. Lowry. "Moisture Evolution from Sealing Glasses Dry
CERDIP Packages," 18th Annual Proceedings, Reliability Physics Symposium, pp.
1-9, 1980.
50. R. K. Lowry, C. J. Van Leeuwen, B. L. Kennimer, and L. A. Miller. "A Reliable,
Dry Ceramic Dual-In-Line Package CERDIP," 16th Annual Proceedings, Reliability
Physics Symposium, pp. 207-212, 1978.
51. D. E. Erickson. "Hybrid Circuit Sealing-Problem Prevention Clinic," Electron.
Packaging Production, 22(11): pp. 133-137, 1982.
52. T. H. Ramsey. "Critical Parameters in Glass Sealed Ceramic Packages," Solid State
Technol., 17(9): pp. 51-59, 1974.
53. P. S. Burggraff. "Semiconductor Package Sealing," Semicond. Int., 2(7): pp. 35-
49,1979.
54. W. K. Denby. "Current Trends in Hermetic Package Sealing Equipment," Microelec-
tron. Manf. Testing, 4(10): pp. 14-19, 1981.
55. J. Wallach. "Surface Requirements for Seals," Proceedings International Conference
on Surface Technology, pp. 606-620, 1973.
56. 1. G. Davy. "Hermetic Packaging Protects Circuitry from Moisture," Electron. Pack-
aging Production, 24(10): pp. 58-62, 1984.
57. A. J. Blodgett and D. R. Barbour. "Thermal Conduction Module: A High-Perfor-
mance Multilayer Ceramic Package," IBM J. Res. Devel., 26(1): pp. 30-36, 1982.
58. K. A. Jewett. "Practical Methods for Hermetic Seal Testing," in Electronic Packag-
ing: Materials and Processes, ed. by J. A. Sartell, pp. 85-88, ASM, Metals Park,
OH,1986.
59. H. A. Schafft. Hermeticity Testing for Integrated Circuits, NBS Special Publication,
400-9, p. 43, u.S. GPO, Washington, DC, 1974.
60. R. W. Thomas. "Microcircuit Package Gas Analysis," 14th Annual Proceedings,
Reliability Physics Symposium, pp. 283-294, 1976.
61. T. W. Carr, E. A. Corl, C. L. Lin, and C. G. Majtenji. "Quantitative H 20 Determina-
tion in Components Using Plasma Chromatograph-Mass Spectrometer," 16th Annual
Proceedings, Reliability Physics Symposium, pp. 59-63, 1978.
62. M. G. Kovac, D. Chleck, and P. Goodman. "A New Moisture Sensor for In-Site
Monitoring of Sealed Packages," 15th Annual Proceedings, Reliability Physics Sym-
posium, pp. 85-91, 1977.
63. J. Finn and V. Fong. Recent Advances in Al2D3 In-Situ Moisture Monitoring Chips
for CERDIP Package Application," IEEE, New York, 1980.
64. N. Bakker. "In-Line Measurement of Moisture in Sealed IC Packages," Phillips
Telecommun. Rev., 37(1): pp. 11-19, 1979.
11-930 PACKAGE SEALING AND ENCAPSULATION

65. R. P. Merret, S. P. Sim, and 1. P. Bryant. "A Simple Method of Using the Die of
Integrated Circuit to Measure the Relative Humidity Inside Its Encapsulation," 18th
Annual Proceeding, Reliability Physics Symposium, pp. 17-25, 1980.
66. D. Chang, P. Crowford, J. Fulton, R. McBribe, M. Schmidt, R. Sinitiski, and C. P.
Wong. "An Overview and Evaluation of Anisotropically Conductive Adhesive Films
for Fine Pitch Electronic Assembly," IEEE Trans. Components Hybrids Manu!
Technol., CHMT 16(8): p. 828, 1993.
67. D. Chang, J. Fulton, H. Ling, M. Schmidt, R. Sinitski, and C. P. Wong. "Accelerated
Life Test of Z-Axis Conductive Adhesives," IEEE Trans. Components Hybrids
Manu! Technol., CHMP 18(8): p. 836, 1993.
68. C. P. Wong and R. Mcbride. "Robust Titanate-modified Encapsulants for High Volt-
age Potting Application of Multichip Module/Hybrid IC," IEEE Trans. Components
Hybrids Manu! Technol., CHMT 16(8): p. 868, 1993.
69. C. P. Wong and P. D' Ambra. "Embedment of Electronic Devices," in Kirk-Othmer
Encyclopedia of Chemical Technology, 4th ed., vol. 9, pp. 377-393, John Wiley &
Sons, New York, 1994.
70. C. P. Wong, 1. M. Segelken, and C. N. Robinson. "Encapsulation of Chip-on-Board",
in Handbook of Wire Bonding, Tape Automated Bonding, and Flip-Chip on Board
for Multichip Module Applications, ed. by 1. Lau, pp. 470-503, Van Nostrand
Reinhold, New York, 1994.
71. C. P. Wong. "Recent Advances in Hermetic Equivalent IC Packaging of Microelec-
tronics", Research Society, Advanced Electroceramics and Packaging Technology,
Proceedings of the 1994 International Conference on Electronic Materials, p. 78,
1994.
72. C. P. Wong and R. McBride. "Preencapsulation Cleaning Method and Control for
Microelectronics Packaging," IEEE Trans. Components Hybrids Manu! Technol.,
CPMT 17(4): p. 542, 1994.
73. C. P. Wong. "Overview of Microelectronic Packaging: Materials and Processes,"
Proceedings of Polymat '94, p. 369, 1994.
74. C. P. Wong. "Materials for Electronic Packaging," in Materialsfor Electronic Pack-
aging, ed. by D. Chung, p. 270, Butterworth Publisher, Boston, 1995.
75. C. P. Wong. "Recent Advances in Low Cost Plastic Packaging of Flip-chip Multichip
Module of Microelectronics," Proceedings on the 1994 International Conference on
Electronic Materials, p. 73, 1995.
76. C. P. Wong. "Thermal Mechanical Behavior of High Performance Encapsulants in
Microelectronic Packaging", IEEE Trans. Components Packaging Manu! Technol.,
CPMT-18: p. 270, 1995.
GLOSSARY AND SYMBOLS

The ISHM Glossary of Hybrid-Circuit Terms (Updated edition-


Summer 1987) by G.S. Szekely provided the foundation for this glossary.
Definitions have in some cases been modified to fit this work. Other terms
used by the authors and the packaging profession have been added and
some terms in the original work have not been deleted. Material chemical
symbols are found at the end of this glossary.

ACCELERATED STRESS TEST. A test conducted at a stress, e.g., chemical


or physical, higher than that encountered in normal operation, for the purpose
of producing a measurable effect, such as a fatigue failure, in a shorter time than
experienced at operating stresses.

ACCELERATOR. An organic compound which is added to an epoxy resin


to shorten the cure time.

ACTIVE COMPONENTS. Electronic components, such as transistors, diodes,


electron tubes, thyristors, etc., which can operate on an applied electrical signal
so as to change its basic characteristics, i.e., rectification, amplification, switch-
ing, etc.

ADDITIVE PLATING. Processing a hybrid circuit substrate by sequentially


plating conductive, resistive, and insulative materials, each through a mask, thus
defining the areas of traces, pads, and elements.

ADVANCED STATISTICAL ANALYSIS PROGRAM (ASTAP). The "Ad-


vanced Statistical Analysis Program" is the mM circuit analysis simulation
program. It performs DC, time domain, and frequency domain simulations. Statis-
tics can be applied to all simulations to predict operating tolerances. Among its
many features is a transmission line analysis program. Also see SPICE.
11-932 GLOSSARY AND SYMBOLS

ALLOY. A solid-state solution or compound formation of two or more metals.


Alternatively, a combination of metals resulting in a phase or phases containing
some of each constituent.

ALPHA PARTICLE. Decay product of some radioactive isotopes. It is a high-


energy (mv range) helium nucleus capable of generating electronlhole pairs in
microelectronic devices and switching cells, causing soft errors in some devices.

APPLICATION SPECIFIC INTEGRATED CIRCUIT (ASIC). Applica-


tion Specific Integrated Circuit is an integrated circuit chip with personalization
customized for a specific product. Personalization refers to wiring on the inte-
grated circuit chip.

AREA ARRAY TAB. Tape automated bonding where edge-located pads and
additional pads on the inner surface area of a chip are addressed in the bonding
scheme. This is practiced with extremely complex dice, VLSI etc. Also for use
with ICs where peripheral pad pitch cannot be further reduced and all 1I0s must
be accommodated.

ARRAY. A group of elements (pads, pins) or circuits arranged in rows and


columns on one substrate.

ASPECT RATIO. The ratio of the length of hole to the diameter of hole in
a board.

ASSEMBLY. A hybrid circuit which includes discrete or integrated compo-


nents that have been attached to the next level of package, usually a card.

ASSEMBLYIREWORK. Terms denoting joining and replacement processes


of microelectronic components. Assembly refers to the initial attachment of
device and interconnections to the package. Rework refers to the removal of a
device including interconnections, preparation of the joining site for a new device,
and rejoining of the new device. Rework is necessary for either repair or engineer-
ing change.

BACKBONDING. Bonding active chips to the substrate using the back of the
chip, leaving the face, with its circuitry face up. The opposite of backbonding
is face down bonding.

BACK-END-OF-THE-LINE (BEOL). That portion of the integrated circuit


fabrication where the active components (transistors, resistors, etc.) are intercon-
GLOSSARY AND SYMBOLS 11-933

nected with wiring on the wafer. It includes contacts, insulator, metal levels, and
bonding sites for chip-to-package connections. Dicing the wafer into individual
integrated circuit chips is also a BEOL process. The front-end-of-the-line (FEOL)
denotes the first portion ofthe fabrication where the individual devices (transistors,
resistors, etc.) are patterned in the semiconductor.

BACK PANEL. A planar package component holding plugged-in lower-level


package components (e.g., cards) as well as discrete wires and cables interconnect-
ing these components.

BACKSIDE METALLURGY (BSM). A metallization pad electrically con-


nected to internal conductors within a multilayered ceramic package, to which
pins are brazed.

BALL GRID ARRAY (BGA). A Ball Grid Array is an area array of solder
balls joined to a SCM or MCM and used to electrically and physically connect
the package to the next level of package, usually a printed circuit board.

BALL LIMITING METALLURGY (BLM). The solder wettable terminal


metallurgy which defines the size and area of a soldered connection, such as C4
and a chip. The BLM limits the flow of the solder ball to the desired area, and
provides adhesion and contact to the chip wiring.

BANDWIDTH. The maximum pulse rate or frequency that can reliably propa-
gate through a transmission line. For a data bus, bandwidth is commonly used
to describe the maximum data rate which is the single line pulse rate multiplied
by the number of parallel bus bit lines.

BIFET. The combination of bipolar and PET transistors integrated together on


the same piece of silicon for enhanced performance and cost.

BINDER. Materials (organic or inorganic) added to thick-film compositions


and to unfired substrate materials to give sufficient strength temporarily for
prefire handling.

BIPOLAR TRANSISTOR. Original transistor design in which two semicon-


ductor junctions (regions of opposite polarity doping) are separated by a narrow
region, called the base. Minority carriers are injected in the base from the emitter
across the base-emitter junction, travel through the base, and are attracted to the
collector through the base-collector junction. These transistors consume more
power than Field-Effect Transistors (PET), but also achieve higher performance.

BLOCK COPOLYMER. A copolymer compound resulting from the chemical


reaction between n number of molecules, which are a block of one monomer,
11-934 GLOSSARY AND SYMBOLS

and .on" number of molecules, which are a block of another monomer. Example:
stearine (rigid) with silicone (elastic).

BOARD. This package element can best be defined as an organic printed-


circuit card or board on which smaller cards or modules can be mounted. Its
connections to the next higher level involve discrete wire or cables.

BOILING. Phase change and formation of bubbles in a superheated liquid.

BONDABILITY. Those surface characteristics and conditions of cleanliness


of a bonding area which must exist in order to provide a capability for successfully
bonding an interconnection material by one of several methods, such as ultrasonic
or thermocompression wire bonding.

BRAZE. A joint formed between two different materials by formation of liquid


at the interface.

BRAZING. Joining of metals by melting a non-ferrous, filler brazing metal,


such as eutectic gold-tin alloy, having a melting point lower than that of the base
metals. Also known as hard soldering.

BTAB. The acronym for tape automated bonding when the raised bump for
each bond site is prepared on the tape material as opposed to the bump being
on the chip.

BUMPED TAPE. A tape for the TAB process where the inner-lead bond sites
have been formed into raised metal bumps on the tape rather than on the chip.
This ensures mechanical and electrical separation between inner lead bonds and
the non-pad areas of the chip (die) being bonded.

BURN·IN. The process of electrically stressing a device (usually at an elevated


temperature and voltage environment) for an adequate period of time to cause
failure of marginal devices.

BURN· OFF. Removal of unwanted materials-typically organics from green-


sheets or organic contamination from substrates.

CAMBER. A term that describes the amount of overall warpage present in


a substrate.
GLOSSARY AND SYMBOLS 11-935

CAPACITANCE. The electrostatic element that stores charge. In packaging


systems, it is used in lumped equivalent circuits to represent part of a line
discontinuity. It is also used in a distributed system to represent the electrostatic
storage property of a transmission line. Because it delivers current in response
to a change in voltage, another use is to filter powering systems.

CARD. A printed-circuit panel (usually multilayer) that provides the intercon-


nection and power distribution to the electronics on the panel, and provides
interconnect capability to the next level package. It is also known as a daughter
board. It plugs into a mother printed-circuit board.

CARD·ON·BOARD. Packaging technology in which multiple printed-circuit


panels (cards) are connected to printed-circuit panel (board) at 90° angles.

CELL DESIGN, STANDARD. A semicustom product implemented from a


fully diffused or ion implanted semiconductor wafer carrying horizontal rows of
primary cells, interlaced with wiring channels (bays). Vertical wiring is supplied
by additional processed layers which may use the cell areas or lie in channels
on an overhead layer. Channel widths may vary to suit particular chip logic, so
that chip sizes are not fixed for all products of a family.

CENTRAL PROCESSOR (CP). Computer processor responsible for fetching,


interpreting, and executing program instructions. Also called Processor Unit (PU)
and Central Processing Unit (CPU).

CERAMIC. Inorganic, nonmetallic material, such as alumina, beryllia, or


glass-ceramic, whose final characteristics are produced by subjection to high
temperatures. Often used in forming ceramic-substrates for packaging semicon-
ductor chips.

CERAMIC BALL GRID ARRAY (CBGA). A ceramic package using ball


grid array technology. See ball gird array technology.

CERAMIC COLUMN GRID ARRAY (CCGA). A ceramic package using


ball grid array technology. See ceramic gird array technology.

CERAMIC DUAL·IN·LINE PACKAGE (DIP), Dual-in-line package in ce-


ramic. See Dual-in-line Package.

CERAMIC QUAD FLAT PACK (CQFP). Quad Flat Pack in ceramic. See
Quad Flat Pack.
11-936 GLOSSARY AND SYMBOLS

CERMET. A solid homogeneous material usually consisting of a finely divided


admixture of a metal and ceramic in intimate contact. Cermet thin films are
normally combinations of dielectric materials and metals.

CHANNELS. Provide communications paths for input to and output from the
computer system.

CHARACTERISTIC IMPEDANCE (Zo). The voltage-to-current ratio of an


electric signal propagating through an infinitely long transmission line. If L
denotes the inductance per unit length and C denotes the capacitance per unit
length, then ~ = (LlC)O.5.

CHEMICAL VAPOR DEPOSITION (CVD). Depositing circuit elements on


a substrate by chemical reduction of vapor of volatile chemical in contact with
the substrate.

CHEMORHEOLOGY. The study of the processability or flow (rheology) and


the chemistry of the polymer system. Processability parameters include, for
instance, heating rates, hold temperatures, injection speeds, and compaction pres-
sures. The chemical aspect, on the other hand, involves the rate of reaction, the
mechanisms, the kinetics, and the cessation of the chemical reaction at the end
of the polymerization.

CHIP. The uncased and normally leadless form of an electronic component


part, either passive or active, discrete or integrated. Also referred to as a die.

CHIP CARRIER. A special type of enclosure or package to house a semicon-


ductor device. It has electrical terminations around its perimeter, or solderpads
on its underside, rather than an extended lead frame or plug-in pins.

CHIP DESIGN, DEPOPULATED. A gate array or standard cell array chip


in which the wiring capacity (and hence chip area) is deliberately chosen to make
automatic wiring possible only for those chips having some amount less than
their maximum possible logic cell occupancy. This increases wafer productivity
and circuit placement flexibility.

CIDP-ON-BOARD (COB). One of many configurations in which a chip is


directly bonded to a circuit board or substrate. These approaches include wirebon-
ding, TAB, or solder interconnections, similar to the C4 structure. In low-end
and consumer systems, chip-on-board generally refers to wirebonding of chips
directly to board. See also Direct Chip Attach (DCA).

CIRCUIT-BOARD PACKAGING. Packaging of chips by the use of organic


printed-circuit boards. See Printed-Circuit Board.
GLOSSARY AND SYMBOLS 11-937

CLADDING. Thin layer of a corrosion-resistant metal coating bonded to a


metal core, usually by heating and rolling. Typical examples are steels clad with
stainless steel, nickel alloys, or copper alloys. Copper cladding on both sides of
invar is accomplished this way.

CLOCK SKEW. A cycle time adder caused by the amount of tolerance associ-
ated with the clock signal arrival times at all of the system latch inputs.

CMOS. See Complementary Metal-Oxide Semiconductor.

COATED-METAL CORE SUBSTRATE. A substrate consisting of an or-


ganic or inorganic insulation coating bonded to metal. Insulated surface or surfaces
are used for circuit deposition.

COEFFICIENT OF THERMAL EXPANSION (CTE). The ratio of the


change in dimensions to the change in temperature-per-unit starting length, usually
expressed in cm/cm/DC. The acronyms TCE and CTE are synomous.

COFFIN-MANSON EQUATION. A commonly used formula, first proposed


by S.S. Manson and L.F. Coffin, relating the fatigue lifetime of a metal to the
imposed strain amplitude. Others have extended the formula to incorporate time
and temperature dependent phenomena.

COFIRING. Processing thick-film conductors and dielectrics through the firing


cycle at the same time to form multilayer structures.

COLORANT. An inorganic or organic compound that is added to a polymeric


resin to impart a desired color.

COLUMN GRID ARRAY (CGA). A Column Grid Array is an area array of


solder columns joined to an SCM or MCM and used to electrically and physically
connect the package to the next level of package, usually a printed circuit board.
A Column Grid Array is used when the package performance requires a higher
riliability that provided with the similiar Ball Grid Array.

COMPLIANT BOND. A bond which uses an elastically and/or plastically


deformable member to impart the required energy to the lead.

COMPLEMENTARY METAL-OXIDE SEMICONDUCTOR (CMOS).


This refers to logic in which cascaded field effect transistors (PET) of opposite
polarity are used to minimize power consumption.
11-938 GLOSSARY AND SYMBOLS

COMPRESSION SEAL. A seal between an electronic package and its leads.


The seal is formed as the heated metal, when cooled, shrinks around the glass
insulator, thereby forming a tight compression joint.

CONDUCTION. Thermal transmission of heat energy from a hotter region to


a cooler region in a conducting medium.

CONDUCTIVE ADHESIVE. An adhesive material, usually epoxy, that has


metal powder added to increase electrical conductivity. Usual conductor added
is silver.

CONDUCTIVE EPOXY. An epoxy material (polymer resin) that has been


made conductive by the addition of a metal powder, usually gold or silver. Best
common conductors are silver, copper, and gold. See also Superconductor, and
Conductor Adhesive.

CONDUCTOR, ELECTRICAL. A class of materials that conduct electricity


easily. They have very low resistivity which is usually expressed in micro-ohm-
cm. The best conductors include silver, copper, gold, and superconducting-ce-
ramics.

CONDUCTOR, THERMAL. A class of materials, such as copper, aluminum,


and beryllia, that conduct heat.

CONFORMAL COATING. A thin nonconductive coating, either plastic or


inorganic, applied to a circuit for environmental and/or mechanical protection.

CONNECTIONS. The connections belonging to nets interconnecting logic


units on a given package level-including connections to terminals on that level-
connecting it to the next higher package level.

CONNECTIVITY. See Wiring Density.

CONTACT ANGLE. The angle between the bonding material, usually a liq-
uid-like solder, and the bonding pad. Also called wetting angle.

CONTACT RESISTANCE. Excess electrical resistance in series with the


bulk conductor resistance of two contacting electrical conductors arising from the
nature of the contact geometry and surface properties of the contacting surfaces.

CONTROLLED COLLAPSE CHIP CONNECTION (C4). A solder joint


connecting a substrate and a flip chip, where the surface tension forces of the
liquid solder supports the weight of the chip and controls the height (collapse)
of the joint.
GLOSSARY AND SYMBOLS 11-939

CONTROLLING COLLAPSE. Controlling the reduction in height of the


solder balls in a flip-chip processing operation.

CONVECTION. Transmission of thermal energy from a hotter to a cooler


region through a moving medium, such as air or water.

COPLANAR LEADS (FLAT LEADS). Ribbon-type leads extending from


the sides of the circuit package, all lying in the same plane for surface mount appli-
cations.

COPOLYMER. A compound, resulting from the chemical reaction and poly-


merization of two chemically different monomers. The resulting larger molecules
contain repeating structural units of the original molecules.

CORDIERITE. A crystalline ceramic material of composition 2MgO-2A120r


5Si02 that can be crystallized from glass of same composition or sintered
from powders.

COUPLED NOISE. (Same as Cross talk) The electromagnetic and electrostatic


linkages between two nearby conductors that allow one line to induce a signal
on the other. See also Cross Talk.

COUPLER. A chemical agent, frequently an organosilane, used to enhance


the bond between a resin and a glass reinforcement.

COUPLING CAPACITOR. A capacitor used to block dc signals, and to pass


high-frequency signals between parts of an electronic circuit.

CRAZING. Fine cracks which may extend on or through layers of plastic or


glass materials.

CREEP. Nonrecoverable deformation proceeding at relatively low strain rates,


less than about lQ-6/sec, usually associated with sufficiently high temperature
to allow significant rates of diffusion.

CROSSOVER. The transverse crossing of metallization paths without mutual


electrical contact. This is achieved by the deposition of an insulating layer between
the conducting paths at the area of crossing.

CROSS TALK. Signals from one line leaking into another nearby conductor
because of capacitance or inductive coupling or both (i.e., owing to the capacitance
of a thick-film crossover.).
11-940 GLOSSARY AND SYMBOLS

CRYSTALLIZATION. Formation of crystalline phase out of amorphous ma-


terial during high-temperature processing. Undesirable or uncontrollable crystalli-
zation is called divitrification.

CUMULATIVE DISTRIBUTION FUNCTION (CDF). Distribution of a pa-


rameter as a fraction of the total number of measurements with respect to a
statistic, e.g. "probits" or standard deviations relative to a particular statistical
distribution, i.e., normal, etc.

CURING AGENT. An inorganic or organic compound which initiates the


polymerization of a resin.

CURING CYCLE. For a thermosetting material, commonly a resin compound


such as a bonding adhesive, it is the combination of total time-temperature profile
to achieve the desired result; for example, the complete irreversible hardening
of the material, resulting in a strong bond.

CURRENT CARRYING CAPACITY. The maximum current that can be


continuously carried by a circuit without causing objectionable degradation in
electrical or mechanical properties.

CURRENT SLEW RATE. The rate of change in current with respect to time
(dildt).

CUSTOM DESIGN. A form of design in which the choice and arrangement


of components and wiring on a package may vary arbitrarily within tolerances
from a regular array.

CYCLE TIME. Unit of time in which elements of the central processor com-
plete their logical functions. Some elements will require more than one cycle to
complete a function. See Cycles per Instruction.

CYCLES PER INSTRUCTION. The number of cycles required to process


an instruction.

DECOUPLING CAPACITOR. A shunt-placed capacitance that is used to


filter transients on a power distribution system.

DELAY EQUATIONS. A set of mathematical terms that are used to predict the
propagation times between driving and receiving circuits that are interconnected
GLOSSARY AND SYMBOLS 11-941

through signal wires. These equations are usually derived from simulation data
using numerical curve fitting techniques.

DELTA-I NOISE (AI). See Switching Noise.

DESIGN LIMITS. The fail points that are incorporated into the hardware
design rules that drive the computer-aided design system. See noise rules and
wiring rules.

DEVITRIFICATION. The undesirable formation of crystals in glass during


firing. The desirable process is called crystallization.

DEW POINT. The temperature at which moisture at a given partial pressure


becomes saturated, and when cooled below which, it condenses.

DIE. Integrated circuit chip as cut (diced) from finished wafer. See Chip.

DIE BOND. Mechanical attachment of silicon to substrate usually by solder,


epoxy, or gold-silicon eutectic, including interface metallurgies on chip and
substrate. The die bond is made to the back (inactive) side of the chip with the
circuit side (face) up.

DIELECTRIC. Material that does not conduct electricity. Generally used for
making capacitors, insulating conductors (as in crossover and multilayered cir-
cuits), and for encapsulating circuits.

DIELECTRIC CONSTANT. The term used to describe a material's ability


to store charge when used as a capacitor dielectric. It is the ratio of the charge
that would be stored with free space to that stored with the material in question
as the dielectric.

DIELECTRIC LOSS. The power dissipated by a dielectric as the friction of its


molecules opposes the molecular motion produced by an alternative electric field.

DIFFERENTIAL SCANNING CALORIMETRY (DSC). A technique for


measuring the physical transitions of a polymer as a function of temperature
compared to another material undergoing a similar heating process but not under-
going any transitions or reactions. DSC uses a servo system to supply energy at
a varying rate to both sample and reference so as to keep their temperatures
equal. A DSC output plots energy supplied vs. average temperature.

DIRECT ACCESS STORAGE DEVICE (DASD). Computer storage hard-


ware subsystem that uses magnetic recording on a rotating disk surface. Access
11-942 GLOSSARY AND SYMBOLS

to the information is accomplished with the use of a moveable arm which positions
one or more read/write heads along the radius of the disk to the desired track.

DIRECT CHIP ATTACH (DCA). A name applied to any of the chip-to-


substrate connections used to eliminate the first level of packaging: see also
Chip-on-Board.

DISCRETE COMPONENT. Individual components or elements, such as re-


sistors, capacitors, transistors, diodes, inductors, and others, as self-contained
entities.

DISTANCE TO NEUTRAL POINT (DNP). The separation of a joint from


the neutral point on a chip. This dimension controls the strain on the joint imposed
by expansion mismatch between chip and substrate. The neutral point is usually
the geometric center of an array of pads and defines the point at which there is
no relative motion of chip and substrate in the X-Yplane during thermal cycling.

DOCTOR BLADE. A method of casting slurry into a thin sheet by the use
of knife blade placed over moving carrier to control slurry thickness.

DOUBLE-SIDED SUBSTRATE. A substrate carrying active circuitry on both


its topside and bottomside, electrically connected by means of metallized through-
holes or edge metallization or both.

DRIVER. The off chip circuit that supplies the signal voltage and current to
the package lines. Also called an output buffer circuit.

DRY FILM PHOTORESIST. Photoresist material that is processed dry, usu-


ally by lamination of prefabricated film.

DRY PRESSING. Pressing and compacting together of dry powdered materials


with additives in rigid die molds under heat and pressure to form a solid mass,
usually followed by sintering to form shapes.

DUAL-IN-LINE PACKAGE (DIP). A package having two rows of leads


extending at right angles from the base and having standard spacings between
leads and between rows of leads. DIPs are made of ceramic (Cerdip) and plas-
tic (Pdip).

DYNAMIC FLEX. A form of flexible circuitry developed for applications


where continued flexure is necessary. In contrast, static (flex), once installed,
remains fixed.
GLOSSARY AND SYMBOLS /1-943

DYNAMIC RANDOM ACCESS MEMORY (DRAM). Electronic informa-


tion storage that employs transient phenomena, typically charge stored in a leaky
capacitor. Refresh cycles are required to restore and thus maintain the information.
DRAM is the simplest and least expensive of electronic memories, but it is also
the least impressive performer.

EFFECTIVE INDUCTANCE (LEFF). A simplified characterization of the


goodness of an AC power distribution system. It consists of a single lumped
inductance that when multiplied by the total current slew rate predicts the total
switching noise across the circuit load.

ELECTRICALLY LONG TRANSMISSION LINE. One in which the delay


from the near-end to the far-end is greater than one-half of the near-end signal's
transition time. When this occurs, the reflections from the far-end do not distort
the near-end signal during its transition time.

ELECTRICALLY SHORT TRANSMISSION LINE. One in which the delay


from the near-end to the far-end is less than one-half of the near-end signal's
transition time. For this case, reflections from the far-end interfere with the near-
end transition waveform causing distortion that usually increases the net's delay.

ELECTROLESS PLATING. Metal deposition, usually in an aqueous me-


dium, which proceeds by an exchange reaction between metal complexes in the
solution and the particular metal to be coated; the reaction does not require
externally applied electric current.

ELECTROPLATING. Deposition of an adherent metallic coating onto a con-


ductive object placed into an electrolytic bath composed of a solution of the salt
of the metal to be plated. Using the terminal as the anode (possibly of the same
metal as the one used for plating), a DC current is passed through the solution
affecting transfer of metals ions onto the cathodic surface.

ELECTROSTATIC DISCHARGE (ESD). Discharge of static charge on a


surface or body through a conductive path to ground. An electronic component or
higher-level assembly may suffer damage when it is included in the discharge path.

ELONGATION. The ratio of the increase in wire length at rupture, in a tensile


test, to the initial length, expressed in percent.

EMITTER-COUPLED LOGIC (ECL). Emitter-coupled logic is also known


as current-switch (logic) circuits. In it, a current source feeds emitters of several
11-'944 GLOSSARY AND SYMBOLS

transistors. The base of all but one acts as input tenninals; the last base is
connected to a reference voltage. Very popular circuit for high-performance
applications, it is often combined with an emitter-follower output stage to further
enhance its performance. It is then called SeEF, for current-switch emitter fol-
lower.

ENAMELING. A process that produces pore-free glass dielectric coating over


a metal-core substrate.

ENCAPSULATION. Sealing up or covering an element or circuit for mechani-


cal and environmental protection.

END OF LIFE (EOL). The end of the useful operating life of a component
or equipment detennined by a "wear-out" or life tenninating mechanism measured
in units of time. EOL is usually specified as an objective in reliability calculations.

ENGINEERING CHANGE (EC). A change in design. An electrical design


change is frequently implanted by cutting out or adding an electrical path to
the manufactured hardware, e.g., laser deleting a line or adding a wire on a
ceramic substrate.

ENTITY. A group of circuits separated from other circuits by a physical


package boundary and associated input and output connections.

EUTECTIC. A term applied to the mixture of two or more substances with


the the lowest melting point possible between those components.

EXTERNAL RESISTANCE. A term used to represent thermal resistance


from a convenient point on the outside surface of an electronic package to an
ambient reference point.

FAILURE. The temporary or permanent impairment of device function caused


by physical, chemical, mechanical, electrical, or electromagnetic disturbance
or damage.

FAILURE RATE. The rate at which devices from a given population can
be expected (or were found) to fail as a function of time (e.g., %/1000 hr.
of operation).

FAST WAVE PROPAGATION. The transmission of energy along a signal


line at the speed-of-light (velocity) expected for the dielectric structure. In the
GLOSSARY AND SYMBOLS 11-945

case of a low-loss line, the fast wave refers to the portion of the signal that
travels at the velocity expected for the dielectric medium.

FATIGUE. Used to describe the failure of any structure caused by repeated


application of stress over a period of time.

FERROELECTRIC. A crystalline dielectric that exhibits dielectric hysteresis;


an electrostatic analogy to ferromagnetic materials.

FIELD EFFECT TRANSISTOR (FET). A transistor in which a voltage


applied to a thin conductor over a thin insulator controls current flow in a
semiconductor region (gate) or one polar type. This component originates and
terminates in two regions of the opposite polar type located at either end of the
gate region.

FIELD REPLACEABLE UNIT (FRU). A component or sub-system of an


electronic assembly which may be replaced at the site of installation. A first- or
second-level package is commonly an FRU for most computers.

FILLER. A substance, usually ceramic or metal powder, used to modify the


properties of fluids or polymers.

FINITE ELEMENT MODELING. A computationally intensive numerical


modeling tool in which the body is discretized into small regularly shaped ele-
ments.

FIRST-INCIDENT SWITCHING. The case that occurs when all of the re-
ceivers on a multi-drop net switch at the first time the signal arrives from the
driver. Nets that are not first-incident are referred to as multi-reflection nets.

FLAME RETARDER. An inorganic or organic compound added to a polymer


mixture that causes the resulting plastic to self-extinguish after a flame is removed.

FLAT PAC. An integrated circuit package having its leads extending from all
four sides and parallel to the base.

FLEXIBLE CIRCUIT CARRIER. Printed circuits employing flexible sub-


strates, processed by patterning copper onto thin flexible Kapton or polyimide
films. Originally used only as connector; now employed for multilayers.

FLEXIBLE COATING. A plastic coating that is still flexible after curing.

FLEXURAL STRENGTH. Strength of a material measured by bending, typi-


cally used for brittle materials, such as glasses and ceramics; expressed in MPa.
11-946 GLOSSARY AND SYMBOLS

FLIP-CHIP. A leadless, monolithic structure containing circuit elements,


which is designed to electrically and mechanically interconnect to the hybrid
circuit by means of an appropriate number of bumps, which are covered with a
conductive bonding agent, located on its face. Alternatively, bonding of chips
with contact pads, face down by solder connection. See also Controlled Collapse
Chip Connection (C4).

FLOOR PLANNING. A procedure in physical design which permits approxi-


mate shaping and placement of logic and memory circuit groupings on a package
before final placement and wiring.

FLOW REGIME, LAMINAR. Flow where fluid layers are undisturbed and
smooth.

FLOW REGIME, TURBULENT. Flow where fluid particles are disturbed


and fluctuate.

FLUX. In soldering, a material that chemically attacks surface oxides so that


molten solder can wet the surface to be soldered, or an inert liquid which excludes
oxygen during the soldering process.

FRACTURE TOUGHNESS. A basic property of a material, or of an interface


between dissimilar materials, describing its crack resistance in a mechanical or
thermomechanical stress field.

FRIT. Glass composition ground up into a powder form and used in thick-
film compositions as the portion of the composition that melts upon firing to
give adhesion to the substrate and hold the conductive composition together.

FRONT-END-OF-THE-LINE (FEOL). See back-end-of-the-line.

FR-4. Electronic Industries Association's designation for a fireretardant epoxy


resin/glass cloth laminate. By common usage, the resin for such a laminate.

G-I0. It is a grade of epoxy-impregnated glass cloth printed-circuit board


material per NEMA (National Electrical Manufacturers Assoc.).

GATE ARRAY. A semicustom product, implemented from a fully diffused


or ion-implanted semiconductor wafer carrying a matrix of identical primary
cells arranged into columns with routing channels between them in the X and
Y directions.
GLOSSARY AND SYMBOLS 11-947

GATE, LOGIC. Usually an electric circuit which combines information of its


two inputs to form its output signal in accordance to the logic function it performs.

GATE, SEA OF (see also Gate Array). A form of custom chip layout in
which the wiring tracks required for interconnecting a fixed array of logic cells
are disposed in rows and columns having widths measured in numbers of wiring
tracks per row or column channel which vary to suit the local wiring demand,
from point to point and from one logic product to another.

GATE, STRUCTURAL. Term used to designate a hinged frame which con-


tains a number of boards and can swing out for servicing and/or access to the
interior of a fixed frame.

GLASS. Inorganic, nonmetallic and amorphous material obtained by melting


oxide(s) into glass and retaining the structure by fast cooling.

GLASS-CERAMIC. Inorganic, nonmetallic material obtained by controlled


crystallization of glass into nonporous and fine microstructure.

GLASS + CERAMIC. Inorganic, nonmetallic material obtained by admixing


crystalline ceramic with glass and sintering composite.

GLASS ± CERAMIC. Refers to family of glass-ceramic and glass + ceramic.

GLASS FABRIC. Cloth woven from glass yarns which are made of filaments.

GLASS TRANSITION TEMPERATURE (TG). In polymer or glass chemis-


try, the temperature corresponding to the glass-to-liquid transition, below which
the thermal expansion coefficient is low and nearly constant, and above which
it is very high.

GLAZED SUBSTRATE. A glass coated ceramic substrate that effects a


smooth and nonporous surface.

GLAZING. In the present context, glazing refers to the coating of a smooth,


adherent layer of glass, such as a sealing glass, by melting the glass over metal
or ceramic surfaces.

GLOB TOP. A glob of encapsulant material surrounding a chip in the chip-


on-board assembly process. The attached chip must have already passed pretest
and inspection because rework after final curing of the epoxy or silicone globs
is virtually impossible.
11-948 GLOSSARY AND SYMBOLS

GLOBAL WIRING. Wiring interconnecting components mounted on a pack-


age (as opposed to the wiring inside the components). Also refers to that wiring
independent of its detailed allocation to wiring tracks within a channel.

GREEN. A term-unrelated to the actual material color- used in ceramic


technology meaning unfired. For example a "green" substrate is one that has
been formed, but has not been fired.

GREEN-SHEET. A composite organic-inorganic, flexible sheet ready for met-


allization, if desired, and lamination to form green substrates which upon removal
of organics results in ceramic substrate. Green refers to the unfired state.

GROUND PLANE. A conductive layer on a substrate or buried within a


substrate that connects a number of points to one or more grounding electrodes.

GULLWING. A common lead form used to interconnect surface mounted


packages to the printed-circuit board. The leads, normally 100 to 250 J.1m thick,
are bent outward, downward, then again outward from the package body, provid-
ing feet for solder interconnection, and some degree of mechanical compliance.

HARD GLASS. Glasses having a high softening temperature (>700°C), such


as the borosilicate glasses used to seal feedthrough leads into metal packages.

HEAT FLUX. The rate of flow of heat energy across or through a surface,
measured in watts/cm2•

HEAT SINK. The supporting member to which electronic components or their


substrate or their package bottom are attached. This is usually a heat conductive
metal with the ability to rapidly transmit heat from the generating source (com-
ponent).

HERMETIC. Sealed so that the object is gastight. The test for hermeticity is
to fill the object with a test gas-often helium-and observe leak rates when
placed in a vacuum. A plastic encapsulation cannot be hermetic as it allows gases
to permeate.

HIGH-LEVEL NOISE TOLERANCE (NTH). The receiver noise tolerance


that occurs when the input signal is in its UP state.

HOMOGENEOUS MEDIUM. A signal propagation structure in which only


a single dielectric is present.
GLOSSARY AND SYMBOLS 11-949

HOT-GAS REFLOW. The technique in which a heated gas, including air, is


impinged on a site to be solder-reflowed, usually to form a solder interconnection.

HOT-KNIFE SOLDERING. The technique in which a heated blade (heated


electrically or conductively) is used as a heat source for melting solder during
package joining. The blade may be used to force mechanical contact throughout
the joining process.

HYBRID MODULE. A special carrier of hybrid microcircuits and other com-


ponents interconnected as a unit, or as a component of an electronic subsystem.
The module may be of single construction or made up of submodules, each
usually with a compartment to house hermetically packaged hybrids and discrete
passive component parts, such as transformers, axle-lead resistors, etc. Nonher-
metic hybrid modules generally are parylene coated. In this book, hybrid module
is also referred to a module containing a combination of thick and thin films.

IMPREGNATION. The process of coating a substrate-say glass cloth-with


a resin solution and drying. The dried product is called prepreg.

INDUCTANCE. The electromagnetic element that stores flux lines. In packag-


ing systems, it is used in lumped equivalent circuits to represent part of a line
discontinuity. It is also used in a distributed system to represent the electromag-
netic storage property of a transmission line. Because it induces an opposing
voltage in response to a change in current, it causes package delta I (LlI) noise.

INERT ATMOSPHERE. A gas atmosphere such as helium or nitrogen that


is nonoxidizing or nonreducing to metals.

INFRARED REFLOW (IR). The technique in which primarily long wave-


length light is used to heat solder joints to the melting temperature. Normally,
a circuit board having prepositioned packages is transported through an IR re-
flow furnace.

INHOMOGENEOUS MEDIUM. A signal propagation structure in which


multiple dielectrics are present.

INJECTION MOLDED. Molding by injecting liquefied plastic into a mold


of desired shape.

INJECTION MOLDED CARD (IMC). Card for electronic packages made


by injection molding of plastics into a mold cavity of desired shape.
11-950 GLOSSARY AND SYMBOLS

INPUT/OUTPUT TERMINAL (110). A chip or package connector (terminal)


acting to interconnect the chip to the package or one package level to the physically
adjacent level in the hierarchy. Usually refers to the number of contacts necessary
to wire to or interconnect an assembly. Pin out, connections, and terminals are
other common words to describe the same. Care must be taken to differentiate
between the total number of lias between levels, signal lias, lias used to
distribute power, and reference lias.

INSULATION RESISTANCE (IR). The resistance to current flow when a


potential is applied. IR is measured in megohms.

INSULATOR METAL SUBSTRATE TECHNOLOGY (IMST). A sub-


strate, such as one made of porcelainized steel, which is not subject to size
limitations and may have superior thermal dissipation characteristics. IMST refers
to insulated-metal substrate technology of Sanyo. It is a single-sided aluminum
core with epoxy coating and etched copper wiring.

INSULATORS. A class of materials with high resistivity. Materials that do


not conduct electricity. Materials with resistivity values of over 106 Q cm are
generally classified as insulators.

INTEGRATED CIRCUIT. A microcircuit (monolithic) consisting of inter-


connected elements inseparably associated and formed in situ on or within a
single substrate (usually silicon) to perform an electronic circuit function.

INTERCHIP WIRING. The conducting wiring path connecting circuits on


one chip with those on other chips to perform a function.

INTERCONNECTION. The conductive path required to achieve connection


from a circuit element to the rest of the circuit.

INTERFACE. The boundary between dissimilar materials, such as between a


film and substrate, or between two films.

INTERNAL RESISTANCE. A term used to represent thermal resistance from


the junction of a device, inside an electronic package, to a convenient point on
the outside surface of the package.

INTERPENETRATING POLYMER NETWORK (IPN). A polymer alloy


made up of two or more crosslinked polymers. The networks interact with each
other only through permanent physical entanglements rather than through covalent
bonding as in a copolymer.
GLOSSARY AND SYMBOLS 11-951

INVAR. A trademark of International Nickel Co., Inc. for a very low thermal
expansion alloy of nickel and iron.

ION MIGRATION. The movement of free ions within a material or across


the boundary between two materials under the influence of an applied electric field.

ISOPAK. An unique pin-grid array consisting of Kovar pins sealed in glass-


to-Klovar plate flush for chip bonding.

J-LEAD. An I.C. package terminal lead configuration that resembles in crossec-


tion, the letter "J".

JOSEPHSON SUPERCONDUCTING DEVICE. Superconducting ceramics


acting as Josephson devices, at very low temperatures, typically at 4°K.

KEEPER BAR, TAB. A strip of dielectric material-such as polyimide-that


remains attached to each row of outer-TAB leads following excise operation. It
helps to reduce misalignment, non-planarity and damage during outer lead
bonding.

KIRKENDAHL VOIDING. Voids induced in a diffusion couple between two


metals with different interdiffusion coefficients.

KNOWN GOOD DIE (KGD). IC semiconductor chips that have been tested
before being packaged and are known to function as required.

KOVAR. An alloy of iron (53%), Cobalt (17%), and nickel (29%) with thermal
expansion matching alumina substrate and certain sealing glasses. Most common
lead frame and pin material.

LAMINATION. The process of consolidating sheets of prepreg under heat


and pressure to form a solid product. Applied also to the consolidation of prepregs
and precircuitized subcomposites to form a composite.
11-952 GLOSSARY AND SYMBOLS

LARGE-SCALE INTEGRATION (LSI). Large-scale integration. Term used


to designate chips with more than one thousand transistors.

LASER SOLDERING. The technique in which heat to reflow a solder inter-


connection is provided by a laser, usually a longer wavelength YAG or CO2
laser. The joints are heated sequentially, and cooled rapidly.

LEAD FRAME. A sheet metal framework on which a chip is attached, wire-


bonded, and then molded with plastic.

LEADED-CHIP CARRIER PLASTIC (PLCC). A plastic package contain-


ing a chip that has terminal leads emanating from four sides. Each lead has a J-
configuration and is designed for surface mounting to a printed-circuit board.

LEADLESS-CHIP CARRIER (LLCC). A surface mounted package having


metallized contacts at its periphery (rather than wire leads) which are soldered
to metallized contacts on the printed-circuit board or substrate.

LIFT-OFF. Pattering of metal by lift off materials around usually in a solvent.

LINE DISCONTINUITY. A load point, consisting of a lumped equivalent


circuit of resistance, capacitance, and inductance anywhere on a transmission
line that produces spurious reflections.

LINE LOADING. Externally connected resistance, inductance, and capaci-


tance, or combination of these on a transmission line.

LINE RESISTANCE. Resistance of conductor lines in a package, measured


in ohms per unit length or for a given cross section, ohms per square.

LINES PER CHANNEL. The number of conductive lines between through


holes in an organic board or ceramic substrate.

LIQUID CRYSTAL DISPLAY (LCD). Display technology based on liquid


crystal materials whose light transmission is changeable by the application of an
electrical field. LCD devices are used in numeric read outs and for flat screen
television receivers.

LOGIC DESIGN. The process of determining the choice and interconnection


of logic units (e.g., nands or nors) to accomplish logical functions in an overall
digital system.

LOGIC PART. A physical implementation of an interconnected and packaged


group of logic circuits used in general more than once in a digital system.
GLOSSARY AND SYMBOLS 11-953

LOGIC PRIMITIVE. A basic logic function embodied as a single unit.

LOGIC SERVICE TERMINAL (LST). A terminal (on a package or package


component) carrying logic signals as opposed to one used only for electrical
power.

LOSS-LESS TRANSMISSION LINE. A signal path where the total series


resistance is less than 10% of the characteristic impedance of the line. For this
case, the signal level attenuation is approximately 10% (VoutlVin) when the line
is terminated in it's characteristic impedance.

LOSSY TRANSMISSION LINE. A signal path with total series resistance


that exceeds two times the characteristic impedance of the line. When this occurs,
the fast-wave portion of the input signal is attenuated by more than "lie" or 63%
(VoutlVin) when the line is terminated in it's characteristic impedance.

LOW LEVEL NOISE TOLERANCE (NTL). The receiver noise tolerance


that occurs when the input signal is in its DOWN state.

LOW-LOSS TRANSMISSION LINE. A signal path where the total series


resistance is greater than 0.1 times but less than two times the characteristic
impedance of the line. In this range, the signal level attenuation is between 10%
and 63% (VoutlVin) when the line is terminated in it's characteristic impedance.

MACRO. A collection of continuous cells defined to be placed as a group


within a chip image.

MANHATTAN DISTANCE (LENGTH). Wire length between terminals of


a net or connection measured in X or Y directions on a package wiring plane
as on city blocks.

MASK. The photographic negative that serves as the master for making thick-
film screens and thin-film patterns.

MCM-C. A multi-chip module with a structure that consists of multi-layer co-


fired ceramic.

MCM-D. A multi-chip module structure that consists of deposited thin film


organic layers.
11-954 GLOSSARY AND SYMBOLS

MCM-L. A multi-chip module structure that consists of laminated organic


layers.

MEAN TIME TO FAILURE (MTTF). Applicable to individual parts or


devices in reliability technology. It is the arithmetic average of the lengths of
time-to-failure registered for parts or devices of the same type, operated as a
group under identical conditions.

MEMORY ADDER. An adder, to the basic cycle-per-instruction rate of the


Central Processor (CP), due to requirements for data or instructions, not available
in the CP when necessary.

METAL MIGRATION. An undesirable phenomenon whereby metal ions,


notably silver, are transmitted through another metal or across an insulated
surface, in the presence of moisture and an electrical potential.

METALLIZATION. A film pattern (single or multilayer) of conductive mate-


rial deposited on a substrate to interconnect electronic components.

METALLIZED CERAMIC (MC). Ceramic (fired) substrate metallized with


thick and thin films of metals. (In IBM, metallized ceramic refers to thin-film
metallization on fired alumina substrate.)

MICRON. An obsolete unit of length equal to a micrometer (Ilm).

MICROSTRIP LINE. A signal line on the surface of a dielectric with air


above it and a reference plane on the opposite side of the dielectric.

MICROSTRUCTURE. Structural features, such as crystal or phase bound-


aries, or defects or inhomogeneities within a solid, usually resolved at high magni-
fication.

MINER'S RULE. An outgrowth of the cumulative damage concept, this for-


mula predicts the fatigue lifetime of a structural element when the load history
encompasses multiple types or various amplitudes of stress.

MODULE. A chip carrier on which the chip terminals are fed out by various
means to terminals spaced to suit the spacing and dimensions of wires on the
next higher level of package (i.e., card or board). It may also contain wiring
planes and power planes interconnecting several of its chips, and thus be used
as a card.

MOIRE. A technique to measure in-plane surface displacement, whereby a


precision grid of fine lines attached to the surface is deformed relative to a
GLOSSARY AND SYMBOLS 11-955

stationary or reference grid. The resulting fringe pattern can be analyzed to


provide displacement values throughout the field.

MOLD RELEASE. An organic compound added to a molding compound or


powder that migrates to the mold surface to form a waxy layer between the
plastic and mold metal and to allow easy removal of the part from the mold.

MONOLITIDC SYSTEMS TECHNOLOGY (MST). Ceramic package


manufactured in IBM by screen printing Ag-Pd conductors onto alumina substrate
swagged with pins and chips bonded with solder connection (C4).

MONTE CARLO ANALYSIS. A statistical analysis method whereby the


resultant distribution is generated by using a random number generator to pick
a large number of cases from the input distributions.

MUL TICHIP MODULE (MCM). A module or package capable of supporting


several chips on a single package. Most multichip packages are made of ceramic.

MULTICHIP PACKAGE. An electronic package that carries a number of


chips and interconnects them through several layers of conductive patterns. Each
one is separated by insulative layer and interconnected via holes.

MULTILA YER CERAMIC (MLC). Ceramic substrate consisting of multiple


layers of metals and ceramics interconnected with vias. All with thick film.

MULTILAYER SUBSTRATES. Substrates that have buried conductors so


that complex circuitry can be handled, using assembly processes similar to those
used in multilayer ceramic capacitors. Fabricated either as a conventional MLC,
or a cofired multilayer ceramic (CMC) hybrid structure, in high- and low-tempera-
ture versions.

MULTI-REFLECTION SWITCIDNG. A network that is not first incident.


See First-Incident Switching.

NEAR INFRARED REFLECTANCE ANALYSIS (NIRA). Infrared spec-


troscopy that covers the region from 0.75 Jlm to 2.5 Jlm. Regular infrared spectro-
photometers, on the other hand, use a glowing light source to provide light with
wavelengths from 2.5 to about 15 Jlm.

NET. A group of terminals interconnected to have a common dc electrical


potential in a package.
11-956 GLOSSARY AND SYMBOLS

NOISE. In a digital system, noise is any undesirable parasitic effect that causes
signal waveform distortion, excessive delay, or false switching. Common types
of noise are: reflection, coupled and switching.

NOISE SATURATION. A phenomenon in which switching noise does not


increase as the number of simultaneously switching circuits increases beyond an
observed number. This effect is caused by a negative feedback between the large
noise generated and the current drawn by each of the switching circuits. An
increase in circuit delay occurs when noise saturation occurs.

NONLINEAR DIELECTRIC. A capacitor material that has a nonlinear capac-


itance-to-voltage relationship ..hw Ti-ta-nates (usually barium titanate) ceramic
capacitors (Class II) are nonlinear dielectrics. NPO and Class I capacitors are
linear by definition.

o
OPTICAL INTERCONNECTS. Composed of the basic optoelectronic de-
vices, these are components and modules used as circuit building blocks. The
light-emitting diode (LED) converts electrical energy to light where junction
electroluminescence occurs as a result of the application of direct current at low
voltage to a suitably doped crystal when forward biased. The light from this
source then is detected by the reverse-biased pn-junction photodiode and/or
phototransistor. Light, of the proper wavelength, creates a current flow, a photo-
current, in the external circuit proportional to the effective irradiance on the
device.

OUTER LEAD BONDING. The process of joining the outer leads of a pack-
age, typically TAB (Tape Automated Bonding) to the next level of assembly
(usually card or board). The inner leads on the tape are joined to the chip by the
process know as inner lead bonding.

OVERCOAT. A thin film of insulating material, either plastic or inorganic


(e.g., glass or silicon nitride), applied over integrated circuit elements for the
purposes of mechanical protection and prevention of contamination.

OVERFLOW WIRE. See Wiring Overflow.

OVERGLAZE. A glass coating that is grown, or deposited, over another


element, normally for physical or electrical protection purposes.

OVERLAY. One material applied over another material.


GLOSSARY AND SYMBOLS 11-957

PACKAGE CROSSING. An interconnection which connects a terminal on


one package with that on another.

PACKAGE DELAY. The time delays associated with the interconnections


between components that complete logical make-up functions. Values depend
on materials and distance.

PACKAGING LEVEL. A member of a nested interconnected packaging hier-


archy (e.g., chip, chip carrier, card, board in order of low to high level).

PAD-GRID ARRAY PACKAGE (FOR VHSICS). A package embodying a


rather novel technology, where solder-contact pads are not just around the package
periphery (as with chip carriers) but cover the entire bottom surface in checker-
board fashion.

PASSIVATION. The formation of an insulating layer directly over a circuit


or circuit element to protect the surface from contaminants, moisture, or particles.

PASSIVE COMPONENTS (ELEMENTS). Elements or components such as


resistors, capacitors, and inductors which do not change their basic character
when an electrical signal is applied. In contrast transistors, diodes, and electron
tubes are active components.

PASTE. Synonymous with "composition" and "ink" when relating to screen-


able thick-film materials, usually consisting of metal or ceramic powders dispersed
in organic vehicles.

PEEL STRENGTH (PEEL TEST). A measure of adhesion between a conduc-


tor and the substrate. The test is performed by pulling or peeling the conductor
off the substrate and observing the force required.

PERMEABILITY. The property of a solid plastic material that allows penetra-


tion by a liquid or gas.

PHASE DIAGRAM. State of a metal alloy or ceramic over a wide temperature


range. The phase diagram is used to identify phases as a function of composition
and temperature.

PHOSPHOSILICATE GLASS (PSG). Phosphorus-doped silicon dioxide


(also known as P-glass). It is often used as a dielectric material for insulation
between conducting layers, for inhibiting the diffusion of sodium impurities, and
11-958 GLOSSARY AND SYMBOLS

for planarization since it softens and flows at 1,000 to 1,1 OO°C to create a smooth
topography for subsequent metallization.

PHOTOLITHOGRAPHY. The generation of a pattern through a sequence of


rubylith, photo-reduction, step-and-repeat, computer-aided design, or finally, stat-
of-the-art electron-beam technique. This procedure will generate a product (mask
or otherwise) to become the primary tool in transferring an image onto a micro-
electronic substrate.

PHYSICAL DESIGN. The process of allocating chip or package components


and their interconnections to their appropriate spatial locations or sockets in an
overall system arrangement.

PICK-AND-PLACE. The manufacturing process whereby chips are selected


and placed on the correct substrate site in preparation for joining (or interconnect-
ing) the chip to the substrate.

PIN. Round, cross-sectional electrical terminal and/or mechanical support.


Used in plug-in type packages, either straight or modified as nail-head, upset,
pierced, or bent variety. A pin's primary functions are, internally, to support a
wirebond or other joint, and, externally, to plug into a second-level package con-
nector.

PIN-GRID ARRAY (PGA). A package or interconnect scheme featuring a


multiplicity of plug-in type electrical terminals arranged in a prescribed matrix
format or array.

PIN-THROUGH-HOLE (PTH). A term referring to the class of packages or


modules that are soldered into plated through holes within the second-level
package (printed-circuit board).

PITCH. The center-to-center spacing, between pads, rows of bumps, pins,


posts, exit leads, etc. Sometimes also the distance as measured point to correspond-
ing point between two adjacent images in a device matrix, on a semiconductor
wafer or its photomask.

PLACEMENT. The manual or automatic placing of chip circuits, chips, chip


carriers, and cards in their actual locations on corresponding images at a given
package level.

PLANAR MOTOR. Typically a brushless DC servo motor of a flat planar


configuration constructed using printed-circuit assembly methods.
GLOSSARY AND SYMBOLS 11-959

PLANAR MOTOR VOICE COIL SERVO. A galvanometer type mechanical


positioner similar in design to the conventional audio loudspeaker voice coil
assembly. Provides limited excursion capability, but extremely high speed, and
with suitable feedback, very high accuracy.

PLASMA. An electrically conductive gas, composed of ionized atoms or mole-


cules, used for dry-etching in the fabrication of devices.

PLASMA ETCHING. The action of an electrically conductive gas, (composed


of ionized gas or molecules), to remove unwanted portion of conductive or
insulative pattern.

PLASTIC. A polymeric material, either organic (e.g., epoxy, polyimide) or


inorganic (e.g., silicone) used for conformal coating, encapsulation, or over-
coating.

PLASTIC BALL GRID ARRAY (PBGA). Ball grid array technology on a


plastic carrier. See Ball Grid Array.

PLASTIC QUAD FLAT PACK (PQFP). See Quad Flat Pack.

PLATING. A condensation of the word Electroplating or Electroless Plating


that describes the coating of a metal on plastic or other surfaces with metal that
is electrolytically or chemically deposited from a bath.

POLYIMIDES. A class of resin compounds containing the NH group which


are derived from ammonia and are "imidized" from polyamic acid at temperatures
high enough to initiate and complete the imide ring closure. Polyimides are useful
as organic dielectric interlevel layers in VLSI technologies. They are mostly
thermosetting ring-chain polymers, whose useful characteristics include perfect
planarity as a (spun) film; high temperature tolerance; excellent weathering and
mechanical-wear characteristics; and a low dielectric constant, a decided advan-
tage in reducing propagation delays in multilayer hybrid circuits (faster
switching).

PORCELAIN. A mixture of borosilicate glass with minor quantities of zirconia


and other ingredients. It is used synonymous with enamel in this book.

PORCELAIN ENAMEL TECHNOLOGY (PET). The technology of coating


glass on metal. See Porcelain.

POWER CYCLING. A method of imposing a cyclic stress on an assembly


of microelectronic components by applying cyclic power to a heat generating
11-960 GLOSSARY AND SYMBOLS

component in the assembly. It is used for accelerated reliability testing of assem-


blies.

POWER DISTRIBUTION. The network of conductors throughout the pack-


age that supplies the operating voltages and currents to the circuits.

PRESSURE CONTACT. Mode of interconnection where the contact points


are not fully bonded (as in soldered) but maintain electrical contact by means
of a continuously applied force (such as a spring or rubber).

PRINTED-CIRCUIT BOARD (PCB). A composite of organic and inorganic


material with external and internal wiring allowing electronic components to be
mechanically supported and electrically connected.

PROBABILITY DENSITY FUNCTION (PDF). The normalized frequency


of occurrence with respect to a particular statistical distribution, e.g., normal,
binomial, Weibull, etc.

PURPLE PLAGUE. One of several gold-aluminum compounds formed when


bonding gold to aluminum and activated by re-exposure to moisture. High temper-
ature Purple Plague is purplish in color and is very brittle, potentially leading to
time-based failure of the bonds. Its growth is highly enhanced by the presence
of silicon to form ternary compounds.

PYROLYZED (BURNED). A material that has gained its final form by the
action of heat is said to be pyrolyzed.

QUAD FLAT PAC (QFP). Ceramic or plastic chip carrier with leads projecting
down and away from all four sides of a square package.

QUAD IN-LINE PACKAGE (QUIP). A diplike plastic package with leads


coming out on 1.27 mm centers. Half of the leads bent close to the body, and
the other half projected out for additional 1.27 mm before bent-down.

RADIAL-SPREAD COATING. Also known as glob top. A coating process


whereby a calibrated amount of resin is dispensed on top of a surface to be
encapsulated. The surface can be either a chip or a circuit board. The resin fans out,
reacts (by heat input or on contact with air), and forms a solid protective coating.
GLOSSARY AND SYMBOLS 11-961

RADIATION. The combined process of emission, transmission, and absorption


of thermal energy between bodies separated by empty space.

REACTION ETCHING. A process wherein a printed pattern is formed by


reaction (chemical/plasma ion) removal of the unwanted portion of conductive
or insulative pattern.

REACTION INJECTION MOLDING (RIM). A molding process where two


(or more) streams of reactants are metered into a small mixing chamber where
turbulent mixing breaks up the fluids into finely interspersed striations for faster
reaction. The mixture is then delivered to a mold to complete the polymerization.

RECEIVER. The off-chip circuit that accepts the signal voltages and currents
from the package lines. Also called an input buffer circuit.

RECEIVER INPUT STABILITY. The assurance that package reactances and


high frequency circuit gain and phase shifts do not interact such that excessive
ringing or oscillations occur at the receiver's output.

REFLECTION NOISE. Spurious voltage and current wavelets on a transmis-


sion lines that are caused by series or shunt networks that disrupt the continuous
nature of the characteristic impedance of the line. These wavelets initially travel
in a direction that is opposite to the wave that stimulates them. Typical discontinu-
ities that cause reflections are stubs, connectors, vias, missing ground planes and
improper terminations.

RELATIVE HUMIDITY (rh). The -ratio of partial pressure of water in any


gas at a particular temperature to the saturated vapor pressure of water in the
same gas at the same temperature, usually expressed in percent.

RELIABILITY. The probability of survival of a component, or assembly, for


the expected period of use. Expressed mathematically, R =one minus the probabil-
ity of failure during the expected life.

REFLOW SOLDERING. A method of soldering involving application of


solder prior to the actual joining. To solder, the parts are joined and heated,
causing the solder to remelt or reflow.

RENT's RULE. An empirical relation, first recorded by E. Rent of IBM, which


states that the number of used input/output terminals on a logic package is
proportional to a fractional power of the number of subpackages interconnected
in the package.
/1-962 GLOSSARY AND SYMBOLS

RESIN. A term used for an organic polymer that when mixed with a curing
agent crosslinks to form a thermosetting plastic.

RESIST. A protective coating that will keep another material from attaching
or coating something, as in solder resist, plating resist, or photoresist.

RESISTANCE. The property of a conductor that opposes the flow of current


by dissipating energy as heat. In packages, it causes voltage and current losses
in signal and power distribution systems.

RESISTIVITY (p). A proportionality factor that is characteristic of different


substances, equal to the resistance that a centimeter cube of the substance offers
to the passage of electricity. Expressed R = pLlA where R is the resistance of
a uniform conductor, L its length, A its cross-sectional area, and p its resistivity.
Resistivity is usually expressed in ohm-centimeters.

RHEOLOGY. The science dealing with deformation and flow of matter.

ROSIN FLUX. A flux having a rosin base that becomes interactive after being
subjected to the soldering temperature.

ROUTING PROGRAM. An automatic program emboding algorithm with


prescribed wiring.

RULES DRIVEN DESIGN SYSTEM. The use of sufficiently accurate mathe-


matical expressions that are derived from a limited number of judiciously chosen
circuit simulation results using curve fitting techniques to properly design a
hardware system in conjunction with a design aids (DA) program. Since these
formulas are computationally fast, they can be applied to the design of every
data path in the system.

SELF-GENERATED NOISE TOLERANCE. A set of differential pulse am-


plitudes and pulse widths for spurious signals that are generated by the simultane-
ous switching of internal circuits that can be impressed across a internal circuit's
power terminals without falsely setting a downstream latch circuit.

SCREENING. The process whereby the desired film-circuit patterns and con-
figurations are transferred to the surface of the substrate during manufacture by
forcing a material through the open areas of the screen using the wiping action
of a soft squeegee.
GLOSSARY AND SYMBOLS 11-963

SEALING. Joining the package case header (or chip carrier base or substrate)
with its cover or lid into a sealed unit. For hybrids, sealing connotes an important
finishing operation in fabricating a hybrid microcircuit, signaling the stage when
the assembly, in the form of a populated package, becomes a bona fide hermetic
(or nonhermetic) device.

SELF STRETCHING SOLDERING TECHNOLOGY (SST). The acronym


for C4 solder joining where two different solder alloys or sized bumps are used
so that surface tension forces of the nonfunctional bumps are used to stretch or
increase the height of the functional solder joints. Taller connections can withstand
higher thermal cycle or power cycle strains.

SHEET RESISTANCE. The electrical resistance of thin sheet of a material


with uniform thickness as measured across opposite sides of a unit square pattern.
Expressed in ohms per square.

SIGNAL DISTRmUTION. The network of package conductors that intercon-


nects the drivers and receivers.

SIGNAL WIRING. A conductive path carrying an electric signal.

SILICON EFFICIENCY. The ratio of sum total of area of all silicon chips
to the total packaging area-primarily at board level.

SIMULTANEOUSLY SWITCHING DRIVER. A driver circuit that changes


state in unison with other drivers on the same chip or nearby chips thereby
creating switching noise.

SINGLE-CHIP CARRIER. An electronic package that connects single-chip


terminals to second-level package by having a different number of terminations
than the chip itself.

SINGLE-CHIP MODULE (SCM). Module or package supporting one chip,


as opposed to multichip which supports several.

SINGLE-IN-LINE (SIP). DIP-like package with single line of leads as op-


posed to two for DIP.

SINGLE-LAYER METALLIZED PACKAGE (SLAM). Ceramic leadless


package without cavity, sealed by ceramic or glass to a ceramic cap.

SINTERING. Heating a metal or ceramic powder, thereby causing the particles


to bond together to form monolithic body.
11-964 GLOSSARY AND SYMBOLS

SKIN EFFECT. A high frequency effect that causes the resistance of a conduc-
tor to increase. This phenomenon occurs because the magnetic fields within the
conductor force the current to flow on the outer surface or skin as frequency of
the signal increases.

SLOW WAVE PROPAGATION. Energy that travels at less than the expected
velocity for a dielectric structure because of series resistance in the signal line
or return path.

SLURRY. A thick mixture of liquid and solids. The solids are in suspension
in the liquid.

SMALL OUTLINE (SOP). Also called SOIC. Small outline integrated circuit
package. It is a rectangular DIP-like package except that it is smaller and leads
on 1.27 mm, 1.0 mm, or 0.85 mm spacing. It is meant for surface mounting.

SOFT ERROR. In memory device technology, a memory state error induced


by a process which produces no permanent alteration of the physical condition
of the device.

SOFT GLASS. Glasses, typically high-lead content glasses, having low soften-
ing points that could be used to seal ceramic or metal lids to packages below
about 450°C. Also called solder glasses because of their ability to wet most
metal surfaces.

SOFTENING POINT. Refers to the temperature at which the log viscosity


of glass is 7.6 poises, as defined and measured to ASTM specification.

SOLDER. A low melting-point alloy used in numerous joining applications


in microelectronics. The most common solders are lead-tin alloys.

SOLDER DAM. A dielectric composition screened across a conductor to limit


molten solder from spreading further onto solderable conductors.

SOLDER GLASSES. Glasses used in package sealing that have a low melting
point and tend to wet metal and ceramic surfaces.

SOLDERABILITY. The ability of a conductor to be wetted by solder and to


form a strong bond with the solder.

SOLDERING. The process of joining metals by fusion and solidification of


an adherent alloy having a melting point below about 300°C.
GLOSSARY AND SYMBOLS 11-965

SOLID LOGIC TECHNOLOGY (SLT). Ceramic package technology prac-


ticed by IBM in 1960s by firing Ag-Pd conductors on to dry-pressed and fired
alumina substrate.

SPACE TRANSFORMER. A package transforming a spatially dense set of


chip connections to a less dense set of connection points on package.

SPICE. The "Simulation Program for Integrated Circuit Emphasis" is the indus-
try standard for circuit simulation. It contains many of the features inherent in
ASTAP. See ASTAP.

SPUTTER CLEANING. Bombardment of a surface with energetic argon or


other noble gas ions to clean the surface of oxide films and residues that could
interfere with subsequent electrical or mechanical contact layers. The bombard-
ment knocks off (or sputters) surface atoms to render the surface clean.

SPUTTERING. The removal of atoms from a source by energetic ion bombard-


ment. The ions are supplied by a plasma. The sputtering process is used to deposit
films for various thin-film applications.

STATIC FLEX. Flexible wiring circuit carrier, which once installed, re-
mains fixed.

STEINER TREE. See stub.

STENCIL. A planar patterned mask used to transfer images on a surface.


Usually metallized patterns on an insulating surface.

STORAGE CONTROL ELEMENT (SCE). Controls the data transfer paths


and the interface between the channels, processor storage, and central processor.
Also called System Control Element.

STORAGE HIERARCHY. The collection of memory elements (cache, main


storage, etc.) and their controls that make up the memory for the processor.

STRESS CORROSION. Refers to the degradation of mechanical properties


of brittle materials by crack propagation due to the acceleration of applied stress
in the presence of corroding atmospheres such as water.

STRIPLINE. A transmission line that is embedded within a single dielectric


medium and sandwiched between two reference planes.

STUB. A short wire which interconnects input at a circuit with the (main)
signal line.
11-966 GLOSSARY AND SYMBOLS

STUD. The conductive path that runs vertically from one level of conductors
to another in a multilayer substrate.

SUBTRACTIVE PATTERNING. The processing sequence generally fol-


lowed in producing thin-film networks or circuits. Films are area-deposited (by
vacuum evaporation, CVD, or sputtering) and the desired conductive, resistive,
etc., pattern is etched into each layer through mask, using appropriate selective
etchant fluids.

SUPERCONDUCTOR. Material offering no resistance to the flow of current.


In addition to metals, ceramics have been recently discovered to have this
property.

SURFACE MOUNT TECHNOLOGY (SMT). A method of assembling hy-


brid circuits and printed wiring boards, where component parts are mounted
onto, rather than into, the printed-wiring board, as in the mounting of components
on substrates in hybrid technology.

SURFACE TENSION. An effect of the forces of attraction existing between


the molecules of a liquid. It exists only on the boundary surface.

SWITCHING NOISE. An induced voltage on the power distribution system


at the circuit terminals caused by the rapidly changing current caused by the
simultaneous switching of many drivers.

TAPE AUTOMATED BONDING (TAB). The process where silicon chips


are joined to patterned metal on polymer tape (e.g., copper on polyimide) using
thermocompression bonding, and subsequently attached to a substrate or board
by outer lead bonding. Intermediate processing may be carried out in strip form
through operations such as testing, encapsulation, bum-in, and excising the indi-
vidual packages from the tape.

TAPE BALL GRID ARRAY (TBGA). Ball grid array technology on TAB.
See TAB.

TEMPERATURE CYCLING. An environmental test where the film circuit


is subjected to several temperature changes from a low temperature to a high
temperature over a period of time.

TENSILE STRENGTH. The pulling stress which has to be applied to a


material to break it, usually measured in MPa.
GLOSSARY AND SYMBOLS 11-967

TERMINAL. A metallic connector or pad to a circuit within a chip or package


that permits electrical interconnection to external circuits.

THERMAL COEFFICIENT OF EXPANSION (TCE). The ratio of the


change in dimensions to the change in temperature-per-unit starting length, usually
expressed in cm/cm/DC. The acronyms TCE and CTE are synomous.

THERMAL CONDUCTION MODULE (TCM). An IBM multichip (100


chips or more) module that is cooled by thermal conduction of pistons in contact
with chips.

THERMAL CONDUCTIVITY. The rate with which a material is capable of


transferring a given amount of heat through itself.

THERMAL CYCLING. A method to impose a cyclic stress on an assembly


of microelectronic components by alternately heating and cooling in an oven. It
is used for accelerated reliability testing of assemblies.

THERMAL FATIGUE. Failure of a structural element from repeated tempera-


ture excursions, wherein the load develops from thermal expansion mismatch of
dissimilar materials.

THERMAL GRADIENT. The plot of temperature variances across the bulk


thickness of a material being heated.

THERMAL MISMATCH. Difference in thermal coefficients of expansion of


materials which are bonded together.

THERMAL NETWORK. Representation of a thermal space by a collection


of conveniently divided smaller parts-each representing the thermal property
of its own part and connected to others in a prescribed manner so as not to
violate the thermal property of the total system.

THERMAL RESISTANCE (DCIW). The opposition offered by a medium to


the passage through it of thennal energy.

THERMOCOMPRESSION BONDING (TIC). A process involving the use


of pressure and temperature to join two materials by interdiffusion across the
boundary.

THERMOGRA VIMETRIC ANALYSIS (TGA). A technique that measures


material weight change as a function of increasing temperature.
11-968 GLOSSARY AND SYMBOLS

THERMOMECHANICAL ANALYSIS (TMA). A technque that measures


the linear expansion or other deformations of a material with respect to changes
in temperature.

THERMOPLASTIC. A substance that becomes plastic (malleable) on being


heated; a plastic material that can be repeatedly melted or softened by heat
without change of properties.

THERMOSETTING. The property of some organic materials to irreversibly


polymerize and set or harden when heated to some appropriate temperature.

THERMOSONIC BONDING (TIS). A bonding process which uses a combi-


nation of thermocompression (TC) bonding and ultrasonic bonding. It is done
on what amounts to a gold-wire TC bonder with ultrasonic power applied to
the capillary.

THEVENIN EQUIVALENT. Electrical model describing voltage-current be-


havior of an electrical network between any two of its nodes. In its simplest
form it can be a constant voltage source with a series impedance (resistance in
case of direct current) or a constant current source shunted by an impedance.
These impedances are often called source impedances.

THICK FILM. A film deposited by screen printing processes and fired at high
temperature to fuse into its final form. The basic processes of thick-film technology
are screen printing and firing.

THIN FILM. Thin film refers to a coating layer of thickness in the range of
from a few (2-3) atomic layers to a few (1-5) microns (micrometers). The
important feature distinguishing thin films from thick films, though, is not so
much the difference in thickness as the method of deposition which takes place
by a variety oftechniques such as chemical vapor deposition, evaporation, or sput-
tering.

THIN-FILM PACKAGING. An electronic package in which the conductors


and/or insulators are fabricated using deposition and patterning techniques similar
to those used for integrated circuit chips.

THREE-LAYER TAPE. An interconnection medium used in tape automated


bonding (TAB), where the tape is comprised of three layers of metallization
(usually copper), with polymer and adhesive in between.

THROUGH HOLE. A hole connecting the two surfaces of a printed-circuit


structure.
GLOSSARY AND SYMBOLS 11-969

TIME·DOMAIN REFLECTION. A time varying voltage and current distur-


bance created at a discontinuity on a transmission line that travels in a direction
opposite to its stimuli, thereby causing spurious line noises and signal distortions.

TINNED. Literally, coated with tin, but commonly used to indicate coating
with solder.

TINNING. To coat metallic surfaces with a thin layer of solder.

TOP SIDE METALLURGY (TSM). An acronym referring to the metalliza-


tion on the top side of a substrate to which a chip is joined (such as a C4
solder connection).

TOPOGRAPHY. The surface condition of a film; bumps, craters, etc.

TRANSFER MOLDING. An automated type of compression molding in


which a preform of plastic (usually an epoxy-based resin) is poured from a pot
into a hot mold cavity.

TRANSFER UNITY GAIN POINT. The point on a logic circuit's Vout vs.
Vin transfer curve where the output voltage equals the input voltage. It determines
the input signal swing at which noise will propagate and amplify through cascaded
logic circuits.

TRANSIENT MISMATCH. Thermal mismatch between elements of a struc-


ture which, because of thermal lag, varies with time until reaching a steady-
state value.

TRANSISTOR OUTLINE (TO) PACKAGE. An industry standard package


designation established by JEDEC of the EIA.

TRANSISTOR·TRANSISTOR LOGIC (TTL). A Nand logic function is


implemented by the switching of voltage changes on distinct emitter inputs of
bipolar transistors sharing common base and collector voltages.

TRANSMISSION LINE. A conductor that is inductively and capacitively


coupled to a nearby return path forming a uniform itemative distributed network
with specific properties.

TRANSMITTED NOISE TOLERANCE. A set of pulse amplitudes and pulse


widths for spurious signals at the input to a receiver circuit that will not falsely
set a downstream latch.

TRI-PLATE LINE. Same as a stripline. See stripline.


11-970 GLOSSARY AND SYMBOLS

TWO-LAYER TAPE. A primary form of tape fabrication for tape automated


bonding (TAB), starting with the metallic sputtering and subsequent pattern-
plating on Kapton carrier tape. No adhesive is used in bonding copper to Kapton.

ULTRA LARGE SCALE INTEGRATION (ULSI). So far, an extreme in


the circuit integration, used to indicate presence of one hundred million transistors
(or more) on a single semiconductor chip.

ULTRASONIC BONDING. A process involving the use of ultrasonic energy


and pressure to join two materials.

v
VACUUM DEPOSITION. Deposition of a metal film onto a substrate in
vacuum by metal evaporation techniques.

VAPOR PHASE REFLOW. The technique for solder reflow to form package
interconnections. The solder joint is heated by the heat of condensation of an
inert vapor. The most common material of choice is a perfluorocarbon.

VERY HIGH SPEED INTEGRATED CHIP (VHSIC). Very high speed


integrated circuit, originally referring to 1.0 )lm ground rules.

VERY LARGE SCALE INTEGRATION (VLSI). Level of integration with


more than approximately ten thousand transistors on a single semiconductor chip.
Upper boundary not well defined.

VIA. An opening in the dielectric layer(s) through which a riser passes, or else
whose walls are made conductive.

VIA, FIXED. A via built into a package on a predetermined grid, and in


general, interconnecting both adjacent and nonadjacent planes.

VIA, PROGRAMMABLE. A via interconnecting adjacent wires on two adja-


cent wiring planes. Location does not correspond to the same grid locations as
fixed vias.

VIA, SEGMENTED. A fixed via interconnection. A predetermined subset of


all wiring planes.
GLOSSARY AND SYMBOLS 11-971

VIA, THROUGH. A fixed via passing through all wiring planes.

VISCOSITY. The intrisic property of a fluid that resists internal flow by


offering counteracting forces.

VOLTAGE SLEW RATE. The rate of change in voltage with respect to time
(dv/dt).

WAFER. Commonly, a slice of a semiconductor crystalline ingot used for


substrate material when modified by the addition, as applicable, of impurity
diffusion (doping), ion implantation, epitaxy, etc., and whose active surface has
been processed into arrays of discrete devices or ICs by metallization and passiv-
ation.

WAVE SOLDERING. The technique for solder application and reflow in


which a jet of liquid solder is directed at the two metallic points to be intercon-
nected. The technique usually involves processing steps to apply flux and remove
excess solder.

WEAROUT. The time following the stable failure-rate period during which
the expected, or observed, failure rate of an item increases and exceeds a spe-
cific value.

WELDING. Joining two metals by applying heat to melt and fuse them with
or without a filler metal.

WETTING. The spreading of molten solder or glass on a metallic or nonmetal-


lic surface, with proper application of heat and flux.

WIRE LENGTH, AVERAGE. The average length measured in logic unit


pitches, of all connections in a given package level.

WIREABILITY. The capability of a package to permit the interconnection of


subpackages mounted on it and terminals attached to it measured as the probability
of wiring success. It is near one when sufficient wiring capacity, via availability
and terminal access, are present.

WIREBOND. A completed wire connection whose constituents provide electri-


cal continuity between the semiconductor die (pad) and a terminal. These constit-
uents are the fine wire; metal bonding surfaces like die pad and package land; and
metallurgical interfaces between wire, and metals on both the chip and substrate.
11-972 GLOSSARY AND SYMBOLS

WIREBONDING. The method used to attach very fine wire to semiconductor


components in order to interconnect these components with each other or with
package leads.

WIRING (ALSO ROUTING). The manual or automatic prescription of routes


(portions of tracks) for wires interconnecting package components or logic cells
on chips.

WIRING ASSIGNMENT. The manual or automatic prescription of particular


pads, pins, connectors, or terminals to which corresponding wires are to be
attached.

WIRING CAPACITY. The total available length of wiring tracks in a package


(before any wires are prescribed on the wiring image).

WIRING CHANNEL. A linear region on a package wiring plane containing


space for at least one wiring track.

WIRING DEMAND. The product of wiring connection count and average


connection length, either locally in a limited region or total over an entire package.

WIRING DENSITY. Total wire length contained within a unit square. Mea-
sured in inches per square inch or centimeters per square centimeter.

WIRING OVERFLOW. A wiring connection called for by logic design but


not inserted in a proposed package wiring image during prior automatic wiring
use of package.

WIRING RULES. A set of electrical constraints that are used in conjunction


with a design aids program to control the topological parameters of an interconnec-
tion network to assure proper functionality. Wiring rules are usually derived from
electrical simulation results.

WIRING TRACK. A linear extent of space in a wiring channel used to contain


one (or more if collinear) conducting wires used to interconnect package compo-
nents.

YELLOW WIRE. Discrete wires that are yellow in color interconnecting


terminals on a package. Originally used in reference to all back panel wiring
on early electronic assemblies.
GLOSSARY AND SYMBOLS 11-973

ZERO·INSERTION·FORCE CONNECTION (ZIF). A form of connector


that allows the connector pins to be brought together under very low force, then
wiped and pressed together during cam activation.

SYMBOLS

See Chapter 4, "Heat Transfer in Electronic Packages," section 4.7, "Nomencla-


ture," for symbols used in the field of heat transfer.

AgN03: Silver nitrate


AgO/PdO: Silver oxide/palladium oxide
AgPd: Silver-palladium alloy
A + Pd: Silver + palladium
Ag-Pd: Silver-palladium
AgPd Au: Silver-palladium-gold ternary alloy
NPdO: Silver/palladium oxide
Al-Cu: Aluminum-copper
A1N: Aluminum nitride
A120 3: Aluminum oxide
3Al20 3 2Si02: Mullite
AuA12: Gold-aluminum alloy
AusAl2: Gold-aluminum alloy
AuGe: Gold-Germanium alloy
AuPt: Gold-platinum alloy
AuSi: Gold silicon alloy
AuSn: Gold-tin alloy
AuSn2: Gold-tin alloy compound
AuSn4: Gold-tin alloy compound
AU3Sn4: Gold-tin alloy compound
BN: Boron nitride
B 20 3: Boron oxide
B20 3 + Si0 2 + (AI20 3): Borosilicate glass + alumina
B203-Si02-AI203nashNa20: Alumino borosilicate glass
BaCI2: Barium chloride
BeO: Beryllium oxide
CPs-CH3: Toulene
CH3CHO+: Butyraldehyde
C2H 30 2: Acetate
CO 2: Carbon dioxide
C20 4: Oxalate ion
CaCI2: Calcium carbide
11-974 GLOSSARY AND SYMBOLS

CaOclon: Calcium oxide


CaO + Al20 3 + Si02 + B20 3: calcia-alumino-borosilicate
Ca(OH)2: Calcium hydroxide
Cu-Ag-Ti-Sn: Copper-silver-titanium-tin
CuAl2: Copper-aluminum alloy compound
Cu-Mo(20% Cu): Copper-molybdenuim alloy
CUS04: Copper sulphate
Cu-Sil: Copper-silver alloy
CU3Sn: Copper-tin alloy compound
CU6Sn5: Copper-tin alloy compound
Cu-W(20%Cu): Copper-tungsten alloy (20% copper)
Fe(CNh: Ferrous cyanide
FeNi: Iron-nickel alloy
GaAs: Gallium arsenide
Ge02: Germanium oxide
HCl: Hydrochloric acid
In-Cu-Sil: Indium-copper-silver
KBr: Potassium bronide
KCl: Potassium chloride
KCI0 4: Potassium chlorate
K3Fe(CNk Potassium ferrocyanide
K 4Fe(CNk Potassium ferricyanide
KHC0 3: Potassium bicarbonate
KI: Potassium iodide
KI04: Potassium iodate
KN0 3: Potassium nitrate
KRe04: Potassium perrhenate
LaCl3: Lanthanum chloride
LiCl: Lithium chloride
LiCI04: Lithium chlorate
Li20 + Al20 3 + Si0 2A120 3: Lithium alumino
Li20-AI203-Si02-B203: Alumino borosilicate
Li20 + Si0 2 + MgO + Al20 3 + Si0 2: Magnesium alumino silicate glass
MgCI 2: Magnesium chloride
MgO: Magnesium oxide
2Mg02A1203.5Si02: Cordierite ceramic
MgO + Al20 3 + Si02 + B 20 3 + (AI 20 3): Magnesium alumino borosilicate glass
+ alumina
MgO-Al203-Si02-B203-P205: Magnesium alumino borosilicate glass con-
taining boron oxide and phosphorous oxide
NaCl: Sodium chloride
NaCI0 4 : Sodium perchlorate
NaI: Sodium iodide
NaOOCCH3: Sodium acetate
GLOSSARY AND SYMBOLS 11-975

NaOOCC2HS: Sodium propionate


NaOOCCP7: Sodium butyrate
NaOH: Sodium hydroxide
Na 2S04: Soldium sulphate
Nb 20 S: Niobium oxide
Nd-Yag: Neodium-yag laser
NiCr: Nichrome
NiCr-Pd-Au: Nichrome-palladium-gold
NH4: Ammonia radical
NH4Cl: Ammonium chloride
NH4C104: Ammonium perchlorate
Ni 3Sn4: Nickel-tin alloy compound
N0 2: Nitrous oxide
N0 3: Nitrate ion
PbIn: Lead indium alloy
Pb-In: Lead-indium
SOPb SOIn: Lead (SO%)-Indium (SO%) alloy
PbO: Lead oxide
PbO + B20 3 + Si0 2 + (AI20 3): Lead borosilicate
PbO-B203-Si02-AI203-ZnO: Zinc lead borosilicate glass
PbO-ZnO-AI203-B303-SI02: Lead-zinc borosilicate-glass (solder glass)
P20 S: Phosphorous pentoxide
9S PBIS Sn: Lead (9S%-tin (S%) alloy
RU02: Ruthenium oxide
Sg: Free sulphur
Si-Au-Sn: Silicon-gold-tin
SiC: Silicon carbide
Si~4: Silicon nitride
Si02: Silicon dioxide
9SSn 3.SAg 1.0 Cd, 0.5 Sb: Tin (9S% )-Silver (3S% )cadmium (1 %)-antimony
(S%) alloy
42Sn S8Bi: Tin (42%)-bismuth (S8%) alloy
SOSn SOIn: Tin (SO%)-Indium (SO%) alloy
40Sn 60Pb: Tin (40%-lead (60%) alloy
6OSn-40Pb: Tin (6O%)-lead (40%) alloy
63Sn 37Pb: Tin (63%)-lead (37%) alloy
9SSn SSb: Tin (9S% )-antimony (S%) alloy
S02: Sulphur dioxide
S04: Sulfate ion
SrCI2: Strontium chloride
Ta2N: Tantalum nitride
TiC: Titanium carbide
Ti02: Titanium dioxide
Ti-Pd-Au: Titanium-palladium-gold
11-976 GLOSSARY AND SYMBOLS

fBa2CU307: Yttria-baria-copper oxide superconductor


f 20 3: Yttrium oxide
ZnCI2: Zinc chloride
ZnO: Zinc oxide
ZnS04: Zinc sulphate
Zr02: Zirconium dioxide
AUTHORS' BIOGRAPHIES

Nanda G. Aakalu-Advisory Engineer-IBM, Poughkeepsie, NY. Mr. Aakalu


is currently involved with electronic packaging at the card and board level, and
with cooling-related development work in mainframe computers. Employed by
IBM for the last 19 years, he has been involved with the advanced technology
area of direct immersion cooling of VLSI chips, development of thermal com-
pounds for internal thermal enhancement of multichip packages, reliability study
of air moving devices, and mechanical analysis of power components. He also
directed the development of terminator resistor circuits on porcelain steel sub-
strates. His fields of expertise are heat transfer, stress/strain analysis, and process
engineering of thick-film packages. He received his BSME from the University
of Mysore, India, and an MSME from Purdue University. He is a member of
ISHM and has an ASME publication on condensation heat transfer as well as
thirty invention publications in the IBM technical disclosure bulletins related to
cooling and packaging of electronics and two patents on cooling and packaging
of electronics. (A current biography was not available at the time of publication.)

Vincent W. Antonetti-Professor and chair; department of mechanical engi-


neering-Manhattan College, Riverdale, NY. Prior to joining Manhattan College,
he was a senior engineer at IBM, where as manager of the IBM Poughkeepsie
Thermal Engineering Laboratory, he was responsible for the thermal design of
high-end computers. His areas of expertise are heat transfer in electronic equip-
ment and thermal contact resistance. He has a BME from the City College of
the City University of New York, a MSME from Columbia University, and a
PhD from the University of Waterloo in Canada. He is a fellow of the American
Society of Mechanical Engineers, and a licensed professional engineer. He has
four IBM Invention Achievement awards, and is the author of 32 publications.

Masaya Asano-Manager, Electronic & Imaging Materials Research Labs-


Toray Industries, Inc., Shiga, Japan. During his 25 years at Toray, Mr. Asano
has been involved with imaging and electronic materials areas including waterless
planographic printing plates and polyimides for microelectronics. He received
the awards from the Japan Institute of Invention and Innovation for Positive
Acting, Waterless Planographic Printing Plate, and from the Society of Polymer
Science, Japan, for the development and commercialization of photosensitive
polyimide. He received his BS and MS in chemistry form Kyoto University.
11-978 AUTHORS' BIOGRAPHIES

Donald E. Barr-Senior Technical Staff Member and Manager of Site Technical


Assurance and Material Engineering-IBM Endicott, NY. Prior to joining IBM,
he was the Technical Director of Research and Development at GAP Photo and
Repro Division in Binghamton, New York. He received his Ph.D. from the
University of Massachusetts, where he was awarded an NDEA Fellowship, is a
member of the American Chemical Society and has publications and patents in
the area of electronic applications and fundamental science.

J. Richard Behun, P.E.-Development Engineering Manager-IBM, Essex


Junction, VT. Mr. Behun is presently charged with definition and implementation
of module test handling equipment for memory, logic, and mixed signal products.
Prior to this, much of his career was devoted to surface-mount development for
multi-layer ceramic products, where he earned the bulk of his 16 inventions
including four patents.
Rick was awarded first and second level IBM Invention Achievement awards,
an IBM Division award, as well an IBM Achievement Award.
Rick is a member in the American Society of Mechanical Engineers, where
he is presently the chairperson of the Design for Manufacturability Committee
within the Design Division. Rick is a licensed professional engineer in Vermont
and New York. Rick joined IBM in 1980 after earning his B.S. and Masters in
Mechanical Engineering at Rensselaer Polytechnic Institute.

Robert L. Beliveau-Senior Technical Specialist-DuPont High Performance


Films, Circleville, Ohio. During his 22 years with DuPont; Bob has been involved
with both materials processing and product development activities in the areas
of Flexible Printed Circuits, Tape Automated Bonding and Multi-Chip Modules.
Bob is a member of the IPC, the Society of Vacuum Coaters and the Surface
Mount Technology Association.
Bob received his BS in Mechanical Engineering from the University of Massa-
chusetts
Frank J. Bolda-Advisory Engineer (retired)-IBM. Frank J. Bolda joined the
company in 1956 in the Technical Services Laboratory in Poughkeepsie, where
he worked on various plating and heat treating processes, failure analysis, and
the development and evaluation of tape slitting devices. From 1960-1965, he
worked in the Mechanical Memory Group in the development and packaging of
ferrite core memories. In 1966 he became a process engineer in the Research
Triangle Park facility in North Carolina supporting the site manufacturing pro-
cesses. In 1972 he joined the Development Laboratory and was involved in
the development of honeycombed cabinetry, keyboard technology, and flexible
substrate packaging. More recently he was involved in the development of glass
substrate for the scanner products, coated metal substrates applications, and
connector development.
He is a co-inventor of six patents and is a member of the American Society
for Metals and the Electronic Connector Study Group. He has four publications
in the field of coated metal substrates and electronic packaging.

Thomas Caulfield-Senior Engineer-IBM, East Fishkill, NY. Dr. Caulfield


manages product development, applications, design and analysis for Ceramic
AUTHORS' BIOGRAPHIES 11-979

Chip Carriers. Over his 7 year career at IBM he has held various engineering
and management positions related to flip-chip and BGA packaging. He holds
numerous US patents and is considered a subject-matter expert in electronic
packaging. Dr. Caulfield has a Doctorate of Engineering degree from Columbia
University (1986) and prior to his work at IBM, he was a Senior Member of the
Technical Staff at Philips Laboratories, in Briarcliff Manor, NY.

Abhijit Chatterjee-Assistant Professor; School of Electrical and Computer


Engineering-Georgia Institute of Technology, Atlanta, GA. Chatterjee received
the Ph.D. degree in electrical and computer engineering from the University of
Illinois at Urbana-Champaign in 1990 and had ten years experience with General
Electric Company before joining Georgia Tech. His research interests are in the
fields of mixed-signal MCM testing, fault-tolerant computing, low-power circuit
design, computer algorithms and design automation. He is a collaborating partner
in NASA's New Millenium Project, is the author of one US patent and has over
fifty publications in refereed journals and conferences. He has received two Best
Paper awards and has twice been nominated for Best Paper awards. He received
the NSF Research Initiation Award in 1993, the NSF CAREER Award in 1995
and is a Senior Member of the IEEE.

William T. Chen-Senior Technical Staff Member-IBM, Endicott, New York.


He joined IBM in Endicott in 1963. He has worked in various technical and
management positions in mathematical sciences, materials laboratory and in the
last 17 years in the electronic packaging area. He is a graduate of Queen Mary
College, University of London. He received a M.Sc. from Brown University,
and a Ph.D. from Cornell University. He is a fellow of the American Society
for Mechanical Engineers.

Tsuneyo Chiba-Senior Chief Researcher, Central Research Laboratory-Hi-


tachi Ltd., Kokubunnji, Tokyo 185, Japan. He received the B.S. degree in E.E.
in 1962 and the Ph.D. degree in I.S. in 1988, both from Kyoto University, Japan.
He joined the CRL, Hitachi Ltd. in 1962. Since then, he has been engaged in
the development of many generations of Hitachi's mainframe computers, such
as HITAC 5020, 8800, M-180, M-200Hl280H, M-680, M-880, and the latest
MP5800. He has taken charge of the corporation's large-scale projects to develop
LSI-based high- end mainframe computers, leading the development of hardware
technologies such as very high-speed ECL gate arrays and their packaging system
as well as design technologies. He has also directed and managed the develop-
ments of disk files, telecommunication LSIs and network systems, as well as
computer systems, serving as department manager in the CRL from 1981 to 1988.
He is currently providing directional guidance in the development of advanced
technologies and future computer systems.
Dr. Chiba is a member of the IEEE, the IEICE of Japan, and the IPSJ.

David B. Clegg (A current biography was not available at the time of publication).

Donald S. Cleverley-Owner and principal consultant-DSC Quality Consul-


tancy, Poughkeepsie, NY. Mr. Cleverley is a NY Quality Consultant and Certified
11-980 AUTHORS' BIOGRAPHIES

ISO-9000 Auditor He assists client companies in ISO-9000 documentation, train-


ing, registrar selection, and pre-registration audits. Previous to this, he was an
Advisory Engineer with the mM Microelectronics Division, East Fishkill, NY.
He was responsible for Reliability and Quality projections, measurements and
improvement actions, including achieving mM's "six-sigma" quality improve-
ment objectives for semiconductor devices. Previously in mM, he managed
Analog Circuit Test Engineering and introduced the Manufacturing Accelerated
Release System (MARS) for new chips. He holds a BSEE degree from Northeast-
ern University and an MSEE degree from Syracuse University. He is a Senior
Member of IEEE and an ASQC Certified Quality Engineer and Quality Auditor.
He is the author of seventeen papers and conference presentations dealing with
reliability and quality improvement and instructs quality courses at the college
level.

Marie S. Cole-Advisory Engineer-IBM, East Fishkill, NY. Ms. Cole has a


B.S.Ch.E. from Rensselaer Polytechnic Institute and an M.S. in Materials Science
from Columbia University. She joined mM in 1984, working until 1988 in
Corporate Component Procurement-Assurance and Test qualifying new SMT
plastic packages and setting the standards for testing the impacts of SMT assembly
on packaging. She earned a Division Award for that work. The author of numerous
publications and one patent related to BGA packaging, she has worked on the
development of CBGA and CCGA packaging since 1988. Her current assignment
is in New Product Applications in the Ceramic Chip Carrier Business Unit,
specifically in MLC Packaging Applications.

Frank Crnie-Brand Acquisition Manager: Mobile Products-IBM, Research


Triangle Park, NC. Frank Crnic joined IBM in 1981 after receiving his Bachelor
of Science degree in Materials and Metallurgical Engineering from the University
of Michigan. He was involved in a variety of process development efforts in the
East Fishkill, NY Semiconductor Laboratory. In March, 1985, he joined the East
Fishkill Multi-Layer Ceramic (MLC) Packaging Plant as Manager of Sintering
Process Engineering. He became involved in automated optical inspection as
Manager of Inspection Systems Engineering in 1987. In 1988 he became Manager
of Test Systems Engineering. In this assignment, he was responsible for imple-
menting the Electrical Module Test (EMT) for screening latent opens defects.
In May, 1993, Mr. Crnic was assigned to mMs Kiosk Solutions group in Atlanta,
GA as a Project Manager. Since July, 1995, he has been Mobile Computing
Brand Acquisition Manager with the IBM Personal Computer Company in Re-
search Triangle Park, NC.

George Czornyj-Senior Engineer-IBM, Hopewell Junction, NY. During his


20 years at mM he has been involved with many technology areas utilizing his
expertise in high temperature polymers (polyimides), photosensitive polyimides,
plasma polymerization of organo and organometallic films, and adhesion of
polymer/metal/ceramic materials to develop materials for thin film packaging
applications. He has authored over 40 professional articles and obtained over 20
patents and invention disclosures in the field of thin film packaging, processing
and lithography. He has participated on numerous ACS Polymer Division Execu-
AUTHORS' BIOGRAPHIES 11-981

tive Committees, as a past Chairman of the Speakers Bureau and as a member


of the ACS Organic Coatings & Plastics Division, ACS Polymer Division and
Materials Research Society. He received the Arthur K. Doolittle Award in 1976
with Dr. B. Wunderlich.
He received his undergraduate schooling in chemistry and physics from Rens-
selaer Polytechnic Institute, graduate schooling in Polymer Physical Chemistry
from Rensselaer Polytechnic Institute and was a Post Doctoral Scientist at IBM
Research at San Jose, CA working on Plasma Polymerization.

Evan E. Davidson-Senior Technical Staff Member-IBM, Poughkeepsie, NY.


Mr Davidson is currently a technology applications engineer for OEM products.
Previously, he was the Manager of High Performance Technology for IBM's
General Technology Division located in East Fishkill, NY. While in that capacity,
he specialized in the electrical package design for the high performance MCMs
used in mainframe computers. Before that he was a circuit designer for logic
and memory circuits at both IBM and Bell Labs. He received his BEE from
Rensselaer Polytechnic Institute and his MSEE form New York University. He
is a Senior Member of the IEEE and a member of the Eta Kappa Nu and
Tau Beta Pi engineering honor societies. He has written numerous articles and
presented many papers on the subject of electrical package design. He holds
many patents in the areas of digital circuit design and package design.

Philip E. Garrou-Chief Scientist-Microelectronics, Dow Chemical, Central


Research & New Businesses, Research Triangle Park, NC. During his 20+ years
at Dow, Dr. Garrou has been involved with both electroceramics and polymers
for microelectronic packaging and interconnect applications. He has authored
50+ professional articles and co-edited the book Thin Film Multichip Modules.
Dr. Garrou is a senior member of IEEE CPMT where he has served as Chairman
of TC-5 (Materials) and was elected to the Board of Governors.
He was elected Technical VP of ISHM and has served as Chairman of the
MCMlAdvanced Packaging subcommittee of the ISHM National Technical Com-
mittee and Chairman of the ISHM Materials Division. He is co-founder of the
ISHMlEEE Ojai Workshop on Advanced Materials and was Technical Chair of
the 4th International MCM Conference in Denver.
He received his B.S. and Ph.D. in Chemistry from North Carolina State Univer-
sity and Indiana University respectively.

John B. Gillett-Senior Technical Staff Member (retired)- IBM, Kingston, NY.


At the time of his retirement, Mr. Gillett was the manager of Advanced Systems
Packaging working in a wide range of advanced technology activities, from
semiconductor devices to total system packaging. Prior to joining IBM, he was
with the Automatic Telephone and Electric Company, United Kingdom, designing
telephony switching equipment and early electronic computers. His field of exper-
tise is high-performance computer hardware technologies. He received his MSEE
in London. He has twelve patents issued in the fields of semiconductor devices
and circuits, data storage systems, connectors and power supplies.

Lewis S. Goldmann-Senior Engineer-IBM, East Fishkill NY. For the last 15


years, he has been involved in the mechanical design, testing and modeling of
11-982 AUTHORS' BIOGRAPHIES

microelectronic chips, packages and materials. Previously, he worked in the


development of flip chip interconnections and the IBM Thermal Conduction
Module. He has published widely in these areas, including many papers on flip
chip joining. Mr. Goldmann's formal education includes a pre-engineering liberal
arts degree from Queens College, CUNY; a B.S. in Mechanical Engineering
from Columbia Univ.; and an M.S.M.E. from MIT. He is a member of ASME
and serves as Associate Technical Editor of its Journal of Electronic Packaging.

Dimitry G. Grabbe-Director, Electronic Interconnection Research-AMP Inc.,


Harrisburg, PA. An AMP Fellow since 1989, Dimitry G. Grabbe comes from a
background that includes a military-academy education in the former Yugoslavia
and machine design education in Germany. He joined AMP in 1973 after having
worked for Photocircuits Corporation, the Main Research Corporation (of which
he was founder and president) and Rockwell International Autonetics. During
his career at AMP, Mr. Grabbe has held the positions of Manager of Advanced
Product and Manufacturing Technologies; Director of Applied Technology; Di-
rector of Interconnection Research and Applied Technology; and Director of
Electronic Interconnection Research. He has won many awards, including the
Man of the Year Award from the International Technology Institute, the Leonardo
da Vinci Award from the American Society of Mechanical Engineers and the
Symposium Honor Award from the International Institute of Connector and
Interconnection Technology. He is a member of the International Technology
Institute Hall of Fame, and an IEEE Fellow. Mr. Grabbe is AMPs most prolific
holder of patents, with a total of some 127 (US and 2 Foreign) of which 116
are with AMP.

Tapan K. Gupta-Consultant Engineer, Electronic Sensors & Systems Divi-


sion-Northrop Grumman Corporation, Baltimore, MD. From 1985 to 1993, he
was a Senior Fellow and the Manager of the Electronic Packaging Center at
Pittsburgh in charge of packaging development activities for Alcoa Electronic
Packaging, Inc., San Diego. From 1967 to 1985, he was an Advisory Scientist
at Westinghouse Research Laboratories at Pittsburgh. He earned his MS and
Sc.D. from Massachusetts Institute of Technology in Ceramics in 1964 and 1966,
respectively. A Fellow of the American Ceramic Society, he has authored or co-
authored more than 100 papers, three book chapters, and been issued more
than 30 US patents. His present activities lies in the manufacture of advanced
microelectronics packaging and other functional devices for digital and micro-
wave applications.

William R. Heller-Senior Physicist (deceased)-IBM, Poughkeepsie, NY. Dr.


Heller passed away in June of 1994. He was working on this update at the time.
His material was passed on to Dr. Rose who incorporated it into this work and
completed the revision. Dr. Heller has been missed, but his work continues to
show through in this revision. He was a member of the IBM Fellow Department
in Poughkeepsie. His expertise was in computer-aided design tools, packaging,
and wireability analysis for digital systems. Prior to joining IBM, he was a
research associate at the University of Illinois, an assistant professor at Yale
University, and worked at the Shell Development Corporation. In 1979, he was
AUTHORS' BIOGRAPHIES 11-983

a visiting professor of computer science at Caltech. He received a BS in physics


from Queens College, City University of New York, an ScM in applied math
from Brown University, and a PhD in solid state physics from Washington
University. He was a fellow of the American Physical Society, a fellow of the
IEEE (Computer Society), and a member of the New York Academy of Sciences.
He had approximately forty publications and patents in solid state physics, materi-
als sciences, computer-aided physical design, testing techniques, and wireabil-
ity analysis.

Hiroo Hiramoto-Director and General Manager of Electronic and Imaging


Materials Research Laboratories-Toray Industries, Inc., Shiga, Japan. Mr. Hi-
ramoto obtained an B.A.Sc. and M.A.S.c. in Applied Chemistry from the Univer-
sity of Tokyo. He joined Toray Industries, Inc. in 1965. His principal research
interests are in electronic materials based on polymers, especially high tempera-
ture polymers. He is one of the inventors of ionic-type photo-definable polyimides.
In 1991, he received an award from the Society of Polymer Science, Japan for
the development and commercialization of photosensitive polyimide.

Robert T. Howard-Senior Engineer (retired)-IBM, Burlington, VT. Dr. How-


ard specialized in Reliability and materials applications and manufacturing proce-
dures for microelectronic packaging. He joined the Federal Systems Division in
Huntsville, Alabama where he was lead engineer for establishment of the labora-
tory for microelectronics packaging. He also pioneered studies for LASER trim-
ming of thick and thin film resistors. In Burlington he was a lead engineer in
development of chip joining studies and quality and reliability studies on solders
and chemical degradation of packages. Dr. Howard has co-authored several
articles on creep and fatigue of solders employed in chip joining.
Prior to joining, IBM, Dr. Howard was Professor of Metallurgy and Materials
Engineering at the University of Kansas. Other teaching positions were at Wichita
State University and the University of Missouri at Kansas City. He received
BS and Sc.D. degrees, both in metallurgy, from the Massachusetts Institute of
Technology. He was now consultant in materials, processes and reliability in
microelectronics. He was a Registered Professional Engineer in Missouri and
Kansas.
Dr. Howard was a senior member of the Institute of Electrical and Electronic
Engineers. And he served as Vice-President for Education of the IEEE Compo-
nents, Packaging, and Manufacturing Technology Society (CPMT). In addition,
he has been a member of the Reliability Program Committee of the Electronic
Components and Technology Conference (ECTC) for several years, and has
chaired many Reliability sessions of ECTC. He was a member of the International
Society for Hybrid Microelectronics (ISHM) the American Society for Materials,
and the American Welding Society. Dr. Howard has published many papers and
articles reporting, research in technical and trade journals.

Akihiko Ikeda-General Manager, Corporate Research Laboratories, Electron-


ics Materials & Devices Laboratories- Asahi Chemical Industry Co., LTD. Fuji,
Japan. During his 21 Years at Asahi, Dr. Ikeda has been involved with research
& development of functional polymers, especially ion exchange resin and photo-
11-984 AUTHORS' BIOGRAPHIES

polymers. He was engaged in research & development of dry film resist for
printed circuit board and ester type photosensitive polyimide for microelectronics
use. He was co-author of the book "Development in UV curable resin" (in Jap-
anese)
He received his Ph.D. in Engineering from Kyoto University in 1975.

Dexter A. Jeannotte-Senior Engineer (retired)-IBM, East Fishkill, NY. Dr.


Jeannotte was responsible for materials reliability issues as they relate to lifetime
assessment of electronic assemblies, focusing on failure mechanism studies. His
concentration had been on solder fatigue and in environmental corrosion studies
and electrical contact quality studies. He received a BME degree from Marquette
University, an MS in metallurgical engineering from Columbia University, and
a PhD in metallurgy from Columbia University. He was chairman of ASTM
Committee B04 on Metallic Materials for Thermostats and for Electrical Resis-
tance, Heating and Contacts, and chairs the ASTM B04.04.07 Task Group on
an Environmental Testing Standard for Aluminum Electrical Connection Systems.
He was vice-chairman of IEEE/CHMT Society TC-l Committee on Electrical
Contacts, and chaired the TC-l Task Group 4 on Environmental Standards for
Electrical Contacts. He was also a member of NACE, ISHM, MRS, IEPS and
AAAS. He has coauthored four patents and 15 publications and presentations.

Nan Marie Jokerst-Associate Professor of Electrical and Computer Engineer-


ing-Georgia Institute of Technology, Atlanta, GA. She received her Ph.D. from
the University of Southern California in 1989 in the area of semiconductor
nonlinearities. She was a Hewlett Packard Fellow and a Newport Research Award
Winner while at USC, and a Summer Intern at the IBM Watson Laboratories in
1982 and 1983. She joined the Electrical Engineering faculty at the Georgia
Institute of Technology in 1989, and won a DuPont Young Faculty Award, a
National Science Foundation Presidential Young Investigator Award, and three
teaching awards. She is co-leader of the Optoelectronics and High Speed Electron-
ics Thrust Area in the Georgia Tech National Science Foundation Engineering
Research Center on Packaging. She has published and presented over 70 papers,
two book chapters, and has 3 patents. She is on the IEEE Lasers and Electro-
Optic Society Board of Governors and the Optical Society of America Engineering
Council. Her current research is highly collaborative and interdisciplinary, and
focuses on the optimization and integration and packaging of thin film optoelec-
tronic and high speed electronic devices with host substrates such as silicon
integrated circuits to form cost effective smart pixel systems. Her professional
activities include Chair (1993, 1994) and Programs Chair (1991, 1992: Most
Improved Chapter Award) of Atlanta IEEE LEOS, and Secretary and Treasurer
of Atlanta IEEE (1995 and 1996). She was elected to the IEEE LEOS Board of
Governors in 1995. She has served on the program committee for the IEEE LEOS
Annual Meeting (1993, 1994, 1995, 1996); the Topical Meeting on Integrated
Optoelectronics (1994); and the Topical Meeting on Smart Pixels (1994). She is
a member of the IEEE LEOS Subcommittee on Optical Interconnects and Process-
ing Systems, was Chair of the Optical and Optoelectronic Interconnect, Switching,
Processing, and Storage Committee of the Conference on Lasers and Electro-
Optics (1995, 1996), and the Program Chair for the LEOS Topical Meeting on
AUTHORS' BIOGRAPHIES 11-985

Smart Pixels (1996). She is also a Member of the Engineering Council of the
Optical Society of America (1995-1998); Member of the OSA Advisory Board
for Optics and Photonics News (1992-1996); Chair (1992) and Member (1991)
of the Newport Research Award Committee; Chair (1995) and Member (1996)
of the New Focus Award Committee; on the Program Committee of the OSA
Optical Computing Conference (1994); and Chair of the Symposium on Smart
Pixels, OSA Annual Meeting (1995). She was the Optoelectronics Representative
on the National Technical Council of the International Society for Hybrid Microe-
lectronics (ISHM, 1995), was Co-Chair of the Optoelectronics and Sensors Com-
mittee of the ISHM Annual Meeting (1994); Conference Chair, ISHM Optoelec-
tronics II (1995); and Co-Chair of the Optoelectronic Materials Committee of
the ISHM Materials Packaging Conference (1995). She was also the coordinator
of the CO-OP Short Course "Optoelectronics Integrated Onto Silicon VLSI:
Devices, Circuits, Systems" (1995).

George A. Katopis-Senior Engineer-IBM, Poughkeepsie, NY. Mr. Katopis is


the technical leader for High Performance Servers 1st Level Package design. His
fields of expertise are electrical package design, electrical noise containment,
and signal integrity. He received his MS and MPh in Electrical Engineering and
Computer Science from Columbia University in 1972 and 1978 respectively. He
has written numerous articles and presented many papers on electrical noise
characterization and containment as well as on signal integrity issues. He holds
several patents in the area of noise containment and prediction.

Robert W. Keyes-Research Staff Member Emeritus-IBM T. J. Watson Re-


search Center, Yorktown Heights, NY. Dr. Keyes has long engaged in research
and development activity in modem optical and electronic technologies and their
application in digital electronics and electronic devices. He joined the IBM
Research Division after several years at the Westinghouse Research Laboratory
in Pittsburgh. He received BS, MS, and PhD degrees in physics from the Univer-
sity of Chicago. Dr. Keyes is a member of the National Academy of Engineering
and is a Fellow of the Institute of Electrical and Electronic Engineers and of the
American Physical Society. He received the IEEE W. R. G. Baker Prize in 1976.
He is the author of 140 papers dealing with solid-state electronics and digital
information processing, eight issued patents, and a book, Physics of VLSI
Systems.

Subash Khadpe-President-Semiconductor Technology Center, Neffs, PA.


Semiconductor Technology Center is a consulting firm specializing in assembly
and packaging. Dr. Kbadpe is also the editor and publisher of "Semiconductor
Packaging Update", a subscription newsletter. Prior to founding the firm in 1985,
he was a member of the technical staff of AT&T Bell Laboratories for seven
years and Motorola for four years.
Dr. Kbadpe received his M.S.E.E. and Ph.D. degrees from Drexel University
in Philadelphia, PA. He also holds B.Sc. (with honors) and M.Sc. degrees from
the University of Bombay, India, and the M.Tech. degree from the Indian Institute
of Technology, Bombay.
He is a senior member of the IEEE, past associate and guest editor of the
11-986 AUTHORS' BIOGRAPHIES

IEEE CHMT Transactions, and past member of the executive board of the
International Electronics Packaging Society. He is a member of Tau Beta Pi, Eta
Kappa Nu, CPMT, IEPS, ISHM, and SMTA. He has published over 100 papers
and articles, and chaired numerous international conferences and workshops.

Bruce C. Kim-Assistant Professor; Electrical Engineering Department-Tufts


University, Medford, MA. Dr. Kim is currently an Assistant Professor with the
Department of Electrical Engineering and Computer Science, Tufts University,
Medford, MA. He is a member of International Society for Hybrid Microelectron-
ics (ISHM) and International Electrical and Electronic Engineers (IEEE). He
received his Ph.D. in Electrical Engineering from the Georgia Institute of Technol-
ogy, an MSEE from the University of Arizona, and a BSEE from the University
of California at Irvine. He has developed a novel test technique to test Multi-
Chip Module substrate interconnections while pursuing his Ph.D. at Georgia
Tech. He was previously employed by Georgia Tech Research Institute as a
Research Engineer before starting his Ph.D. program at Georgia Tech. He has
published 13 conference and journal papers in the areas of testing Multi-Chip
Module substrates and mixed-signal circuits. His research interests are in the
areas of testing multichip modules and mixed-signal circuits.

Alan (AI) G. Klopfenstein-President-AGK Enterprises, Hopewell Junction,


NY. Mr. Klopfenstein is involved with the development of strategies for semicon-
ductors and associated packaging for electronic systems. Prior to retiring as a
Senior Engineer, he held management, technical, and marketing positions at IBM
in semiconductor, packaging, and manufacturing-equipment development. He
received a BSME from the University of Connecticut and an MSME from the
University of Alabama in Huntsville. He is a member of the National Society
of Professional Engineers (NSPE), the International Society for Hybrid Microelec-
tronics, (ISHM), and the International Electronic Packaging Society (IEPS). He
is one of the three principal editors of this book.

Nick G. Koopman-Fluxless Soldering Program Manager- MCNC Center for


Microelectronics Systems Technologies, Research Triangle Park, North Carolina.
Dr. Koopman received a BS in Metallurgical Engineering from Lafayette College,
and an MS and Ph.D. in Metallurgy and Materials Science from Massachusetts
Institute of Technology. He was employed for 23 years with the IBM General
Technology Division in East Fishkill, NY, specializing in computer microelec-
tronics interconnection metallurgy development. The last position held was as
manager of Bonding and Interconnections where the major responsibility was
flip chip development. Since 1960 Nick has been with MCNC developing ad-
vanced solder and interconnection technologies. The current assignment involves
the development and marketing of the PADS fluxless/no-clean soldering process.
He has published extensively in the area of interconnections with 44 papers and
holds 61 inventions. He has served as a guest lecturer at several universities and
is an associate editor of the International Journal of Microelectronics Packaging.

J. Richard Kraycir-Board of Directors and Director of Operations-BST


Corporation, LaGrange, NY. Mr. Kraycir has a BS in Engineering, an MS in
AUTHORS' BIOGRAPHIES 11-987

Industrial Administration and is a licensed Professional Engineer. Currently, at


BST Corporation he specializes in precision plastic injection molding. Formerly,
he was Program Manager for IDM Microelectronics in their East Fishkill, NY,
MultiLayer Ceramic facility. His group was responsible for designing, tooling,
implementing, and qualifying the production facility for the 90 mm-33 layer
alumina-based ceramic substrates and later the three-level thin-film wiring pro-
gram for the glass ceramic package. After product qualification, his group had
responsibility for the continuous improvement program required to attain the
yield, cost and reliability requirements of the product. He co-chaired the team
that accomplished IS09001 registration of the IDM Microelectronics Multi-Layer
Ceramic facility.

John (Jack) A. Kreuz-DuPont Fellow; High Performance Films,-Dupont,


Circleville Research Laboratory, Circleville, Ohio. Jack has done research on
polyimides for 35 years. Most of his efforts have been connected with Kapton®
polyimide film, where he has contributed to improvements in monomers, polymer
backbones, specifications, adhesion, processing, thermal durability, and hydroly-
sis resistance, as well as to understandings of imidizationlcyclization chemistry.
He has been a participative corporate member of the Institute of Printed Circuits
(IPC) and of the Materials Research Society (MRS). He is presently a member
of the American Chemical Society (ACS). His research has been documented
by 14 papers, over 20 US and foreign patents, and numerous seminars.
Jack received his BS from St. Bonaventure University and his Ph.D. in organic
chemistry from the University of Notre Dame.

Ananda H. Kumar-Member of Technical Staff-Applied Materials, Santa


Clara, CA. Ananda has worked in the areas of ceramic and thin film packaging
for nearly twenty years, first at IDM Microelectronics for 15 years, and later at
the David Sarnoff Research Center, Princeton, NJ. Ananda has written many
papers and has earned thirty patents related to many areas of packaging technol-
ogy. Ananda was the recipient of the top IBM Corporate Technology Award in
1991 for his valuable contributions to the development of glass-ceramic multilayer
substrates used in the Company's most advanced computers. He has a Ph.D.
degree in Ceramic Engineering from University of Illinois in Urbana.

Nobuyuki Kuramoto-General Manager-Fujisawa Research Laboratory, To-


kuyama Corp., Fujisawa, Japan. He received the master's degree of science from
the Kyushu University, Japan in 1971 and joined Tokuyama Corporation. Since
1977, he has been engaged in the research and development of fine ceramics,
especially aluminum nitride. He received the R.M. Fulrath Pacific Award on
"Development of Translucent Aluminum Nitride Ceramics" form the American
Ceramic Society in 1988, and the Prize of Technologies from the Ceramic Society
of Japan in 1996 on "Development of High Purity AIN Powder and It's Related
Products". He has 54 Japanese patents, 9 US patents, and 28 from other countries.

Joy Laskar-Assistant Professor, School of Electrical and Computer Engineer-


ing-Georgia Institute of Technology, Atlanta, GA. Prof. Laskar received the
PhD degree in Electrical Engineering from the University of Illinois at Urbana-
11-988 AUTHORS' BIOGRAPHIES

Champaign in 1991 and is currently an assistant professor in the School of


Electrical and Computer Engineering at Georgia Tech. His research interests
include characterization and design techniques with applications to wireless elec-
tronics and high speed packages. He has served as a research engineer at IBM's
TJ Watson Research Center, Visiting Assistant Professor at the University of
Illinois, and Assistant Professor at the University of Hawaii. He serves on the
technical program committee for IEEE's MTT symposia and is a co-organizer
of the Advanced Heterostructure Workshop. He is a 1995 recipient of the Army
Research Office's Young Investigator Award and a 1996 recipient of the National
Science Foundation CAREER Award.

Richard F. Levine-Program Manager, Memory BAT World Wide Manufactur-


ing Program Office-IBM, Fast Fishkill, NY. Dr. Levine has worked for IBM
in various manufacturing and development positions in semiconductor, and pack-
aging. His work has included ceramic packaging sintering and thin films manufac-
turing engineering, product transfer leadership as well as product engineering.
He received his BS and PhD from Rutgers University in Ceramic Engineering
and has presented to ISHM and ASME on thin films and product transfer. He
holds two patents on semiconductor reliability.

Weiping Li-Ph.D. student, Packaging Research Center (PRC)-Georgia Insti-


tute of Technology, Atlanta, Georgia. Mr. Li is engaged in via formation processes
in photosensitive polymers for the construction of the single level integrated
module (SLIM) proposed by Packaging Research Center at Georgia Tech for the
next generation electronic package. From 1990 to 1993, he was a visiting scholar
in the Institute for Materials Research, GKSS-Research Center, Geesthacht, Ger-
many working on high temperature fracture mechanics of metallic materials.
From 1987 to 1990, he was an assistant researcher in the Institute of Materials
Science and Engineering, Shanghai Jiao Tong University, China, working on
intermetallic compounds. He received both his BS and MS in materials science
from Shanghai Jiao Tong University. He is a member of ISHM-the microelec-
tronics society.

Jerry P. Lorenzen-owner and principal consultant-Lorenzen Consulting,


Stone Ridge, NY. He supports methods to improve quality and reduce cost for
all phases of design, development, and production. Dr. Lorenzen consults in
the application of Taguchi design of experiments, quality function deployment,
statistical process control, basic quality tools, total quality management, and team
building. He also teaches classes, seminars, and workshops on a wide range of
quality topics for industry, colleges, and government.
Jerry has a Ph.D. in chemistry from Oklahoma State University. He has more
than 25 years of academic and industrial quality engineering experience in elec-
tronics, chemicals, automotive, ceramics, and data processing. He has written
numerous journal articles, is a contributing author to three books, and has two
patents. Dr. Lorenzen is an ASQC Certified Quality Engineer.

Wadie F. Mikhail-Senior Statistician (retired)-IBM, Charlotte, NC. Dr. Mik-


hail was involved in the implementation of the design of experiments and data
AUTHORS' BIOGRAPHIES 11-989

analysis to the manufacturing process with an emphasis on quality control. He


developed models for evaluating the reliability and maintenance strategies for
the first semiconductors memories with error correction used at IBM. He had
worked on models for circuit and product delay, yield analysis, and physical
design, where he collaborated in providing methods for estimating wiring space
requirements. He was considered an expert in the area of wireability. Dr. Mikhail
received his BSc (First Class Honors) and MSc in mathematics from Cairo
University, Egypt, and his PhD in statistics from the University of North Carolina.
He was a senior member of IEEE, and was the official IBM representative to
the American Statistical Association. He has 13 journal and proceedings publica-
tions in the areas of physical design, reliability, and statistics, four patents, and
11 publications in the IBM Technical Disclosure Bulletin.

Wataru Nakayama-Professor; Department of Mechanical & Intelligent Sys-


tems Engineering-Tokyo Institute of Technology, Japan. Wataru Nakayama
received a Doctor of Engineering degree from Tokyo Institute of Technology in
1966. From 1970 to 1989 he was with Mechanical Engineering Research Labora-
tory, Hitachi, Ltd., as a heat transfer specialist. His work experience includes
semiconductor packaging and cooling of computers. Since he moved to Tokyo
Institute of Technology, he has been teaching thermal management of electronic
equipment and modeling of flow and heat transfer processes in industrial equip-
ment. Dr. Nakayama is Past Chairman of the Heat Transfer Society of Japan, a
Fellow of ASME, a Senior Member of IEEE, and the recipient of ASME Heat
Transfer Memorial Award in 1992.

Luu T. Nguyen-Engineering Manager, Strategic Planning & Development


Group-National Semiconductor, Santa Clara, CA Dr. Nguyen is with the
Package Technology Group working on various aspects of packaging reliability
issues and design-for-manufacturability. He is currently in charge of external
leveraging with the Government, industry consortia, and research universities.
He is also the Focused Technical Advisory Board (FTAB) member of the
Sematech Assembly & Packaging Thrust, and the Chair of the Packaging
Sciences Technical Advisory Board of the Semiconductor Research Corp. He
obtained his Ph.D. in Mechanical Engineering from MIT on a Hertz Foundation
Fellowship, and has worked for IBM Research and Philips Research. He has
co-edited 2 books on packaging, and has 10 patents, over 15 invention
disclosures, and over 100 publications.

Koichi Niwa-Board Director-Fujitsu Laboratories Ltd., Kanagawa, Japan.


Dr. Niwa received his BS in physics from Chiba University, Japan in 1964. and
his Ph.D. in ceramic engineering from Tokyo Institute Technology in 1988.
He has pioneered the development of glass/ceramic composites which combine
a low sintering temperature, a low dielectric constant and a low thermal expansion
coefficient. He has contributed to the technology of copper co-firing with glass/
ceramic sheets. One of the contributions is the organic binder which can be
burned out completely in non-oxidizing atmospheres. He discovered a special
firing procedure for multilayer ceramic circuit board with copper conductors.
He has more than 50 published papers and more than 10 US patents. He
11-990 AUTHORS' BIOGRAPHIES

received 1992 Ohkouchi Technical Award for the development of multilayer


glass/ceramic circuit board for super-high-density mounting, and Technical
Award from Japanese Ceramic Society for the development of the ceramic circuit
board for high speed computer, in 1995. He is a member of the Electronics
Division in ACerS. He has been the Fellow of ACerS since 1991. He is a Vice
President of Ceramic Society of Japan.

Sevgin Oktay-founder and president-Oktay Enterprises International, Pough-


keepsie, NY. Mr. Olaay is currently involved in projects ranging from forming
business partnerships abroad to preparing patent applications for clients in com-
puter technology and semiconductor manufacturing. During his more than thirty
year career with IBM, he has held various technical and management positions,
published extensively in scientific and professional journals and books in the
areas of heat transfer, microelectronics and computer systems technology. He
has more than forty inventions of which ten are US Patents. He has conducted
workshops and seminars on computer technology world-wide, including Japan,
China and Turkey. He is a founding member and an Honorary Member of MIM,
the Society of Turkish American Architects, Engineers and Scientists, Inc., and
served as its president in 1994. He is a member of ASME, American Society of
Mechanical Engineers, where he served as chairperson of various committees
and is currently the Secretary and Treasurer in the Executive Committee of the
Electrical and Electronic Packaging Division. He is also a member of the Institute
of Electrical, Electronic Engineers, and an elected Fellow of ASME. He is listed
in Who's Who in Engineering. Mr. Oktay received his BS in Engineering Science
from Antioch College, Yellow Springs, Ohio in 1959, and MS and Professional
ME in Mechanical Engineering in 1960 and 1963, respectively, from Columbia
University, New York, NY.

Kanji Otsuka-Professor of Meisei University, College of Informatices, Depart-


ment of Electronics and Computer Science-Tokyo, Japan. He graduated with a
BS in ceramic engineering from Kyoto Institute of Technology in 1958, and
received a doctor grade in ceramic material science from Tokyo Institute of
Technology in 1987.
He was with Hitachi, Ltd for 34 years, beginning in 1935. He began his career
in Ge and Si transistor manufacturing and packaging technology. From period
of IC and LSI, he worked in LSI package design and its production issues that
included all kind of technology. The job function shifted to the design of system
packaging for mainframe computers, especially in high densitylhigh speed pack-
aging and MCMs for coming VLSI era. Final job function was system packaging
design in RISC processors for workstation.
He transferred his job from Hitachi to Meisei University in 1993 as the
professor in the class of microelectronics. He has taught microelectronics,
computer design, electronics circuit and electromagnetic phenomena in bus
signal transmission.
He is a board member of the Japan Institute for Interconnection and Packaging
Electronic Circuit, and Society for Hybrid Microelectronics, a councilor of The
Japan Federation of Engineering Society, a professional member of The Institute
of Electronics, Information and Communication Engineers, and a senior member
AUTHORS' BIOGRAPHIES 11-991

of IEEE engaged in CPMT including TC-6 committee and ECTC program com-
mittee.

Burhan Ozmat-Senior Scientist-Harris Power R&D, Latham, NY. Since Feb-


ruary 1994, Dr. Ozmat has been the leader of the advanced packaging group.
His current responsibilities are in advanced packaging materials, processes and
cooling technologies for power electronics. He worked for Texas Instruments
from February 1987 to 1994. As a Member of Technical Staff, his responsibilities
at TI included doing applied research and development for the advanced packag-
ing materials and technologies. MCMs, and multilayer interconnects for micro-
wave and digital applications, light weight constraining core materials for SMT
applications, Finite Element Analysis of vibration and thermal fatigue for the
solder joints of leaded and leadless ceramic packages, thermal technology develo-
ment for high density high performance systems were among his areas of interest.
Previously he worked for IBM in East Fishkill, New York where he was responsi-
ble for doing applied research and development on the advanced thermal techno-
logies, packaging materials and bonding for MCMs. He received his Ph.D. in
mechanical engineering from Massachusetts Institute of Technology in 1984. His
concentration was in the mechanical behavior of materials field. His main area
of interest was in the high temperature deformation and failure mechanisms of
engineering materials. Burhan has several publications and patents and is a
member of ASME, ASM, Sigma Xi and IEPS.

Michael G. Pecht-Professor and Director; CALCE Electronic Packaging Re-


search Center (EPRC )-University ofMaryland, College Park, MD. The CALCE
EPRC is sponsored by over 35 organizations, and conducts reliability research
to support the development of competitive electronic products in timely manner.
Dr. Pecht has a BS in Acoustics, a MS in Electrical Engineering and a MS
and PhD in Engineering Mechanics from the University of Wisconsin. He is a
Professional Engineer, an IEEE Fellow, an ASME Fellow and a Westinghouse
Fellow. He is the chief editor of the IEEE Transactions on Reliability, an associate
editor for the SAE Reliability, Maintainability and Supportability Journal, an
associate editor on the International Microelectronics Journal, and on the advisory
board of IEEE Spectrum and the Journal of Electronics Manufacturing. He
serves on the board of advisors for various companies and consults for the U.S.
government, providing expertise in strategic planning in the area of electronics
packaging.

Timothy C. Reiley-Research Staff Member-IBM Almaden Research Center,


San Jose, CA. Dr. Reiley, currently working in the area of micromechanics for
data storage, has held several staff and management positions at IBM-Yorktown
and IBM-Almaden in the areas of packaging, electrophotographic printing and
magnetic storage. Before joining IBM, he worked at the Oak Ridge National
Laboratory in experimental solid state physics associated with radiation effects
on mechanical behavior. He received a Sc. B. Degree from Brown University
in Materials Science Engineering, and M. S. and Ph. D. degrees from Stanford
University in Materials Science and Engineering. His current research focus is
on the micromechanical miniaturization and integration of disk drive components.
11-992 AUTHORS' BIOGRAPHIES

He has co-authored over 50 external publications and has 20 patents awarded


or pending.

Ottmar Rohde-head of the laboratory for research and development of photo-


sensitive polymides (deceased)-Ciba-Geigy, Basel, Switzerland. Dr. Rohde
passed away in December of 1994. He was working on updating the section on
preimidized photosensitive polyimides in the polymer chapter at the time. His
material was passed on to Dr. Richard Hopla for completion. Dr. Rohde has been
missed, but his work continues to show through his contribution to this revision.
He received his MS. degree in Chemistry from the University of Oregon and his
Ph.D. from the Institute of Polymer Chemistry at the University of Freiburg in
Germany. His experience included spin-labeling, photopolymer chemistry, and
a visiting professorship in Polymer Chemistry in Taipei, Taiwan. He was a
member of the Materiel Research Society, and is the author of several patents
and publications related to polyimides and photosensitive polymers.

Kenneth Rose-Professor; Electrical, Computer, and Systems Engineering De-


partment-Rensselaer Polytechnic Institute, Troy, NY. He received the BS degree
in engineering physics in 1955 and the MS and Ph.D. degrees in electrical
engineering in 1957 and 1961, all from the University of Illinois in Urbana-Cham-
paign.
He joined the General Electric Research Laboratory in 1961 and Rensselaer
Polytechnic Institute as an Associate Professor in 1965, becoming a Full Professor
in 1971. At Rensselaer his research has included the use of superconductors for
radiation detection and high performance packaging, the growth and characteriza-
tion of nitrides and silicon-rich oxides, and the development of CAD tools for
VLSI design. A recent interest has been the development of CAD tools for the
early estimation of interconnect requirements.
He is a senior member of the IEEE, serving as co-editor of a special issue of
the Proceedings of the IEEE on Thick and Thin Films for Electronic Applications
in 1971, and a member of the Materials Research Society, serving as a co-chair
of MRS Symposia on the role of interfaces in microelectronics processing in
1982 and 1993. He is a recipient of Rensselaer's distinguished faculty award
and a founding member of Rensselaer's Center for Integrated Electronics.

Eugene J. Rymaszewski-Research Professor in Materials Science & Engineer-


ing and Associate Director of the Center for Integrated Electronics-Rensselaer
Polytechnic Institute, Troy, NY. His current interests include teaching and research
in the high data-rate packaging structures, signal and power distribution systems,
thermal management and stress analysis. In 1956 he joined the IBM Research
laboratory in Poughkeepsie, NY to work on leading-edge computer technologies,
starting with Project STRETCH. Subsequently, as engineer and engineering man-
ager, he had contributed to product development of semiconductor chips, their
packaging and design and use of test equipment for many generations of IBM
mainframes, notably IBM 7000 series, Systems 360, 370, 308X, 3033, 3090 and,
lastly, 390.
From 1950 to 56 he worked in the Microwave Laboratory of C. Lorenz AG,
subsidiary of ITT, Pforzheim, Germany on crystal-controlled transmitter section
AUTHORS' BIOGRAPHIES 11-993

of a microwave TV link and on microwave stages of their 120 telephone channel


microwave link. He is a Senior member (now Senior Life Member) of IRE-IEEE
since 1957 and a member of Research Society of America-Sigma Xi.

Koppolu Sasidhar-Georgia Tech, Atlanta, GA. Mr. Sasidhar received the Btech
degree in Computer Science and Engineering from the Indian Institute of Techno1-
ogy, Kharagpur, India, in 1993 and the MS degree in Electrical Engineering from
Georgia Institute of Technology, in 1996. Presently, he is working towards his
Ph.D. at Georgia Tech. His main research interests include Multi-Chip Module
Testing, Parallel and Distributed Algorithms, Graph Theory and Built-In-Self-
Test (BIST).

George P. Schmitt-Senior Engineer (on leaveY-IBM Microelectronics Divi-


sion, Enidcott, NY. Mr. Schmitt has been involved with the development of
materials for advanced second-level packaging. He has worked for a number of
corporations in thermosetting polymer systems; the last 23 years have been with
IBM. His fields of interest are in the chemorheology of thermosets, chiefly
laminating resins, and in photo-patternable coatings. He received a BS in chemis-
try from Gettysburg College. He is a member of SPE, ACS and Sigma Xi. He has
six patents and as many publications, in the area of laminating resins, continuous
systems for resin manufacture and photosensitive systems.

Donald P. Seraphim-IBM Fellow (retired), Consultant-Vestal, NY. Dr. Sera-


phim is a consultant in the field of Electronic Packaging and in materials applied
science. Previously he headed IBM's Systems Technology Division's Materials
Science and Engineering Function which developed advanced flexible and printed
circuit board designs including direct-chip attach applications. The function had
strong collaborations with IBM Research and Universities. He has been a member
of the IBM Corporate Technology Committee reporting to the IBM Chief Scien-
tist. He managed the design and development of IBM's first bipolar integrated
circuits and first CMOS circuits. He also managed the development of PCBs and
connector packaging for the 4300 and 3091 large systems. He has a masters
degree in Applied Science from the University of B.C., Canada and a doctorate
in engineering degree from Yale University. He has over 50 published papers
and a substantial list of patents. Recently he has been applying his background
in packaging to applications in flat panel displays in an entrepreneurial role.

Michael J. Sheaffer-Director of Technical Support, Packaging Materials


Group-Kulicke & Soffa Industries Inc., Willow Grove, PA. Mr. Sheaffer is
responsible for the operation of the K&S worldwide customer support network
for packaging materials. This includes regional laboratories and on-site engineers
that provide complete solutions to customer wire bonded packages, which includes
bonding parameters, wire, tools, and material handling parts. He joined K&S in
1981 and has been actively involved with wirebonder manufacturing and wire
bonding processes since 1977. He is the author of several technical papers and
articles and holds patents in low looping trajectories and bonding dynamics. He
received his MS degrees from Drexel University and Millersville University and
his MS degree from Ball State University.
11-994 AUTHORS' BIOGRAPHIES

Yuzo Shimada-Senior Manager, Material Development Center-NEC Corp.


Kanagawa, Japan. Yuzo Shimada received the M.S. degree in chemistry from
Kyoto University in 1979. He joined the NEC Corporation in 1979 as inorganic
chemist. He engaged in material and process development for high density packag-
ing technology at Material Development Center and Computers Division. Interests
of his research and development are in the areas of electrical materials, process
technology of multilayer packaging substrate for large scale computer systems
and advanced packaging and interconnect technologies for high speed VLSI. Mr.
Shimada is a member of the Ceramic Society of Japan and the Institute of
Electronic, Information, and Communication Engineers of Japan. He is also a
member of the American Ceramic Society.

Robert E. Simons-Consultant-Electronics Cooling Applications, Poughkeep-


sie, NY. Mr. Simons' consulting firm provides thermal analysis, design, and
instructional services. Prior to retiring from IBM in 1993, he was a manager in
the Poughkeepsie Advanced Thermal Laboratory. He joined IBM in 1966 and
throughout his career he participated in the development of new and advanced
cooling techniques for computer electronics. As a co-inventor of the cooling
scheme for the IBM Thermal Conduction Module (TCM), he received an IBM
Outstanding Innovation Award and a Corporate Award. He is also the recipient
of ten IBM Invention Achievement awards, holds 18 patents, and has published
over 35 papers and book chapters on cooling electronic packages and systems.
While at IBM he was elected a member of the IBM Academy of Technology.
He has been an active participant and organizer of SEMI-THERM Symposia
serving in the capacities of session, program, and general chairman. In recognition
of his contributions to the art and science of thermal management for electronic
systems, he was awarded the 1995 SEMI-THERM Symposium significant con-
tributor award. He is a past chairman of the ASME Heat Transfer Division
K-16 Committee on Heat Transfer in Electronic Equipment and is a member of
the IEEE.

Pratap Singh-President-RAMP Labs, Round Rock, TX. RAMP Labs is a


consulting service specializing in pin-in-hole and surface mount assembly manu-
facturing processes, packaging reliability and failure analysis.
Prior to this, he worked at IBM Endicott, NY and IBM Austin, TX where he
had responsibility in the development of high density muti-Iayer printed circuit
boards, zero insertion-force connectors, BGA sockets and stress-test reliability
testing. He has 27 years experience in PCB manufacturing, SMT and PIR assem-
bly processes, statistical process controls and electronics packaging reliability
testing.
Pratap has one patent and published 21 invention disclosures in the IBM
Technical Disclosure Bulletin. He has authored and co-authored 5 IBM technical
reports and 10 technical papers. Pratap was awarded first and second level IBM
Invention Awards and two levels of Technical Author Recognition awards.
He has chaired and co-chaired sessions at IEPS in 1982 and at NEPCON in
1993, 1994 and 1995. Pratap was also the member ofNEPCON Advisory Council
from 1991 to 1995 and is active in Central Texas Electronics Association for
the last four years. He is also a member of IEEE since 1970. Pratap received
AUTHORS' BIOGRAPHIES 11-995

his BS in mechanical engineering from University of Ujjain, India in 1962 and


MS in industrial engineering from the University of Iowa, USA in 1969 before
joining IBM at Endicott, NY.

Toshio Sudo-Senior Research Scientist-Toshiba Corporation, Manufacturing


Engineering Research Center, Yokoham, Japan. He is currently developing the
board-level electrical design methodology for high-speed digital/analog systems.
Mr. Sudo joined the Research and Development Center, Toshiba Corporation,
in 1975, where he was engaged in the research of high-speed GaAs packaging and
high-density multichip module technology. In 1991, he moved to Semiconductor
Device Engineering Laboratory, where he has been engaged in the development
of high-performance packages and flip-chip technology for advanced CMOS
VLSIs until March 1996. He published several articles for the evaluation of
CMOS simultaneous switching noise using a test chip. His research interests
include the electrical modeling and characterization of high-speed interconnec-
tions and high-performance packages. He received the RE. and M.E. degrees in
electrical engineering from Tohoku University, Sendai, Japan, in 1973 and 1975,
respectively. He is a member of IEEE CPMT society, ISHM and IEICE (Institute
of Electrical, Information, and Communication Engineering) in Japan.

Madhavan Swaminathan-Manager, Design and Simulation, Packaging Re-


search Center-Georgia Institute of Technology, Atlanta, GA. Madhavan Swami-
nathan received the M.S. and Ph.D. degrees in electrical engineering from Syra-
cuse University in 1989 and 1991, respectively. In 1990, he joined the Advanced
Technology Division of the Packaging Laboratory at IBM, E.Fishkill, New York
where he was involved with the design, analysis, measurement and characteriza-
tion of packages for high performance systems. At IBM, he was part of a team
that was instrumental in the design, development and prototyping of IBM's low-
cost multilayer thin-film technology. He joined the Packaging Research Center
at Georgia Tech in 1994 to pursue unique challenges arising in low cost packaging.
His research interests are in package design, modeling, measurement and testing.
Dr. Swaminathan has 45+ publications in refereed journals and conferences, six
inventions, two issued patents and has taught several short courses in packaging.

Masami Terasawa-General Manager, Technology Strategy Division, Corpo-


rate Semiconductor Components Group-Kyocera, Kyoto, Japan. Dr. Terasawa
was born in Gifu, Japan on March 24, 1951. He received the BS, MS, and Ph.D.
degrees from Osaka University, Osaka Japan in 1973, 1976, 1979 respectively.
In 1979 he joined Kyocera Corporation, Kyoto, Japan and has been engaged in
research and development of advanced packages. He developed a CulPolyimide
package for which he is regarded as a pioneer in the industry. In 1983, he received
the best paper award from ISHM, lEPS in the United States. Further he developed
various new ceramics-AlN, Mullite and Glass Ceramic.

Theodore (Ted) G. Tessier-Manager, New Materials and Processes Group,


Semiconductor Products Sector-Motorola Inc., Tempe, AZ. Ted is Manager of
the New Materials and Processes Group within Motorola's Semiconductor Prod-
ucts Sector working in the area of chip scale packaging technologies. He has a
11-996 AUTHORS' BIOGRAPHIES

Bachelor of Science degree in Organic Chemistry from Laurentian University,


Sudbury, Canada and a MSc. degree in Applied Polymer Chemistry from the
University of Ottawa. Ted has published more than 45 papers and articles in the
area of high density substrate technologies and interconnection.

Paul A. Totta-IBM Fellow-IBM, East Fishkill, NY. Mr. Totta is currently


an advisor to the General Manager of IBM's packaging business within the
Microelectronics Division. Previously, and for many years, he was project man-
ager for the development of chip and package thin film interconnections including
IBM's flip chip solder bumps. He is the recipient ofISHM's Outstanding Techni-
cal Achievement Award for these activities. In addition he is an ASM Fellow
and has over 25 issued US patents in his field.

Rao R. Tummala-Petit Chair Professor and Director of PRC-Georgia Tech,


Atlanta, GA. Dr. Tummala received the BS degree in Physics, Mathematics and
Chemistry from Loyola College, India, the BE degree in Metallurgical Engineer-
ing from the Indian Institute of Science, Banglore, India, the MS degree in
Metallurgical Engineering from Queen's University, and the Ph.D. degree in
Materials Science and Engineering from the University of Illinois. He joined the
faculty at Georgia Tech in 1993 as a Pettit Chair Professor in Electronics Packag-
ing and as Georgia State Research Scholar. He is also the Director of the Low-Cost
Electronic Packaging Research Center funded by NSF (as one of its Engineering
Research Centers), the state of Georgia, and US electronics industry. Prior to
joining Georgia Tech, he was an IBM Fellow at the IBM Corporation, where he
invented a number of major technologies for IBM's products for displaying,
printing, magnetic storage and multichip packaging. He is both a fellow of IEEE
and the American Ceramic Society, a member of the National Academy of
Engineering, 1996 General Chair of IEEE-ECTC, and 1996 President of ISHM.
He was recently named by Industry Week as one of the 50 Stars in the US, for
improving US competitiveness.
He is co-editor of widely-used Microelectronics Packaging Handbook. He
published 90 technical papers and holds 21 US patents and forty four other
inventions. He received a number of awards including: David Sarnoff award,
sustained technical achievement award from IEEE, John Wagnon's award from
ISHM, Materials Engineering achievements award from ASM-I, distinguished
alumni award from University of Illinois, and the Arthur Friedberg Memorial
award from American Ceramic Society.
Dr. Tummala's current research interests include packaging materials (metals,
ceramics, and polymers) and processes, mechanical properties of materials, thin
and thick MCMs, thermal and electrical designs, and integrated passive compo-
nents.

Elizabeth J. Twyford-Sr. Member of the Technical Staff-TRW, Redondo


Beach, CA. Dr. Twyford received a BA degree in Philosophy from St. JohnTs
College in Annapolis MD, a BS in Electrical Engineering from Florida State
University, and a MS and Ph.D. from Georgia Institute of Technology with
support from a Kodak Fellowship (1990-92) and an NSF Fellowship (1995). At
Georgia Tech, she studied under Dr. Nan Marie Jokerst, focusing on optical
interconnects, hybrid integration of III-V semiconductor devices and glass waveg-
AUTHORS' BIOGRAPHIES 11-997

uides, and photoelectrochemical etching of diffraction gratings on GaAs/AIGaAs.


The Sigma Xi chapter of Georgia Tech awarded her the Doctoral Dissertation
Award for 1995-1996. At TRW she is working on fiber-optic analog RF links
using linearized optical modulators.

Puligandla Viswanadham-member of the technology development team in the


Circuit Card Assembly unit-Texas Instruments Inc., Lewisville TX. He is cur-
rently involved in the Ball Grid Array and other technology implementation
projects. Prior to joining Texas Instruments he worked at the International Busi-
ness Machines Corporation in Austin TX, Endicott NY, and Rochester MN
facilities. He was involved in the process development and qualification of Surface
Laminar Circuitry, Assembly and Reliability of fine-pitch quad flat-packs, Thin
Small-Outline Packages, and Tape Automated Bonding.
While at IBM Austin he was also site analytical laboratories manager during
1989-1990. As a member of the Materials and Process Engineering group at IBM
Rochester Viswanadham was involved in corrosion studies, analytical methods
development, plating and contamination control. Prior to joining IBM his research
activities included high temperature chemistry and thermodynamics of binary
and ternary chalcogenides, atomic absorption, slag-seed equilibria in coal fired
magnetohydrodynamics energy generation, and astrophysics.
He has authored or co-authored over 65 technical publications in journals,
symposium proceedings, and trade magazines. He has authored or co-authored
four book chapters in the areas of microelectronic packaging, tape automated
bonding, fine pitch surface mount technology and ball grid array technology.
He received the first and second IBM Invention Achievement awards, an IBM
Excellence award, and fourth level Technical Author Recognition award. During
1974-78 he was on the faculty of Ohio Dominican College, Columbus, Ohio as
Assistant Professor and taught physics and chemistry.
Puligandla Viswanadham has a Ph.D., degree in chemistry from University of
Toledo, Ohio, and an M.Sc., degree in chemistry from Saugor University, Saugor,
India. He co-authored two patents and 15 invention disclosures.

B.D. Washo-Member Technical Staff, Technology-AMP Incorporated, Harris-


burg, PA. Formerly he was with the IBM Development Laboratory for 271/2 years.
His speciality was cable/connector engineering. Dr. Washo works in connector
mechanics and teaches "The Electrical Engineering of Signal Connectors", an
internal course at AMP Incorporated. He has a BS in Chemical Engineering,
Pennsylvania State University, and a Ph.D. in Chemical Engineering/Polymer
Science from Rensselaer Polytechnic Institute. He is a member of the American
Society of Metals, the Society of Plastics Engineers, and Tau Beta Pi and Sigma
Xi honorary scientific societies. He has published 18 papers in the areas of
polymer science, electrocoating, plasma polymerized thin films, spin coatings,
and the electrical modeling and characterization of cablelconnector packages. He
has 37 patents pUblications and holds 5 issued US patents.

Thomas A. Wassick-Senior Engineer-IBM, East Fishkill, NY. Mr. Was sick


is currently involved in the development of thin film technology for advanced
high-performance computer packaging. He is responsible for thin film repair
technologies and methodologies and electrical test diagnostics for IBMs thin
11-998 AUTHORS' BIOGRAPHIES

film packaging programs. His field of expertise is in laser-based processing for


deposition and etching of microelectronic materials. He received his BS degree
in Biomedical Engineering from Rensselaer Polytechnic Institute and an MS
degree in Materials Engineering, also from RPI. He is a member IEEE, ISHM,
SPIE, and OSA, He holds a number of patents in the laser processing area, has
authored several technical articles involving MCM-D repair technologies and is
the co-author of a book chapter on repair and rework of microelectronics.

Toshihiko Watari-Director and General Manager-PPC Engineering and


Production Division, NEC-Niigata, Kashiwazaki City. Niigata, Japan. Mr. Watari
is the director and general manager of PPC Engineering and Production Division,
NEC- Niigata. He joined NEC in Tokyo in 1967. Since 1976 to 1994 he was
responsible to develop the circuit and packaging technologies for high perfor-
mance mainframes and Supercomputers. He is a member of IEEE and IEICE
Society of Japan. He is a graduate of Hiroshima University, Hiroshima Japan,
and he received a BE degree of Electrical Engineering. He has written many
articles and presented many papers in packaging technology development. He
holds various patents in the areas of computer circuit and packaging design.
He is now managing the development and production of Electro-Photography
for printers.

Scott Wills-Assistant Professor; Electrical and Computer Engineering-Geor-


gia Institute Of Technology, Atlanta, GA. He received his BS in Physics from
Georgia Tech in 1983, and his SM, EE, and ScD degrees in Electrical Engineering
& Computer Science from MIT in 1985, 1987, and 1990, respectively. His
research interests include High Throughput Parallel Architectures, System Inte-
gration, Interconnection Networks, and VLSI. He is a member of the IEEE and
the Computer Society.

C.P. Wong-Professor of Materials Science and Engineering and Assistant


Director of the Packaging Research Center-Georgia Institute of Technology,
Atlanta, GA. Dr. Wong received the BS. degree in chemistry from Purdue Univer-
sity, and the Ph.D. degree in organic/inorganic chemistry from the Pennsylvania
State University. After his doctoral study, he was awarded two years as a postdoc-
toral scholar with Nobel Laureate Professor Henry Taube at Stanford University.
He joined AT&T Bell Laboratories in January 1977 as a member of the
technical staff. He was appointed an AT&T Bell Laboratories Fellow in 1992.
In January, 1996 he joined the Georgia Institute of Technology. He is responsible
for the Packaging Research Center's Assembly, Reliability, and Thermal Manage-
ment Areas. His research interests lie in the fields of polymeric materials, high Tc
ceramics, materials reaction mechanism, IC encapsulation, in particular, hermetic
equivalent plastic packaging, and electronic manufacturing packaging processes.
He holds over 38 US patents, numerous international patents, has published over
90 technical papers and 100 key-notes and presentations in the related area.
Dr. Wong is both a Fellow of the IEEE and AT&T Bell Labs. He was president
of the IEEE-CPMT Society (1992, 1993). He currently chairs the IEEE Technical
Activities Board, Steering Committee on Design and Manufacturing Engineer-
ing (1995-).
INDEX
A Advanced SLT (ASLT), II-294-95
Aerospace systems, card packaging for, III-259
Accelerated testing, 1-522 A-I films, 1-455
Acceleration factors, and CBGAlCCGA Air delivery system, and external thermal
interconnection reliability, III-201-202 resistance, 1-373, 1-374
Accelerators, and plastic packaging, II-415-16 Air flow, and cooling systems, 1-376-78,
Acid-copper baths, III-317-18 1-393-95
Acoustic microscopy, 1-469 Airlliquid cooling capabilities, 1-36, II-36,
Active terminators, 1-230-32, 1-267-68 III-36, 1-122, II-122, III-122
Actuation, of connectors, III-423-24 Alignment
Additive electroless plating, of copper, II-711 of connectors, III-420-23
Additive electroplating, II-699, II-700-701, of optic fiber, III-507
II-705-706 Aliphatic amines, II-416
Additive process enhancement, 1-118, II-1I8, Alkaline etching, II-745
III-1I8 Alkyl substituents, and Probimide 400
Adhesive bonding, and chip-to-package polyimides, II-586
interconnections, II -266-68 Allied-Apical Co., II-534
Adhesives and adhesion Alloy 42, 1-284-86, III-362, III-363
chip bonding and, II-501, II-502 Alpha emission rates (AERs), and plastic
conductive for package-to-board packages, II-415
interconnections, III-223-33 Alpha-particle emission, and flip-chip solder-
fatigue and, 1-486-87 bump connections, II-179, II-181-82
films and coated-metal substrates, III-369-70 Alpha-particle protection layers, as chip
of high temperature non-polyimide coating, II-609
dielectrics, II-545 Alpha quartz, II-411
linear laminates, III-300-301 Alumina ceramic packaging, II-303, II-305-24,
molding compounds and, II-447 II-347, II-355, II-359-61
nonhermetic packaging and, II-883-84 Aluminum
photoneece and molding compound, II-611 chemical corrosion of, II-885
photosensitive benzocyclobutene, II-595-96 coated-metal packaging and, III-350-54
polyimide-to-glass-ceramic, II-525-26 ultrasonic welding and, II-190-93
polyimide-to-metal, II-526-27 wirebonding and, II-203-205
polyimide-to-polyimide, II-525 Aluminum-magnesium wire, II-204-205
Probimide 400 polyimides, II-587 Aluminum nitride (AIN), II-326-32
thin-film interfaces and polyimides, Aluminum-polyimide multilevel thin-film on
II-524-25 silicon, II-734-36
three-layer tape and, II-225 Amide-imide passivation, 1-455
wear of, 1-520 Arnkor Electronics, II-495
wirebonding and, II-207-208 Amoco Corp., II-713
11-1000 INDEX

AMPFLAT contact sockets, for area-array Backward coupling, in signal lines, III-283
MCMs, III-433-35 Backward cross-talk, III-408
AMP Inc., III-418 Back-wipe, of connectors, III-425
Amplifiers, optical, III-497-98 Balanced coaxial cable, III-449-50, III-451
Anisotropic conductive adhesives, III-228-33 Ball bonding, 11-188, 11-189, 11-190, 11-211-12
Anisotropic conductive film, 11-255-56, Ball and column attach, and CBGAlCCGA
11-268 interconnections, III-194-96
Anode reactions, 1-436-39 Ball-grid array (BGA)
Antisolder, 11-491 ceramic, 11-383
Aperture-plate molds, 11-453-54 development of, 1-66, 11-66, III-66, 1-93,
Apical polyimide films, 11-534-35 11-93, III-93, 1-94, 11-94, III-94, 1-95-96,
Application-specific integrated circuits 11-95-96, III-95-96
(ASICs), 1-18, 11-18, III-18, 1-235 market for, 1-87-88, 11-87-88, III-87-88
Area array C4 configuration, 11-141 package-to-board interconnections,
Area-array sockets, III-430--31 III-187-210
Area-array TAB (ATAB), 11-258 second-level assembly of, III-212-16
Area-bonding conductive (ABC) adhesive, tape automated bonding and, 11-258-59
III-232 thermal failure of, 1-504-506
Argon ion sputter cleaning, 11-148, 11-150 wireability of, 1-187-91
Aromatic amine curing agents, 11-416 Ball-limiting metallurgy (BLM) pad, 11-138,
Asahi Chemicals, 11-555, 11-558, 11-570 11-747
Asperity contact, 1-428 Bandwidth
A-spot contacts, III-400 connectors and, III-405
Assembled-MCM test, and diagnostic in high-speed data processing, 1-20, 11-20,
procedure, 11-866-68 III-20
Assembly Bar-code labels, 11-469, 11-470
ball-grid arrays and second-level, III-212-16 Bar printing, and solder deposition, III-168
ceramic ball-grid and ceramic column-grid Barrier metals, 11-697
arrays, III-191-92 Base-catalyzed photosensitive polymers,
fine-pitch components, III-I71-74 11-572-73
lead as hazardous material, III-218-21 Bell Laboratories, 11-132, 11-196, 11-217,
trends in packaging, 1-114, 11-114, III-114 11-272, 11-552, III-342-43
Assembly drawings, for coated-metal Benchmark processes, in manufacturing,
packaging, III-378-79, III-380 1-595-96
Assembly test chips (ATCs), 11-929-30 Benzocyclobutenes (BCB), 11-536, 11-593-606,
ASX (Advanced Statistical analysis program 11-664, III-324-25
eXtended), 1-208 Beryllium oxide, 11-333
ATARI computer games, 11-266 Beta-Phase Inc., III-424
AT&T Bell Labs, 11-302, 11-606, 11-787 BF Goodrich, 11-550
Attachment, of connectors, III-418-20 BICMOS, 1-44, 11-44, III-44
Attenuation, of cables, III-461-63 Binders
Audits for alumina packaging, 11-312
for ISO 9000 registration, 1-609-10 for glass-alumina composite, 11-360, 11-361
of process control, 1-605-606 removal of from glass-ceramic/copper
Automotive industry, 1-62, 11-62, III-62, 1-63, substrate, 11-351-53
11-63, III-63 Bipolar junction transistor (BIT) detector,
Avatrel Dielectric Polymers, 11-550 III-495
Average interconnection length, 1-140 Bisazide sensitizer, III-306
Bismaleimide triazine (BT), and tape
automated bonding, 11-225
B Blowers, selection of, 1-376-78
Backplane, and optical interconnects, III-507 Board characteristic impedance, 1-299-302
Backsputtering, during reactive ion etching, Board-level rework, of moisture-sensitive
11-693 components, 11-472-73
INDEX 11-1001

Boeing Company, 11-613, 11-787, III-362 CALCE Electronic Package Research Center
Boiling, and heat transfer, 1-359-68 (University of Maryland), 11-495
Bonding agents, and linear laminates, Calibration, of equipment monitors, 1-601
III-300--30 I Can and header transistor, 11-395
Bonding tools, and wirebonding, 11-200, Capacitance testing, 11-825-29, 11-830
11-208-209 Capacitors, and integrated passives in thin-fihn
Bond pad, contamination of, 11-481-82, 11-483 packages, 11-788-90
Borosilicate Carbon black, 11-420
glass and silica substrate, 11-373-74, 11-378 Carbon dioxide lasers, III-316
viscosity-temperature relation for, 11-343 Carbon oxidation, and glass-ceramic/copper
Bottomless-markless technology, and plastic substrate, 11-351-53
packages, 11-492 Carbon steel, and coated-metal packaging,
Bottom-side gating, transfer molding with, III-349-50
11-396 Card and board technologies
Bounce diagram, voltages for low-loss line, package hierarchy for, III-258
1-265 in second-level packaging, 1-1l5-16,
Boundary scan test, 11-861--63 11-115-16, III-1l5-16
Braided shield cable, III-468 Card connectors, III-474
Brazing, and package sealing, 1-125, 11-125, Card pad, surface-mount array and design of,
III-125, 11-902, 11-903 III-21 0--11
Brinnel hardness test, 1-428 Carrier size, and fatigue, 1-503
Broadcast infrared transmission, III-476 Case temperature rise, 1-356-57
B-stage epoxy prepreg, 11-778 Cast CCGA packages, III-I96
Buckling beam, cross section of, 11-843 Castro-Macosko model, 11-459
Buffer coating, for stress relief, 11-609 Cathode control, 1-442, 1-443
Buffer inventories, elimination or reduction of, Cavity-down PBGA packages, III-209
1-37, 11-37, III-37 Cavity gates, 11-480
Built-in self-test (BIST), 11-860-61, 11-863--64, Celest Patents, III-308
11-865-66, 11-867 Cellular phones, trends in mass and size of,
Built-in termination resistors, 11-746, 11-748 1-45, 11-45, III-45
Bulk conductor spring, III-410-11 Central electronic complex (CEC), connections
Bulk resistance, III-399 between direct-access storage devices and,
Bump bonding III-394
for flip chips, 11-266-68 Ceramic ball-grid array (CBGA), 11-383,
for optoelectronic devices, III-512 III-188-205, III-212-16
Bumped tape, and tape automated bonding, Ceramic carriers, stacked memory on, III-323
11-226, 11-231-32 Ceramic column-grid array (CCGA),
Bump-fabrication process, 1-78, 11-78, III-78 III-188-205, III-212-16
Bump integration technology (BIT), 1-77, Ceramic dual-in-line packages (CERDIPS),
11-77, III-77 11-404, 11-889, 11-895-96, 11-909
Burndy Corp., III-390 Ceramic packaging
Burn-in advanced SLT, 11-294-95
CBGNCCGA module level, III-194 alumina, 11-303, 11-305-24
connectors and, III-394, III-396 chip attachment and thermal dissipation of,
cycle for flip chip, 11-182-83 11-369-71
screens for, 11-492-93 chip carriers, flat packs, and hybrid
packages, 11-300--305

c defect detection for, 11-823


development of, 1-91-92, 11-91-92,
Cables, packaging of III-91-92
classification of, III-446-50 drawbacks of, 11-290--91
future of, III-476-77 dual-in-line packages (DIPS) and, 11-292-94
introduction to, III-385-88, III-445--46 future of, 11-372-87
signal theory and, III-450--72 heat transfer and, 1-393
11-1002 INDEX

hermetic types of, 11-895-98 Chip-logic cell, schematic of, 1-163


history of, 11-291-92 Chip-mounting layer, 11-745
introduction to, 11-284-85 Chip-to-package interconnections. See
laser writing for, 11-725 Interconnections
low-temperature, 11-335-69 Chip passivation integrity, plastic packages and
monolithic systems technology, 11-295 loss of, 11-477-78
multilayer ceramic technology, 11-297-300 Chip-scale packages (CSP), 1-82-85, 11-82-85,
plastic packaging versus, 11-399, 11-401-406, III-82-85, 11-383
11-489-90 Chip size, and fatigue performance, 1-491
recent developments in, 11-324-35 Chip temperature, 1-33, 11-33, III-33, 1-369-70
reliability of, 11-371-72 Chip-up PBGA packages, III-208
resistance and continuity testing, 11-842-50 Chlorofluorocarbons, ban on, 11-498
solid logic technology (SLT) and, 11-292 Ciba Geigy, 11-555, 11-608
substrate materials, 1-94, 11-94, III-94 Circuit, electrical design and attributes of,
thermal expansion of materials in, 11-421 1-205-16. See also Equivalent circuit
thick-film (crossover) technology, 11-295-96 Circuit card assemblies, trends in, 11-498
trends in, 1-93, 11-93, III-93 Circuit count
unique attributes of, 11-285-90 cycles per instruction and, 1-47, 11-47
Ceramic PGAs, 1-86, 11-86, III-86 number of connections per pin and, I-lSI
CERDIPs (ceramic dual-in-line packages), Circuit density, and media delay, I-59, II-59,
11-294, 11-404, 11-889, 11-895-96, III-59
11-909 Circuit drivers
C4 mechanics, 1-488-90 definition of, 1-199
Chang, F. Y., 1-209 electrical design of, 1-212-16
Characteristic impedance. See also Impedance Circuitization, materials and processes for,
cables and, III-454-56 III-304-20
connectors and, 111-405-407 Circuit receiver
delay adder and line, 1-298-99 definition of, 1-199
noise and board, 1-299-302 electrical design of, 1-206-12
Checkpoint process, three-step, 1-557 Circuit-simulation program, and package
Chemical failure, mechanisms of, 1-434-35, modeling, 1-270
11-885 Clad materials, for lead frames, 11-426
Chemically vapor deposited (CVD) diamond, Cleaning, of fine-pitch components, 111-178
11-242 Clock skew, 1-305
Chemical properties Closed-loop management review, 1-606-607
of high temperature non-polyimide Closed wire, III-447, III-449-50
dielectrics, 11-540-41 Cluster probe tester, 11-845
of molding compounds, 11-451 Coated metal, and second-level packages,
Chemical thermodynamics, 1-439-46 1-116,11-116, III-1l6, 1-117, 11-117,
ChemicaVwet processing, of dielectrics, III-I17
11-693-96 Coated-metal packaging
Chip attachment, and thermal dissipation of crossovers and vias, III-344
ceramic substrates, 11-369-71 design rules for, III-378-79, III-380
Chip-on-board future of, III-380-81
interconnections, III-428-29 ground plane and shield, III-343-44
thermal failure of, 1-506-507 insulators and, III-362-65
Chip carriers introduction to, III-340-43
ceramic, 11-300-305 metallization methods, III-365-67
connectors for, 111-473-74 metal substrates and, III-348-62
Chip-to-component connection, 111-391 reliability and, III-345-47
Chip joining, and alumina packaging, properties of coatings, III-372-78
11-323-24 substrate fabrication and assembly processes,
Chip-level interconnections, 1-71-78, 11-71-78, III-367-72
III-71-78 substrate as structure and heat sink, III-343
INDEX 11-1003

Coating introduction to, III-385-88


photosensitive benzocyclobutene, II-596-97 MCM level-2 interconnections and,
photosensitive polyimides, II-609-1O III-426-37
Coaxial cables, III-449-50, II-456, III-462 optical-fiber, III-441-42
Cofired ceramic modules, 11-142 reliability and, III-443-44
Coffin-manson-type relationships, 1-473-76 requirements for, III-395, III-397-426
Coherence domain multiplexing (CDM), standards for, III-444-45
III-511 types of, III-388-94
Coloring agents, 11-420 Constriction resistance, 1-424-27, 1-429-30,
Column-grid array, and package-to-board III-400
interconnections, III-187-210 Consumer Electronics, Inc., 11-364, 11-365
Common opens/shorts test (COST), 11-850 Contact conductance, 1-339, III-410
Communications, and optoelectronic Contact geometry, of connectors, III-424-25
technology, III-484-85 Contact-interface materials, and connectors,
Compatibility, of thin-film polymer structures, III-411-15
11-513-15 Contact micromotion, 1-524
Complementary metal on silicon (CMOS) Contact resistance
technology connectors and, III-399-400
design of PCs and, 1-44, 11-44, III-44 fretting cycles and, III-413
non-linear terminator, 1-231 normal force versus, III-401
trends in packaging and, 1-14-16, 11-14-16, reliability and, 1-423-31, 1-523
III-14-16 Contact sequencing, of connectors, III -402-
Component density, trends in, 1-107, 11-107, 403
III-I07, 1-110, 11-110, III-110 Contact target area, III-420-21
Composite noise tolerance, 1-210-11 Continuity testing, II-829, 11-831-33
Computational fluid dynamics modeling Continuous-flow manufacturing (CFM),
(CFD), 1-396 1-560-61, 1-590-91
Computer-on-a-chip, II-143 Continuous-wave (CW) lasers, 11-904-905
Computer-aided design (CAD) Controlled collapse chip connections. See Flip-
as design tool, 1-194 chip solder-bump connections
heat transfer and, 1-396 Convective heat transfer, 1-320, 1-322,
package electrical design and, 1-204-205 1-342-59, 1-528
rules driven, 1-308-309 Conventional flux-assisted joining, 11-160-61
Computer Aided Design of Microelectronic Cooling, of packaging, 1-119-23, 11-119-23,
Packages (CADMP II), 11-495 III-119-23. See also Heat removal; Heat
Computers. See Computer-aided design; transfer
Personal computers; Software Cooling fins, 11-758
Conducting plane, current flow in, 1-273 Copper
Conduction, and heat transfer, 1-322-42 additive electroplating, 11-701
Conduction-belt reflow soldering, III-368 alloys of and coated-metal packaging,
Conductive adhesives, for package-to-board III-354, III-357-61
interconnections, III-223-33 drilled holes plated with, III-317-20
Conductors, and properties of package electroless plating and, 11-804-808
materials, 1-86, II-86, III-86 foil and linear laminates, III-299-300
Conformal coatings, 1-454-56 glass-ceramic substrate, 11-345-48
Conformal multichip-flex, II-776-77 as gold wire substitute, 11-207
Conformal via copper-polyimide, II-738-40 lead frames and alloys of, 1-286, II-427
Connectivity, and wiring, 1-139-40 metallization with glass-ceramic, 11-338-39
Connector blocks, for multilayer printed 60/40 solder and reaction rate of, III -14 7
circuits, III -257 thermosonic welding and, II-J93-95
Connectors, packaging of. See also three-layer tape for tape automated bonding,
Interconnections, chip-to-package; 11-224
Interconnections, package-to-board Cordierite
cuture of, III-472-76 differential thermal analysis of, 11-343
11-1004 INDEX

glass-ceramic substrates and, II-34O, Critical processes, in manufacturing, 1-595


11-358-59 Crosby, Philip, 1-558-59
properties of, II-344 Cross-coupling noise, 1-27-28, 11-27-28,
viscosity-temperature relation for, 11-343 III-27-28
Corrosion Crossovers, and coated-metal packaging,
chemical failure and, 1-434-35, 11-885 III-344
chemical thermodynamics and, 1-439-46 Cross-talk, 1-204, 1-232, 1-237-41, 1-252-56,
control of, 1-453-46 III-408. See also Noise
die metallization and, II-478-79 Cryogenics, and C4 interconnected structures,
electrochemical processes and, 1-436-39, 11-184-85
1-453 Crystal growth inhibitor, 11-374
flip-chip solder-bump connections, 11-179 Crystalline silica, 11-411
of lead frame, 11-483-84 Cumulative distribution function (CDF), 1-411
nonhermetic packages and, 11-885 Curing
wire bond and bond-pad, 11-483 of high temperature non-polyimide
Cost dielectrics, II-546-47
design of PCs and, 1-42-43, 11-42-43, of photosensitive benzocyclobutene, 11-605
III-42-43, 1-271 of photosensitive polymers, 11-563, 11-565
glass-ceramic substrates and large-area plastic packaging and agents for, 11-415-16
processing, 11-380-82 of Probimide 400 polyimides, 11-589
of interconnections, 11-129, 11-130 Current-carrying capacity, of coated-metal
multichip packaging, 1-104-106, 11-104-106, substrates, III-376-78
III-I04-106 Current density, on ground plane, 1-276
of plastic versus ceramic packaging, Current flow, in conducting plane, 1-273
11-401-403 Current rating, of connectors, 111-40 1
single-chip packaging, 1-96, 11-96, III-96 Custom design, and wireability, 1-164-65
of socketing, III-430 Customer expectations, 1-6, II-6, III-6
thick-film packaging and, 11-649-50 Cuvar, and coated-metal packaging, III-362,
thin-film packaging and, 11-649-50, III-363
11-711-29 Cyanate ester-glass, as insulation, III-292,
Coupled-line-length rule, 1-257 III-294-95
Coupled noise Cycles per instruction, 1-46, II-46, III-47, 1-47,
board characteristic impedance and, 1-299 II-47, III-47
cables and, III-464 Cycle time, 1-46, 11-46, 111-46
connectors and, III-403 Cyclotene, II-536, II-539, 11-542, 11-543,
definition of, 1-204 11-544, 11-545-48, 11-593-94
electical design and, 1-236, 1-250, 1-252-56
parallel signal lines, III-283-84
switching noise and, 1-256-58 D
thin-film structures and, 1-268-70
Coupled transmission lines, 1-252-53 Dam-barless technology, 11-493
Couplers, and insulation, III-297-98 David Sarnoff Research Center, 11-365-66,
Coupling agents, and epoxy resins, 11-410--15 III-343
Coupling efficiency, of optoelectronic emitters, Deadhesion, of lead frames, 11-484-85
III-493-94 Decoupling capacitance and capacitors, 1-232,
Covalently bonded ester-type PSPI, 11558-73, 1-239-40, 1-270
11-611-12 Defect detection
Cracking potential, of plastic packaging, ceramic substrates and, 11-823
11-441-42,11-485-86 thin-film packaging and, 11-716, 11-719-20
Cratering, of wire bonds, 11-481-83 Deformation, and solder microstructure,
Cray-2 supercomputer, 1-389 III-151-58
Creep strain, and fatigue life, III-141, III-142 Dehaine, Gerard, 11-217
Crescent bond, 11-214 Delamination, of lead frames, 11-484-85
INDEX 11-1005

Delay, and connectors, III-404. See also package density and, II-643
Propagation delay polymers, II-799
Delay adder, transmission-line, 1-298-99 thin-film structures and, II-663-70
Delay equation, generation of, 1-229-30 Die metallization, and corrosion, II-4 78-79
Delidding, of welded packages, II-905 Differential scanning calorimetry
Dendrites, 1-442-43 degradation temperature and, 1-457
Densification, of glass-ceramic/copper molding compounds and, II-448
substrate, II-353 Differential thermal analysis (DTA), II-910
Density, of connections and connectors, Diffractive lenses, III-504-506
II-269-70, III-416-18 Digital Equipment Corporation (DEC), II-218
Descum process, for photosensitive ceramic packaging and, II-381
benzocyclobutene, II-605 VAX 9000 multichip package, 1-384-85,
Design, of microelectronics packaging. See II-757-68
also Electrical design Digital noise, 1-249
coated-metal packaging and, III-378-79, Digital signal transmission nets, 1-220-32
III-380 Diglycidyl ethers of bisphenol A (DGEBA),
of flip-chip solder-bump connections, II-418
II-146-48, II-173-74 Dimensional control, of glass-ceramic/copper
package wiring and logic changes during, substrate, II-353-55
1-159 Dimensionless numbers, and heat-transfer
plastic packages and, II-494-95 calculations, 1-342-44
quality function deployment (QFD) and, Diode-pumped fiber amplifiers, III-498
1-565-67, 1-578-84· Diode terminator, 1-231-32
statistical approaches to thermal, 1-336-39 Direct-access storage devices (DASD),
as technology driver, 1-24-29, ll-24-29, connections between central electronic
III-24-29 complexes and, III-394
thermal mismatch and, II-173-74 Direct bandgap semiconductors, for
wireability and, 1-164-65, 1-194-95 optoelectronic devices, III-490-92
Design of experiments (DOE), 1-567-78 Direct chip attach carriers (DCA)
Design limits, 1-228 definition of, III-246
Dessicants, and moisture-barrier bags, II-468 PWB with underfill, II-178-79
Detectors, optical, III-494-97 structures and processes, 111-263-65
Deutsche Aerospace, 11-788 Direct current (DC) potentials, 1-432
Development, and package manufacture, Direct current (DC) power supply, III-289-90
1-556-57 Direct current (DC) resistance, of signal lines,
Development phase, in product qualification, III-284, III-286
1-587-89 Discontinuities, in transmission path,
Dew point, measurements of, II-918-19 III-407-408
Diamine, and polymides, II-516-18 Discretely loaded and distributed nets,
Dianhydride, and polymides, II-516-18 1-224-27
DICY-cured BPA-based epoxies, III-292-93 Display capability, and design of PCs, 1-44,
Die bonds II-44, III-44
materials for, II-207-208 Distributed Bragg reflector (DBR) laser,
moisture and, II-889-90 III-511
plastic packaging and, II-500-502 Distributed feedback (DFB) lasers, III-511
thermal stress in, 1-462, 1-483-87 Distributional analysis, of thermal stress,
Die cracking, 1-484-86 1-465-66
Die fracture, in plastic packaging, II-476-77 Documentation, for ISO standard, 1-611-12
Dielectrics Documented process, 1-594-96
breakdown of, II-708 Dogbone card pad, III-211
cable materials and, III-460-61 Donath, W. E., 1-135
chemical/wet processing of, II-693-96 Double heterostructure P-I-N device, III-495,
dielectric constant, 1-220 III-496
11-1006 INDEX

Double-row solder bump TAB, 11-264 Electrical overstress (EOS), and plastic
Double-sided board process, III-311 packaging, 11-486-87
Dow Chemical, 11-536 Electrical parameters, of chip-to-package
DriClad epoxy-glass, as insulation, III-292 interconnections, 11-268-69
Drilling, of holes Electrical performance. See also Performance
laser, III-315-16 of multichip packages, 1-102-103,
mechanical, III-313-15 11-102-103, III-102-103
Drill smear, III-315 of thin-film packages, 11-641-45
Driver output impedance, 1-215-16 Electrical properties
Driver slew rate, 1-214-15 of connectors, III-399-409
Dry etching, 11-547 of high temperature non-polyimide
Dry-film photoresist, III-307-10 dielectrics, 11-538, 11-539
Dry packing, of moisture-sensitive PEMs, of metal coatings, III-372-74
11-462-63,11-470-73 of molding compounds, 11-449-51
Dual-level metal (DLM), 11-702 of photo-BCB, 11-594
Dual-in-line packaging (DIPS), 1-64, 11-64, Electrical reliability, of package-to-board
III-64, 11-292-94 interconnections, III-180-81
Ductility, of copper plating, III-319, III-320 Electrochemical failure
DuPont Chemical, 11-313, 11-382, 11-519, processes of, 1-436-39
11-533, 11-555 statistical distribution of rate of, 1-453
DYCOstrate thin-film technology, 11-777-80, Electrochemical planarization, 11-654
11-781, III-266 Electrocorrosion, 11-892
Dynamic loads, 1-508 Electrodeposited (ED) foil, 11-224
Dynamic random-access memories (DRAMs), Electrodeposition (ED) method, of photoresist,
requirements for packaging of, 1-18, 11-18, III-308-309, III-31O, III-311
III-18, 1-40-42, 11-40-42, III-40-42 Electroless copper plating, 11-804-808,
Dynamic stress, of polymers, 11-528 III-319-20
Electroless seeding, 11-774
Electrolytes, and film formation, 1-446-50
E Electrolytic conduction, 11-885-86
Edge-emitting lasers, III-493-94 Electromagnetic compatibility (EMC), and
Efficiency, of multichip packages, 1-101-102, cable, III-466-69
11-101-102, III-101-02 Electromagnetic interference (EMI)
E-field coupling, and cable signal theory, cables and, III-466-69
III-451, III-452 cooling systems and, 1-380
Eight subpackage wiring board, wiring results Electromagnetic modeling, and connectors,
for, 1-179 III-409
Electrical design. See also Design Electromigration
basic concepts, 1-199-205 fine-pitch lead interspaces, III-181
CBGAlCCGA packages, III-196-99 thin-film technology, 11-708
circuit attributes and, 1-205-16 Electromotive force series, 1-438
design space, 1-298-302 Electron beam testing, 11-833-37, 11-904
low-loss transmission systems, 1-258-70 Electronic Industries Association (EIA),
noise containment, 1-235-38 committee on standards, III -445
package improvements, 1-306-308 Electroplated bumps, 11-152-54
package modeling, 1-270-97 Electroplating, and contact reliability, III-443.
power distribution, 1-232-35 See also Additive electroplating;
signal distribution, 1-216-32 Plating
signal lines and interconnections, III-279 Electrostatic discharge (ESD), 1-231, 1-432,
system impact, 1-302-305 1-433, 11-236-37, 11-464-65, 11-486-87
Electrically long mainline, 1-223 tape automated bonding devices and
Electrically short cluster, 1-223 protection from, 11-236-37
Electrical module test, 11-837-40 Electrostatic spray, 11-744
INDEX 11-1007

Emitter coupled logic (ECL), 1-15, 11-15, ETA Systems, Inc., 11-263-64
ill-15, 1-206, 1-231 Ethylene glycol (EOME), and epoxy-glass
Emitters, and optoelectronics, ill-492-94 insulation, ill-293
Encapsulation EurocardIDIN, ill-398
CBOAICCOA packages, ill-193-94 European Economic Community (EEC), 1-608
failure mechanisms, 11-884-91 Evaluation
future developments in, 11-930--31 of inner lead bonding, 11-250
of hermetic versus nonhermetic packages, of thermal fatigue, 1-460--83
11-880--84 of wirebonding, 11-213-16
introduction to, 11-877-80 Excimer laser, 11-687, ill-316
material requirements, 11-892-94 Expanded PTFE (OORE-TEX), ill-461
plastic packaging and, 11-485, 11-490--94 Expansion cards, for PC configuration, 1-43,
principles of, 1-123-25, 11-123-25, ill-23-25 11-43, ill-43
recent advances in, 11-925-30 Expected automatic utilization, of total wiring
reliability testing, 11-920--25 capacity, 1-186
testing of hermetic packages and, 11-914-20 Extended surfaces, and heat transfer, 1-331-32
of types of hermetic packages, 11-894-914 External thermal resistance, 1-317-18, 1-373
End of system life (EOL), 1-408 Extra-fine pitch (XFP), ill-134
Energy methods, of fatigue calculation, Extrusion coating, and polymers, 11-680--81
1-477-78
Environment
plastic packaging ancd considerations of,
F
11-473-74 Fabric epoxy resin based PCB, ill-296
portable electronic equipment and operating, FACOM 790, ill-261, ill-262
1-44-45, 11-44-45, ill-44-45 Failure, definition of, 1-407
reliability and, 1-432-57 Failure mechanisms
Epoxy ceramic and plastic packaging compared,
die attach material and moisture, 11-889-90 11-489-90
failure mechanisms in, 1-456 moisture-induced, 11-884-91
insulators for coated-metal packaging, plastic packaging and classification of,
ill-364-65 11-474-89
moisture and hardening of, 11-449 solder microstructure, ill-151-58
organic dielectrics and, 11-664 Failure rate
photosensitive, 11-606-608 components of "bathtub" curve, 1-408,
plastic packaging and, 11-408-20, 11-490--94 1-409, 1-415-16
silicone-modified, 11-927 cumulative distribution of, 1-408, 1-410
wirebonding and curing of, 11-208 electrochemical processes and, 1-453
Epoxy-filled ceramic, 11-368, 11-369, 11-370 evolution of reductions in, 1-35-36,
Epoxy-glass, as insulation, ill-290--94 11-35-36, ill-35-36, 1-37, 11-37, ill-37
Epoxy-glass printed circuit boards, ill-252, ffiM logic-circuit reliability and, 1-405
ill-253 polymers and, 1-456-57
Epoxy-molded compounds, 1-90, 11-90, ill-9O prediction of, 1-450--52
Equipment, calibration and monitoring of, time to fail vs. reciprocal temperature, 1-421
1-600--601 Weibull distribution, 1-412-15, 1-416, 1-417
Equivalent circuit Fanout, and optical interconnects, ill-507
for distributed net, 1-227 Fans, selection of, 1-376-78
simplified for SCM, 1-243 Faraday's law, 1-443, 1-445, 1-452
Erodible etch masks, 11-692-93 Far-end coupled noise, 1-254, 1-269
Error signals, failure to function, 1-36, 11-36, Fast wave, of low-loss waveform, 1-261-64
ill-36 Fatigue
Ester type photosensitive polyimides, 11-555, encapsulant fracture and, 11-485
11-559-60, 11-565, 11-567, 11-593, SMT interconnections, ill-134-45
11-611-12 theories of, 1-471-81
11-1008 INDEX

Feasibility phase, of product qualification, plastic PGA technology, 1-90, 11-90, III-90
1-586-87 single-chip packages, 1-78-82, 11-78-82,
Ferranti Computer Systems (U. K.), III-342 III-78-82, 1-86-88, 11-86-88, III-86-88
Few-chip module (FCM), 11-740 Fixed vias, 1-180, 1-182-85
Fiberglass, insulating properties of, III-295 Flame retardants, and plastic packaging,
Fiber-optic connectors, III-441-42 11-418,11-419
Field-effect transistor (FET), 1-433, III-509 Flammability, of plastic packages, 11-451
Field-effect-transistor-self-elect rooptic effect Flex-film connector, III-396
devices (FET-SEED), III-509 Flexibilizers, and plastic packaging, 11-416,
Field performance, and product qualification, 11-418
1-589-90 Flexible circuits
Fillers carriers, 1-69, 11-69, III-69, 1-116, 11-116,
collision of with wire bonds, 11-481 III-116, III-246
epoxy resins and, 11-410-15 multilayer printed-circuit board processes,
Film resistance, 1-424, 1-430, III-400 III-268-72
Fine leak testing, 11-916 Flexible-film extension, of printed-circuit
Fine-line technology, and ceramic packaging, board, III-420
11-382 Flexible leads, 1-461-62
Fine-pitch ball bonding, 11-201 Flex-line system, 11-210
Fine-pitch technology Flexural strength, of plastic packaging,
connectors for, III-440 11-440-41
fine-pitch quad flat packages (FPQFP), Flexure tests, for silicon multilayer films,
1-108-109,11-108-109, III-108-109, 11-755
11-494, III-135 Flip chip
package-to-board interconnections, advances in bumping, 1-79, 11-79, III-79
III-133-34, III-158-63 chip-level interconnections, 1-76-78,
Finishing 11-76-78, III-76-78
of glass-ceramic/copper substrate, 11-355-56 gold bump bonding for, 11-266-68
of lead material in fine-pitch technology, solder joining and thermal fatigue,
III-159-62 1-487- 94
Finite element analysis Flip-chip decoupling capacitor, 11-143
for heat-transfer equations, 1-333 Flip-chip encapsulation, capless and capped
modeling for thermal stress, 1-462-65 approaches, III-194
Fins, and heat transfer, 1-331-32 Flip-chip solder-bump connections (C4)
Fire-retardant properties, of metal-core cards, acronyms for, II-136-37
III-378 assembly/rework of, 11-154-70
First-incident switching, 1-203, 1-221 design of, 11-146-48
First-level packages future trends in, II-185
ball-grid array packaging, 1-93, 11-93, III-93, history of, 11-137-44
1-94, 11-94, III-94, 1-95-96, 11-95-96, materials for, II-144-46
III-95-96 processes and tools, 11-148-54
ceramic packaging, 1-91-92, 11-91-92, Flip-chip solder joint, fatigue life of, 11-336
III-91-92 Floating capacitors, 11-788
chip-scale packages (CSP), 1-82-85, Floor planning, 1-165
11-82-85, III-82-85 Flowing Mixed Gas test (FMG), 11-929
definition of, 1-12, 11-12, III-12 Fluid-bed method, for protective coatings,
interconnection technology considerations, III-37 I
III-276-78 Fluid flow, and heat-transfer degradation,
lead-frame fabrication, 1-90-91, 11-90-91, 1-527-29
III-90-91 Fluids, PrandtI numbers for selected, 1-345
multi-chip packaging, 1-99-106, 11-99-106, Fluorocarbon liquids
III-99-106 gross leak testing and, 11-915
plastic packages, 1-88-89, 11-88-89, thermal properties of, 1-351
III-88-89, 11-465-67 Fluoropolymers, 11-664, III-289
INDEX 11-1009

Fluxless flip-chip joined MEMS device, G


11-169
Fluxless plasma-assisted dry soldering (PADS) GaAs chips, and flexible circuit, 11-266, 11-267
processing, 11-161-65 Gallium arsenide, thermal expansion of, 11-341
Fluxless pretreatment PADS tool, 11-162 Gang bonding, and process flow, 11-242
Focused ion beam (FIB), and thin-film Gang-pot molds, 11-454
packaging, 11-721-22 Gap conductance, 1-340
Foil gauges, 1-466 Gardner-Denver Company, 1-132
Folded-path light slab, ill-500-501 Gases, thermal conductivity of, 1-325
Footprint Gasket seals, 11-911-14
design of PCs, 1-43, 11-43, ill-43 Gate-array chips, and package wiring,
surface-mount dimensions, ill-173 1-159-61
Forced-air-cooled arrays of vertical PWBs, Gate count, versus track requirement for
1-351-52 package wiring, 1-156
Forced convection, 1-342, 1-349-59, 1-380 Gel time, of molding compounds, 11-446
Forward cross-talk, ill-408 General Dynamics, 11-302
Four-level wiring, 11-133, 11-134 General Electric
Four-pin net, varieties of, 1-150 polymide flexible circuit, ill-271
Fracture mechanics, 1-478-81 tape automated bonding, 11-217
Fracture test, for plastic packages, 11-442 thin-film technology and, 11-776-77
Fracture toughness, 1-480 GenRad tester, with LIF connector, ill-397
Free-abrasive lapping process, 11-357 Georgia Institute of Technology, 11-869
Free convection, 1-342 Giga-scale integration (GSI), 1-19, 11-19, ill-19
Free-convection air-cooled vertical parallel Glass, and package sealing, 1-125, 11-125,
plates, 1-345-46 ill-I 25
Free-space transmission, ill-499-501 Glass adhesives, 11-208
Frequency Glass-ceramic on metal (LTCC-M), 11-365-66,
impedance for power distribution system 11-375
and,1-234 Glass-ceramic packaging
package modeling and, 1-283-86 adhesion of polyirnide to, 11-525-26
power-supply impedance at high, substrates for, 11-335-69
1-241-44 Glass fabrics, insulating characteristics of,
Frequency division multiplexing (FDM), ill-295-96, ill-297
ill-511 Glass sealing, 11-905-11
Fretting corrosion, in tin-based contacts, Glass-transition temperature, and thermal
ill-41 1-12 expansion, 11-439-40
FR4 dielectric base, ill-276 Glass yams, as insulation, ill-296-97
Fuji-kura, Ltd., ill-342 Global tracks, 1-177
Fujimetex substrate, ill-373 Gold
Fujitsu bonding wire, 11-205, 11-207
glass-ceramic substrate and, 11-358-62, bump bonding for flip chips and, 11-266-68
11-381 as contact-interface material for connectors,
laminate technology, ill-265 ill-411, ill-414
liquid-cooled bellows, 1-388-89 lubrication of contacts, ill-415
VP 2000 module with surface-mount AIN metallization with glass-ceramic,
PGA,II-330 11-338-39
Full-wave rectifier, transfer molding of, 11-395, TAB wafer bumping, 11-239-40
11-396 thermosonic welding and, 11-193-95
Fully additive plating process, ill-366 Goldthwaite plot, 1-420
Functional photosensitive polymers, GORE-TEX, expanded PTFE, ill-461
11-567-68 Graded viscosity seal, 11-911, 11-912
Functional tests, 11-859-69, 11-871 Grain-boundary sliding (GBS), and eutectic
Furnace sealing, 11-910-11 solder, ill-156-58, ill-159
Fused metal seals, 11-898 Graphite, as lubricant, ill-415
11-1010 INDEX

Graphite-core composites, and coated-metal Hermetic packages


packaging, III-362 ceramic packaging and, 11-288-90
Grashof number, 1-342, 1-343-44 DIPS and, ll-293-94
Greensheet, ceramic, ll-311-12, ll-318, ll-320, package environment and, 1-434
ll-324, ll-339, ll-344-45, ll-347, ll-349 plastic packaging and, 11-406
Griffith cracks, and ceramic substrates, ll-372 sealing of, 1-124-25, ll-124-25, III-124-25,
Gross leak testing, ll-915-16 11-880-84
Ground bonds, and wirebonding, ll-203 testing of, 11-914-20
Ground plane types of, 11-894-914
coated-metal packaging and, III-343-44 Hertz dot, III-424
current density pattern on, 1-276 Heterostructure-Bipolar Transistors (HBTs),
Ground plane TAB tape (GTAB), ll-234 III-509, III-510
Ground-return DC resistance, III-464 Heterostructure Field-Effect Transistors
G-1O dielectric base, III-276 (HFETs), III-509, III-510
Hewlett-Packard Corporation, 11-218,
11-777-80, III-266
H H-field coupling, and cable signal theory,
III-451, III-452
Halogens, as flame retardants, ll-418, ll-419 Hierarchical printed circuits, description of,
Hamada, Hiroshi, 1-562 III-255
Handling, of plastic packages, ll-460-74 High-altitude large optics (HALO), ll-422
Hardening, of molding compounds, ll-446, High Density Electronics Center (University of
ll-448-49 Arkansas), 11-788
Hard glass packages, ll-896-97 High density signal carrier (HOSC), 11-758-61
Hartman sensor, III-497 High-dielectric-constant ceramics, 11-384
Heat-affected zone, and wirebonding, ll-206 High-Electron -Mobility Field-Effect Transistor
Heat-bondable anisotropically adhesive films, (HEMT), III-509, III-510
III-229 High-end applications, and cable signal theory,
Heat dissipation III-469-72
coated-metal substrates and, III-343 High-energy laser optics (HELO), 11-422
with C4, ll-183-85 Higher-level packages, wireability of, 1-165-91
Heat equation, 1-333 High-flow velocity, and wire sweep, 11-480
Heat flux, 1-315 High-frequency package, design of, 1-272-83
Heat removal. See also Heat transfer High insertion force (HIF) connectors, IlI-416
package cooling and, 1-119-23, ll-119-23, High-I/O TAB, 11-230
III-119-23 High-level noise tolerance (NTH), 1-206
as technology driver, 1-33-35, ll-33-35, High loss, 1-260
III-33-35 Highly Accelerated Stress Tests (HAST),
thin-film technology and, ll-639-41 1-452, 11-880, 11-920-25, 11-928-29
Heat sink, III-343 High-performance carrier (HPC), 11-778, 11-781
Heat transfer High-performance interposer, IlI-418
basic concepts, 1-314-22 High-performance packaging, 11-761-67
cooling system designs, 1-378-92 High-performance printed-circuit boards
experimental techniques and instromentation, cut-away of, III-287
1-369-78 pattern methods used for, IlI-312
fundamentals of, 1-322-69 High-performance processors, as technology
nomenclature, 1-397-98 drivers, 1-45-48, 11-45-48, IlI-45-48
plastic packages and, 11-499-500 High-pin-count packages, 1-98, 11-98, IlI-98
recent developments in, 1-392-96 High-precision, high-speed cables, III-476
reliability and degradation of, 1-524-30 High-reliability printed-circuit laminates,
research on, 1-396-97 IlI-302-303
Helium leak testing, 11-890-91, ll-916 High-speed edge-emitting lasers, III-493
Hermeticity Testing for Integrated Circuits High temperature non-polyimide dielectrics,
(National Bureau of Standards), ll-915 11-536-51
INDEX 11-1011

Hinged connector, III-393 thin-film packaging and, 11-651, 11-684,


Hitachi 11-713,11-730-40,11-773-74,11-778,
ceramic packaging and, 11-381, 11-383 11-781-84
M-880 processor series, 11-330, 11-332 I-line photoimageable ester-type PSPI, 11-572
photosensitive polymers, 11-552 Immersion cooling, 1-360
plastic packaging and, 11-466, 11-473 Impedance. See also Characteristic impedance;
silicon-carbide RAM module, 1-382 Transfer impedance
thin-film packages, 11-651, 11-711, 11-746-49 coaxial cables and, III-456
Hitt Plate, III-354 design of low-impedance power-distribution
Holm, Ragnar, III-387 system, 1-272-83
Holm conferences, III-387 driver output, 1-215-16
Holographic optical elements (HOEs), frequency for power distribution system and,
III-504-505 1-234
Homogeneous transmission structure, 1-221, high frequency and power-supply, 1-241-44
1-255, 1-256 signal line, III-280--83
Honeywell Corporation, 11-264 Impregnation, processes for multilayer printed-
Horizontal shipping magazines, 11-466 circuits, III-301-302
Hot bar outer lead bonding, 11-252-55 Improved price per performance ($IMIPS),
Hot-gas dress tool, 11-167 1-37, 11-37, III-37
Hot-gas jet refiow soldering, III-368 Indirect gap semiconductors, III-491-92
Hot plugging, of connectors, III-402-403 Inductance
Humidity-indicator card (HIC), and moisture- definition of, 1-272
barrier bags, II-469 high-inductance SCM, 1-277
Hybrid connectors, III-419-20 lead frames and, 1-285
Hybrid integration, III-511-18 Inductors, and integrated passives in thin-film
Hybrid thin- and thick-film package, 11-368 packages, II-790--92
Hydrogen overvoltage, 1-442 Inert fiexibilizers, 11-418
Hypertronics Y series of connectors, III-416, Infrared moving-belt refiow soldering, III-368
III-417 Infrared radiation sealing, 11-911
Infrared transmission, as alternative to cables,
III-476
I Inhomogeneous transmission, 1-221
ffilS Associates, 11-713 Injection-molded cards (IMC), 1-116, 11-116,
ffiM III-116, III-273, III-275
alumina packaging, 11-309 Injection-molded boards, III-275-76, III-285
ceramic packaging, 11-287, 11-289, 11-308, Inkjet printers, 11-265
11-381,11-383 Inner lead bonding (ILB), and tape automated
chip-level interconnections, 11-13 2 bonding, 11-217, 11-220, 11-225, 11-240--50
4381 air-cooled module, 1-383 Inorganic board, as second-level package,
gasket seal, 11-913-14 1-116,11-116, III-116
glass-ceramic substrate, 11-340-58 Input/output connections, in thin-film modules,
Journal of Research and Development, 11-638-39
III-320 Insertion force, and connectors, III-416
packaging development laboratory, 1-565--67 Inspection
package electrical testing, 11-829, 11-837, of copper plating, III-318
11-842-50 of screened greensheets, 11-350
photosensitive epoxy mixture, 11-606, 11-608 Instantaneous failure, as reliability parameter,
polyimide polyester development for thin- 1-417
film applications, 11-518, 11-520--21 Institute for Interconnecting and Packaging
polymers, 11-511, 11-520 Electronic Circuits (IPC), III-133-34,
surface laminar circuits (SLC), III-266, III-445
III-267, III-268 Insulation, materials for, III-290-98
tape automated bonding, 11-217 Insulation displacement connectors (IDC),
thermal-conduction module, 1-385-86 1-430, III-394, III-395
11-1012 INDEX

Insulation resistance solder paste, ill-168-78


of connectors, ill-402 solder and solder joining, ill-151-58
test, ill-180 surface-mount array connections,
Insulator materials, properties of, 1-85, ll-85, ill-187-217
III-85 surface-mounted packages, ill-132-51
Insulator-metal substrate technology (IMST), Interdigitated dew-point sensor, ll-919
ill-350-54, ill-355, III-356 Interface metallurgy, 1-492
Insulators, and coated-metal packaging, Interference. See Electromagnetic interference
ill-362-65 Interlayer applications, of polyimides,
Integral tabs, on flexible-film extension of 11-610-14
printed-circuit board, III-420 Intermetallic growth, 11-464
Integrated ceramic packaging with passives, Internal audits, 1-605-606
ll-384-87 Internal plated holes, III-289
Integrated-circuit packaging, 1-135 Internal thermal resistance, 1-317-18, 1-370-73
Integrated passives, and thin-film technology, International Electrical and Electronics
11-784-92 Engineers (IEEE), Computer Society
Integrated rigid-flex thin-film package, ll-778, Technical Committee on Packaging,
11-781 ill-444
Integrated Teflon Flex, III-270 International Organization for Standards (ISO),
Intel Corporation, 11-265. See also Pentium 1-608
Intelligent vehicle highway systems (IVHS), Interposer connectors, ill-420, III-431
1-62, 11-62, III-62 Interposer MCM sockets, III-432
Interchip signal interconnections, 1-25, ll-25, Intrinsically conductive polymers, III-233
ill-25 Intrinsically photodefinable epoxy, 11-608
Interconnection nets, design of, 1-203 lon-getter additives, 11-420
Interconnections, chip-to-package. See also Ionic bonded-type PSPls, 11-573-78, 11-612-13
Connectors ISO 9000, as worldwide quality standard,
adhesive bonding and, II-266-68 1-608-14
anisotropic conductive adhesives, II-268 ISO 9001, elements of, 1-617
basic concepts, 11-129-32 ISOPAK pin-grid array, 11-302
CBGA and CCGA packages and, ill-192-93 Isotropic conductive adhesives, ill-225-28
density of connections, 11-269-70
electrical paramiters of, II-268-69
evolution of, II-132-36 J
flip-chip solder-bump connections,
11-136-85 Japan
MCM level-2, III-426-37 ceramic packaging and, 11-285, 11-286,
pressure connects, 11-265-66 11-307
tape automated bonding (TAB), 11-217-65 consumer product component density trend
wirebonding,II-186-209 in, ll-385
Interconnections, package-to-board. See also mass production of low-cost electronic
Connectors products, 11-385
basic concepts, ill-129-32 molding compound development in, 1-89,
conductive adhesives, 111-223-33 11-89, 111-89
fine-pitch technology, ill-158-63 packaging production by type, 1-87, 11-87,
future developments in, ill-233-34 III-87
introduction to, 1-106-14, 11-106-14, quality and manufacturing in, 1-558
ill-106-14 soldering technology trends in, 1-114, 11-114,
lead-free solders, III-217-23 111-114
multiwire technology, 111-276 standards for plastic packaging in, 11-499
printed-wiring boards and technology tape automated bonding and, 11-218, 11-272
considerations, III-276-90 thin-film technology and, 11-772-73
reliability, III-178-87 Joint conductance, 1-339
INDEX 11-1013

Joint geometry thin-film packaging and ablation, 11-686-90,


fatigue performance and, 1-491-92 11-801-803
thermal mismatch and, 11-173-74 welding, 11-904-905
Joint Test Action Group (ITAG), 11-860-61 writing for thin films, 11-725
Just-in-time inventory control (JIT), 1-590 Latent open testing, 11-837, 11-839, 11-840,
II-846-47
Late packing, and wire sweep, 11-481
K Layer factor, and wireability of BGAs, 1-188
LCD industry, and tape automated bonding,
Kanegafuchi Chemical Industry Co., Ltd.,
11-265
11-533, 11-534-35, 11-614
Lead borosilicate-alumina glass & ceramic,
Kapton polyimide films, 11-533-34, 11-540-41
11-362-63
Killer defects, elimination of causes of, 1-37,
Leaded chip carriers, thermal failure of,
11-37, III-37
1-500-504
Kneading, of molding compounds, 11-493
Lead-on-chip technology (LOC), 11-567
Knoop hardness test, 1-428
Leaded multichip module connector (LMMC),
Known good die (KGD), pretested chips,
sockets for peripheral MCMs, III-436-37
1-39-40, 11-39-40, III-39-40
Lead frame
Knudsen Effusion-Mass Spectrometry, 11-320
Alloy 42 and, 1-284-86
Kovar
corrosion of, II-483-84
coated-metal packaging, III-362, III-363
deadhesion and delamination, 11-484-85
preformed package, 11-395
fabrication of, 1-90-91, 11-90-91, III-90-91
seal frame, 11-896, 11-897
plastic packaging and materials for,
Kyocera Corp., 11-307
11-407-37
Lead-free solders, and package-to-board
L interconnections, ill-217-23
Leadless chip carriers
Laminated ceramic package, 11-294 ceramic, 11-304-305
Laminated multilayer ceramic, 11-319 fatigue behavior, 1-498-500
Laminate technology thermal resistance model of, III -136
glass-ceramic packaging, 11-348, 11-350-51 Leadless MCM substrates, sockets for,
low-dielectric-constant materials, III-322, III-431-37
I1I-324 Leads, See also Flexible leads; Lead frame
low-thermal-expansion, III-327 damage to in handling, 11-463-64
processes for, III-301, III-302-303 fatigue and, 1-503-504
thin film on PCB, III-265-68 fine-pitch technology and materials and
LAMPAC, III-342-43 finish for, III-159-62
Langmuir-Blodgett (LB) membrane, 11-614, inductance of, 1-31-33, 11-31-33, III-31-33
11-616 Lead-zinc-borate glasses, II-908
Large-area polymer deposition, 11-798, Leak testing, of hermetic packages, 11-915-16
11-800-801 Leverages, and multichip packaging, 1-104,
Large-area processing (LAP) technology, cost 11-104, III- 104
models of, 11-711-13 Lift-off technology, 11-706-707
Lasers Light-emitting diodes, III-492, ill-493-94
bonding, 11-244-46 Linear laminate, as insulation, III -296,
Cyc\otene and ablation, 11-547-48 III-298-301
drilling, III-315-16 Line characteristic impedance, 1-298-99
laser chemical vapor deposition (LCVD), Line discontinuities, 11-745
11-722-27 Line modeling and simulation, 1-590-92
optoelectronic devices and, III-493-94, Liquid-cooled modules (LCM), III-262
III-510-11 Liquid-crystal composite laminates, III-324-26
retlow systems and fine-pitch components, Liquid-crystal polymers (LCP), II-493-94,
III-175 III-329
11-1014 INDEX

Liquid crystals, as optical modulators, III-498 M


Liquid immersion cooling, 11-184
Liquid photoresist, III-305-307 Machining, of glass-ceramic/copper substrate,
Liquids, thermal conductivity of, 1-325. See 11-355-56
also Moisture Macros, on chips and package wiring,
Lithographic properties, of ionic bonded-type 1-161-62
PSPls, 11-575-78 Magnetic energy, 1-274
Local-stress soft error, 11-493 Mainframes, multilayer printed-circuit boards
Logic interconnections, topology and for, III-261-62
constraints on, 1-149-52 Management
Logic LSI package, 1-92, 11-92, III-92 ISO 9000 registration and, 1-610-11
Logic wiring images, wireability of, 1-155 process control and, 1-606-607
Lognormality, 1-481-83 Manhattan distance, 1-140, 1-179, 1-180
Lorenz constant, 1-427 Manhattan wires, wire length distribution for,
Lossless, definition of, 1-258-59 1-169
Loss tangent, dielectric, 1-220 Manufacturers, of coated metal substrates,
Lossy signal lines, 1-218-20, 1-260 III-366
Low-affinity surface finishes, 11-484 Manufacturing, of packaging. See also
Low-cost bumping, 1-80, 11-80, III-80 Assembly
Low-cost, fine-line, thin-film process, 1-119, costs of and trends in, 1-19, 11-19, III-19
11-119, III-1I9 ISO 9000 registration and, 1-608-14
Low-cycle fatigue cracking, of copper plating, key concepts in, 1-556-61
III-318 lead-free solder alloys, III-219-20
Low-dielectric-constant laminate materials, line modeling and simulation, 1-590-92
III-322, III-324 manufacturability as technology driver,
Low-end applications, and cable signal theory, 1-37-38, 11-37-38, III-37-38
III-469-72 plastic packaging and, 11-495-98
Low-impedance power-distribution system, poJyimides and, 11-532-33
1-272-83 process control, 1-592-608
Low-insertion-force contact, III-416 product qualification and, 1-584-90
Low-IIO tab, 11-229 quality in product and process design,
Low-level noise tolerance (NTL), 1-206 1-561-84
Low-loop process, and wirebonding, 11-198, Markets and marketing
11-199,11-202 of optoelectronics, III-485, III-488
Low-loss transmission systems, design of, for single-chip packages, 1-86-88, 11-86-88
1-258-70 worldwide for printed-circuit boards, III-250
Low-melting solder alloys, 11-145 Material properties, of connectors, III-409-15
Low-noise differential cables, III-476 Mating electrical contacts, model of, 1-521
Low-stress molding compounds, 11-420-22 Matrix reduction, mathematical method of,
Low-temperature ceramic or glass-ceramic 1-293
packaging, 11-335-69 Matsushita Electric, 11-232, 11-233
Low-temperature firing, of aluminum nitride, Maze runner, 1-178
11-327 Mean time to fail (MTTF), 1-417
Low-thermal expansion laminates, III-327 Measurement, for ISO 9000 standard, 1-613-14
Low-thermal-expansion photosensitive Mechanical cycling, 1-471
polymers, 11-568-72 Mechanical drilling, of plated-through holes,
Low-transmission-Ioss waveguides, III-504 III-313-15
LSI technology, progress in and cost-reduction Mechanical loading, and reliability, 1-508-19
of digital systems, 1-50-58, 11-50-58, Mechanical properties
III-50-58 of connectors, III-415-26
LTCC packaging, 11-365-66, 11-375 of high temperature non-polyimide
Lubricants dielectrics, 11-537
contact-interface materials and, III-415 of photo-BCB, 11-594
contact wear processes and, 1-524 of plastic packaging, 11-440-43
INDEX 11-1015

of polymers, 11-528-30 METLAM, III-342


of Probimide 400 curved films, 11-588 Microcapsule adhesive (MCA), III-231-32
Mechanical splice, for optical fiber, III-442 Microcarrier BGA assembly, 1-77, 11-77,
Mechanical structure, of connectors, III-395, III-77
III-397-98 Microelectromechanical structure (MEMS)
Mechanical thermal contact, 1-526-27 device, 11-168, 11-170
Media delay factor, 1-58-62, 11-58-62, Microelectronic Center (North Carolina),
III-58-62 11-153, 11-154
Melt viscosity, of plastic packaging, 11-442 Microelectronics packaging. See also Wiring
Memory chips, stacked on ceramic carrier, chip-level interconnections, 1-71-78,
III-323 11-71-78, III-71-78
Memory packaging, as technology driver, first-level packages, 1-78-106, 11-78-106,
1-40-42, 11-40-42, III-40-42 III-78-106
Meniscus coating, 11-679 introduction to basic concepts, 1-3-21,
Merck Chemical, 11-555 11-3-21, III-3-21
Metal ball-grid-array packages, and package- package-to-board interconnections, 1-106-14,
to-board interconnections, III-209-1 0 11-106-14, 111-106-14
Metal-core boards, III-246 package sealing and encapsulation,
Metal-insulator combinations, for coated-metal 1-123-25,11-123-25, III-123-25
packaging, III-341 packaging cooling, 1-119-23, 11-119-23,
Metal-insulator-semiconductor (MIS) FETs, III-119-23
III-509 packaging technologies, 1-62-71, 11-62-71,
Metallization III-62-71
of aluminum nitride, 11-329 second-level packaging, 1-114-19,
coated-metal packaging and methods of, 11-114-19, III-114-19
III-365-67 technology drivers, 1-21-62, 11-21-62,
corrosion of, 11-885 III-21-62
plastic packages and deformation, 11-479 Microelectronic substrates, desired properties
thin-film technology and, 11-697-707 for, 11-531-32
Metal mask evaporation technology, Micro-interposer contact sockets, for peripheral
11-148-52 and area-array MCMs, III-432-33
Metal migration, and high temperature non- Microjet air cooling, 1-122-23, 11-122-23,
polyimide dielectrics, 11-545 III-122-23
Metal-oxide semiconductor (MOS), 11-884-85 Micro Module Systems
Metal packages, sealing of hermetic, capacitance testing and, 11-829
11-894-95 thin-film packaging and, 11-771-72
Metals and metallurgy. See also Specific Micro-Moire interferometry, of underfilled
metals chips, 11-182
adhesion of polyimide to, 11-526-27 Microstrip lines, 1-255, 1-257, 1-280
Brinnel hardness test, 1-428 Microstructure, of solder, III-151-58
gasket seals and, 11-913 Microwiring thin-film technology, 11-740-45
skin depths for various, 1-284 MICTOR connectors, III-439, III-440, III-441
as substrates for coated-metal packaging, Military systems, card packaging for,
III-348-61 III-259-60
terminal design and, 11-138 Million instructions per second (MIP), 1-44,
thermal conductivity, 1-323-24 II-44, III-44
thermal expansion of, 11-421 Miner's rule, 1-476-77
typical cofired, 11-322 Miniature z-axis actuator, 11-847
Metal-Semiconductor-Field-Effect Transistor Miniaturization, of surface-mount technology,
(MESFET), III, 509, III-510 III-288
Metal-semiconductor-metal (MSM) detectors, MiniMod process, of tape automated bonding,
III-495, III-496-97 11-217
Methylethylketone (MEK), as insulation, Mismatch stress, and ceramic substrates, 11-372
III-293 Mitsubishi Electric, 1-382-83, 11-592
11-1016 INDEX

Mixed-signal equipment Multichip packages (MCPs)


as technology driver, 1-44-45, 1I-~5, applications of, 1-106, 11-106, III-106, 1-109,
III-44-45 11-109, III-I09
wireability and, 1-193-94 cost of, 1-104-106, 11-104-106, III-I04-106
Mocute ceramic, mainframe hardware development of, 1-84, 11-84, III-84,
technology based on, 11-334 1-99-100,11-99-100, III-99-1oo
Modeling efficiency of, 1-101-102, 11-101-102,
chip circuit placement and package wiring, III-IOI-102
1-155-65 electrical performance of, 1-102-103,
electrical design and, 1-270--97 11-102-103, III-102-103
manufacturing and, 1-590-92 functions of, 1-100-101, 11-100-101,
Modularity, of connectors, III-426 III-IOO-IOI
Modulators, optical, III-497-98 reliability of, 1-103-104, 11-103-104,
Moire interferometry, 1-467-68, 469 III-103-104
Moisture single-chip packages compared to, I-52,
in CERDIP packages, 11-909 II-52, III-52
epoxy hardening process and, 11-449 substrate materials, 1-108, 11-108, III-108
failure in nonhermetic packages and, testing of, 1-39-40, 11-39-40, III-39-40
11-884-91 typical package parameters of, 1-107, 11-107,
in hermetic packages, 11-916-20 III-I07
plastic packages and, 11-445, 11-446, Multilayer ceramic (MLC) technology
11-451-52,11-461-63 alumina powder for, 11-311
Moisture-barrier bag (MBB), 11-468, 11-472-73 ball-grid arrays and, III-191
Moisture-film conductivity, 1-446-50 development and applications of,
Moisture-induced reliability, of package-to- 11-297-300,11-319
board interconnections, III-184-86 multilayer PCB technology compared to,
Molded circuit boards (MCBs), III-247, III-254
III-272-76 prediction of noise and cross-talk, 1-237-41
Molding compounds, and plastic packaging, reliability of, 11-290
11-407-37 Multilayer LTCC-M, metallization in,
Molding equipment, for transfer molding, III-365-66
11-453-54 Multilayer printed-circuits
Mold-killed steel, III-350 materials and processes, III-290--320
Mold-release agents, 11-419-20 reliability and controls, III-320--21
Molybdenum, and coated-metal packaging, structure and processes, III-251-76
III-362 Multilayer thin-film structures, types of,
Monomaterial optoelectronic integration, 11-652-59
III-508-509 Multilevel wiring
Monomers, for polymide synthesis, 11-516 chip-level interconnections and, 11-133,
Monte Carlo method, 1-293, 1-465-66 11-134
Moore's Law of Packaging, 1-66, 11-66, chip wireability and, 1-162-64
III-66 Multimaterial hybrid DEle integration, III-513
Motorola Corporation Multiple flexible Teflon-based cables, III-272
pen-sized pager, 11-181 Multiplunger molds, 11-453, 11-454, 11-455,
tape automated bonding, 11-234-35 11-456
thin-film technology and, 11-713, 11-762 Multispindle drill machines, III-313
Mullite, 11-333-35 Multiwire interconnections, III-276
Multichip modules (MCM) Murata, ceramic integrated capacitors by,
connectors for, III-426-37 11-387
definition of, III-246
introduction of dense liquid-cooled, 1-18-19,
11-18-19, III-18-19
N
testing of, 1-39-40, 11-39-40, III-39-40, National Semiconductor, 11-218
11-864-65 Natural convection, 1-342, 1-345-49, 1-380
INDEX 11-1017

Nd-YAG lasers, III-316 Normal force, of connectors, III-400-401


Near-end coupled noise, 1-254, 1-269 NOSC (Navan Ocean Systems Center)
NEC package, 11-332
future of ceramic packaging and, 11-381, NTK Electronics, 11-787
11-382 Numerical analysis, of thermal stress, 1-462-65
glass-ceramic substrates, 11-362-63, 11-368, Nusselt number, 1-344, 1-346, 1-349, 1-350,
11-378 1-353, 1-355, 1-356, 1-357, 1-358, 1-359
liquid-cooling module, 1-386-88
printed-circuit interconnections in SX-31
SX-X PCB, III-262
o
R-3000 workstation MCM, 11-606 OCD balancing, and switching noise, 1-281
thin-film packaging and, 11-768-71 Off-chip interconnection networks, 1-222
Negative-acting photoresists, III-304, Off-chip signal paths, 11-633
III-305-306, III-308 Off-contact printing, and solder deposition,
Negative ion detection (NID), 11-915-16 III-166-67
Nemst equation, 1-439-40, 1-449 Offset cube topology, III-520, III-521
Neural networks, 1-194 Ohm's law, 1-452
Next-generation electronics, attributes of, 1-15, Oki Electric Industry Co., Ltd., 11-711
11-15 Olefin dielectrics, 11-550-51
Nickel, reaction rate of with 60/40 solder, Olin Interconnect Technologies, 11-494-95,
III-147 III-209-10
Nickel-iron alloys, and lead frames, 11-426 On-chip circuits, and decoupling capacitors,
Nippon Electric Company, 11-466, 11-467 1-239-40
Nitride waveguides, III-504 On-chip narrow conductors, 1-260-61
Nitto Denko products On-chip noise, 1-236-37
application-specific materials (ASMATs), On-contact printing, and solder deposition,
11-493 III-167-68
development plan of semiconductor-related One-layer tape, for tape automated bonding,
materials, 11-492 11-222
TAB process and, 1-74-76, 11-74-76, One-part connectors, III-419
III-74-76 Opens and shorts test (OST), 11-845, 11-850-58
Noise. See also Coupled noise; Cross-talk; Open wire, III-446-47
Noise tolerance; Switching noise Operators, certification of, 1-599
board characteristic impedance and, Opposed electrode welding, 11-904, 11-905
1-299-302 Optical cables, III-476-77
connectors and, III-403-404 Optical detectors, III-494-97
electrical design and containment of, Optical fiber
1-235-38 connectors and, III-441-42
electrical design and types of, 1-204 as transmission medium, III-501-502,
thin-film structures and, 1-268-70 III-506-507
Noise penalty, 1-305, 1-306 Optical inspection, for defect detection in
Noise saturation, 1-246, 1-248, 1-277 ceramic substrates, 11-823
Noise tolerance Optical interconnects, 1-48, 11-48, III-48
composite, 1-210-11 Optical properties, of high temperature non-
coupled-line-length rule, 1-257-58 polyimide dielectrics, 11-542
self-generated, 1-209-10 Optical signal routing elements, III-504-506
transmitted, 1-206-209 Optical waveguides, 11-613, III-502-504
Noncritical processes, in manufacturing, 1-595 Optoelectronic integrated circuits (OEICs),
Nonhermetic packages III-507-18
hermetic packages compared to, 11-880-84 Optoelectronic receivers, III-509-10
moisture and failures in, 11-884-91 Optoelectronics
reliability of, 11-879 integration of with electronics, III-507-18
Nonhomogeneous structures, 1-255 as interconnection media, III-490,
Nonshrinking preimidized PSPI, 11-613 III-498-507
11-1018 INDEX

introduction to, ill-481-9O input-output datat-rate dependency on, 1-47,


systems applications of, ill-518-23 11-47, ill-47
types of devices, ill-490-98 package electrical design and, 1-302-305
Organic adhesives, and wirebonding, packaging considerations and evolution of,
11-207-208 1-48-62, 11-48-62, ill-48-62
Organic dielectrics, materials for, 11-664-70 of plastic versus ceramic packaging, 11-401
Organosilane couplers, ill-298 power distribution as technology driver,
Outer lead bonding (OLB), and tape automated 1-29-33, 11-29-33, ill-29-33
bonding, 11-221, 11-225, 11-250-58 trends in packaging and, 1-19-20, 11-19-20,
Overflows. See Wiring overflows ill-19-20
Oxidation, and chemical failure, 1-434-35 Peripheral bonding, verticaIJhorizontal stresses
Oxidative stability, of Probimide 400 causing, 11-194
polyimides, 11-586-87 Peripheral sockets, ill-430-31
Oxygen index, and plastic packages, 11-451 Personalization, and glass-ceramic packaging,
11-348
Personal Computer Memory Card Induced by
p Association (PCMCIA), 11-387
Personal computers
Package-to-board interconnections. See electronics assembly of, 1-406
Interconnections, package-to-board price threshold for, 1-271
Package configurations, and environmental as technology drivers, 1-42-44, 11-42-44,
interactions, 1-433-34 ill-42-44
Package electrical testing. See Testing Personal digital assistant (PDA), ill-250
Package weight, versus pin count, 1-97, 11-97, Personalized opens/shorts test (POST), 11-851,
ill-97 11-852
Packaging hierarchy, 1-131-32 Phase diagrams, and soldering, 11-146
Packaging Research Center (Georgia Institute Phosphor-bronze springs, 1-429-30
of Technology), 11-789-90, 11-796 Photo-cross-linking, of Probimide-related
Packing, of plastic packages, 11-460-74 polyimides, 11-584-90
Pad-array optimization, ill-423 Photodefinable epoxy mixture, 11-606, 11-608
Pad density, and flip-chip solder-bump Photoformed ceramic modules (PCM), 11-317,
connections, 11-149 11-382
Pad tilt, and molding compounds, 11-493 Photoform fully additive process, ill-367
Palladium, as contact-interface material for Photolithography
connectors, ill-411 circuitization and, ill-304, ill-305
Panasonic Corp., 11-363-64 PIMEL and process for, 11-561-63
Parallel seam welding, 11-904 Photoneece, 11-573, 11-574-78, 11-579, 11-580,
Partial inductance matrix, 1-272 11-581,11-609-10,11-611,11-612
Passivation layer, plastic packages and failure Photoselective copper-reduction process,
of,II-477-78 ill-367
Passive integration Photosensitive benzocyclobutene, 11-593-606
ceramic packaging and, 11-384-87 Photosensitive epoxy, 11-606-608
thin-film packaging and, 11-784-92 Photosensitive polymers, 11-552-73,
PCB spacing, 1-348-49 11-803-804
PCK Corporation, ill-367 Photosensitive polymides (PSPls), 11-515,
Pelletizing, of molding compounds, 11-493 11-552-73, 11-575, 11-608-14, 11-616,
Pentium, and tape automated bonding, 11-694-95, 11-702
11-260-63, 11-265 Photosensitive thin-film dielectric process,
Performance 11-671
cycle times and cycles per instruction, 1-46, Photospeed, of preimidized photosensitive
11-46, ill-46 polyimides, 11-581
definition of, 1-3-4, 11-3-4, ill-3-4 Photo via making, ill-317
design of PCs and, 1-43, 11-43, ill-43 Physical -design software, and package wiring,
heat transfer and criteria for, 1-315-16 1-136-37
INDEX 11-1019

Physically separated systems, connections quality and reliability of, 11-474-90


between, 1II-394 surface-mounted devices, 11-399
Piezoresistive gauges, 1-466 through-hole-mounted devices, 11-398-99
PIMEL, 11-555, 11-560, 11-562-65, 11-570 transfer-molding process, 11-452-60
PINBELL,II-302 Plastic PGA technology, 1-90, 11-90, 111-90,
Pin count, versus package weight, 1-97, 11-97, 1-91, 11-91, 111-91
1II-97 Plastic pin-grid array (PPGA), 11-398-99,
P-i-n detectors, 1II-495-96 11-400
Pin-grid arrays (PGA) Plastic quad fiat package (PQFP), 1-86-87,
silicon gels and, 11-501-502 1-86-87, 111-86-87, 11-399, 11-401
sockets for pinned area-array MCMs, Plated-through-holes
1II-438-39 mechanical drilling of, 111-313-15
Pin-in-hole, and thermal fatigue, 1-494-97 structures of, 1II-286
Pin-parallel composita building, 1II-311 Plating. See also Additive electroplating;
Pin-to-pin interconnections, and wiring Electroplating
demand, 1-130 of alumina packaging, 11-321-23
Pitch card edge connector, 111-390 of lead frames for plastic packaging, 11-435,
Placement 11-437
of fine-pitch components, 1II-169-71 loss of wire bond and, 11-482
wiring assignment and routing programs, Plug-in card, wireability of, 1-172
1-137 Polishing, of glass-ceramic/copper substrate,
Planarization 11-358
of glass-ceramic/copper substrate, 11-357 Polybenzoxazole (PBO) and
of high temperature non-polyimide polybenzobisthiazole (PBZT), as liquid-
dielectrics, 11-543-45 crystal materials, 111-325-26
thin-film technology and, 11-707-11, 11-804 Polyisoprene resin, 1II-306
Planars, for multilayer printed-circuit boards, Polymer Hybrid Integrated Circuit (POLY-
111-256-60 CffiC), 11-606
Planar stacked studs, 11-653-55 Polymerization, molding compounds and rate
Plaskon Electronic Materials, 11-418, 11-419 of, 11-447-48
Plasma-enhanced chemical-vapor-deposited Polymers
(PECVD) silicon nitride, 11-477 commercial preimidized polyimide films,
Plastic ball-grid-array packages, and package- 11-530-36
to-board interconnections, 111-206-209 common encapsulants for plastic packages,
Plastic dual-in-line (PDIP) package, 11-398-99, 11-409
11-400, 11-401, 11-494 die adhesives and plastic packaging,
Plastic encapsulated microelectronics (PEMs), 11-500-502
11-396-98 dielectrics for thin-film packaging, 11-515
Plastic-encapsulated packages, reliability of, failure mechanisms and, 1-456-57
1-123-24,11-123-24,111-123-24,1-434 future directions of thin-film packaging and,
Plastic optic fiber, 111-502 11-797-801
Plastic packaging historical perspective on, 11-510-12
ceramic packaging versus, 11-399, as insulation, 111-293, 111-294
11-401-406 intrinsically conductive, 1II-233
characterization of molding-compound introduction to, 11-509-10
properties for, 11-437-52 liquid crystal, 11-493-94, 111-329
development of, 1-88-89, 11-88-89, microelectronics applications of, 11-510
111-88-89 nonhermetic packaging and, 11-883-84
die adhesives and, 11-500-502 photosensitive benzocyclobutene,
future of, 11-490--99 11-593-606
heat transfer in, 1-392-93, 11-499-500 photosensitive epoxy, 11-606-608
introduction to, 11-394-98 photosensitive polymers, 11-552-73,
molding compounds and lead frame 11-608-14
materials for, 11-407-37 polymides and, 11-516-30
11-1020 INDEX

positive photosensitive polyimides, Power connections and connectors, III-404,


11-590-93 III-475-76
preimidized nonshrinking photosensitive Power cycling, 1-470
polyimides, 11-578-90 Power density
research developments and future of, definition of, 1-33, II-33, III-33
11-614-17,11-928-29 requirements, 1-34, II-34, III-34
thermal expansion of, 11-421 Power distribution
thin-film packaging and, 11-512-15, design of, 1-201, 1-203, 1-232-35, 1-272-83
11-675-97 as technology driver, 1-29-33, 11-29-33,
Polyimide-ceramic substrate, 11-768-70 III-29-33
Polyimides, and insulators for coated-metal Power supply
packaging, III-364-65 de-coupling of, III-290
Polymide-glass, as insulation, III-292, III-294 direct current (DC), III-289-90
Polymides impedance of, 1-241-44
chemistry, features, and applications of, for thin-film packages, 11-639-41
11-517 Power-time product, theory of, 1-48-50,
commercial FET chip with, 11-511 11-48-50, III-48-50
commercial preimidized films, 11-530-36 Prandtl numbers, 1-344, 1-345, 1-350-51
dielectrics for thin-film packaging, 11-515, Pre-exposure bake, of photosensitive
11-664 benzocyclobutene, 11-597-98
flexible circuits, III-269 Preimidized nonshrinking photosensitive
passivation layer and, 11-478 polyimides, 11-578-90
polymers and, 11-516-30 Preimidized polyimide films, 11-530-36
tape automated bonding and, II -225 Preplacement and pre wiring, influence on
thermal expansion of, 11-665, 11-666 wireability, 1-193
Polymide siloxanes, 11-927 Pressed ceramic (Cerdip), II-293
Polynorbomene polymers, 11-550-51 Pressure connects, and chip-to-package
Polyolefins, 11-664 interconnections, 11-265-66
Polyquinolines, 11-548-50 Pressure drop, and heat transfer, 1-376-78
Polystyrene, and shipping trays, II -466 Printed-circuit boards (PBC). See also
Polyurethane coatings, 1-455 Multilayer printed-circuits; Printed-wiring
Polyvinyl butyral (PVB), as binder, 11-312, boards
11-313,11-314,11-315 connections between, III-392-93
Polyvinyl chloride, and shipping trays, 11-466 connections between components and,
Poon, Ron K., 1-26, 11-26, III-26 III-391
Popcoming, of plastic packages, 11-485-86 definition of, III-246
Porcelain enamel technology (PET), III-340 future trends in, III-328-30
Porcelain insulators, for coated-metal integral tabs on flexible-film extension of,
packaging, III-362-64 III-420
Porous ceramics and glasses, 11-376, 11-378-80 recent developments in, III-321-28
Portable electronics and portability worldwide market for, III-250
design of PCs and, 1-43, 11-43, III-43 Printed-wiring boards. See also Printed-circuit
size reduction and, l-llO, 1I-110, III-I 10 boards
as technology drivers, 1-44-45, 11-44-45, ambient air environment and, 1-432
III-44-45 basic concepts in packaging of, III-245-51
types of, 1-7, 11-7, III-7 cable replacement and, III-477
Positive photosensitive polyimides, 11-590-93 direct chip attach (DCA) to, 11-178-79
Positive working PSPI, 11-591-93 interconnection technology considerations,
Postcure, of epoxy molding compounds, III-276-90
11-449 terminology and, III-247
Postsinter sizing, of glass-ceramic/copper trends in materials and fabrication
substrate, 11-356 technology, 11-498, III-248
Powder coating, of epoxy on steel or worldwide production totals for, III-251
aluminum, III-365 Printing machines, requirements for, III-168
INDEX 11-1021

Probability curves, for wiring success, Quarter-size small-outline package (QSSOP),


1-154-55 11-494
Probability density function (PDF), 1-411 Quasi-hermetic, metallized ceramic (MC)
Probability density table, 1-533-37 package, 1-434
Probimide, 11-555, 11-578-90, 11-613
Process control
R
in manufacturing line, 1-592-608
for plastic packages, 1-89, 11-89, III-89 Radiated noise, and connectors, 111-404
Process monitors, 1-599-600 Radiation
Product integrity, disruptions of, 1-36, 11-36, heat transfer and, 1-322, 1-368--69
III-36 in plastic packages, 11-415
Product qualification, 1-584-90 Radio-frequency (RF)
Programmable vias. See also Plated-through integrated ceramic packaging and, 11-385-86
holes mixed-signal applications environment, 1-45,
concept of, 1-175-76, 1-182-85 11-45, III-45
interconnection technology considerations printed-circuit assemblies for, III-329
and, III-289 Rail-to-rail noise, 1-282
Projection laser ablation, and thin-film Random access memory (RAM), number of
packaging, 11-684, 11-686, 11-688-90, terminals required for, 1-23, 11-23, III-23
11-802-803 Random logic, and Rent's Rule, 1-144
Propagation delay Raw-card drawings, for coated-metal
cables and velocity of, III-459--60 packaging, III-378, 111-379
connectors and, III-408-409 Rayleigh number, 1-343, 1-344, 1-347, 1-348
Protective coatings, and soldering of coated- RCA Laboratories, 11-297-300, III-342
metal substrates, III-370--72 Reactive flexibilizers, 11-416, 11-418
Pulse bonding thermodes, 11-241, 11-242 Reactive ion etching (RIE), 11-690, 11-692-93,
Pulsed thermode method, of TAB rework, III-504
III-177 Reactive liquid-phase sintering, 11-374
Punching, of glass-ceramic packaging, Receiver input stability, 1-211-12
11-348-50 Receivers, optoelectronic, III-509-10
Pyralin PO, 11-555 Recycling, of package materials, 11-474, 11-492
Pyro-deposited copper, III-319 Redistribution wires, 1-131, 1-132
Redundancy
product design and, 1-559
Q thin-film technology and, 11-727-28
Quad flat package (QFP) Redundant connectors, III-422-23
development of, 1-64, 11-64, III-64, 1-95-99, Reflected noise, and connectors, 111-403-404
11-95-99, III-95-99 Reflection
sockets for peripheral MCMs, III-437-38 cables and coefficient for, III-456, III-458
wireability of, 1-187-91 connectors and, III-407-408
Quality. See also Reliability signal degradation and, 1-27-29, 11-27-29,
inner lead bonding and, 11-246-49 III-27-29, 1-228, 1-236
international standards of and ISO 9000, Reflow, and fine-pitch components, III-174-76
1-608-14 Refractive lenses, 111-504, III-505
key concepts, 1-558--61 Registrars, for ISO 9000 standards, 1-609,
of plastic packaging, 11-474-90 1-614
in product and process design, 1-561-84 Registration Audit, for ISO 9000, 1-609
Quality function deployment (QFD), 1-563-67, Regular package image, 1-137-38
1-578-84 Reinforcement
Quality leverage properties of materials for, III-327
concept of, 1-562 of substrates, 11-743-44
as technology driver, 1-37-38, 1I-37-38, Relative humidity, and moisture-related
111-37-38 damage, 1-432, 1-434, 1-435, 1-446, 1-450,
Quality manual, for ISO 9000, 1-611 1-451,11-462
11-1022 INDEX

Relative motion, 1-520 Resistance


Releasing agents, and delamination of plastic of connectors, 111-399-400
packages, 11-484-85 DC resistance, 111-464
Reliability. See also Quality lead frames and, 1-285
ceramic packaging and, 11-288-90, signal degradation and, 1-27, 11-27, 111-27
11-371-72, 111-200-205 testing for, 11-829, 11-831-33
coated-metal packaging and, 111-345-47 Resistors, and integrated passives in thin-film
conductive adhesives, 111-224 packages, 11-787-88
connectors and, 111-443-44 Resonant cavity enhanced (RCE) detectors,
contact resistance and, 1-423-31 111-496
copper plating and, 111-318 Retention, of connectors, 111-425
definition of, 1-404, 1-407 Return paths, and connectors, 111-404
design of PCs and, 1-43, 11-43, 111-43 Reworking
environmental interactions, 1-432-57 of conductive adhesive interconnections,
of equipment monitors, 1-601 111-225
flip-chip solder-bump connections and, of fine-chip components, 111-176-78
11-170-85 of moisture-sensitive components, 11-472-73
heat-transfer degradation and, 1-524-30 of multichip assemblies, 11-165-70
mechanical loading and, 1-508-19 of photosensitive benzocyc1obutene, 11-605
metrology of, 1-407-23 Reynolds number, 1-342, 1-343, 1-350
of multichip package, 1-103-104, RF magnetron sputtering, 11-748
11-103-104,111-103-104 Rheological compatibility, of molding
multilayer printed-wire boards and, compounds, 11-443-46
111-320-21 Rigid flex circuits, 111-269, III-270
overview of, 1-404-407 Rigorous analysis, of noise in power
package-to-board interconnections, distribution system, 1-30-31, 11-30-31,
111-178-87 III-30-31
of plastic packaging, 11-403-405, 11-474-90 Rimmed steels, I1I-350
as technology driver, 1-35-36, 11-35-36, Risetime, of connectors, III-405
111-35-36, 1-37, 11-37, 111-37 Robust process design, 1-582-84
testing of sealing and encapsulation, Rogers Corporation, 11-778
11-920-25 Roller coating, and polymers, 11-678
thermal fatigue and, 1-457-508 Roll-to-roll process, 11-774, 11-776
of thin-film packages, 11-729-30 Round-trip delay, in high-speed data
thin-film polymer structures, 11-513-15 processing, 1-20, 11-20, III-20
tribological degradation, 1-519-24 R3 bum-in methodology, for C4 flip chips,
of wirebonding, 11-186 11-183
Reliability function, 1-407
Reliable on-chip nets performance, 1-25, 11-25,
111-25
s
Rent's Rule, 1-23-24, 11-23-24, 111-23-24, SAMOS (Silicon Aluminum Metal Oxide
1-130,1-142-48,1-192,11-633 Semiconductor), 11-510, 11-511
Repair, of thin-film packaging, 11-716-27 Sandia National Labs, 11-929-30
Requalification process, for equipment Sandwich probe guide, 11-846
monitors, 1-601-602 Sandwich seal, 11-908
Rerouting, and thin-film technology, 11-728 Sanyo Electronics, III-342
Residual gas analyzers (RGA), 11-916-17 Sapphire wafer, 11-770-71
Residual stress, in polyimides, 11-527-28, Sarnoff, David, 11-365-66
11-529 Scanning electron micrography (SEM), 1-468,
Resin-glass fiber fracture, and mechanical 11-214
drilling, 111-315 Scanning laser ablation, 11-687-88
Resins. See also Epoxy Schottky barrier diodes, 1-230
for plastic packaging, 11-407-20, 11-445 Screening
thermosetting, 111-301 alumina packaging and, 11-316-19
INDEX 11-1023

copper paste and glass-ceramic substrate, trends in packaging and, 1-16-17, 11-16-17,
11-350 III-16-17, 1-18, 11-18, III-18
design of experiments (DOE) and, 1-570 Semishielded flat cable, III-449
Screen printing Separable connections, III-429-30
polymers, 11-679-80 Sequential multilayer thin-film process, 11-672,
solder paste deposition, III-I66-67 11-741-45
Sealant materials, effectiveness of, 11-892 Serial history file (SHF), 11-858
Seal flange grind, of glass-ceramic/copper Series resistance, in signal lines, III-284,
substrate, 11-357-58 III-286
Sealing, of packages Series welding, 11-904
description of process, 1-123-25, 11-123-25, Sharp Corporation, 11-218, 11-272-73
III-123-25 Shear-thinning behavior, and moisture in
failure mechanisms, 11-884-91 plastic packages, 11-445, 11-446
future developments in, 11-930-31 Sheldahl Electronics, 11-774, 11-776
of hermetic versus nonhermetic packages, Shelf life, of PIMEL, 11-561
11-880-84 Shindo Company Ltd., 11-218
introduction to, 11-877-80 Shipping trays, for plastic packages, 11-466
material requirements for, 11-892-94 Shrinkage stresses, in molded plastic packages,
recent advances in, 11-925-30 11-437-39
reliability testing, 11-920-25 Shrink DIP (SDIP), 11-494
testing of hermetic packages and, Shrink quad flat pack (SQPF), 11-494
11-914-20 Shrink small-outline package (SSOP), 11-494
of types of hermetic packages, 11-894-914 Siemens Chemical, 11-558, 11-740-45
Second-level assembly, and ball-grid-array Signal connectors, III-475-76
packages, III-212-16 Signal degradation, 1-27-29, 11-27-29,
Second-level packaging III-27-29
definition of, 1-13, 11-13, III-13 Signal delay
insulating polymers, III-293 electrical packaging and, 1-287
interconnection technology for, III-277 switching noise and, 1-247
plastic packages and, 11-467 Signal distribution
technologies of, 1-114-19, 11-114-19, digital signal transmission nets, 1-220-32
111-114-19 electrical design and, 1-203
Segmented vias, 1-182 lossy signal lines, 1-218-20
Seiko Epson, Inc., 11-268 slow package design, 1-216-18
Selectilux HTR, 11-555 Signal lines
Selective reflow methods, for fine-pitch characteristic impedance of, III-280-82
components, III-175 connectors and, III-404
Self-alignment, of flip-chip solder-bump coupled noises in parallel, III-283-84
connections, 11-156-60 electrical design of, III-279
Self-electro-optic effect device (SEED), III-498 interconnections and structure of, III-279
Self-generated noise tolerance, 1-209-10, low-loss transmission systems and design of,
1-251-52 1-259-68
Self-induced repair, of thin films, 11-726 matched-load impedance of, III-282-83
Self-stretching soldering technology, 11-174 series resistance, III-284, III-286
Semens-Nixdorf,II-381 Signal path, of connectors, III-395
Semiconductor Industry Association (SIA), Signal reflections, 1-29, 11-29, III-29
11-216, III-328 Signal routing, optical elements, III-504-506
Semiconductors Signal theory, and cables, III-450-72
modulators, III-498 Signal wire
molding components for, 11-424, 11-425 DC resistance, III-464
multichip approach to, 1-21, 11-21, III-21 electromagnetic properties of, 1-26-27,
polymers and, 11-512-15 11-26-27, III-26-27
reliability of, 1-103-104, 11-103-104, Silane, as protective coating for porcelain
III-103-104,1-405 surfaces, III-371-72
11-1024 INDEX

Silica, and chip passivation layer, 11-477 Skin depth, with frequency for conductor
Silicon chips, energy level of gates on and materials, m-467
processor design, I-53, II-53, m-53 Skin effect, as frequency-dependent distortion,
Silicon Circuit Board (SICB), 11-749-57 1-283-84, 1-286
Silicone coatings, 1-455-56 Slew rate, driver, 1-214-15
Silicone gels, 11-500-502, 11-927 Slow package design, 1-216-18
Silicone-modified epoxies, 11-927 Slow wave, 1-220, 1-261-64
Silicon/gold eutectic bonding, 11-500 Slump test, for solder paste, m-165
Silicon-on-silicon MCM with AIN, 11-333 Slurry, and alumina packaging, 11-312-13
Silvar alloy, as coated-metal substrate, m-362, Small Outline Integrated Circuit (SOIC),
m-363 m-133, m-137
Silver Small-outline I-headed (SOl) package, 11-399
as contact material, m-414-15 Small-outline packages (SOP), 11-399, 11-401
glass-ceramic packaging and, 11-339 Smart cards, 11-265
lead frames and, 1-286 Sockets
Silver-epoxy die attach compound, 11-889-90 cost of, m-430
Silver-filled conductive adhesives, m-227-28 EIA committee on, m-445
Silver-palladium, 11-339 electromagnetic modeling of, m-409
Silver spot plating, 11-435, 11-437 for MCMs, m-430, m-431-37
Simulation. See also Modeling peripheral versus area-array, m-430-31
manufacturing and, 1-590-92 Soft errors, and plastic packaging, 11-487
of transfer-molding process, 11-459-60 Software. See also Computer-aided design
Single-chip packages heat transfer, 1-333-34
evolution of, 1-65, 11-65, m-65, 1-84, 11-84, Timberwolf,I-191-92
m-84 Solder-bump interconnections. See Flip-chip
as first-level package, 1-78-82, 11-78-82, solder-bump connections
m-78-82 Solder joints
markets for, 1-86-88, 11-86-88, ill-86-88 fatigue and, 11-487
multichip packages compared to, I-52, II-52, reliability of, m-181-84
m-52 SMT interconnections, 11-145-51
Single-device burn-in, 11-492-93 stress analysis of, 1-461-62
Single-layer metallized (SLAM) chip carrier, Solder paste, and package-to-board
11-302 interconnections, m-163-78
Single-layer thin-film packages, 11-650-52 Solder and soldering
Single-level integrated module (SLIM) ball connections and thermal failure,
package, 1-64, 11-64, m-64 1-505-506
Single-in-line memory modules (SIMMs), to coated-metal substrate, m-367-68
connectors for, m-391, m-392, composition of and fatigue performance,
m-435-36 1-492-93
Single-in-line (SIP) package, 11-398-99, 11-400 encapsulation and strain, 1-81, 11-81
Single-point bonding, and tape automated flip-chip solder-bump connections and
bonding, 11-243-44 materials for, 11-144-46
Single tier PGA, die-up configuration for, hermetic sealing, 11-898, 11-900-902
m-265 lead-free compounds for, m-217-23
Sintering MCM package interconnections, m-427-28
alumina packaging and, 11-319-21 miniaturization and, 1-113, 11-113, m -113
glass-ceramic/copper technology and, package-to-board interconnections,
11-351-55 m-151-58
Site-to-site personality open/shorts test, 11-852 package sealing, 1-125, 11-125, m-125
6-sigma concept, 1-559 residual after reworking, 11-166
63-layer glass-ceramic/copper substrate, thermal mismatch and composition of,
11-340-58 11-174-76
Sizing, of glass-ceramic/copper substrate, Solid logic technology (SLT), and ceramic
11-356 packaging, 11-292
INDEX 11-1025

Solvent cleansers, environmentally acceptable, Substrates. See also Ceramic packaging;


II-498 Coated-metal packaging; Glass-ceramic
Solvents, for photosensitive benzocyclobutene, packaging
II-600-603 definition of, 1-13, 11-13, III-13
Space division multiplexing (SOM), III-51 1 photosensitive benzocyclobutene and dicing
Spatial attentuation, 1-248--49 of,II-605
Speckle pattern photography, 1-468 reinforcement, 11-743--44
Spectral sensitivity, of preimidized silicon, 11-749-57
nonshrinking photosensitive polyimides, testing of, 11-821--40, 11-858-59, 11-869-71
II-58 1 thin-film, 11-660, 11-662-63
SPICE (Simulation Program for Integrated Subtractive etching, 11-699-700
Circuit Emphasis), 1-208 Subtractive metallization, 11-697
Spin coating, and polymers, II-596-97, Subtractive process, of coated-metal
II-676-78 packaging, III-367
Spray coating, and polymers, II-678 Sumitomo Bakelite (Japan), 11-592
Spiral flow length, of molding compounds, SuperBGA, III-209
II-443, II-444 Supercomputers
Spontaneous emission devices, optoelectronic, multilayer printed-circuit boards for,
III-492 III-261-62
Spring-type connectors, III-417-18 tape automated bonding and, 11-263-64
Stacked solid logic technology, II-295 Superconductors, and ceramics, 11-289
Stacked stud copper-polyimide, 11-737-38 Super slim TAB LCD tape design, 11-228
Stacking, of glass-ceramic packaging, 11-348, Supplier processes, in manufacturing, 1-602-604
11-350-51 Suppliers
Staggered bonds, 11-201 of ceramic packaging to U.S. market, 11-286
Stainless steel, and coated-metal packaging, of coated metal substrates, III -366
III-354, III-357 Surface laminar circuitry (SLC), 11-178,
Stair-stepped via structures, II-703 11-773-74, III-266
Standards Surface-mount array connections, and package-
for connectors, III-444--45 to-board interconnections, III-187-217
for plastic packaging, 11-499 Surface-mounted packages and surface-mount
Static flex circuits, 111-269 technology (SMT)
Static random-access memories (SRAMs), design factors in, III-286-89
1-18, II-18, III-18 evolution of packaging and, 1-66, II-66,
Statistical process control (SPC), 1-607-608 III-66, 1-107-108, II-107-108,
Statistics, and evaluation of failure, 1-481-83. III-107-108
See also Monte Carlo method package-to-board interconnections, III-132-51
Steel-ball drop test, III-369 plastic packaging and, II-399
Steiner tree, 1-150 thermal failure of, 1-497-507
Stencilling, and solder deposition, III-167-68 types of, 1-83, 11-83, III-83
Stimulated emission devices, optoelectronic, Surface preparation, for photosensitive
III-492 benzocyclobutene, 11-595-96
Storage stability, of preimidized nonshrinking Surveillance Audits, for ISO 9000, 1-610
photosensitive polyimides, 11-580 Switching noise
Strain gauges, 1-466, 1-467 causes of, 1-28-29, 11-28-29, III-28-29
Stress analysis, classical, 1-461-62 coupled noise and, 1-256-58
Stress-corrosion cracking, 1-435 definition of, 1-204
Stress-release additives, and plastic packaging, electrical design and, 1-236, 1-245-52
11-416, 11-418 high-inductance SCM, 1-277
Stress relief, buffer coating for, 11-609 OCO balancing, 1-281
Strip line, 1-255 thin-film structures, 1-270
Structured testability approach, 11-865-66 waveforms for SCM with high frequency,
Subassemblies, connections between, 1-278
III-393-94 System-level thermal resistance, 1-317
11-1026 INDEX

T Terminators, for transmission lines, 1-230--32


Terminology
Tackiness test, for solder paste, Ill-165 flip-chip solder-bump connections,
Taguchi, Genichi, 1-568 11-136-37
Taguchi robust design of experiments (DOE), heat transfer, 1-397-98
1-567-78 package wiring and standardization of,
Tape automated bonding (TAB) 1-138-39
chip-level interconnections and, 1-74-76, Testing
11-74-76, Ill-74-76 comparison of methods, 11-840-42
future trends in, 11-265 connectors used for, Ill-394, Ill-396
inner lead bonding, 11-240--50 contacts and probes, 11-842-50
introduction to, 11-217-22 of copper plating, Ill-318
outer lead bonding, 11-250--58 functional tests, 11-859-69, 11-871
package applications, 11-258-64 of hermetic packages, 11-914-20
properties of ideal substrate for, 11-531, introduction to, 11-818-21
11-532 opens and shorts test implementation,
reliability of solder joints, Ill-182-83 11-850--58
types of tape for, 11-222-37 reliability, 1-522, 11-920--25, Ill-179
wafer bumping, 11-238-40 substrate testing, 11-821-40, 11-858-59,
Tape ball-grid array (TBGA), 11-258-59, 11-869-71
Ill-205-206 testability as technology driver, 1-38-40,
TapePak, 11-259-60, 11-261 11-38-40, Ill-38-40
Tape reels, 11-466-67 Texas Instruments, 11-467, 11-490
Technology drivers, in microelectronics Thermal aging, and solder joint strength, Ill-149
packaging, 1-21-62, 11-21-62, Ill-21-62 Thermal characteristics, of CBGA and CCGA
Teflon packages, Ill-199-200
dielectric constant of, Ill-460--61 Thermal-conduction modules (TCM), 11-308,
flexible circuits and, III-270, III-272 Ill-262
Teflon (R) card, Ill-270 Thermal conductivity
Tektronics, Inc., 11-266 basic concepts in, 1-323-24
Telephone, plug/jack combination for modular, of epoxy resins, 11-412
Ill-386-87 of glass-ceramic packaging, 11-336-37, 11-339
TEM mode cutoff frequency, Ill-464-65 of lead frame materials, 11-427
Temperature. See also Thermal conductivity; of molding compounds, 11-449
Thermal expansion; Thermal mismatch; of representative packaging materials, 1-324
Thermal resistance of silicon circuit board, 11-751-52
chips and measurement of, 1-369-70 Thermal contact resistance, 1-339-42
heat flux and, 1-315 Thermal cross-linking
mechanical drilling and, Ill-315 of preimidized nonshrinking photosensitive
Tensile strength/stress polyimides, 11-582-83
of ceramic packaging, 11-372 of Probimide 400 polyimides, 11-585-86
copper plating and, Ill-318, Ill-319 Thermal cycling
of high temperature non-polyimide characteristics of failures, 1-488-90, 1-494-96
dielectrics, 11-545 stress testing and, 1-469-70
of plastic packaging, 11-441 Thermal degradation, and failure mechanisms,
of polyimide films, 11-569 1-456
of polymers, 11-523, 11-524 Thermal design, as technology driver, 1-33-35,
of silicon films, 11-753-55 11-33-35, Ill-33-35
Terminals Thermal displacements, and SMT
counts of and Rent's Rule, 1-146-48 interconnections, Ill-134-45
metals and processing of, 11-703-707 Thermal dissipation
number of as technology driver, 1-22-24, chip attachment and ceramic substrates,
11-22-24, Ill-22-24 11-369-71s
INDEX 11-1027

coated-metal substrates and, III-374-76 module wiring of, 11-638


requirements for, 1-35, II-35, III-35 thin-film and, II-629, II-645-59
Thermal expansion Thin film bonded optoelectronic devices,
coefficients of, (CTE), 1-458 III-512-18
of copper alloys, II-427, III-358, III-359-61 Thin-film packaging
glass-transition temperature and, II-439-40 ceramic packages with, 11-297
of polyimides, II-569, II-570, II-665, II-666 commercial applications of, II-730-73
of representative materials, 11-421 electrical performance of, II-641-45
of silvar alloy, III-363 emerging technologies in, II-773-84
Thermal fatigue, and reliability, 1-457-508 future directions of, II-792-808
Thermal management. See Heat transfer integrated passives in, II-784-92
Thermal mismatch introduction to, II-628-34
flip-chip solder-bump connections and, 110 connections, II-638-39
11-172-85 low-k packaging parameters and, 11-379
polymers for thin-film applications and, materials and processes for, II-660-716
II-665 micromodule systems and, II-771-72
reliability and, 1-457, 1-483-507 noise design for, 1-268-70
Thermal networks, 1-334 photoneece and, II-609-10
Thermal ratcheting, 1-501 photosensitive polyimides and, 11-555-58
Thermal resistance polymers and, 11-512-15, II-522-30
definition of, 1-33, 11-33, III-33 polyimide polyester development for, 11-518,
heat transfer and concept of, 1-324-31 II-520-21
types of, 1-317-18 power supply and heat removal, II-639-41
Thermal shock PCB laminate technology, III-265-68
copper-plating and, III-318, III-319 reliability of, II-729-30
fatigue life of LLCC packages, III-143 thick-film technology compared to,
Thermal stability, of high temperature non- 11-645-59
polyimide dielectrics, II-543 wiring of, 11-634-38
Thermal stress, evaluation of, 1-461-66 yield/cost considerations, II-716-29
Thermocompression bonding, and tape Thin-film transfer technology, 11-781-84
automated bonding, II-241-42 Thin moisture films, 1-446-50
Thermocompression welding, II-195-96 Thin quad flat pack (TQFP), II-494
Thermodes, and tape automated bonding, Thin sealed small-outline (TSSOP), II-494
II-241-42 Thin small-outline packages (TSOPs), 1-501,
Thermoelectric coolers, 1-389-90 11-401, II-494, III-183
Thermomechanical reliability, of plastic Third-level packaging
packaging, II-405 definition of, 1-13-14, 11-13-14, III-13-14
Thermomigration, and flip-chip solder-bump plastic packages and, II-469
connections, II-179 Third wiring level, benefit of, 1-162
Thermoplastics Thirty-six subpackage, 1-181
future trends in, III-329 Three dimensional optoelectronic
temperature and softening of, 1-456-57 interconnects, III-519-22
Thermosetting resins, III-301 Three-layer tape, for tape automated bonding,
Thermosonic welding, 11-193-95 11-224-25
Thermosyphon device, 1-390, 1-391 3M Electronic Product Division, II-915
Thermotropic liquid crystalline polymers, Through-hole-mounted devices, and plastic
III-329 packaging, II-398-99
Thick-film technology Through hole packages, types of, 1-83, II-83,
assumptions concerning, II-63 7 III-83
ceramic packaging and, II-295-96 Through vias, wiring results using, 1-181, 1-182
coated-metal substrates and adhesion of, Timberwolf (software), 1-191-92
ill-370 Time-domain waveform distortion, and
copper and, 11-346-48 package modeling, 1-286-88
11-1028 INDEX

Tin, as contact-interface material for Ultra large-scale integration (ULSI), and


connectors, III-411-14 package wiring, 1-195
Tin-bismuth solder, III-221 Ultralow-stress epoxy, 11-410
Tioga drill machine, III-314 Ultrasonic wedge bonding, 11-754
Topology Ultrasonic welding, 11-190-93, 11-481
of connectors, III-418-20 Ultrathin PSPls, 11-614, 11-616
logic interconnections and, 1-149-52 Ultraviolet light radiation, and damage to
Toshiba Corporation, 11-240, 11-267, 11-592 electronic devices, 11-893
Total estimated interconnect cost (TEIC), 1-192 Unbalanced flows, in cavities, 11-480
Total resistance, 1-318 Underfill, and fatigue performance, 1-493-94,
Total track count, of wiring, 1-154 1-497
Track requirement, and package wiring, 1-156, Uniglobe Kisco, Inc., 11-535
1-157,1-158 U.S. Army Signal Corps, 11-297-298
Transfer impedance, of cable, III-468 Unit operation analysis, 11-711
Transfer molding, 11-395-96, 11-452-60, Upilex polyimide films, 11-535, III-269
III-274 User friendliness, and design of Pes, 1-44,
Transferred bump TAB, 11-232-33 11-44, III-44
Transfer unity-gain point, 1-206
Transistor outline (TO) package, 11-291-92
Transmission-line delay adder, 1-298-99 v
Transmission-line terminator, 1-230-32
Transmission modulators, III-498 Vapor-phase reflow soldering
Transmission properties, of porcelain-coated- coated-metal substrates and, III-368
metal substrates, III-373 fine-pitch components and, III-174-76
Transmitted-noise tolerance, 1-206-209, 1-210, Variation, and quality, 1-578
1-251-52 Vertical cavity surface-emitting lasers
Transmitters, optoelectronic, III-510-11 (VCSELs), III-493, III-494
Trends, in microelectronics packaging, Very Large Scale Integration (VLSI)
1-14-21,11-14-21, III-I4-21 package wiring and, 1-195
Triagine-based epoxy, 11-607 packaging hierarchy, 1-136
Triaxial cable, III-449, III-450 Very-small-outline package (VSOP), 11-494
Tribological degradation, and reliability, Very small peripheral array (VSPA) package,
1-519-24 1-95-99, 11-95-99, 11-399
Triplate line, 1-255 Vias
Triplate wiring structure, 11-658, 11-659, alumina packaging and punching of, 11-313
III-280 ceramic packaging and punching of, 11-382
Tunneling resistance, 1-424 coated-metal packaging and, III-344
Twelve-subpackage wiring board, 1-180 metallization processes, 11-702-703
Twin-lead cable, III-448 nonplanar staggered, 11-655-57
Two-conductor tape, and tape automated package wireability and availability of,
bonding, 11-233-36 1-177-85
Two-layer tape, for tape automated bonding, thin-film package substrates and, 11-681-97,
11-222, 11-224 11-801-804
Two level smart camera, III-522-23 Viscosity
Two-part connectors, III-419 of molding compounds, 11-442-43
Two-point probe tester, 11-848 of solder paste, III -164-65
Tyvek,II-468 wire sweep and, 11-480
Visual examination, of solder paste, III-I64

u Void transport, and wire sweep, 11-480-81


Voltage contrast electron beam test, 11-834
Ube Industries, Ltd., 11-533, 11-535 Voltage standing-wave ratio (VSWR),
ULaminate family, of PBGA packages, III-209 III-407
Ultrafine pitch (UFP), III-134 Vycor glass, 11-378-79
INDEX 11-1029

w Wirebonding
applications of, II-198-203
Wafer bumping, and tape automated bonding, basic concepts, 11-186-87
11-238-40 as chip-bonding technology, 1-72-74,
Wafer scale substrate, for GaAs packaging, II-72-74,11I-72-74
11-770-71 evaluating/optimizing process of, 11-213-16
Warpage, of coated-metal substrates, III-344, future of, 11-216-17
III-345, III-346, III-347 history of, 11-196-98
Water, permeability of through organic and joining technology and, 11-187-96
inorganic materials, 1-123, 11-123, III-123 materials for, 11-203-209
Water absorption, of high temperature non- plastic packages and failure of, II-481-83
polyimide dielectrics, 11-542-43, 11-544 process automation and, 11-209-10
Water-vapor transmission rate (WVlR), and process choice decisions/guidelines,
moisture-barrier bags, II-468 II-210-13
Waveforms Wirebond metal BGA package, III-210
low-loss transmission forms and, Wire braid, shield coverage of, III-468
1-261-64 Wire CCGA packages, III-I96
near-end coupled-noise, 1-257 Wire length, versus track requirement, 1-157
switching noise, 1-278 Wire sweep, as failure mechanism, 11-479-80
time-domain distortion of, 1-286-88 Wiring
Waveguides, optical, 11-613, III-502-504 basic concepts, 1-129-39
Wavelength division multiplexing (WDMs), chip circuit placement and model
III-511 predictions, 1-155-65
Wave soldering, and coated-metal substrates, future challenges in, 1-191-95
III-368 of thin-film packages, II-634-38
Wedge bonding, 11-188, 11-191, 11-195, 11-201, variables in, 1-139-55
II-212-13 wireability of higher-level packages,
Weibull distribution, of failure rate, 1-412-15, 1-165-91
1-416, 1-417 Wiring board, model of, 1-168
Welding, and package sealing, 1-125, 11-125, Wiring capacity, 1-130, 1-139, 1-166, 1-179,
III-125, 11-902, 11-904-905 1-180,1-185, III-278
Western Electric Company, III-342, III-367 Wiring demand, 1-130, 1-140, 1-142-48, 1-166,
Wet etching, of polyimides, II-695-96 1-179,1-180,1-185
Wicldess heat pipe, 1-390 Wiring density
Wipe, and connectors, III-425-26, III-443-44 ceramic substrates and, II-287-88
Wireability cost of ceramic, thin-film, and PWB
basic analysis of, 1-139-42 technologies and, II-381
of chips, 1-158 thin-film technology and, 11-630
custom design and, 1-164-65 Wiring distribution, 1-148-49
definition of, 1-129-30 Wiring efficiency, 1-130, 1-139
design tools and, 1-194-95 Wiring length, 1-141
of higher-level packages, 1-165-91 Wiring-level metallization, 11-699-702
macros on chips and, 1-161-62 Wiring overflows
mixed-signal wiring, 1-193-94 definition of, 1-131
multilevel wiring and chip, 1-162-64 overflow count, 1-178
overall value of analysis, 1-191-93 quantitative model for prediction of,
packaging hierarchy and, 1-131 1-185-87
preplacement and prewiring, 1-193 for through-hole constraints, 1-171
probabilistic model for, 1-152-55 wiring effectiveness and, 1-159
as technology driver, 1-22-24, 11-22-24, Wiring rules, 1-203, 1-214, 1-227-29
III-22-24 Wiring-track accessibility, and package
wiring distribution, 1-148-49 wireability,I-177-85
Wirebonded wafer bumps, 11-206 Withdrawal force, and connectors, III-425
11-1030 INDEX

W. L. Gore and Associates, 11-787 y


Work instructions, for manufacturing, 1-597-99
Work-in-process (WIP) control, 1-604-605 Yield, and cost of thin-film packaging,
Workstations
11-716-29
planar PCB with surface laminar circuitry,
III-257
as technology drivers, 1-42-44, 11-42-44,
z
III-42-44 Zero-defect manufacturing, 1-38, 11-38, III-38
Zero-insertion-force (ZJF) systems, III-258,
III-416
x Zigzag in-line package, connectors for, III-392
Zinc, in lead-free solder alloys, 111-221, 111-223
X-ray diffraction, 1-468 Z-Link, 111-270

You might also like