'THE LARGE SCALE STRUCTURE OF THE UNIVERSE

INTERNATIONAL ASTRONOMICAL UNION

UNION ASTRONOMIQUE INTERNATlONALE

SYMPOSIUM No. 79
HELD IN TALLINN, ESTONIA, U.S.S.R., SEPTEMBER 12 -16,1977

THE LARGE SCALE

STRUCTURE
OF THE UNIVERSE
EDITED BY

'M. S. LONGAIR
Mullard Radio Astronomy Observatory,
Cavendish Laboratory, Cambridge, England

AND

J. EINASTO
W. Struve Astrophysical Observatory of Tartu, Toravere, Estonia, U.S.S.R.

D. REIDEL PUBLISHING COMPANY

DORDRECHT : HOLLAND / BOSTON: U.S.A.

1978
 Library of Congress Cataloging in Publication Data

Main entry under title:

The large scale structure of the universe.

(Symposium - International Astronomical Union; no. 79)

Bibliography: p.
Includes index.
1. Cosmology - Congresses. 2. Galaxies - Congresses. 3. Stars - Clusters -
Congresses. I. Longair, M. S., 1941- II. Einasto, Jaan. III. Series:
International Astronomical Union. Symposium; no. 79.
QB981.L3 523.1 78-7032
ISBN-13: 978-90-277-0896-0 e-ISBN-13: 978-94-009-9843-8
DOl: 10.1007/978-94-009-9843-8

Published on behalf of
the International Astronomical Union
by
D. Reidel Publishing Company, P.o. Box 17, Dordrecht, Holland

All Rights Reserved

Copyright © 1978 by the International Astronomical Union

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

by D. Reidel Publishing Company, Inc.
Lincoln Building, 160 Old Derby Street, Hingham,
Mass. 02043, U.S.A.

No part of the material protected by this copyright notice may be reproducea or

utilized in any form or by any means, electronic or mechanical, including
photocopying, recording or by any informational storage and retrieval system,
without written permission from the publisher
 fA U Symposium 79 (edited by M S. Longair and 1. Einasto)

ERRATA

PAGE READS AT PRESENT SHOULD READ

vi 25th and 26th line

A. G. D. PhIlip and A. G. Davis: Fine Structure in the Magellanic A. G. D. Philip: Fine Structure in the Magellanic Stream
Stream

xi 10th line from below

L. Lund L. Luud

xiii 9th line from below

L. Lund L. Luud

52 13th line
companions M 32 and NGC 205, M 81 and NGC 3077. Dwarf companions M 32 and NGC 205, M 81 with M 82 and NGC 3077.
galaxies of very Dwarf galaxies of very

55 Caption Figure 3, line 7

triangles - respective num bers of hypergalaxies. 90% confidence triangles - respective num bers of hypergalaxies if the M 81 group is
limits have been shown. considered as an aggregate of four hypergalaxies. 90% confidence
limits have been shown.
56 21 st and 22nd line
This process can Increase the mass and luminosity of the main This process can increase the mass and luminosity of the main
galaxies and of the clusters of galaxies of the clusters of

57 Table 1
M M
(M"')6 (M )
Disk (stars, galactIc clusters) 6 x 10 Disk (stars, galactic clusters) 6 x 1'0 10

p.t.o.
PAGE READS AT PRESENT SIIOULD READ

131 14th line

A. G. D. Philip and A. G. Davis A. G. D. Philip

245 Figure 3. Distribution of avaUable redshifts of galaxies In areas A, FIgure 3. Distribution of available redshifts along 'the Perseus chain.
Band C (see Fi~ure 2).

404 24th line

is hI> 2 x 10-4 n -5/4 h-l/2 0 + -3nlh, is hi> 2 x 10-4 n -5/4 h-I/2(1 + 3nh 2 ),
-I -1 -I -1
where h = 110/75 km s Mpc where 11 = HO / 75 km s Mpc .

427 9th line

(il) We do not know which forces were most Important for the (ii) We do not know which forces and processes were most
origln important for the origin

429 4th line from below

before recombination and those produced by undamped motions before recombmation (marked B) and those produced by undamped
after reeom- motions after rccom-

431 9th line

complexes of massive galaxies), which are first formed with complexes of massive galaxies), which are first fonmed due to the
spectrum of spectrum of

434 20th line

some klnd of "space-time curvature turbulence" (Tomita 1972). Some kind of "space-time-curvature turbulence" (Tomita 1972).
The The

434 28th line

that the hypothesis of Chibisov (1976), who proposed that near the the hypothesis of Chibisov (1976), who proposed that near the
 TABLE OF CONTENTS

Foreword xi

Organising Committees xiii

List of Participants xv

In the table of contents Invited Lectures are shown in block

capitals, Contributed Papers in lower case, and Short Communications
in indented lower case format.

I. GALAXIES IN SMALL GROUPS

B. A. VORONTSOV-VELYAMINOV: THE PHOTOGRAPHY OF GROUPS

OF GALAXIES 3

I. D. KARACHENTSEV: VELOCITY DISPERSION IN SMALL SYSTEMS

OF GALAXIES II

E. L. TURNER: OPTICAL STUDIES OF SMALL GROUPS OF GALAXIES 21

R. B. TULLY and J. R. FISHER: NEARBY SUALL GROUPS OF GALAXIES 31

W. K. Huchtmeier: Observations of high velocity HI clouds in

the Loc;> 1 Group 49

J. EINASTO: HYPERGALAXIES 51

J. R. GOTT, III: N-BODY SIMULATIONS AND THE VALUE OF n 63

G. A. TAMMANN and R. KRAAN: THE GALACTIC NEIGHBOURHOOD 71

J. Materne: Three dimensional analysis of groups of galaxies 93

D. J. Hegyi: Detection of an optical halo surrounding
the spiral galaxy NGC 4565 95
J. Einasto: The scatter in mass-to-luminosity ratios 96
L. M. Ozernoy and M. Reinhardt: Stabilization of systems
of galaxies by subclustering 98
C. Casini and J. Heidmann: Morphological investigation of
pairs containing Markarian galaxies 100
vi TABLE OF CONTENTS

K.-H. Schmidt: Cluster membership of Seyfert galaxies 101

A. Tomov: The results of observations of double galaxies
in the UBV system 102
I. Pronik and L. Metik: On the origin and evolution of
pairs of galaxies that have different physical
and kinematic characteristics 103
E. Khachikian: Markarian galaxies with double and
multiple nuclei 105
W. Hagner and N. Richter: Equidensitometric determination
of angular diameters and mean surface magnitudes
of compact galaxies 105

A. TOOMRE: INTERACTING SYSTEMS 109

B. Vorontsov-Velyaminov: On the tidal origin of MSl-type

systems 117
V. Dostal and V. Metlov: The frequency of ring galaxies
and the probability of their formation by
collisions 117

P. Biermann and R. Wielen: Encounters of spherical galaxies:

N-body simulations and comparison with
theoretical predictions 121

D. LYNDEN-BELL: ON THE MAGELLANIC STREAM, THE MASS OF THE

GALAXY AND THE AGE OF THE UNIVERSE 123

W. E. Kunkel: The stellar component of the Magellanic

Stream 130
A. G. D. Philip and A. G. Davis: Fine structure in the
Magellanic Stream 131

II. CLUSTERS OF GALAXIES

J. Melnick, S. White and J. Hoessel: Photoelectric surface

photometry of the Coma Cluster 135

C. LARI and G. C. PEROLA: RADIO PROPERTIES OF ABELL CLUSTERS 137

R. D. EKERS: RADIO OBSERVATIONS OF CLUSTERS OF GALAXIES:
THE TAIL SOURCES 149

E. A. Valentijn and H. van der Laan: Extended radio emission

in clusters of galaxies: recent Westerbork
observations 153

R. Wielebinski: Clusters with extended radio emission at high

frequencies 157
TABLE OF CONTENTS vii

V. L. Ginzburg: Radio haloes around galaxies and in clusters 161

J. L. CULHANE: X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES 165

J. SILK: GAS IN GALAXY CLUSTERS 179

S. J. AARSETH: COMPUTER SIMULATIONS OF GALAXY CLUSTERING 189

G. Paal: An analogy between simulated and actual clusters

of different kinds 197
F. W. Baier: Structures and number-density distributions
in clusters of galaxies 198
J. Burczyk and A. Zieba: Mathematical approach to the
problem of clustering 199
A. Michalec and J. Machalski: Radio emission of Abell
clusters in the GB and GB2 regions 200

III. LARGE SCALE SYSTEMS

G. DE VAUCOULEURS: THE LOCAL SUPERCLUSTER 205

R. B. Tully and J. R. Fisher: A tour of the Local

Supercluster 214

P. J. E. PEEBLES: LARGE SCALE CLUSTERING IN THE UNIVERSE 217

L. Grishchuk: The isotropy of the universe on scales

exceeding the horizon 226

K. RUDNICKI and S. ZIEBA: RESULTS ON THE LARGE SCALE DISTRIBUTION

OF EXTRAGALACTIC OBJECTS OBTAINED BY THE METHOD OF
STATISTICAL REDUCTION 229
"
M. JOEVEER and J. EINASTO: HAS THE UNIVERSE THE CELL STRUCTURE? 241

G. O. ABELL: THE PHYSICAL PROPERTIES OF LARGE SCALE SYSTEHS

FROM OPTICAL OBSERVATIONS 253

M. Tarenghi, W. G. Tifft, G. Chincarini, H. J. Rood and

L. A. Thompson: The structure of the Hercules
supercluster 263

W. G. Tifft and S. A. Gregory: Observations of the large scale

distribution of galaxies 267

J. Huchra: The field luminosity function and nearby groups of

galaxies 271
viii TABLE OF CONTENTS

M. A. Arakelian: Markarian galaxies in the vicinity

of the Coma cluster 274
P. Flin and M. Urbanik: Remarks on the angular
distribution of Markarian galaxies 275
M. Kalinkov, V. Dermendjiev, B. Staikov, I. Kaneva,
B. Tomov and K. Stavrev: Superclustering of
galaxies 276
B. I. Fessenko: On the methods of discovering groups and
clusters of galaxies 279
S. Phillipps: The autocorrelation analysis of deep
galaxy samples 280

DISCUSSION OF METHODS ON DETERMINING THE ~ffiAN MATTER DENSITY

OF THE UNIVERSE 281

IV. OBSERVATIONAL EVIDENCE FOR COSMOLOGICAL EVOLUTION

M. SCHMIDT: QUASARS AND COSMOLOGICAL EVOLUTION 289

J. G. BOLTON and A. SAVAGE: THE SURFACE DENSITY OF QUASARS 295

H. Lorenz: A survey programme for QSO and related objects 303

M. S. LONGAIR: THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES 305

Y. N. PARIJSKIJ: SEARCH FOR PRIMORDIAL PERTURBATIONS OF THE

UNIVERSE: OBSERVATIONS WITH RATAN-600 RADIO
TELESCOPE 315

P. E. BOYNTON: THE QUEST FOR FINE-SCALE ANISOTROPY IN THE

RELICT RADIATION 317

H. GURSKY: COSMOLOGICAL INFORMATION FROM X-RAY OBSERVATIONS 327

I. D. Karachentsev and A. I. Kopylov: Counts of faint galaxies 339

B. M. TINSLEY: THE EVOLUTION OF GALAXIES: EVIDENCE FROM

OPTICAL OBSERVATIONS 343

J. P. OSTRlKER: ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF

GALAXIES 357

B. J. T. JONES and M. J. REES: THE EPOCH OF GALAXY FORMATION 377

B. A. Peterson: QSO absorption lines and intergalactic

hydrogen clouds 389
TABLE OF CONTENTS IX

R. A. SUNYAEV: FLUCTUATIONS OF THE MICROWAVE BACKGROUND

RADIATION 393

A. A. Kurskov and L. M. Ozernoy: Constraints on the mean

density of the Universe which follow from the
theories of adiabatic and whirl perturbations 404
P. Boynton: Recent advances in microwave cosmology 405

V. THE FORMATION OF STRUCTURE IN THE UNIVERSE

YA. B. ZELDOVICH: THE THEORY OF THE LARGE SCALE STRUCTURE

OF THE UNIVERSE 409

A. G. Doroshkevich, E. M. Saar and S.F. Shandarin: The internal

structure of protoclusters and the formation
of galaxies 423

1. M. OZERNOY: THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE

IN THE UNIVERSE 427

A. D. Chernin: Origin of protogalactic eddies 439

M. Davis: Evidence for the gravitational instability picture
in a dense Universe 441

E. M. Kellogg: Primeval clusters of galaxies and the X-ray

background 445

R. D. Davies: Observational limits on neutral hydrogen in

primordial protoclusters 447

J. Jaaniste and E. Saar: Orientation of spiral galaxies

as a test of theories of galaxy formation 448
A. M. Fridman: Non-linear effects in flat gravitating
systems 450

M. S. LONGAIR: PERSONAL VIEW- THE LARGE SCALE STRUCTURE

OF THE UNIVERSE 451

AUTHOR INDEX 463

 FOREWORD

The significance of the present IAU symposium, "The Large Scale

Structure of the Universe", fortunately requires no elaboration by the
editors. The quality of the wide range of observational and theoretical
astrophysics contained in this volume speaks for itself. The published
version of the proceedings contains all the contributions presented at
the symposium with the exception of the introductory lecture by V. A.
Ambartsumian. Contributed papers, short contributions and discussions
have been included according to the recommendations of the IAU.

Many people contributed to the success of the symposium. First of

all, thanks are due to the USSR Academy of Sciences and to the Estonian
Academy of Sciences for sponsoring this symposium in Tallinn. The
efforts of Academician K. Rebane, President of the Estonian Academy of
Sciences, are particularly appreciated. The astronomical hosts of the
symposium were the members of the W. Struve Astrophysical Observatory of
Tartu who made outstanding efforts to lavish participants with Estonian
hospitality which was greatly appreciated and enjoyed by them and their
guests.

The members of the Scientific and Local Organising Committees are

listed below and we thank all of them for their contributions which were
central to the success of the symposium. In addition are listed members
of the Technical Organising Committee who were responsible for all
details of the organisation and whose vigilance ensured that all aspects
of the symposium ran smoothly and efficiently. Their contributions are
all gratefully acknowledged.

Certain individuals made particularly important contributions to

the symposium. Dr V. Unt as chairman of the Local Organising Committee
bore a heavy responsibility for the management of the symposium.
Drs L. Lund and T. Viik had the unenviable task of attending to all
detailed practical matters of the symposium and they devoted many weeks
and months to ensuring everything ran smoothly. Miss Kaidi Ahero was
responsible for the secretarial work in connection with all aspects of
the symposium in Estonia, including correspondence with foreign partici-
pants. Her enormous efforts before and during the symposium were
greatly appreciated by everyone, especially the organisers.

The organisation of the Scientific Programme was done in Cambridge

and Miss Shirley Fieldhouse acted as secretary to the Scientific
Organising Committee. This involved an enormous amount of correspon-
ri
xu FOREWORD
dence which she organised with her customary efficiency and enthusiasm.
She was also responsible for the production of the present volume in
camera-ready typescript. This huge task included retyping all the
papers by participants whose native language was not English and all the
short contributions and discussions. The Editors express their warm
gratitude to her for the many hours she devoted to the symposium.
Finally, the photographic reduction of diagrams for many of the papers
was expertly done by Mr W. H. Andrews of the Cavendish Laboratory.

M. S. Longair J. Einasto
Editors
 SCIENTIFIC ORGANISING COMMITTEE

M. S. Longair United Kingdom

(Chairman)
J. Einasto USSR
K. C. Freeman Australia
E. B. Holmberg Sweden
J. P. Ostriker USA
G. A. Tammann Switzerland
G. de Vaucouleurs USA
S. van den Bergh Canada
Y. B. Zeldovich USSR

LOCAL ORGANISING COMMITTEE

V. Unt (Chairman)
J. Einasto (Vice-Chairman)
H. Eelsalu (Secretary)
E. A. Dibai
E. Ergma
G. S. Khromov
G. Naan
I. D. Novikov
Y. N. Parij skij
G. K. Tikhonov
H. M. Tovmassian

TECHNICAL ORGANISING COMMITTEE

L. Lund (Chairman)
T. Viik (Vice-Chairman)
K. Ahero (Secretary)

Interpreters: A. Berezin, T. Johanson, L. Kivim~gi,

H. Kabi, R. Pobul, E. Shedova
Members: A. Cherbanovski, V. Davydov, U. Hanni,
P. Kalamees, A. Kivikas, R. Kivikas,
A. Kivila, R. Koppel, I. Pustylnik,
V. Randmets, H. Silvet, J. Sitska

xiii
 LIST OF PARTICIPANTS

AARSETH, S.J., Institute of Astronomy, Cambridge, England.

ABELL, G.O., Department of Astronomy, University of California,
Los Angeles, USA.
AFANASIEV, A.A., Special Astrophysical Observatory, Stavropolskij Kraj,
USSR.
ALEKSEEV, V.A., Moscow University, USSR.
AMBARTSUMIAN, V.A., Byurakan Astrophysical Observatory, Armenia, USSR.
ARAKELIAN, M.A., Byurakan Astrophysical Observatory, Armenia, USSR.
AUDOUZE, J., Laboratoire Rene Bernas, Orsay, France.
BABADZANIANC, M.K., Astronomical Observatory, Leningrad University,
USSR.
BAIER, F.W., Zentralinstitut fUr Astrophysik, Potsdam-Babelsberg, GDR.
BARABANOV, A.V., Sternberg Institute, Moscow, USSR.
BASINSKA-GRZESIK, E., Copernicus University, Astronomy Center, Polish
Academy of Science, Torun, Poland.
BERTOLA, F., Osservatorio Astronomico, Padova, Italy.
BINNEY, J., Department of Theoretical Physics, University of Oxford,
England.
BISNOVATYI-KOGAN, G.S., Space Research Institute, Moscow, USSR.
BOLTON, J.G., ANRAO, Division of Radiophysics, Parkes, Australia.
BORCHKHADZE, T.M., Abastumani Astrophysical Observatory, Georgia, USSR.
BOTTlNELLI, L., Departement de Radioastronomie, Observatoire de Meudon,
France.
BOYNTON, P.E., Department of Astronomy, University of Washington,
Seattle, USA.
BURCZYK, J., Wyzsza Szkota Pedagogieczna, Kielce, Poland.
CAPACCIOLI, M., Osservatorio Astronomico Padova, Italy.
CHERNIN, A.D., A.F. Ioffe Physical-Technical Institute, Leningrad, USSR.
CHERNOMORDIK, V.V., P.N. Lebedev Institute of Physics, Moscow, USSR.
CHIBISOV, G.V., P.N. Lebedev Physical Institute, Moscow, USSR.
XVI LIST OF PARTICIPANTS

CHINCARINI, G.I., Department of Physics and Astronomy, University of

Oklahoma, USA.
CORWIN, H.G., Department of Astronomy, University of Edinburgh, Scotland.
CULHANE, L., Mullard Space Science Laboratory, Dorking, England.
DAVIES, R.D., Nuffield Radio Astronomy Laboratories, Jodrell Bank,
Cheshire, England.
DAVIS, M., Center for Astrophysics, Cambridge, USA.
DEMIANSKI, M., Uniwersytet Warszawski, Instytut Fizyki Teoretyeznej,
Poland.
DENISYK, E.K., Astrophysical Institute, Alma-Ata, USSR.
DIBAI, E.A., Sternberg Institute, Moscow, USSR.
DOROSHKEVICH, A.G., Institute of Applied Mathematics, Moscow, USSR.
EFSTATHIOU, G., Physics Department, University of Durham, England.
EINASTO, J., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
EKERS, R.D., Kapteyn Laboratory, Groningen, The Netherlands.
EPSTEIN, R.I. Nordita, Copenhagen, Denmark.
ERGMA, E., Astronomical Council, Moscow, USSR.
FALL, S.M., Institute of Astronomy, Cambridge, England.
FESSENKO, B.T., Pskov Pedagogical Institute, USSR.
FLIN, P., Uniwersytet Jagiellonski, Observatorium Astronomicze, Poland.
FRIDMAN, A.M., Siberian Institute of Earth Magnetism, Irkutsk, USSR.
GINZBURG, V.L., Lebedev Physical Institute, Moscow, USSR.
GISLER, G.R., Sterrewacht, Leiden, The Netherlands.
GOTT, III, J.R., Department of Astrophysical Sciences, Princeton
University, USA.
GOUGUENHEIM, L., Departement de Radioastronomie, Observatoire de Meudon,
France.
GRISHCHUK, L.P., Sternberg Institute, Moscow, USSR.
GURSKY, H., Center for Astrophysics, Cambridge, USA.
GYLDENKERNE, K., Copenhagen University Observatory, Bronfelde, T¢l¢se,
Denmark.
HARTWICK, F.D.A., Department of Physics and Astronomy, University of
Victoria, Canada.
HAUD, U., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
HEGYI, D.J., Harrison M. Randall Laboratory of Physics, University of
Hichigan, USA.
LIST OF PARTICIPANTS XVll

HEIDMANN, J., Observatoire de Paris, Section d'Astrophysique, France.

HOLMBERG, E.B., Uppsala Observatory, Sweden.
HUCHRA, J.P., Center for Astrophysics, Cambridge, USA.
INNANEN, K.A., Department of Physics, York University, Ontario, Canada.
JAAKKOLA, T., Observatory, Tahtitornimaki, Helsinki, Finland.
JAANISTE, J., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
JOEVEER, M., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
JONES, B.J.T., Institute of Astronomy, Cambridge, England.
KAASIK, A., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
KALINKOV, M., Department of Astronomy, Sofia, Bulgaria.
KAPLAN, S.A., Radiophysical Research Institute, Gorky, USSR.
KARACHENTSEV, 1.0., Special Astrophysical Observatory, Stavropolskij
Kraj, USSR.
KARACHENTSEVA, V., Special Astrophysical Observatory, Stavropolskij Kraj,
USSR.
KARDASHEV, N.S., Space Research Institute, Moscow, USSR.
KELLOGG, E.M., Center for Astrophysics, Cambridge, USA.
KHACHIKIAl~, E.E., Byurakan Astrophysical Observatory, Armenia, USSR.
KHRISTICH, V.G., Leningrad University, Astronomical Observatory, USSR.
KHROMOV, G.S., Astronomical Council, Moscow, USSR.
KIANG, T., Dunsink Observatory, County Dublin, Ireland.
KIPPER, A., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
KIPPER, T., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
KLYPIN, A.A., Institute of Applied Mathematics, Moscow, USSR.
KOMBERG, B.V., Space Research Institute, Moscow, USSR.
KOPYLOV, A.I., Special Astrophysical Observatory, Stavropolskij Kraj,
USSR.
KOPYLOV, I.M., Special Astrophysical Observatory, Stavropolskij Kraj,
USSR.
KOTOK, E.V., Institute of Applied Mathematics, Moscow, USSR.
KUNKEL, W.E., Observatorio Nacional, Rio de Janeiro, Brasil.
KURSKOV, A.A., Institute of Physics, Minsk, USSR.
KURT, V.G., Space Research Institute, Moscow, USSR.
KURTANIDZE, 0., Astrophysical Observatory, Abastumani, Georgia, USSR.
KUS, A., Institute of Astronomy, Copernicus University, Torun, Poland.
XVill LIST OF PARTICIPANTS

KUZMIN, G., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.

LARI, C., Laboratorio di Radioastronomia, Bologna, Italy.
LIPOVETSKIJ, V.A., Special Astrophysical Observatory, Stavropolskij
Kraj, USSR.
LONGAIR, M.S., Mullard Radio Astronomy Observatory, Cavendish Laboratory,
Cambridge, England.
LORENZ, H., Zentralinstitut fUr Astrophysik, Potsdam Babelsberg, GDR.
LYNDEN-BELL, D., Institute of Astronomy, Cambridge, England.
MAROCHNIK, L.S., Astrophysical Department, Rostov State University, USSR.
MARX, S., Zentralinstitut fUr Astrophysik, Potsdam-Babelsberg, GDR.
MATERNE, J., European Southern Observatory, Geneva, Switzerland.
MAVRIDES, S., Observatoire de Paris, Section d'Astrophysique, Meudon,
France.
MICHALEC, A., Uniwersytet Jagiellonski, Observatorium Astronomicze,
Krakow, Poland.
MOROZOVA, N.D., "Priroda", Moscow, USSR.
MORTON, D.C., Anglo-Australian Observatory, Epping, N.S.W., Australia.
NOTNI, P., Zentralinstitut fur Astrophysik, Potsdam-Babelsberg, GDR.
NOVIKOV, I.D., Space Research Institute, Moscow, USSR.
OBURKA, 0., Technical University, Brno, Czechoslovakia.
OLEAK, H., Zentralinstitut fUr Astrophysik, Potsdam-Babelsberg, GDR.
OOSTERBAAN, C.E., Sterrewacht, Huygens Laboratorium, Leiden,
The Netherlands.
OSTRlKER, J.P., Princeton University Observatory, USA.
OZERNOY, L.M., Lebedev Physical Institute, Moscow, USSR.
PAAL, G., Konkoly Observatory, Budapest, Hungary.
PARIJSKIJ, Yu.N., Special Astrophysical Observatory, Stavropolskij Kraj
USSR.
PEEBLES, P.J.E., Princeton University, USA.
PEROLA, G.C., Instituto di Fisica, Milano, Italy.
PETERSON, B.A., Anglo-Australian Observatory, Epping, N.S.W., Australia.
PETROSIAN, A.P., Byurakan Astrophysical Observatory, Armenia, USSR.
PETROS IAN , V., Institute for Plasma Research, Stanford, USA.
PETROVSKAYA, I.V., Leningrad University, Astronomical Observatory,
Leningrad, USSR.
PHILIP, A.G.D., Dudley Observatory, Latham, USA.
LIST OF PARTICIPANTS XIX

POLJACHENKO, V.L., Siberian Institute of Earth Magnetism, Irkutsk, USSR.

POLJAKOVA, G.D., Astronomical Observatory, Pulkovo, Leningrad, USSR.
PRONIK, 1.1., Crimean Astronomical Observatory, Crimea, USSR.
PRONIK, V.I., Crimean Astronomical Observatory, Crimea, USSR.
REBANE, K., Academy of Sciences of the Estonian SSR, Tallinn, USSR.
REINHARDT, M., Astronomisches Institut der Universitat, Bochum, GFR.
RICHTER, G.M., Zentralinstitut fur Astrophysik, Potsdam-Babelsberg, GDR.
RICHTER, N., Zentralinstitut fur Astrophysik, Potsdam-Babelsberg, GDR.
ROOD, H.J., Department of Astronomy and Astrophysics, Michigan State
University, USA.
RUDNICKI, K., Uniwersytet Jagiellonski, Observatorium Astronomicze,
Krakow, Poland.
SAAR, E., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
SAHAKIAN, K.A., Byurakan Astrophysical Observatory, Armenia, USSR.
SAPAR, A., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
SCHMIDT, K.H., Zentralinstitut fur Astrophysik, Potsdam-Babelsberg, GDR.
SCHMIDT, M., California Institute of Technology, Pasadena, USA.
SHAHBAZIAN, R.K., Byurakan Astrophysical Observatory, Armenia, USSR.
SHANDARIN, S.F., Institute of Applied Mathematics, Moscow, USSR.
SHKLOVSKIJ, V.F., Space Research Institute, Moscow, USSR.
SHVARTSMAN, V.F., Special Astrophysical Observatory, Stavropolskij Kraj,
USSR.
SIGBATULLIN, N.A., Moscow University, Faculty of Mechanics and Mathe-
matics, USSR.
SIGOV, J.S., Institute of Applied Mathematics, Moscow, USSR.
SILK, J., Berkeley Astronomy Department, University of California, USA.
SLYSH, V.I., Space Research Institute, Moscow, USSR.
STRAIZYS, V., Astronomical Observatory, Lithuanian State University,
Vilnius, USSR.
SUCHKOV, A.A., Astrophysical Department, Rostov State University, USSR.
SUNYAEV, R.A., Space Research Institute, Moscow, USSR.
TAGO, E., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
TAMMANN, G.A., Astronomisches Institut der Universit~t Basel, Binningen,
Switzerland.
TARENGHI, M., Laboratorio Fisica Cosmica, Milano, Italy.
TEERIKORPI, P., Observatory and Astrophysics Laboratory, T~htitornim~ki,
Helsinki, Finland.
xx LIST OF PARTICIPANTS

TIFFT, W.C., Steward Observatory, University of Arizona, Tucson, USA.

TINSLEY, B.M., Yale University Observatory, New Haven, USA.
TOMOV, A., Astronomical Department, Academy of Science of Bulgaria,
Sofia, Bulgaria.
TOOMRE, A., Massachusetts Institute of Technology, Cambridge, USA.
TOVMASIAN, H.M., Byurakan Astrophysical Observatory, Armenia, USSR.
TRAAT, P., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
TREMAINE, S.D., California Institute of Technology, Pasadena, USA.
TULLY, R.B., Institute for Astronomy, University of Hawaii at Manoa,
Honolulu, USA.
TURNER, E.L., Harvard College Observatory, Cambridge, USA.
URBANIK, M., Uniwersytet Jagiellonski, Observatorium Astronomicze,
Krakow, Poland.
UUS, U., Tartu Astrophysical Observatory, Toravere, Estonia, USSR.
VALENTIJN, E.A., Sterrewacht Huygens Laboratorium, Leiden,
The Netherlands.
VAN WOERDEN, H., Kapteyn Astronomical Institute, Groningen,
The Netherlands.
VAN DER LAAN, H., Sterrewacht, Huygens Laboratorium, Leiden,
The Netherlands.
VAUCOULEURS, A. DE., University of Texas at Austin, Astronomy Department,
USA.
VAUCOULEURS, G.H. DE, University of Texas at Austin, Astronomy Department,
USA.
VENNIK, J., Tartu Astrophysical Observatory, T6ravere, Estonia, USSR.
VON BORZESZKOWSKI, H.-H., Zentralinstitut fUr Astrophysik, Potsdam-
Babelsberg, GDR.
VORONTSOV-VELYAMINOV, B.A., Moscow University, GAISH, USSR.
WIELEBINSKI, R., Max-Planck-Institut fur Radioastronomie, Bonn, GFR.
WIELEN, R., Astronomisches Rechen-Institut, Heidelberg, GFR.
ZASOV, A.V., Sternberg Institute, Moscow, USSR.
ZELDOVICH, Ya.B., Institute of Applied Mathematics, Moscow, USSR.
ZIEBA, A., Astronomical Observatory, Jagellonian University, Krakow,Poland.
ZIOLKOWSKI, J., Polish Academy of Science, Institute of Astronomy,
Warsaw, Poland.
I

GALAXIES IN SMALL GROUPS

THE PHOTOGRAPHY OF GROUPS OF GALAXIES

B. A. Vorontsov-Velyaminov
Sternberg Astronomical Institute,
Moscow University, U.S.S.R.

1. IMPROVEMENTS IN PHOTOGRAPHY

The method of combining several negatives together is at last

coming into use. Forty years ago this method was used for the first
time by Leontovsky in Leningrad. Putting together many negatives he
was able to do photometry of M3l to fainter magnitudes than was
possible photoelectrically. The method of preliminary baking of
emulsions is more effective when it is done in a nitrogen atmosphere
and still better in hydrogen. Spectacular results with the last method
were obtained by A. Smith (1977). With a 70m exposure on IIIa-J plates
there are no traces of the planetary nebula the Helix. After hydroniz-
ation the same exposure reveals the overexposed image of this object.
Arp and Lorre (1976) obtained striking results on IIIa-J emulsions
using the process of deconvolution which improves the negatives. By
means of some kind of filtration, removing faint stars and those of
medium brightness etc., they improved the resolution of the extended
images and obtained a better contrast. The authors published their
photographs of the Stefan group and of the jet in M87. As a result of
deconvolution it becomes evidence that NGC 7320 is much nearer to us
than its apparent neighbouring galaxies because it is clearly resolved
into HII regions and star clusters.

One has to be extremely cautious about the detection of faint

details on negatives which have suffered strong intensification. The
procedure also intensifies spurious details. Can you imagine what
WQuld have happened if Arp and Lorre forgot to warn the reader that the
dark rings around all condensations in the jet of M87 are but the
result of their processing?

2. THE APPLICATION OF PHOTOGRAPHY TO GROUPS

The methods of colorimetry and photographic photometry through

colour filters, which enable one to discern stellar populations and the
distribution of ionized hydrogen has been used very seldom for groups.
3
M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 3-10. All Rights Reserved.
Copyright © 1978 by the fA U.
4 B. A. VORONTSOV-VELYAMINOV

The evaluation of their distances from the apparent diameters of HII

regions can be used. Unfortunately the total dispersion of their
linear diameters is very much greater than the value found by Sandage
and Tammann if larger volumes of the Metagalaxy and interacting
galaxies are considered.

In determining distances to the Stefan group, Arp used an average

value for the diameters of HII regions of 200 pc for all its members.
However Vorontsov et al. (1974) have drawn attention to the fact that
in the interacting pair NGe 2535-6 the largest HII regions exceed 2000
pc in length. These measurements were made on a photograph taken by Arp
in the light of H8 and we adopted the cosmological redshift which Arp
himself adopted for this object. Our paper, which discussed the
Stephan group, was never published. The referee rejected its publica-
tion advancing arguments of the kind: "the authors apparently mistook
for HII regions in this galaxy the images of the bright field stars far
from NGe 2535" and so on.

It would be desirable to evaluate the masses of the members of

groups from their luminosities and morphological types. However even
the integral magnitudes and the colours in most cases are only known
very crudely or are unknown altogether. Estimates of morphological
types are often uncertain due to the peculiarity of the objects or to
overexposure. It is most difficult to differentiate apparently
spheroidal objects.

The problem of membership of groups is more complicated. The most

recent studies show that apparent proximity alone does not guarantee a
physical relationship. The criterion formerly considered to be best -
the similarity of redshifts - suffers from the lack of red shifts. But
even this criterion is violated by Dr Arp because he believes in very
large non cosmological red shifts and he gives us no optical criteria
for deciding which objects must have non cosmological red shifts
(excluding QSS).

In the group VV 166 one apparent member has red shift 1000 km s-1
smaller than that of the other members. Kormendy and Sargent (1974)
argue that it is a chance projection because it shows no interaction
with its spiral neighbour. But this argument is not convincing because
this elliptical galaxy can only respond feebly to tidal forces. The E
galaxy NGe 71, an undisputable member of this group, does not show
deformations due to its neighbours.

The sole optical peculiarity of E galaxies was our finding

(Vorontsov et al. 1962) that some giant E galaxies (gE), some of them
mUltiple, some in clusters and some outside, have large coronae and are
best denoted by gEH, H meaning Halo or Haze. Their discovery has been
ascribed falsely and persistently in the West to W. Morgan. He denoted
them by cD only three years later. Besides it introduces confusion,
since by D he also means "disk galaxies".
THE PHOTOGRAPHY OF GROUPS OF GALAXIES

The discovery of interacting galaxies in 1958 drew attention to

the importance of taking photographs, even on a small scale, with long
exposures, in order to study these enigmatic bridges and tails. The
Palomar Sky Atlas made it possible to identify, describe and measure
nearly 2000 such objects in the MCG (Morphological Catalogue of
Galaxies). The discovery of extradense groups - nests of galaxies -
indicates the next task: to obtain photographs which might discern the
individual peculiar and irregular objects the images of which have
coalesced or are in contact. This requires large scale and exposures
with a light ratio 1:4 of 4 to 1 minute to resolve the number of nuclei.
This has not been done so far, but in this way a peculiar object can
transform under our eyes into several galaxies. We are convinced that
such a break down in fact happens.

Having little time, I will show a few slides to illustrate how a

series of exposures presents "cuts" at various levels of a galaxy. In
my Atlas (Vorontsov 4 and 5) of interacting galaxies I stress that from
the study of overexposed images of the Palomar Sky Atlas I can only
suspect that a given object is not single, but is a tight system.
Rephotography of such objects by Dr Arp (1966) and by the observers at
the 6-m telescope confirmed this susp~c~on in most cases. There still
exists the uncertainty introduced by the chance projections of fore-
ground stars and of very compact galaxies. I will show first some
nests, then chains (a particular case of nests) and cases where the
structure of the objects is enigmatic, let alone their origin. Below
are shown some of the VV objects, discovered in 1958-1964, and recently
photographed with the 6-m telescope by I. D. Karachentsev, M. F.
Shabanov and J. P. Korvjakovsky. For permission to show them I also
thank cordially the Director of SAO Dr I. M. Kopylov.

VV 644. In compiling the MCG in 1961, I described this object as

two galaxies in a close interaction, nearly coalescent. The 6-m
photograph by Dr Karachentsev shows that in fact it is a nest of two
pairs of compact galaxies. Both consist of a compact elongated galaxy
in contact with a compact elliptical. A 5th member of this nest is
suspected. No bridges or tails are present. Markarian put this
"object" on his list of blue galaxies under catalogue number 8. Only
this made D. Weedman measure its properties: U-B = -0.52 and absolute
magnitude Mv = -21. The components have Mv from -19 to -18.

VV 556. A nest of two groups. The first consists of a minichain

(3 members in contact). The other consists of a large diffuse object
including a spheroidal galaxy and a bridge to a galaxy with a tail.

VV 568. This seems to be a nest of galaxies in a common envelope.

There are two bright E and three faint IrI galaxies. Objects of such
appearance are called "clumpy irregulars" by Heidmann and Casini. The
irregular IrI galaxies have no such compact condensations. We expect
that nests of dwarf irregular galaxies must be more frequent at the
present late stage of evolution of the Metagalaxy. They represent a
late stage of fragmentation of galaxies.
6 B. A. VORONTSOV-VELYAMINOV

VV 600. This is a complicated mixture of arms and filaments. One

spiral arm is seen edgewise next to small galaxies connected by a fila-
ment. In addition there are 7 faint, nearly starlike objects. The
three brightest of them probably form a chance projection of a triple
star, but this must be checked spectrosopically.

VV16 = NGC 5996-94. At first glace, it might be considered to be

a single SB galaxy, but·... To the right of the bar there is an obj ect
akin to a nest of 5 dwarf but compact galaxies in contact, possibly far
away. Note the peculiar structure of the bar (3 parts detached), the
great asymmetry of the matter in the arms and a possible connection
with the smaller early spiral to the left. Is it not all the late
stage of fragmentation of a single system?

VV 470. This is not a minichain, but a triple galaxy of "twice

MSl" type. The central object is a spiral galaxy seen nearly edgewise.
Near its plane it is connected with two symmetrical and symmetrically
situated companions, each with a strong tail. The ratio of masses
here could enable one to construct a tidal theory akin to the one so
ingeniously compiled for MSl by the Toomres, separately for every
companion. However the simultaneous close encounter of two similar
companions moving in different planes with nearly parabolic velocities
is most improbable. The improbability is of much higher order than
that of a chance configuration of sprayed grains. Of 160 systems of
MSl type which we have found, over ten belong to the "twice MSl" type.
Some of the components at the end of spiral arms are dwarfs of the IrI
type, or large HII regions. In two cases their location on a single
arm was spectroscopically recognized by Arp (1970).

VV 20. In this system a pair of E(?) galaxies has a faint tail to

the left, possibly produced by tides. In addition a very strong spiral
bridge stretches from them to the big E-galaxy below. We find in it
3-4 condensations - a minichain in a common envelope. But we stress
the impossibility of extracting so massive a formation from these tiny
galaxies. Besides the latter are also connected to the big E-galaxy by
a straight bridge. In the case of VV 19 too,as well as connections by
means of two arms, the galaxies are connected by a straight bridge as
is found in the cases of M5l and the others. Such bridges are not
predicted by the tidal theory.

VV 523 = NGC 3991 = Haro 5. On the overexposed photograph in the

PSA I mistook this object for a "cudgel" - a single body. Proper
exposures showed that this blue object is a multimembered chain. Six

Page opposite: The pictures present enlargements to 2 arc sec per mm

of the photographs of some interacting galaxies obtained at the prime
focus of the 6-m telescope. North is at the top, West to the right;
otherwise, they are shown by arrows. The numbers correspond to the VV
numbers of the objects. For details, see text.
THE PHOTOGRAPHY OF GROUPS OF GALAXIES 7

568
-w

470

•
if,

,
20 551
..

'~1
4970.. 497b
.*

261 b
8 B. A. VORONTSOV-VELYAMINOV

of the galaxies are in contact and the largest is somewhat apart, but is
connected to the rest. In total M = -21, the members having Mv = -19
and less. They are compact with strong emission lines. According to
Page the difference of velocity between the largest member and the rest
is 265 km/s.

VV 497. The overexposed image of this system on the PSA appears

as three round galaxies connected by short filaments with a blue haze to
the south. The photography with the 6-m telescope transforemd this
mini chain into a multimembered chain, while one of the members was
shown to be a star. As in VV 523 the brightest member, after a gap
containing a thin bridge, begins a chain of 4 or more members in
contact. They are dwarfs of the IrI type. Below, in the haze, there
are a number of faint dwarf members or "isolated HII regions", so that
the massive chain is the main part of a blue nest.

VV 528. Let us compare two slides. On the first slide this system
is a galaxy with two concentric systems of spiral arms and a "jet" with
an amorphous condensation. The shorter exposure shows the object to be
more enigmatic. One sees two superimposed circles with perfectly sharp
edges and a "bridge" leading to a companion. Now it looks less like a
jet. Is it a central projection of a double star or a pair of extremely
compact galaxies with spiral arms?

VV 621. Two "non-flat" galaxies are connected by a short "bridge"

and have unusually bright tails. Is this a chance projection on an
object of much larger dimensions but of faint surface brightness? The
main part of the system consists of an irregular ring with condensations
and a hole in the centre. The ring and its appendages are very differ-
ent from the 2-3 other rings known without nuclei. Differential
spectroscopic observations are needed.

VV 261. "Homunculus", a "man". The shorter exposure shows that the

main body probably consists of two coalescent, very compact spheres and
there is a jet toward the "hat" with a condensation at its end. This is
a supergiant with M = -21.1 and U-B = -0.25, included in Markarian's
lists as number 201.

VV 33 = NGC 5216-18. On the first slide the usual photograph of

this system is shown: below, to the left there is apparently a spiral
galaxy A connected by a very long and very thin bridge with a round
galaxy B. From B emerges a bright spiral tail. I have already mention-
ed this system as a difficulty for the tidal theory of bridges and
tails. Now on the second slide it is evident that B really is an E, and
not an SO galaxy with the flat component which is necessary for the
formation of a spiral tail by means of tides and rotation. Besides it
can be seen that the tail emerges from the main round body and not from
its periphery. Object A provides a further surprise. In fact it is not
~ single spiral galaxy. Instead it is an elongated formation, irregular,
probably breaking up into smaller ones. The main "body" consists of
THE PHOTOGRAPHY OF GROUPS OF GALAXIES 9

3-4 coalescent compact galaxies. They form an excellent nest. The

nest is in contact with the elongated body. From the point of contact
two spiral arms emerge from the nest. To the right above there is a
nebulous condensation and two faint "Brackets" - above and below. The
long bridge toward galaxy B emerges not from the bodies described
above, and not from the spiral arms, but from one of the outer
"brackets". The very complicated and amazing structure of B and the
details of A cannot be accounted for by a purely mechanical picture.

Our knowledge of the nature of galaxies is still incomplete, as

these examples show. They are as important for the understanding of
galaxies as the phenomena of New stars and Supernovae are for the
understanding of stars. It is a shame to neglect any longer the
systematic study of nests and chains o~ galaxies.

After Dr Toomre's lecture it will be shown that our calculations

rule out the chance collisional origin of pure ring galaxies. The
same holds true in connection with the interpretation of dividing pairs
as chance mergers, even if very eccentric closed orbits are supposed.
It is not possible to set aside tight nests and chains of galaxies
because they exist and cannot be considered chance coincidences.

We have witnessed efforts to explain the Sun's energy by collisions

with meteorites, the outbursts of Novae by collision with a planet,
Supernovae by collisions of stars, spiral nebulae by collisions of
stars, planetary systems from the spirals thus formed. These theories
are now discredited. The same fate awaits the hypotheses of collision
of galaxies, as it has failed to explain radiogalaxies.

REFERENCES

Arp, H., 1966, Atlas of Peculiar Galaxies, Astrophys. J. Suppl. 14.

Arp, H., 1970, Astron. & Astrophys., 3, 418.
Arp, H. and Lorre, Z., 1976, Astrophys. J., 210, 58.
Korrnendy, J. and Sargent, W., 1974, Astrophy~J.,193, 19.
Smith, A., 1977, Sky and Telescope, 53, 24. ---
Vorontsov-Velyaminov, B., 1959, "Atlas and Catalogue of Interacting
Galaxies, I", Moscow University.
Vorontsov-Velyaminov, B. and Krasnogorskaja, A., 1962, "Morphological
Catalogue of Galaxies, I", Moscow University.
Vorontsov-Velyaminov, B., Efremov, J. and Kornberg, B., 1974, Preprint
of the Inst. Appl. Mathematics, Moscow.
Vorontsov-Velyaminov, B., 1977, "Atlas of Interacting Galaxies, II",
Astron. & Astrophys. Suppl., ~, 1.
10 B. A. VORONTSOV-VELYAMINOV

DISCUSSION

Heidmann: I would like to emphasise what Dr Vorontsov-Vel'yaminov said

about the importance of measuring radial velocities for his nests in
order to find their absolute magnitudes. According to their intrinsic
luminosities, they may be either classical resolved irregular galaxies,
or giant irregulars of the clumpy type pointed out last year by Casini
and myself or tight groups, i.e. nests, of galaxies.
VELOCITY DISPERSION IN SMALL SYSTEMS OF GALAXIES

I. D. Karachentsev
Special Astrophysical Observatory
Academy of Sciences, USSR

1. GENERAL REMARKS

During recent years the problem of the virial paradox has grown
less acute for many galactic systems, especially for those having few
members. As an example one can take the revision of vi rial mass-to-
luminosity ratio, f = MIL, for de Vaucouleurs' groups performed by
Materne and Tammann (1974). The reduction of the f-estimate is due to
various reasons: an increase in accuracy in measuring radial veloci-
ties of galaxies, especially noticeable for 21-cm surveys of groups
(Fisher and Tully 1975), regard for the sub-structure of systems of
galaxies and improved exclusion of accidental members of groups
(referred to here as "optical" members). Some very scattered systems
proved to be low-contrast fluctuations of the expanding background, and
not physical groups in their usual sense.

The most noteworthy recent feature has been the transition from
episodic virial mass estimates for some selected systems, using data of
different quality by different authors, to planned programmes of
regular investigation of certain kinds of systems of galaxies. I would
like to emphasize the following advantages of such an approach: a
well-defined selection criterion of systems allows one to estimate
quantitatively selection effects and the presence of optical members in
each sample; the uniformity of radial velocity measurements decreases
the probability of errors encountered in heterogeneous compilations of
data.

2. PAIRS OF GALAXIES

Being the simplest type of system, pairs are most suitable for
investigation of the virial paradox. At our Observatory, in coopera-
tion with other astronomers, a programme of systematic radial velocity
measurements has been started for double galaxies. The catalogue of
603 isolated pairs by Karachentsev (1972) has been taken as a basis.
By now, relative motions have been measured only for 258 binaries.
11
M. S. Longair and J. Einasto (eds';. The Large Scale Structure of the Universe. 11':'20. All Rights Reserved.
Copyright © 1978 by the fA U.
12 L D. KARACHENTSEV

That is why inferences made from the observations available at this

stage should be regarded as preliminary.

I would like to dwell on two problems connected closely with the

viria1 mass interpretation of double galaxies: the radial velocity
measurement errors and the presence of optical pairs in the sample.

Figure 1. The distribution of

radial velocity difference for 258
double galaxies. The solid line
and the dashed line show the
distribution of measurement errors.

In Figure 1 is presented the distribution of the radial velocity

difference for 258 double galaxies. Histogram of the I~vl-distribution
has a maximum near zero and also an extended tail up to 1000 km/s. The
solid line drawn through the crosses shows the distribution of measure-
ment errors, o~v. As can be seen the radial velocity differences for
many pairs are comparable with the errors of measurement. Because I~vl
does not exceed 100 km/s for half of the sample, to obtain reliable
information on the kinematics of double galaxies one evidently needs
higher accuracy for the radial velocities. The dashed line in Figure 1
shows the distribution of errors for 63 pairs observed with the 6-metre
telescope using the dispersion ~ 90 R/mm. As the analysis shows, it is
highly desirable to repeat observations of some pairs with higher
dispersion.

For a pair of galaxies with an integral luminosity L, projected

distance between components X, and an orbital eccentricity e, the
unbiased estimate of the orbital mass-to-luminosity ratio is
-1 -1 -1
f M/L 32/3TI(1-2e 2 /3) y X(~V2-6~v) L ,
c
where y is the gravitational constant, and the numerical coefficient
reflects the supposition of an isotropic orientation of double
galaxies. For definiteness let us assume that the pair's components
have strictly circular motions (e = 0), in which case the f-estimates
will be a minimum.

The integral distribution of fc for 258 pairs is presented in

Figure 2 by dots. A logarithmic scale has been chosen to show the wide
range of fc observed. To show the role of radial velocity errors, open
VELOCITY DISPERSION IN SMALL SYSTEMS OF GALAX IrS 13

100

,80
•
D
. ., .. eo

•
. • D

,.
,GO
Figure 2. The integral distribu-
0
~~ tion of the mass-to-luminosity
.... • •
~
0 ratio for 258 pairs. Dots mark
~
~
0 the unbiased estimates; open
2D
0 circles denote the biased estimates.

00
a s ,/10 SlID fII~

".f
fO 10 S6 {(JO
111/"

circles mark thQ distribution of the biased estimate f, calculated for

a~V = O. From these data it can be seen that more than half of the
pairs have fc < 5 f g ; however for 10% of the pairs the mass-to-
luminosity radio exceeds 100 f@. The interpretation of large values
f ~ (10 2 + 105)f 0 is apparently connected with the presence of acciden-
tal optical pairs in the sample.

Table 1

f < 103f 0 f < 102f (')

Sample of
n
CPG-pairs tot <f>
<f> <fc> n <fc> n

Data by different
151 45 33 149 19 8 137
authors
Common with
43 18 9 42 18 9 42
Turner's sample
Observed with the
44 39 31 41 12 9 37
5-metre telescope
Observed with the
63 16 14 59 10 8 58
6-metre telescope

Total 258 37 28 249 16 8 232

Let us consider the inferences made from the data of Table 1. In

this table are presented the mean mass-to-luminosity ratios for different
samples of double galaxies from the Catalogue. The first line collects
together data based on published estimates of radial velocities, often
the radial velocities of different components for one pair being
measured by different authors. The second line contains the data for
43 pairs in the Catalogue common with Turner's (1976) sample. In the
following lines are presented the results of recent 5-metre telescope
14 I. D. KARACHENTSEV

observations (Karachentsev, Sargent and Zimmermann 1977) and those

obtained with the 6-metre telescope (Karachentsev 1977). Alongside the
unbiased means <f c >, in the table there are presented the biased ones,
<f> assuming a~v ~ O. The left part of the table contains the means
calculated after exclusion of undoubted optical pairs with fc > 103f®;
for the right part of the table the optical pair criterion has been
taken to be fc > 10 2 f 0 •

From these data the dependence is seen of the mean orbital mass-to-
luminosity ratio on both the accuracy of the radial velocity measure-
ments and the presence of optical pairs in the sample. Comparison of
the last column data shows that irrespective of the sample the mean
values, <f c > ~ (8 t 9)f@, are practically the same and correspond to
normal mass-to-luminosity ratios in the absence of massive coronae.

The role of optical pairs in the interpretation of <f>-values has

been repeatedly emphasized by Fessenko (1976). According to preliminary
calculations the criterion for isolated pairs used in the Catalogue
allows for the presence in the sample of about 10% of the optical pairs.
This agrees quantitatively with the supposition that all 26 pairs,
having fc > 100f@, out of the total number 258, are optical pairs. The
correlation of apparent magnitudes and radial velocities for these
pairs gives further evidence for this (see Figure 3). The members of
each pair are marked in the figure by dots linked by straight lines and
the dashed lines denote a 90% confidence interval for a distribution of
field galaxies relative to mean dependence m ~ 5 logV. As expected for
optical pairs their vectors are oriented mainly along Hubble's relation.

Figure 3. The correlation

of apparent magnitudes
and radial velocities for
optical pairs.

Considering double galaxies for the components of which radial

velocity differences have been measured repeatedly by different
authors, one finds rather a large number of cases in which the 6V-
estimates are poorly reproduced. For instance, according to RCG the
VELOCITY DISPERSION IN SMALL SYSTEMS OF GALAXIES 15

velocity difference for the interacting pair of spirals ePG 295 = VV 228
= Arp 294 is ~V = (410 ± 148) km/s. New measurements give the value
~V = (60 ± 22) km/s, which reduces the former high estimate fc = 77f@ to
a normal value, 1.7 f@, for spiral galaxies.

There exist a number of effects which may lead to noticeable errors

in measuring radial velocities. Here are some of them:
(1) Blending of absorption lines in galaxy spectra by the night-
sky lines (Simkin 1977).
(2) A possible non-coincidence of the mean velocity of a galaxy
measured from emission and from absorption lines when the object has
large non-circular internal motions.
(3) A composite structure of a number of tight interacting pairs
in which no distinct nucleus is observed as the dynamical centre of the
components. Examples of such pairs have been given by Tifft (1977).
Spectra obtained with scale 8~6 per mm at the slit show that the
splitting of the emission lines is caused by discrete structure of the
emitting clouds.
(4) Some residual effects of image-tube distortion.
(5) Different estimates by different authors of the radial velocity
corrections due to the motion of the Sun and the Earth.

These errors may lead to a total value 0v ~ 50 km/s, and may lead
to systematic overestimation of virial masses, especially for wide
pairs. So, the error in velocity of each component ~ 50 km/s gives a
fictitious value f ~ 80 f for a pair having L = 1 x 10 10 L0 and
X = 200 kpc. Obviously, rf these effects are ignored, evidence may be
found for unobservable coronae in double galaxies.

3. SMALL GROUPS OF GALAXIES

The exclusion of optical members in groups of galaxies is a more

difficult task in comparison with pairs. There are systems, for
example, Stephan's Quintet and Seyfert's Sextet, where the radial
velocities of one or two members differ considerably from the rest.
This leads to a large M/L - value for a system. Exclusion of such
members reduces the virial mass estimate by a large factor.

As another similar example I present the sextet of galaxies in

Cetus in which there is a large redshift difference. This system has
been studied by Karachentsev and Tifft (1977). A reproduction of this
group from the Palomar Sky Survey print is presented in Figure 4. The
radial velocities of galaxies are marked. As can be seen, the radial
velocities of the two northern galaxies (C and D) differ from the rest
by ~ 1000 km/s. The dynamical parameters for the group are shown in
Table 2. The notations of values in it are generally accepted. From
data in the table it follows that the hypothesis of accidental projec-
16 I. D. KARACHENTSEV

Figure 4. The group

of galaxies in Cetus
with a large red-
shift difference.
The reproduction is
from the Palomar Sky
Survey print. Radial
velocities of the
galaxies are marked.

Table 2

-1 -1
<Vo> <IW 2 >! <R· .> L f fc
Combination r 1J
km/s km/s kpc L@ f(\) ffi)

A+B+C+D+E+F +6919 411 70.8 3.72 10 10 848 835

A+B+E+F +7206 82 99.8 3.47 10 10 56 36

C+D +6345 2 14.1 0.54 10 10 0.03

tion of the pair (C and D) on the quartet (A,B,E,F) eliminates the

virial paradox. In spite of the fact that the group in Cetus is
apparently isolated the mathematical expectation of a number of optical
configurations similar to it is n ~ 1 among galaxies contained in
Zwicky's catalogue.

The presence of interacting features in groups of galaxies reduces

considerably the probability of finding optical members in them. A
rich collection of such systems is included in the two parts of the
"Atlas of Interacting Galaxies" by Vorontsov-Velyaminov (1959,1977).
The Moscow State Institute of Astronomy in cooperation with Special
Astrophysical Observatory has started a programme of spectral investi-
gation of different types of interacting systems with the 6-metre tele-
scope. Examples of spectrograms of several VV-objects are presented in
Figure 5. From the results of the first observations (Arkhipova et al.
1977) the following parameters have been estimated for 23 interacting
VELOCITY DISPERSION IN SMALL SYSTEMS OF GALAXIES 17

Figure 5. The spectrograms

of interacting systems:
(a) VV 620, (b) VV 394,
(c) VV 615. The original
dispersion is ~ 90 ~/mm and
the scale on the slit is
8'.'6 per mm.

systems classified by Vorontsov-VTlyaminov as "nests", "minichains",

etc., - radial velocities, <~V~>2, the effective radius R, the virial
mass M, and the luminosity L. The basic data for these systems are
presented in Table 3. The mean virial mass-to-luminosity ratio,
<f c > = (6.9 ± 2.9) fe' with H = 75 km/s Mpc, shows that in the sample
optical members are practically absent. For a wide range of sizes
(0.2 kpc ~ 30 kpc) and luminosities (6 x 10 7 - 1 x lOll) La of inter-
acting systems their mass-to-luminosity ratios are approximately the
same.

4. CONCLUSION

The data presented show that close interacting systems, and also
pairs of galaxies both tight and wide have on the average a normal virial
mass-to-luminosity ratio <f c > ~ 8 f 9 , which can be explained without the
hypothesis of the existence of maSS1ve hidden coronas around galaxies.
18 I. D. KARAC'HENTSEV

Table 3

~
<Vo> <tJ.V2>
r -+ R M/l0 10 L/l0 1U f fc
0 °tJ.V
n VV
km./s km/s kpc ~ Le fE) fE)

1 587 1808 6 ± 23 27.4 0.078 0.83 0.1 -1.3

2 587E 1805 ~ 65 1.0 0.97 0.11 " 8.8
3 555 4320 39 ± 20 7.5 3.1 0.52 5.9 4.4
4 699 4407 31 ± 20 30.8 9.7 2.5 3.8 2.2
5 644 3714 123 ± 20 3.0 2.9 1.7 1.7 1.6

6 141 2751 46 ± 30 4.8 0.8 0.31 2.6 1.5

7 761 12024 71 ± 35 26.6 44.0 4.05 10.9 8.3
8 541 2568 75 ± 20 7.9 3.5 2.16 1.6 1.5
9 620 1659 52 ± 44 17.7 3.8 0.63 6.0 1.7
10 794 575 ~ 20 2.2 0.19 0.010 ~ 19

11 498 1319 23 ± 20 1.8 0.31 0.059 5.2 1.4

12 457 1292 28 ± 20 2.0 0.34 0.14 2.4 1.2
13 126 10427 57 ± 30 25.1 6.4 11.1 0.6 0.4
14 523 3100 116 ± 25 4.9 5.3 1.1 4.8 4.6
15 497 888 ~ 20 2.8 0.25 0.027 ~ 9.3

16 261 2583 6 ± 30 3.2 0.01 1.42 0.0 -0.2

17 432 -245 79 ± 30 2.3 1.14 0.068 16.8 12.4
18 558 188 ~ 30 1.9 0.034 0.024 ~ 1.4
19 48 3570 34 ± 26 26.4 2.18 5.44 0.4 0.2
20 339 7253 240 ± 20 38.3 174. 2.86 60.9 60.5

21 394 438 ~ 15 0.2 0.010 0.0065 ~ 1.5

394+ 19.0 0.79 2.36 0.3 0.2
22 426 23 ± 15
MI0l
23 615 2516 93 ± 20 4.0 11.4 0.63 18.1 17.2
VELOCITY DISPERSION IN SMALL SYSTEMS OF GALAXIES 19

REFERENCES
Arkhipova, V.P., Afanas'ev, V.L., Dostal, V.A., Karachentsev, I.D. and
Metlov, V., 1977 (in press).
Fessenko, B.I., 1976. Soviet Astron. Zh., 53, 112.
Fisher, J.R. and Tully, R.B., 1975. Astron:-& Astrophys., 44,151.
Karachentsev, I.D., 1972. Catalogue of Isolated Pairs of Galaxies in
Northern Hemisphere, Soobsch. Special Astrophys. Obs., 7, 3.
Karachentsev, I.D. and Tifft, W.G., 1977. Astron. & AstrophYs.
(in press).
Karachentsev, I.D., Sargent, W.L.W. and Zimmermann, B., 1977 (in press).
Karachentsev, I.D., 1977 (in press).
Materne, J. and Tammann, G.A., 1974. Astron. & Astrophys., ~, 383.
Simkin, S.M., 1977. Astron. & Astrophys., 55, 369.
Tifft, W.G., 1977. Astrophys.J., 211, 31. --
Turner, E.L., 1976. Astrophys. J.:-208, 20.
Vorontsov-Velyaminov, B.A., 1959. Atlas and Catalogue of Interacting
Galaxies, I, Moscow State Univ., Moscow.
Vorontsov-Velyaminov, B.A., 1977. Atlas of Interacting Galaxies,
Astron. & Astrophys., Suppl., 28, 1.

DISCUSSION

Zeldovich: What is the value of MIL for our own galaxy?

RaPaahentsev: About 5 to 10 f e •

Os tri ker:What is the mean projected separation of your large sample of

binary galaxies?

Karachentsev: 30 kpc assuming a Hubble constant H = 75 km s-1 Mpc- 1 .

Ostriker: In that case the mass you find [(MIL) x ~1 is just what
would be expected, even if there were a great deal of mass at distances
of 100-1000 kpc from the galaxy centres.

Karachentsev: But I find the values of MIL to be independent of the

separation of the galaxies including those with separations> 100 kpc.

Zasov: Is there any difference between the values of MIL for pairs of
elliptical and pairs of spiral galaxies?

Karachentsev: Yes. For pairs of ellipticals MIL is about twice as

large as it is for spiral pairs.

Audouze: What is the value of the declaration parameter that you deduce
or favour from your analysis? I guess that it is a very small one.

Karachentsev: The low mass-to-luminosity ratio for pairs and tight

groups does not contradict any value of qo in the range 0.01-0.2 because
20 I. D. KARACHENTSEV

the bulk of the hidden virial matter may be associated with much richer
systems.

Jaakkola: The virial mass problem is dependent on the real nature of

redshifts. Specifically, I refer to the usual practice of assuming
that a positive magnitude-redshift relation for possible members of
double systems, groups and clusters is taken as evidence for these
systems being chance projection. However, such a relation is also
expected starting from the overall phenomenology of the non-velocity
redshifts. For example, if associations between some quasars and bright
galaxies are physical ones, there is an intrinsic (m,z)-relation in the
system. Such relations appear to be a common effect in systems of
galaxies (Jaakkola, Proc. III European Meeting 3 Tbilisi 1975, p.2l8).
The effect is also present in pairs and groups of Zwicky's galaxies
(Jaakkola, Astrophys. Space Sci. 49, 99, 1977). In the latter systems,
there are in most cases evidence of physical interactions and hence the
relation is not due to projection effects.
OPTICAL S'.lUDIES OF SMALL GROUPS OF GALAXIES

Edwin L. Turner
Harvard College Observatory

INTRODUCTION

Galaxies occur in a wide variety of systems ranging from binary

pairs through small groups to rich clusters. These systems in turn
possess a wide range of densities, with typical separations between
bright (L? L* = 3.4 x 10 10 L0) galaxies varying from~ 10 kpc up to
~ 1 Mpc. Among the most common of these systems are small, loose
groups containing..6 10 bright galaxies with separations 100 kpc. :r
Such systems probably contain a substantial fraction of all galaxies
(de Vaucouleurs 1975; van den Bergh 1962; Karachentseva 1973). Familiar
examples include the Local Group and M8l group.

In this paper, the main results of a statistical study of small

groups (Turner and Gott 1976a, 1976b; Gott and Turner 1977a, 1977b;
hereafter TGI, TGII, GTIII, and GTIV, respectively) are reviewed and
compared to N-body simulations of galaxy clustering.

CATALOG OF GROUPS

The problems in compiling a catalog of small groups arise both

from the uncertainty in any particular group's membership and from the
difficulty in consistently identifying each group's existence.
De Vaucouleurs (1975) has suggested that such groups might be suitably
defined as enhancements in the volume number density of galaxies and
might be identified as enhancements in the surface number density of
galaxies on the sky. Many group catalogs (de Vaucouleurs 1975;
Holmberg 1937; Sandage and Tammann 1975) have been based on a detailed,
but somewhat subjective, consideration of a variety of data (e.g.,
redshift, position, magnitude, appearance) concerning the candidate
galaxies.

In TGI, a new catalog of groups is presented; this catalog, in

contrast to earlier ones, has been generated by the "blind" application
of a precisely defined group identification procedure. This procedure
21

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 21-29. All Rights Reserved.
Copyright © 1978 by the fA U.
22 EDWIN L TURNER

only considers the positions of galaxies in the sky. As a r€'~lt, it

sometimes makes absurd "mistakes" (e.g., assigning a dwarf spheroidal
member of the Local Group to the same group as a galaxy with cz =
4000 km s-l), but these are usually too obvious to be misleading. In
addition, the shortcomings of the groups defined by our naive method
are offset, we feel, by their objectivity (no unconscious observer
biases), homogeneity, and completeness. These attributes are critical
in any statistical study of group propertieso

The sample of galaxies to be searched for groups is defined by:

o> 00 ,

b II > 40 0 ,

m > 14 0, 0 (1)
pg
with all positions and magnitudes taken from the Catalog of Galaxies
and Clusters of Galaxies (Zwicky et ale 1961-1968, hereafter CGCG).
The sample contains 1087 galaxies. This sample is likely to be quite
homogeneous and complete since the CGCG extends well beyond each of the
three limits (1).

The following group identification procedure has been applied to

the sample defined by the limits(l):

1.1 For each galaxy ~~<~~e sample, we consider the surface density
a(8) = 7N2 «8)/(1 - cos8)~~, (2)
IT - TI6

where N(~6) - 1 is the number of galaxies within an angular distance 8

of the galaxy being considered.

2. For each galaxy, we then choose the largest possible angle 6 c

such that
0(6 < 6 ) > fa, (3)
- c -
where 0 is the mean surface density of galaxies in the sample (594
galaxies per steradian for our sample) and f ~~ a surface density
enhancement factor. Here we have used f = 10~3 in hopes of identifying
groups with volume density enhancements >10 as suggested by de Vaucouleurs
(1975)0 For computational reasons 6 has only been determined to an
accuracy of 0 0 .25. c

3. For any galaxy with N(~ Bc » 1, a circle of angular radius 6 c

centered on the galaxy is drawn on a map of the sky. Galaxies whose
nearest neighbor is more distant than ~(TIf0/2)-~2 (about 0 0 .75 here)
have N(~6c) = 1 and have no circle drawn about them.

4. When steps (1) through (3) are completed for each galaxy in
the sample, a map of the sky showing all of the resulting circles is
prepared. The circles fall into many (103) distinct (i.e., nonover-
lapping) clumps; each clump contains from two up to ~200 overlapping
OPTICAL STUDIES OF SMALL GROUPS OF GALAXIFS 23

circles. The outside boundary of each clump of circles roughly approxi-

mates an iso-surface-density-enhancement contour; that is, the mean
surface density of galaxies within the boundary is ~fij. Each of these
distinct clumps of circles is identified as a separate group with a
boundary defined by the perimeter of the region of overlapping circles.

5. All galaxies lying within a particular group's boundary are

considered (at least tentatively) to be members. Any galaxy lying
outside all of the group boundaries is considered a field galaxy and
not assigned to any group.

A total of 737 galaxies are assigned to groups, and 350 to the

field. It should be noted that although the procedure was designed
to locate loose groups, it also identifies large clusters, binary pairs,
and generally any system which has a surface number density of galaxies
>fa. All of these systems will hereafter be referred to as groups.
This sample of small groups is well suited to statistical analyses
because it is complete, well defined, and statistically homogeneous.

LUMINOSITY FUNCTION

Of the 103 groups identified in TGI, 63 have one or more

members with measured radial velocitieso Taking the mean radial
velocity of each group (Table 3 of TGI) to indicate its distance, a
determination of the individual group luminosity functions is possible.
However, because most groups possess rather few members, these
individual luminosity functions are not very informative. Therefore,
in TGII, we have combined the 63 separate group functions into a single
composite luminosity function. It should be remembered, of course,
that by using only the groups with radial velocities, some unknown
biases may have been introduced.

Before proceeding, several conventions should be specified. All

quantities are calculated with H = 50 km s-l Mpc- 1 • Unless otherwise
note, all magnitudes are from th~ CGCG. The accuracy of the CGCG
magnitudes has been verified recently in an extensive study by Huchra
(1976). Since the TGI groups all have z « 1 and b II > 400 , both
absorption and K corrections are neglected. The Sun is assumed to have
an absolute CGCG magnitude of 5.48.

Let ~l(L)dL be the observed luminosity function of the ith group,

that is, the number of galaxies in the ith group with luminosities
between L and L + dL. Also let L be the faintest absolute luminosity
which would be visible in a partigular group. We then construct
absolute luminosity which would be visible in a particular group. We
then construct the function yeLl according to

yeLl = NL- 1 41~lL)dLlf;~(L)dLl (4)

where NL is the number of groups with Lc 2 L. Suppose the brightest

24 EDWIN L. TURNER

galaxy observed in any group has luminosity L'; then the composite
group luminosity function ¢(L)dL is

¢(L> L')dL = 0, (5)

¢ (L') dL = 1, (6)

¢ (L < L' )dL = Y (L)Ir,~ (L)dL. (7)

In practice, the dL's in equations (4) through (7) are replaced by ~

log L = 0.2 (i.e., 1/2 magnitude bins), and equation (7) is solved
numerically. Equations (5) and (6) amount to a normalization of ¢(L)
at the bright end o This procedure is preferable to simply adding the
various ~(L)dL because it gives equal weight to each group. Simple
addition gives more weight to the groups with more members; if
applied to the present data, the result would primarily reflect the
luminosity function of group 57 (the Virgo cluster) alone. Although
our method could be used to determine the luminosity function of field
galaxies, the presently available redshift data (TGI) is too meager
for a good determination; rough consistency with the group luminosity
function is indicated.

A weighted least-squares fit of the data to a functional form

suggested by Schechter (1976),

¢(L/L*)d(L/L*) = ¢*(L/L*)ae-L/L*d(L/L*) (8)

yields a = -0.83 + 0.17 and M * = -20.59 + 0.260 If, for simplicity,

we constrain a = ~l, then theP~it gives M -* = - 20.85 ~ 0.13,
corresponding to L* = 3.4 X 10 10 L0. Bg~h fits give a reduced chi-
square of 0.63 and are, therefore, equally good. Since the analytic
form of equation (8) is particularly convenient if a = -1, the latter
fit is adopted.

Schechter (1976) has fitted equation (8) to a composite luminosity

function constructed from Oemler's (1974) data for rich clusters and
obtained a = -1.25 and M ( = -20.6. These values are in fairly
close agreement with the BJ£6ve results for small groups. The most
significant difference (~ 2cr) is in the value of a (slope of the low-
luminosity tail)0 It is intriguing that some of Oemler's (1974)
clusters seem to have relatively fewer low-luminosity galaxies than
others (i.e., larger a's)o

The composite luminosity function for early (E and SO) and late
(S, SB, and Irr) type galaxies were determined by the same procedure
as the total luminosity function. Fits of equation (8) yield a = -0.79
+ 0.23 and M * = -20 49 + 0.30 for late types and a = -1.27 + 0024
0

ind M * -~~.34+ 0.60 f~r early types. These results are identical
pg
OPTICAL STUDIES OF SMALL GROUPS OF GALAXIES 25

within the errors (20); but it is, again, intriguing that the a value
for early-type galaxies agrees so well with Schechter's result for rich
clusters (in which early-type galaxies are often concentrated).

The evidence for a "universal" luminosity function is sufficiently

convincing to warrant the exploitation of equation (8)'s many convenient
analytic properties in a wide variety of applications.

MASS-TO-LIGHT RATIOS AND CROSSING TIMES

In GT III, a detailed dynamical analysis of 39 TGI groups using

available radial-velocity data has resulted in the following conclusions:
The groups are characterized by typical velocity dispersions and sizes
of ~ 200km s-l and ~ 500 pc, respectively, and have typical total
luminosities of several L*. Those groups contaminated by foreground or
background objects can each be plausibly subdivided into one or more
uncontaminated groups closely resembling the originally uncontaminated
majority of the groups. For all 39 groups (uncorrected for contami-
nation), the median (a very stable estimator) value of MIL is 141.
When contamination correction and possible variation of MIL with tota~
group luminosity are taken into account, a mean MIL of 90 (corresponding
to 200 for L > 10L* and 65 for L < 10L*) is obtained. The uncertainty
in each of these values is roughly a factor of 2, and they all
correspond to ~G t 0.1. For these 39 groups, ~tHo ~ 0.1, indicating
that collapse and virialization have just occurred. Taken together
with the typical group density enhancement y ~ 950, this very crudely
implies ~ % 0.12, in good agreement with the more rigorous MIL
determina~ion. In general, the data examined here offer little hope of
closing the Universe with the mass associated with galaxies.

MULTIPLICITY FUNCTION

The spectrum of galaxy cluster sizes is a valuable cosmological

datum. In GTIV the problem is formalized by defining the multiplicity
function as the luminosity function of groups of galaxies which
satisfy a surface density enhancement criterion, a) u •
- g
The observed function is particularly simple n {L)dL« L- l dL for
L<L*, n {L)dL« L-J3dL for L* < 1< 350L*, n {L)dL ~go for L > 350 L*.
T~e breik in the function at L* (a
typical ~right galaxy luminosity) is
presumably due to astrophysical processes related to galaxy formation.

The form of the mUltiplicity function for L > L* should reflect

the initial conditions at recombination and should not depend on the
specific value of Og. Since the covariance function of galaxies is a
power law, it is reasonable to assume that the original densjty _~_
fluctuation spectrum at recombination was also a power law (1&) « M
~, Where n = 0 is a Poisson spectrum and n = -1 is the spectrum pre-
dicted by standard hot big bang cosmology (Gott and Rees 1975, Peebles
1974,Doroshkevich et al. 1974). Using the Press and Schecter (1974)
theoretical multiplicity formulae, the observed multiplicity tunction
26 EDWIN L TURNER

can be fit to give an estimate of n. The formal best fit is

n = -lo3:t0.3.

A separate analysis of the observed distribution of binaries,

triples, quadruples, etc. yields a crude estimate of n = -1.2. The
relative frequency of Local Group and Coma cluster sized aggregates
also gives a simple estimate of n = -1.4. We have also estimated the
mUltiplicity function by a completely different method, using the
nearest neighbor distribution for galaxies in the Zwicky catalogue and
the hierarchical clustering model of Soniera and Peebles (1977). Here
the best fit is n =-0.9. The latter three values are less certain than
the first but provide independent supportive evidence.

An n = -1 result finds observations andthe standard hot big bang

theory in pleasant agreement, but it must be cautioned that the
mUltiplicity function data and the Press and Schechter theory are
still subject to a number of systematic uncertainties.

COMPARISON TO N-BODY SIMULATIONS

Computer N-body simulations of cosmological galaxy clustering in a

comoving volume have been carried out by Aarseth, Gott, and Turner (1978).
These simulations reproduce many of the observed properties of the
galaxy clustering including the two point correlation function (Gott,
Turner, and Aarseth 1978). Turner, Aarseth, and Gott (1978, hereafter
TAG) have used the endpoints of these calculations to simulated galaxy
catalogs similar to the CGCG and have identified groups in these simu-
lated catalogs by the same procedure used to define the groups dis-
cussed in the preceeding sections of this paper. Examination and
analysis of these simulated groups is instructive since complete infor-
mation (position, velocity, and mass of each point) is available. TAG
show that the simulated groups represent real spatial clusters, that
l/f is a good estimate of average contamination of groups by background
and foreground objects, and that the dynamics of these simulated groups
resembles that of the observed groups. In particular, the median mass
per particle determined by a straightforward virial analysis of each
group(analogous to the M/L = 141 MG /L Q of G T III) is found to be
within a factor of two of the true mean particle mass in every case
examined. These results considerably strengthen one's confidence in
the conclusions described in the previous sections.

ACKNOWLEDGMENTS

The results described above have all been obtained in collaboration

with Svere Aarseth and/or Richard Gott. We have greatly benefited from
conversations with numerous colleagues and from the hospitality and
support of the California Institue of Technology, Harvard College
Observatory, the Institute for Advanced Study, the Institute of
Astronomy, Princeton University Observatory.
OPTIC AL STUDIES OF SMALL GROUPS OF GALAXIFS 27

REFERENCES

Aarseth, S., Gott, J.R., and Turner, E.L., 1978, in preparation.

de Vaucouleurs, G., 1975, in Stars ~ Stellar Systems, vol. 9, ed.
A. Sandage, M. Sandage, and J. Kristian (Chicago: University of
Chicago Press).
Doroshkevich, A.G., Sunyaev, R.A., and Zeldovich, Y.B., 1974, in
Confrontation of Cosmological Theories and Observational ~, ed.
M.S. Longair (Holland: D. Reidel), 213.
Gott, J.R., and Rees, M.J., 1975, Astron. Astrophy., i3, 365.
Gott, J.R., and Turner, E.L., 1977a, Ap. ~., ~, 309 (GTIII).
1977b, Ap.~., ~, 357 (GTIV).
Gott, J.R., Turner, E.L., and Aarseth, S. 1978, in preparation.
Holmberg, E., 1937, Ann. Obs. Lund, fl, l.
Huchra, J.P., 1976, ~.~., ~, 952.
Karachentseva, V.E., 1973, Soobsehch, Spets, Astrofiz. Obs., No. 80
Oemler, A. 1974, Ap. ~., ~, 10
Peebles, P.JoE. 1974, Ap. ~. Lett., ~, L51.
Press, W.H., and Schechter, P. 1974, Ap. ~., ~, 425 0

Sandage, Ao, and Tammann, G.A. 1975, ~. ~., ~, 265.

Schechter, p. 1976, Ap. ~., ~, 297.
Soneira, R.M., and Peebles, P.J.E., 1977, ~. ~., ~, 1.
Turner, E.L., Aarseth, S., and Gott, J.R., 1978, in preparation (TAG).
Turner, E.L., and Gott, JoR., 1976a, Apo ~. Supp., ~, 409 (TGI) 0
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _0 1976b, Ap. ~., ~, 6 (TGII).

van den Bergh, S., 1962, Zs. !. Ap., ~, 21.

zwicky, F., Herzog, E., Wild, p., Karpowicz, M., and Kowal, CoT., 1961-
1968, Catalog sf Galaxies and Clusters of Galaxies, in 6 vols.
(Pasadena: California Institute of Technology) (CGCG).
28 EDWIN L. TURNER

DISCUSSION

Davis: Do you assume the luminosity distribution of galaxies in your

simulations to be a delta function, and if so, what would be the effect
of using a realistic luminosity function?

Turner: Yes. The simulation represents a volume limited sample; the

data are, of course, magnitude limited. We do not expect this effect to
make a qualitative difference because the group identification procedure
guarantees the same average fractional contamination (~ l/f) for simula-
ted and observed groups.

Heidmann: Roughly are the examples of the simulations you showed repre-
sentative of the differences in groupings bet~een radial distance and
radial velocity representations?

Turner: They are reasonably representative. Essentially all of the

simulated groups show a real spatial cluster with more or less back-
ground/foreground contamination. In most cases, the reshift distribu-
tion is misleading in detail.

Zeldovich: What are the details of the simulations?

Turner: These will be given in detail by Dr Aarseth in a later paper.

Ekers: You discussed the expected bias in M/L for groups determined by
various selection criteria by comparing them with simulations using
~g >, 0.1. Presumably, simulations with smaller values of ~g would give
more contamination and consequently even more bias. Shouldn't you also
compare the observed distribution with such simulations?

Turner: Yes, examination of an ~ ~ 0.01 simulation would be useful.

They are computationally more expensive, and we have not yet produced
one. The point here is that ~ = 0.1 or even ~ = 1 models might be
taken for ~ = 0.01 situations if groups are incautiously identified in
redshift space (i.e. defined as having small velocity dispersions).

Rood: What is the mean number of galaxies in the simulated groups?

Does the spread and shape of the histogram in M/L for the simulated
groups depend on~? And if so, why?

Turner: The mean number of members is in the range of 5 to 10. The

shape is affected because the masses of groups whose velocity dispersions
are determined by background/foreground contamination does not depend on
~. The relatively uncontaminated groups have masses proportional to ~,
naturally. These two distributions combine differently for different
values thus giving rise to a variation in shape with n. The observed
M/L distribution resembles that of the ~ = 0.1 model in shape (as well
as median) more than that of the ~ = 1 model; I do not know how seriously
this shape argument should be taken.
OPTICAL STUDIES OF SMALL GROUPS OF GALAXIES 29

Fessenko: According to my calculations, Turner's clusters contain about

40% or more false members which are foreground or background objects.

Turner: The contamination fraction should be llf ~ 20 to 25%. This

number is confirmed by the N-body simulations and, to some extent, by
Kirshner's recent observations of real groups.

Holmberg: Sho'lldn' t you use the ari thmetic mean rather than the median
in estimating average values of MIL?

Turner: True, but the mean is an unstable estimator if there are a few
bad points.

Zeldovich: Kolmogorev taught me that the median is better (!lan the

mean. An example: if somebody says the time is 1 o'clock, another
that it is five past one, and a third one 5 o'clock, obviously the
median is better than the mean.

Turner: A perfect analogy! May I steal it for use on another occasion?

NEARBY SMALL GROUPS OF GALAXIES

R. Brent Tully
Institute for Astronomy, University of Hawaii
and
J. Richard Fisher
National Radio Astronomy Observatory

THE QUESTIONS

To date, our view of the universe has largely been two-dimensional.

Velocity data, the basis for a look in the third dimension, have been
too incomplete and uneven in quality to provide a clear picture. None-
theless, the pioneering work by de Vaucouleurs (1975) has given us a
rough idea of what the universe is like locally. At least a good
fraction of galaxies are improbably close to their nearest neighbours
compared with expectations based on statistical fluctuations of a
random distribution. Our vocabulary to describe these associations
includes the words: binary, group, cloud, cluster and supercluster.
Does the real universe indeed have characteristic scales that make
these terms meaningful? Or, as Peebles and his co-workers (Davis,
Groth & Peebles 1977, and reference therein) would have us believe, is
there structure on all scales, at least up to about 15 Mpc? And asso-
ciated galaxies aside, are there galaxies truly randomly distributed:
are there field galaxies? Looking two-dimensionally, it has been
possible to arrive at remarkably different conclusions. Turner & Gott
(1975) concluded that roughly 40% of all galaxies are randomly distri-
buted while Soneira & Peebles (1977) set an upper limit of 18%. It was
roughly this latter figure that de Vaucouleurs (1975) derived with his
early look into the third dimension. So we ask: (1) what are the
characteristic scales and densities of galaxy associations, and (2)
what are the scales and densities of the voids?

The next set of questions concern the dynamical conditions in what-

ever associations we determine to exist. Do we have evidence that these
entities are stable? If they would dissipate in a time short compared
with the age of the universe then the answer is yes. Then the follow-
up question: is the virial theorem satisfied with "normal" masses
assigned to the constituent galaxies, or must we resort to missing mass
or anomolous redshift explanations. Some well-known cases bracket the
range of possibilities. The rich clusters such as Coma must be bound
and there are large discrepancies between luminous and virial masses

31

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 31-47. All Rights Reserved.
Copyright © 1978 by the fA U.
32 R. BRENT TULLY AND J. RICHARD FISHER

(Zwicky 1933). On the other hand, the Local Supercluster is, at least
to a first approximation, expanding at the Hubble rate (Sandage &
Tammann 1975; Tully & Fisher 1976). Such an entity is hardly bound,
unless in a cosmological sense. In between, the small groups, again
controversy has been possible arising out of studies of very similar
data. Rood, Rothman & Turnrose (1970) found very large virial masses
implied for de Vaucouleurs' (1975) groups while ~1aterne & Tammann (1974)
found much lower values. Turner (1976) found large mass to light
ratios for binaries while Karachentsev (1976) found low ratios.

In this talk we will not be considering the extreme environment of

the rich clusters. But we will look at examples covering the gamut of
galaxy associations to be found near by. Characteristic crossing times
and virial masses will be calculated. These parameters will provide a
pretty clear indication of what is going on.

NEW DATA

Radial velocities are needed for a three-dimensional look at the

nearby universe and velocities we have. Based on a complete survey of
the Palomar Sky Atlas down to 0 = _45°, we compiled an extensive list
of objects without known redshifts with the intent of observing them in
the 2l-cm line of neutral hydrogen. The primary selection criterion
was one akin to a luminosity classification. Candidates were judged
from their structure and size to have redshifts less than 2000 km/s.
No such judgement could be made for very early-type systems which, in
any event, are weak 2l-cm emitters. So our survey was effectively only
of types Sbc and later. In addition there were two minor criteria.
There was a lower size limit of 1 arc min, and the surface brightness
had to exceed the Sky Atlas threshhold.

A check has been provided. Subsequently, we have observed all

entries without velocities in the Uppsala Catalogue (Nilson 1973) not
originally included in our survey and larger than certain limits: for
Sdm and later, 2 arc min; for Sd and earlier, 3 arc min. Many of these
additional galaxies have been detected with velocities in the range
1000-3000 km/s but only about a dozen have velocities under 1000 km/s.
These are mostly unusual cases. Hence we claim to have a high degree of
completeness for systems Sc and later, larger than about 1.5 arc min on
Nilson's scale, and velocities less than 1000 km/s.

Fortunately we have been complemented by optical observations of

mostly early systems. Sandage (private communication) has now obta~ned
velocities for all remaining Shapley-Ames galaxies brighter than 13 .
We now have available some 2000 redshifts over the whole sky out to a
cut-off of 3000 km/s. Roughly 60% of these are our own 2l-cm redshifts
and these have an accuracy of 15-20 km/s. Our observations extend down
to 0 = -45 so cover 80% of the sky. Only a small fraction of these
0 ,

observations have been published (Fisher & Tully 1975). In passing, a

program also involving M. Goss, U. Mebold and H. van Woerden has begun
which will provide consistent coverage of the southern polar region.
NEARBY SMALL GROUPS OF GALAXIES 33

To avoid major complications ar~s~ng from incompleteness, only a

restricted volume will be considered today. The boundaries: veloc-
ities adjusted for solar motion less than 1100 km/s, galactic latitudes
beyond 30° from the plane, exclusion of the Local Group, and (a) in the
north, exclusion of a region of 6° radius centered on the Virgo cluster,
and (b) in the south, exclusion of the southern half of the hemisphere
defined by the great circle lying along ~ = 30° and ~ = 210°. This
latter criterion is roughly a lower declination limit at 0 ~ -25°.
This cut eliminates most of the members of our nearest neighbour the
Sculptor group from consideration so the entire group has been excluded.

In the volume so defined, 5000 Mpc 3 , we have 412 redshifts. For

the analysis which follows, all distances will be taken strictly from
redshifts assuming a Hubble constant of 75 krn/s/Mpc. The results are
not substantially affected by the choice of a Hubble constant or by
modest deviations from the Hubble flow.

THE SOUTH GALACTIC HEMISPHERE

It is, of course, known that there are many more nearby galaxies
north of the galactic plane than south. De Vaucouleurs (cf 1976) pro-
poses that we are at the outer edge of a supercluster associated with
the Virgo cluster. Let us look at the simpler region first then, the
relatively empty region away from the supercluster in the southern
galactic hemisphere.

A third of the total volume we will consider is in the south,

yet there are only 34 redshifts, 8% of the total. Their distribution
is shown in Figure 1. Only 7 of these 34 have integrated magnitudes
exceeding _19m. These seven account for 80% of the mass in galaxies
in this volume.

a) COJVte-ta;t.[OYl Sc.a1.e6

It should be clear from Figure 1 that the galaxies are not randomly
distributed. A majority are in one of three •.•. let us call them asso-
ciations. Moreover these three associations are nearer to one another
than could be expected with a random distribution.

We can graph a close kin to the two-point volume covariance func-

tion introduced by Peebles (1973). Using each galaxy in the sample in
turn as a point of reference, we derive the number of galaxies per
unit volume contained within a shell of given radius, then sum and
normalize. The parameter, n(R)~related to Peebles' parameter, ~(R):
~(R) ~ (n(R) - ~)/W
where W is the mean number density in the sample. Without inspecting
a volume large compared with the correlation scales we have no objec-
tive way of estimating n.
 34 R. BRENT TULLY AND J. RICHARD FISHER

..
.0
,

,', 0" •••••

..
':x"

1.',
'b

.•
•
•
'.
/ •
~"
0-100
• 100-200
.200-300
.300-400
.400-500

• .:••
.500-600
.600-700
.700-800

• •••
.800-900
.900-1000
01000-1100
)( 1100-1500

F-lgWte 1. The dM:tJU.bu:Uon 06 neMby ga..taUe6 -In :the .6Ot.U.:tVtYl gcd.a.c;U.c.

he.m.iJ.,pheJte, exc.lucUng membeJt6 06 :the Loc.a£. gJtoup and :the Sc.uip:tolt gltoup.
TheJte Me 34 ga..taUe6 wah V < 1100 kIn/-6 -In :the he~Uttc.le enc.lO-6-
-lng :the Mea 06 -6ky wah b <0_ 30 0 and 30 0 < 2 < 210. COltlteiatiOn-6
welte made be;tween membeJt6 06 :thM -lV/.V/.eJt volume and all gcd.a~u wUh..[n
a Iteg-lon de6-lned by Vo < 1500 km/-6, b < -20 0 and 00 < 2 < 240 .
The distribution n(R) obtained from the southern volume is shown
in Figure 2. To avoid boundary effects, correlations were made bet-
ween the 34 galaxies inhabiting the volume defined above with all
galaxies in a larger volume, defined by a corrected velocity cut-off
of 1500 km/s, b < _20° and 0° < £ < 240°. Figure 2 shows that there
is a strong correlation between galaxy positions on scales less than
1.5 Mpc and lesser but still significant correlation on scales of 1.5-
4 Mpc. Scales larger than this cannot be studied in such a restricted
volume.

Shown in Figure 3 is the same type of covariance diagram but re-

stricted to the 7 more massive systems and their correlation with mas-
sive galaxies only. There are no close pairs of large galaxies in this
small sample, but there is a significant enhancement in the covariance
function between 2 and 3.5 Mpc. Note that in a homogeneous cubic close
packed universe, a large galaxy in this volume would expect his nearest
large neighbour to be 6 Mpc removed.
NEARBY SMALL GROUPS OF GALAXIES 35

TWO POINT CORRELATION

Soulnem Hemisphere
34 Goloxl's
5

Mean distance of MOresl

neighbor with a
random d"trlbullon

l f I I

2 4 6 8
RADIUS R (Mpc)

FiguJte. 2. UnnoJunilize.d :two-point volume. covcuUance. 6uncUon 60ft aff

gafauv., in :thc J.,Ou;theJLn hcnU.J.,pheJLe. volume. dc6.{ned in UguJte. 7. We
plo:t n(R) vc~Uh R whe.~e. n(R) ~ ~e.fa;te.d :to Pe.eble'h (7973) volume.
C.OVcuU(lY[:t 6uncUon: ~(R) '\, n(R)/n-7. E~~Oft ba~ Me &.

'I
TWO POINT CORRELATION
Southern J-temrsphere

IlL I
7 Galaxies \ifill! M"<-19
011:-

!Pr~
.001 I Mpr: IOMpe

Mean distance of nearest

BIG ~9hbor with a
random distribution

" 2 4
RADIUS R {Mpel
6

FiguJtc 3. Two- point volume covcuUance. 6uncUon 60ft oniy :thO-6 e. gafaUe.-6
wLth Mpg < - 79 m in :the. J.,Ou:theJLn volume..
36 R. BRENT TULLY AND J. RICHARD FISHER

To return to the sample of 34 again, the case for associations bet-

ween galaxies can be put differently. Turner & Gott (1976) have, in the
two-dimensional case, used an algorithm which purportedly would pick out
ten-fold density enhancements above a mean density. We can do similarly.
The mean num~er density of galaxies in our southern volume is 0.02
galaxies/Mpc. A radius is defined such that the volume number density
within this radius about an isolated galaxy is 10 times the above mean
number density. Then any galaxies with such spheres which abut form
pairs or larger units with ten-fold the mean density. In other words,
any galaxy with a neighbour within twice this radius, i.e. within 2.1
Mpc, is in a region of at least ten-fold the mean density. Of 34
galaxies, 29 fulfill this criterion. If the criterion were relaxed to
3 times the mean density then it is satisfied by fully 33 of the 34.

Yes, the rule is proven with one exception; one "field" galaxy.
This system, DDO 215, is a remarkable 7 Mpc from any other with a known
3
red hift. It is the sole galaxy of any significance in a volume of 1440
Mpc ; if "significant" is an appropriate word to describe an irregular
with M ~ _16 m ! DDO 215 is, indeed, at about the center of the most
conspigaous void in the southern volume. Its presence provides us with
3
something ot~~r than an uP P r limit to the density of this vast empti-
ness: 7 x 10 galaxies/Mpc. Yet it is possible to form dwarf
galaxies in utter isolation!

b) CfLOM.,[l1g HmM and V.ifL.ia! Ih6.6 M:

We would like to examine the dynamical properties of galaxy

associations. Although the above discussion leads to the conclusion
that almost all galaxies are positively correlated in position with
other galaxies, here only the more obvious clusterings will be con-
sidered. Entities which stand apart as enhancements on the local mean
density will be referred to as groups, with no a priori restrictions on
dimensions, internal dispersions or space densities.

For the groups that will be considered, characteristic crossing

times and the virial relationship are calculated. The moment of
inertia radius (Jackson 1975) is used:
2 1/2
RI = (Lm.r.
ill
/ Li m.)
l
where m., r. are individual masses and linear distances from the group
center 6f m~ss. Projected R are listed in Table I. A deprojection
adjustment of ~ is includ~d in deriving crossing times as a frac-
tion of the age of the universe. The kinetic and potential energy are
calculated assuming mass to luminosity ratios of 7 for spirals and
irregulars and 15 for ellipticals and lenti~ulars. The.ratio ~T/~
gives the discrepancy between the mass requlred to fulflll the vlrlal
theorem and the assumed mass from the group luminosity.

In the volume south of the galactic plane, unfortunately, there

are no clean groups of significant proportions. Perhaps the most in-
NEARBY SMALL GROUPS OF GALAXIES 37

teresting,Cetus I (group IS in de Vaucouleurs 1975; hereafter DVlS) ,

spills across our volume cut-off of 1100 km/s and careful attention is
required of nearby groupings at yet higher velocity. Indeed, there
is a rich field of galaxies extending some 12 Mpc from Cetus I to the
Fornax cluster that we are only glimpsing the edge of in the volume we
are considering.

The kind of dilemma we regularly face is shown up in a small group

near NGC 7814. There are four galaxies very close to NGC 7814 with a
velocity dispersion among themselves of only 19 km/s. However NGC 7814
has a larger redshift than their mean by 200 km/s! But this is an
optical redshift of low quality, while the rest are 2l-cm measurements.
Is NGC 7814 really at a different distance along the line-of-sight
(though undoubtedly positively correlated in the sense of the co-
variance function)? Is it at the same distance and the virial theorem
is applicable (large masses implied)? Or do we just have a bad
velocity?

NGC 628. Then there is an association in which the largest galaxy is

NGC 628. There are 6 galaxies within a projected radius of 1 Mpc with
2l-cm redshifts measured. As can be seen from the data compiled in
Table I, there is a severe discrepancy between the virial mass and the
luminous mass if the entity is assumed to be bound. With standard
statistical projection factors for radial velocities and radii on the
plane of the sky, the crossing time is less than the age of the uni-
verse by a factor of five. We have two choices: (1) there is a case
for hidden mass or anomalous redshifts, or (2) although surely corre-
lated in the sense of the covariance function, the high velocity pair
(NGC 600 and companion) are removed in the line-of-sight from the others,
and the crossing time is comparable with the age of the universe.

Obviously, in a given instance this ambiguity cannot be resolved.

Were we to consider many such groups we might apply a statistical test.
Assuming line-of-sight distances from redshifts, we could ask if these
associations are elongated or flattened in the line-of-sight or are they
spherical. Elongation recalls the analogy of structure in our own
Galaxy noted to point toward the sun which Bart Bok has called the
Fingers of God telling us we are doing something wrong. In our case,
the implication would be high velocity dispersions and the reality of
large vi rial masses. However in the north and south volumes under con-
sideration, given the exclusion of the Virgo cluster, there are no very
obvious fingers pointing at us.

NGC 1023. The pickings are slim in the southern galactic hemisphere,
but there is one of the cleanest galaxy groups to be found locally ex-
cluded from our volume by its proximity to the galactic plan~: the
NGC 1023 group (DV7). In spite of the low galactic latitude (18° < I bl
<2S0) we have included it in our study (see Table I). Materne (1974)
has pronounced this group stable on the basis of 5 redshifts. We now
have 13 and all of the original velocities have been improved. The
 '"00
TABLE I
===========================================V===========R============iPEi=====KE======================
Group DV#, K# # Mass B cr I TIRo 58 KE/ ~T
====================!~~====~g!===!~~=~@==~~Lg==~~Lg===~g~=============!~===~~~g========~~l==~===~~~~
NGC 628 6 3 825 66 240 0.2 0.3 4. 14. 28. 1
1 74%
NGC 1023 DV 7 13 8 726 59 570 0.5 2.0 8. 4. 8. 2
3 78%
Leo DV9 K31 4 6 661 45 70 0.09 9. 4. 0.4 0.9
-M66 TG 38 4 Similar
Leo DVll K27 6 6 691 60 120 0.1 12. 7. 0.6 1• 1
-M96 TG 27 6 Similar
Leo combined 16 15 675 66 640 0.5 26. 20. 0.8 1.6
-all in vicinitl 12 Similar
M81 DV2 K22 6 2 164 116 30 0.02 0.9 7. 7. 15. 3
-restricted TG 16 2 87%
M81+M82 coupled (6) 2 164 (6) 30 0.3 0.3 0.02 0.07 0.1 4
-restricted
M81+M82 coupled (14) 3 191 43 41 0.5 0.4 1.4 3.5 7. 4
-all in vicinitl 3 76%
M101 DV 5 K46 4 2 389 (18) 20 0.05 0.3 0.2 0.7 1.5 :0
-restricted TG 82 1 91% .,;
~
M101 9 3 391 26 150 0.3 0.4 0.5 1.4 2.8 tTl
-all in vicinity 1 81% :z:
...,
12 5 992 270 0.2 1.7 11.
...,
N5005 83 9. 5. 5 c::
t""
+N5033 TG 67 2 75% t""
N5907 DV30 K54 7 6 871 31 200 0.3 2.2 1.9 0.8 7.6 ><
>
:z:
TG 97 2 87% t:J
N3184 DV12 + 12 3 613 68 1,180 0.9 0.2 4. 21. 42. :-
:0
6 90% ;::;
;J:
DDO 168 7 .05 296 23 290 0.7 3x10- 4 8x10- 3 32. 64. 6 >
:0
dwarfs 1 55% t:J
"!l
Centaurus DV4 K47 9 8 325 28 470 0.9 1.9 2.0 7.. 1.0 U;
;J:
tTl
==============================~====~~~================================================================ ~
NEARBY SMALL GROUPS OF GALAXIES 39

Table I - Description and Notes

Column 2: DV = de Vaucouleurs (1975); K Karachentsev (1970);

TG = Turner & Gott (1976)
Column 4: M =~m.
i 1.
Column 5: VB = L m.V .ILm.
i 1. 1. i 12 1/2
Column 6: 0= ( Lim.
ll
(V . -VB) I 2: m. )
i 1
2 1/2
Column 7 : RI (2:m.r.l2:m.)
ill i 1
Column 8: TIHo = (1l.5 RII /3 0) H0
Column 9: PE -(2/TI)G L m.m./r ..
pairs 1. J 1.J
2
Column 10: KE (3/2) ~ m. (V.-V B)
I 1. 1.
Column 12: ~T/~ = 2 KE/lpEI

Notes
1. NGC 600 + companion only in line-of-sight?
2. 60% of KE in NGC 891.
3. Almost all KE and PE in M81-M82 pair.
4. M8l and M82 considered as a single object with combined mass and
barycentric velocity.
5. NGC 5005 velocity should be checked. Dominates KE.
6. Near CVn I region.

case for stability has been diminished: the ratio KEI/PEI is 4.

This value is probably not significantly different from unity given our
uncertainties in distance, mass, velocity an10projection factors. But
as important, the crossing time, at 0.7 x 10 years, is half the age
of the universe.

THE NORTH GALACTIC HEMISPHERE

The differences between north and south are not subtle. In
Figure 4 there are plotted 378 galaxies with b > 30° and V < 1100 kmls
(i.e., out to but excluding the Virgo cluster). There are o 99 galaxies
with Mpg < _19m. The major regions of concentration are in Leo, in
Virgo (southern extension) and, expecially, in Can uS Venatici-Ursa
Major. There is also a tremendously large region where there are no
galaxies at all.

The two-point covariance distribution is shown in Figure 5 for the

northern material. In the logarithmic plot a featureless power spectrum
is seen on scales up to 6 Mpc. Larger scales cannot be tested in such
a restricted volume. The correlation between big galaxies only, shown
in Figure 6, is qualitatively similar. In the logarithmic plot there
is a suggestion of a demt~of pairings of large systems on scales less
40 R. BRENT TULLY AND J. RICHARD FISHER

~~~~ ----------~ ~~~

~ ~

r!. ..I.
I! ~
"·0.,
0,.
~. ~ ~\ I " ' .. \
\\
\
i • 0

·.f~,:~
• x,"
.:::, •

.. ~.·o
"oJe. .,
'.

~
ow. ,"\.r
,
0 '0 '0 • • 0 'I

II .~. x· . ,... • .. --. •

\
~~.
.•". CLUSTER
•••
VIRGO
.... • •• /~O.~
: ~. 0 !
I /'
0 • • •••

, ,_ '. • 0 0 "'.~ ~<}l Jt'. b ~ .. '

\\ \( ..' • .~, ' •. ~ ~. ~~~~.: J /
.'-!.' / /
. .
• \ ,; •• ;A.
\
\'1 -\. • Po ~.. 0 '>. • / •

.
0 0"

\• • •~ .'.
~.~'.
..' . I.' . /

0.0" . . . .0'/'
\ ~ ' " • 6 •• ~. ~ . ' .• ) /

\~~
\~: .. .' 'jl ~~:, /1• 0-100
'ml,
.' . . . :/ /"/ : ~~~=~~~
-. • ,,/

~~
~ .•
_________
•
~
____---------.
/
. ..0 /
•
: ~g~~~~g
• 300-400

• 600-700
~-------- /' • 700-BOO
eSOO-900
~
~/
~ .900-1000
• ~ 01000-1100
x 1100-1500

F-igWLe. 4. The. dA.MJubiLUon 06 ne.Mbq gal.cvuv., -in .the. nolLtheJLn galactic.

he.miJ.,pheJLe.. TheJLe. Me. 378 galauu w-Lth V < 1100 k.m/;., aYLd b > 30°.
COfULe.iatioM WeJLe. made. betwe.e.n .thue. and ~ galauu w-Lth V < 1500 k.m/;"
and b > 20°, btLt e.xdud-iYlg .tho;., e. wdh-LYl a 6 Jtad-iu.;., a 6 .the. c.~n..teJL 06
.the. V.utgo Uu.;.,.teJL.

than 1 Mpc compared with a power law distribution or the distribution of

all galaxies. This may be an artifact given that line-of-sight dis-
tances come from velocities and dispersions may be high for close pairs.

As for in the south, we can determine the fraction of galaxies

which meet the ten times mean density criterion. Of 378 galaxies, all
but 29 (92%) meet the demanding requirement that their nearest neigh-
bour lie within 1.2 Mpc implying a group of two or more with a local
number density 10 times the mean in the northern volume. However the
mean density in the northern volume must be unusually high, and a more
NEARBY SMALL GROUPS OF GALAXIES 41

TWO POINT CORRELATION I

Northern Hem Isphere
378 Golaxles

I .. ..
~
o
g,
1.5
...

IMpe IOMpe

Mean distance of nearest

nelQhbor w,th CI
random dIStribution

2 4 6 8
RADIUS IMpel

FigWLe 5. Two-poiYLt volume covaJU.ance bunc;ti.on bOlL :the nolL:thV1.n

!l'.1..iac;ti.c he.miJ..phV1.e Jtegion deMned in FigWLe 4.
FigWLe 6. Two-poiYLt
'I volume covaJU.ance
TWO POINT CORRELATION bunc.tio n bwveen
galauu ,1Jil:l M < -19 m
Northern Hemisphere
99 Gcllollo with Mpg <-19 m
pg
--Lvl :the IwJt:thV1.n volume.
I
; 4 I
I.
i 0"
..
'"
iii

t 3
..
~ 01
'M.. "
IOMpe

~
..
5
: .2
I
Mean d •• tenet! of ntoresl
BIG neighbor Wl,h 0
random dd'lb\,lhon

~~
=
I
I I

.
I
~ I
"0 •
•
2 4 6 8
RAOIUS (Mpcl
42 R. BRENT TULLY AND J. RICHARD FISHER

realistic mean that is more universally characteristic might be that

derived from the southern volume. Remarkably, only 3 galaxies in the
north are in a lower density region than ten-fold th; mean southern
density. For all of these, the local density exceeds five-fold the
mean southern density. The most isolated galaxy in the northern hemi-
sphere has a nearest neighbour at 2.6 Mpc! In the regions considered
in both hemispheres only one of 412 galaxies is negatively correlated!

In spite of the large number of galaxies in the northern sky with-

in 1100 km/s, it is still possib1 3 to find a contiguous region devoid
of galaxies of more than 1000 Mpc That is roughly one-third of the
total volume we are considering here! In Table II, some characteristic
number densities from our sample are given:

TABLE II
Volume Galaxies/Mpc 3
3
Virgo cluster 15 Mpc 10
C Vn - U Ma 150 " 1
The void > 1000 " < 10- 3

Let us turn quickly to the details of several individual groups.

There can be no attempt at completion. Instead we will try to explore
the range of possibilities by picking examples that are as clean as
possible. The very difficult CVn-UMa region will be ignored.

Leo. In Leo are two tight groups near to each other which are well
known (DV9 and DVll; in the catalog by Karachentsev, 1970, they are
K3l and K27; in that by Turner and Gott, 1976, they are TG38 and TG 27).
Materne and Tammann (1974) found these two groups considered as a
single entity to be stable assuming conventional masses. Our analysis
is summarized in Table I. For both tight groups, crossing times are
short but the virial theorem is satisfied assuming only conventional
masses. Combined together, the single large entity is still stable
and now crossing times compare with the age of the universe. Including
five more nearby systems, two rather deviant in velocity and three re-
moved spatially, the situation is not substantially changed (as all
five are low mass objects).

These two groups are reminiscent of perhaps the most interesting

of the entities catalogued by Turner and Gott (1976; see Gott and
Turner 1977): condensed groups with several massive galaxies within a
radius of roughly 100 kpc, often early morphological types, and group
crossing times substantially less than the age of the universe. In-
evitably Gott and Turner found severe virial mass discrepancies. The
most noteworthy exceptions were these cases in Leo. So we agree that
NEARBY SMALL GROUPS OF GALAXIES 43

in these nearest examples of tight groups with several massive members

there is no evidence for large masses. Unfortunately, there are no
other clean groups of this type within the volume we are considering.

M81. Although we would put as many as 14 galaxies into an extended

M81 groups, both the kinetic and the potential energy are dominated
by the M81-M82 pair. On this basis alone there is a large virial dis-
crepancy with our assumed masses. An individual case can be rational-
ized: for example, suppose M82 is near perigalacticon orr an eccentric
orbit. Considering M81 and M82 as a single unit, the remaining close
neighbours would be easily bound and for the extended group the crossing
time begins to compare with the age of the Universe.

MI01. In this group, MIOl is by far the dominant galaxy and as a re-
sult the mass weighted velocity dispersion and moment of inertia radius
are unrealistically low. However the virial analysis should apply.
Stability is implied for both the restricted and most general group.

NGC 5005 + NGC 5033. This group contains two galaxies of comparable
mass, plus a host of small companions. The problem is that NGC 5005
has a reported velocity which deviates considerably from most of the
rest in the group and dominates the kinetic energy term. This velocity,
which we have drawn from the Second Reference Catalogue (de Vaucouleurs,
de Vaucouleurs and Corwin 1976), should be checked before drawing any
firm conclusions.

NGC 5907. Here again is a group dominated by two members. In this

case, however, there is no virial discrepancy. There are several other
more distant galaxies which must be associated in the sense of the
covariance function. However, they could not be bound.

NGC 3184. We now consider a very different kind of group. A portion

of this entity was catalogued by de Vaucouleurs (DV12). However it can
now be followed considerably further south with roughly constant density
on the plane of the sky. Not too far away to the north is the group
DV6 and to the south are the two Leo groups, DV9 and 11 already dis-
cussed. It is seen in Table I that the moment of inertia radius is
much larger than for any of the other groups considered and the con-
sequence is that the crossing time becomes camparable with the age of
the universe. Since there is a large virial mass discrepancy, the
implication is that the group is not bound.

DDO 168 It may not be fair to consider this association of dwarf ir-
regular galaxies to be an independent group, as they only stand apart
from the Canus Venatici I region (DV 3) by 1 Mpc. However, the Table I
data show that in themselves they are certainly not bound. There is
no problem with the dissipation time-scale.

Centaurus A. We again leave the strict confines of our working volume

for one last very nice example. Galaxies in the Centaurus group (DV4)
44 R BRENT TULLY AND J. RICHARD FISHER

are in the galactic latitude range 13° < b < 32°, and there is the
danger that some are hiding further south. But this group is nicely
isolated and it contains four significant galaxies. We see in Table I
that the crossing time is comparable with the age of the universe and
there is no appreciable mass discrepancy.

So we have been able to show the existence of groups of galaxies

which satisfy the stability criterion of the virial theorem assuming
only masses typically associated with galaxies. These groups character-
istically have radii RI less than 300 kpc. Gr~yps that are larger
usually have crossing Eimes greater than 0.5 H • For the few excep-
tions that turn up (NGC 628 + NGC 600, M8l + M§2, NGC 5005 + NGC 5033)
it is possible to suggest plausible scenerios which do not require high
mass to light ratios. We feel that the evidence weights strongly
against the existence of a lot of unseen matter distributed like the
galaxies in small groups.

Clearly, the analysis can be taken a lot further. There has been
no sensitivity to the possibility that a fraction of a group may be
bound while the rest is expanding. We are not making effective use of
our velocity data through application of the virial theorem in those
common instances where there are only one or two massive systems but
ten or so "test particles" of insignificant mass making up the group.
These small galaxies offer the means of weighing the groups to sub-
stantial radii.

Finally, the importance of obtaining good redshifts (accuracies

<20 km/s) for all nearby galaxies must be stressed. One may have a
dozen good velocities in a group but a single bad value associated
with a large galaxy can compromise the results.

CONCLUSIONS

1. In agreement with Peebles and co-workers, the volume two-point

covariance function shows a featureless power law spectrum over those
scales that we can meaningfully examine.

2. There is no evidence for a significant number of uncorrelated

"field" galaxies.

3. There are two large voids even in the restricted region we

have surveyed, which includes much of the Local Supercluster. These
voids encompass about half the total volume under discussion and bet-
ween them contain all of one galaxy.

1
4. The number density of galaxies averaged over a fair y large
region in the plane of the Local Supercluster is at least 10 times
the number density in the voids.

5. The moderately compact small groups that have been studied

show little or no virial discrepancy with the optical masses we have
NEARBY SMALL GROUPS OF GALAXIES 45

accepted. Our results suggest virial mass to luminosity ratios in the

range 5-30.

6. The loosely correlated groups have crossing times which ar~l

a large fraction of the age of the universe: typically 0.5 - 1.0 Ho

7. In the volume we have considered there are no convincing

examples of groups such as those identified by Gott & Turner with both
short crossing times and large virial mass discrepancies. If real,
such groups are rare.

References

Davis, M., Groth, E.J. & Peebles, P.J.E.: 1977, Ap.J.(Letters) 2l2,Ll07.
Fisher, J.R. & Tully, R.B.: 1975, Astron. & Astrophys. 44, 151.
Gott, J.R. & Turner, E.L.: 1977, Ap. J. 213, 309.
Jackson, J.C.: 1975, MNRAS 173, 4lP.
Karachentsev, I.D.: 1970, Problems of Cosmic Physics 5, 201. (Kiev
State University).
Karachentsev, I.D.: 1976, IAU Colloq. #37, 321. (Ed. Balkowski &
Westerlund, CNRS, Paris).
Materne, J.: 1974, Astron. & Astrophys. 33, 451.
Materne, J. & Tarnrnann, G.A.: 1974, Astron. & Astrophys. 37, 383.
Nilson, P.: 1973, Uppsala General Catalogue of Galaxies, Uppsala
Astron. Obs. Ann., Band 6.
Peebles, P.J.E.: 1973, Ap. J. 185, 413.
Rood, H.J., Rothman, V.C.A. & Turnrose, B.E.: 1970, Ap. J. 162, 411.
Sandage, A.R. & Tammann, G.A.: 1975, Ap. J. 196, 313.
Soneira, R.M. & Peebles, P.J.E.: 1977, Ap. J. 211, 1.
Tully, R.B. & Fisher, J.R.: 1976, IAU Colloq. #37, 95. (Ed. Balkowski
& Westerlund, CNRS, Paris).
Turner, E.L.: 1976, Ap. J. 208, 304.
Turner, E.L. & Gott, J.R.: 1975, Ap. J. (Letters) 197, L89.
Turner, E.L. & Gott, J.R.: 1976, Ap. J. Suppl. 32, 409.
de Vaucouleurs, G.: 1975, Stars & Stellar Systems IX, Ch. 14.
de Vaucouleurs, G.: 1976, Ap. J. 203, 33.
de Vaucouleurs, G., de Vaucouleurs, A. & Corwin, H.G.: 1976, Second
Reference Catalogue of Bright Galaxies, U. of Texas Press.
Zwicky, F.: 1933, Helvetica Phys. Acta. 6, 110.

DISCUSSION

Davis: I believe there is an inconsistency in your low estimate of the

peculiar RMS velocity of galaxies (a ~ 50 km s-l) and the covariance
analysis. Your covariance coordinate s is the redshift separation of
the pairs, which is the physical separation added to the relative pecu-
liar velocity of the pair. This coordinate produces bias in the
covariance function slope, flattening it from y ~ 2 to roughly y ~ 1.
The covariance function will have the slope y ~ 2 only for redshift
46 R. BRENT TULLY AND J. RICHARD FISHER

separations s ~ 6/Ho' Therefore if the slope y ~ 1 corresponds to

redshift separation s ~ 10 Mpc (as shown in your data in the Northern
sky), then the true peculiar velocity dispersion a must be of the order
of 500 km s-1 (Ho = 50 km s-1 Mpc- l ), i.e. by a factor of 10 higher than
you state. Your group selection procedure unfortunately guarantees that
your estimate of a will be much less than 500 km s-l.

Tully: Indeed, the slope on the covariance function is y = 0.9 which is

quite different from the y = 1.77 derived by the Princeton group. A
very high velocity dispersion would explain this discrepancy. However,
I feel that such a large dispersion is completely excluded by the obser-
vational material. We are aware of the dangers which could lead to an
under-estimation of a and our group selection criteria do not guarantee
such an under-estimation.

Gatt: One way to test the membership of groups is to plot a 3 dimen-

sional map of the galaxies assuming they all lie exactly at their
redshift distances. If the groups in such a "redshift space" look
spherical then the internal velocity dispersions of the groups are
small; if however the groups in the redshift space appear as "fingers"
pointing at the Earth, then this shows that the internal velocity
dispersions are large.

Tully: I agree but I have not yet completed this test. My impression
is that I do not find much evidence for "fingers".

Ostriker: A point concerning nomenclature. It might help in understand-

ing if one did not use the term MLi ht since that depends on an assumed
mass-to-light ratio. It would be cYearer if the observed quantities
(Mvt/L) were plotted directly.

Tully: I agree. I just used a method which conveniently allowed for a

difference in M/L values between ellipticals and spirals.

Ozernoy: Is there a difference between the mass discrepancy in loose

groups and compact groups?

Tully: The only compact groups with several massive members in our
sample are the two in Leo, for which there is no mass discrepancy. The
rest of our discussion has been based on groups which contain only a
couple of massive members or are very loose.

Silk: Have you derived the multiplicity function of your groups?

Tully: We have not derived the multiplicity function, but we intend to

do so.

Silk: It may be worth pointing out that a different shape would be

expected for this function as compared to that obtained by Gott and
Turner because of the different definitions of groups: in particular
the use of a volume as opposed to a surface density enhancement.
NEARBY SMALL GROUPS OF GALAXIES 47

Van Woerden: Dr Tully's observations taken in the northern hemisphere

provide only partial coverage of the sky. We should soon be able to
fill in the remainder of the sky from observations at Parkes (by Van
Woerden, Coss, Mebold and Siegman), where we have measured about 500
galaxies in the 2l-cm line.
OBSERVA TIONS OF HIGH VELOCITY HI CLOUDS IN THE LOCAL
GROUP

W. K. Huchtmeier
Hamburger Sternwarte

High velocity clouds (HVC) of neutral hydrogen in or around our

galaxy and the observations of intergalactic HI in the Local group:
Magellanic stream (Mathewson et al., 1974, Astrophys. J.190, p. 291),
M31 (Davies,R.D., 1975, Mon.Not.R.astr.Soc., 170,p.45P), and in
the Sculptor group of galaxies (Mathewson et al. , 1975, Astrophys. J.
195,p. L97) motivated us to search for HVC-phenomena in a number
of nearby late-type galaxies with the 100 m Effelsberg radio telescope
which has a half power beam width of 8. 5' at the wavelength of 21 cm.
HVC-phenomena have been observed in the direction of M 33,
Sextans A, WLM, and IC 10. Distances of the HVC's have been
assumed to be those of the corresponding galaxies. Observed values
of line widths and surface densities are typical for HVC's. Coinciden-
ces in position and radial velocity are most convincing for M 33 and
IC 10. The HVCin direction ofWLMseems tobepartoftheMagellanic
stream.
Galaxy Distance Extent typical column HI-mass
(km s-l) (Mpc) angular linear line width density (M)
(1018 cm -2) 0
(kpc) (kmLs)
M 33 (-180) 0.7
(HVC-147) 7' 1.4 12 40 3 10 6
5
(HVC-155) 6' 1.2 15 20 3 5 x 10 6
(HVC-225) 9' 1.8 25 75 3 5 x 10
Sextans A(325) 1
(HVC+142) 10 17 25 50 3 10 7
WLM f-130) 0.87
(HVC -200) >3 0 x1.5° > 50 25 60 3 >10 8
IC 10 (-346) 1. 26
(HI-shell) 80' 29 20 ... 50 4 '" 10 8

49
M. S. Longair and J Einasto (eds.), The Large Scale Structure of the Universe, 49. All Rights Reserved.
COPYright © 1978 by the [AU.
HYPERGALAXIES

Jaan Einasto
Tartu Astrophysical Observatory
Toravere 20244
Estonia, U.S.S.R.

"Galaxies are like people: they depend on both

genetics and environment" (van den Bergh 1975)

1. INTRODUCTION

Studies of the dynamics of small aggregates of galaxies have shown

that these systems possess extensive and massive coronas (Einasto 1972,
Ostriker and Peebles 1973, Einasto, Kaasik and Saar 1974, Ostriker,
Peebles and Yahil 1974). The dimensions of massive coronas are so large
that all close companion galaxies as well as high-velocity hydrogen
clouds are situated in their interiors. In other words if massive
coronas were considered extensions of galaxies then giant galaxies with
their coronas would form in fact compact groups of galaxies. It is
evident that it is not suitable to identify a galaxy with a group of
galaxies. For this reason following the suggestion of Chernin we con-
sider galaxies with their massive coronas and all objects moving in the
coronas as distinct building blocks of the Universe as hypergalaxies
(Einasto et al. 1974a). Hypergalaxies form compact subgroups in groups
of galaxies. Examples of hypergalaxies are subcondensations of galaxies
and extragalactic matter around the Galaxy and the Andromeda galaxy in
the Local Group.

In the following we use the term galaxy for the galaxy proper, i.e.
its visible populations without the massive corona. The term group of
galaxies is used in the conventional sense to designate a density
enhancement in the Universe smaller than clusters of galaxies. In the
case of isolated hypergalaxies, there is no real difference between
hypergalaxies and groups. Many hypergalaxies form mUltiple systems -
these aggregates of hypergalaxies are conventional groups of galaxies.

Studies of the spatial distribution of nearby galaxies indicate

that most galaxies form aggregates of varying richness from poor groups
to rich clusters of galaxies (de Vaucouleurs 1971). This re~ult has
51

M. S. Longair and J. Einasto (eds.), The Large Scale Strncture of the Universe, 51-61. AU Rights Reserved.
Copyright © 1978 by the 1AU.
52 JAAN EINASTO

been confirmed by a statistical study conducted by Soneira and Peebles

(1977). They conclude that if a true field component of galaxies does
exist, it amounts to substantially less than 18% in a catalogue selected
to a limiting apparent magnitude. The majority of known groups of
galaxies are located in the disks of superclusters. The mean distance
between the centres of the groups is 3-10 Mpc, and their mean outer
radius ~ 1 Mpc. Thus groups are well separated from each other.

2. SPATIAL STRUCTURE OF HYPERGALAXIES

The distribution of galaxies in hypergalaxies is not random.

Bright companion galaxies together with the main galaxy usually form a
multiplet in the centre of the aggregate. Examples are our Galaxy and
the Magellanic Clouds, the Andromeda galaxy and its two elliptical
companions M 32 and NGC 205, M 81 and NGC 3077. Dwarf galaxies of very
low luminosity form a cloud of much larger radius around the main
galaxy (Einasto et al. 1974a).

The structure of a hypergalaxy can be studied in detail in the case

of our Hypergalaxy. As demonstrated by Lynden-Bell (1976), Kunkel and
Demers (1976), Einasto et al. (1976a) and Haud and Einasto (1977), both
optical companions of our Galaxy as well as the high-velocity hydrogen
streams are all strongly concentrated towards a great circle. In space
all these companions form a flat disk with the Galaxy in its centre. It
is remarkable that the disk of the Hypergalaxy forms almost a right
angle, 70 0 , with the main plane of the Galaxy. This is a clear indica-
tion that the disk of the hypergalaxy is not a simple continuation of
the disk of the main galaxy.

3. DYNAMICS OF HYPERGALAXIES

One of the most remarkable properties of hypergalaxies as well as

of clusters of galaxies is a very close dynamical link between the main
galaxy and the surrounding aggregate of galaxies. In studying the mass
distribution in hypergalaxies Einasto et al. (l974a, 1975) noted that
the cumulative mass distribution, calculated from the motion of compan-
ion galaxies, forms a smooth extrapolation of the mass distribution of
the main galaxies, calculated from the inner motions in these galaxies.
Since M(R)ao r 2 (R)RaR, this relationship means that the velocity disper-
sion of stars in the main galaxies is approximately equal to the disper-
sion of relative velocities of companion galaxies. A more detailed
study (Einasto et al. 1976b) indicated that this equality is valid over
the whole observed range of velocity dispersions from 80 km s- l in dwarf
hypergalaxies to 1000 km s-l in rich clusters of galaxies (Figure 1).
This equality concerns only the main galaxies of aggregates. All com-
panion galaxies have a smaller internal velocity dispersion.
HYPERGALAXIES 53

Figure 1. Internal velocity

dispersion cr, in galaxies versus
external velocity dispersion cr comp '
2000
of galaxies for individual hyper-
.. (km 5-')
galaxies and clusters. Internal
x
1000 " velocity dispersions are given for

""•
the nuclei of galaxies (circles)
for the periphery of galaxies
•• •
500 ~
(squares) or are calculated from
•
• •
the maximum rotational velocities
of spiral galaxies (triangles) or
200
II from X-ray temperatures (crosses).
c Aggregates with spiral main galaxies
lh 68 are designated by open symbols,
100 c C
aggregates with elliptical main
I
galaxies by filled symbols. Data
50 100 200 500 1000 2000
.. comp(km 5~)
from Faber and Jackson (1976) and a
compilation by Einasto et al.
(1976b).

4. MORPHOLOGY AND LUMINOSITY OF GALAXIES IN HYPERGALAXIES

The principal properties of hypergalaxies depend on the luminosity

and morphological type of the main galaxy. All aggregates of galaxies
with an elliptical main galaxy have a velocity dispersion exceeding
200 km s-l; all aggregates with a spiral main galaxy have a smaller
velocity dispersion (Figure 1). The velocity dispersion is proportional
to the mean density of the galaxy. Thus this relationship indicates
that the density is the principal factor determining both the morpholog-
ical type of the main galaxy and the properties of the whole aggregate.

In a given hypergalaxy companion galaxies of different morphologi-

cal types and luminosities are segregated from each other (Einasto et
al. 1974b). All elliptical companions of a given luminosity are located
inside a sphere of a certain radius, all spiral and irregular companions
of a given luminosity are located outside this sphere (Figure 2). The
radius of the segregation sphere is smaller the higher is the luminosity
of the companions.

5. LUMINOSITY FUNCTION OF HYPERGALAXIES

When studying hypergalaxies, we are interested in luminosity func-

tions of three different kinds: (a) the volume density of hypergalaxies
~(Mo), considered as a function of the absolute magnitude of the main
galaxy Mo, (b) the differential luminosity function of hypergalaxies
~(M'), considered as function of the magnitude difference M' = M-Mo , and
(c) the volume density of galaxies ~(M) (i.e. the conventional luminos-
ity function).
54 JAAN EINASTO

O.-----~r_--~----~
Figure 2. Luminosity L versus distance
logL R from the main galaxy for companions of
I;,...
our Galaxy (circles) and the Andromeda
-2 I galaxy (squares). Elliptical companions
have been designated as filled circles or
squares, spiral and irregular companions
as open ones. The full line represents
-4 the radius of the segregation sphere for
companions of different luminosity •
••
-6~----~----~--~~
o 2 log R 3

The functions ~(Mo) and ~(M) are defined as the numbers of hyper-
galaxies or galaxies per volume and per unit interval of absolute magni-
tude (M ± 0.5). The differential luminosity function of hypergalaxies
~(M') is defined as the mean number of galaxies in one hypergalaxy per
unit magnitude interval (M' ± 0.5). These three functions are mutually
connected by the formula (Einasto et al. 1974a)
+00

Using the first list of hypergalaxies (Einasto et al. 1977) and a

preliminary version of the second list of hypergalaxies, Vennik (1977)
determined all three functions of interest. The results are given in
Figures 3 and 4. We note, first of all, that the conventional luminos-
ity function ¢(M), determined by this non-conventional method, is in
good agreement with other recent determinations (Christansen 1975,
Kiang 1976).

The luminosity function of hypergalaxies ~(Mo) is quite similar to

the conventional luminosity function ~(M). Both functions have a
secondary maximum at M* = -20.5, the slope of both functions at high
luminosities (M < M*) is much larger than at low luminosities (M > M*).
The differential luminosity function of galaxies in hypergalaxies
~(M') is completely different from both the conventional luminosity
function and the luminosity function of hypergalaxies (see Figure 4).
It has a maximum at M' = 0, caused by the main galaxy. There follows a
region in which there are few galaxies. Approximately from M' = 2.5
onwards the number of companion galaxies increases, the slope of this
section of the function ~(M') coinciding with the slope of the conven-
tional luminosity function at low luminosities.

Thus the study of hypergalaxies shows that the knee-point in the

conventional luminosity function is due to the presence of the corres-
ponding feature in the distribution of hypergalaxies as a function of
luminosity. The slope of the high luminosity section of the luminosity
HYPERGALAXIES 55

"tH')
6

-/

IJ

-2

-17 -18 -19 -20 -2/ -22 -lJ 1'10 2

Figure 3. Volume density of Figure 4. Mean differential

hypergalaxies versus the absolute luminosity function of hyper-
magnitude of the main galaxy Mo. galaxies, derived by Vennik
Open circles are numbers of (1977) for all nearby hyper-
hyp~rgalaxies, per unit volume, galaxies from the first list.
calculated from the first two Rms error bars have been
lists of hypergalaxies, triangles given.
- respective numbers of hyper-
galaxies. 90% confidence limits
have been shown.

function is determined by the distribution of hypergalaxies, whereas the

slope of the low luminosity section is fixed by the distribution of
galaxies in hypergalaxies.

6. THE MASSES AND MASS-TO-LUMINOSITY RATIOS OF HYPERGALAXIES

According to a recent determination by Einasto et al. (1976c), the

mean mass-to-luminosity ratio of S-hypergalaxies is about 80 in solar
units and that of E-hypergalaxies and clusters of galaxies is about 250.

7. INTERACTION BETWEEN HYPERGALACTIC GAS AND GALAXIES

Hypergalaxies contain some gas. Radio data show the presence of

neutral hydrogen clouds. Interferometric data indicate the presence of
ionized hydrogen, the mass of ionized hydrogen being approximately equal
to the mass of the visible galaxies (Golev and Scheglov 1975). X-ray
56 JAAN EINASTO

data show that hot gas, having a temperature about 10 6 K, also surrounds
the Galaxy (Field 1975). According to presently available data the
gaseous populations fill just the potential well of hypergalaxies asso-
ciated with their massive coronas and have the same density distribution
law p ~ R- 2 (Einasto et al. 1974b). For this reason it is expected that
the interaction between gas and galaxies is confined to the whole volume
of groups of galaxies.

In hypergalaxies at least three kinds of interaction between

hypergalactic gas and galaxies take place. When moving in gaseous
coronas of hypergalaxies, companion galaxies are subject to ram pressure
and dynamical friction; hypergalactic gas clouds may collide with the
main galaxy.

(a) Ram pressure will sweep the gas out of companion galaxies, if
the gravitational field of the companion is insufficient to bind the
gas. This mechanism may explain the segregation of companion galaxies
according to morphological types as suggested by Chernin (Chernin,
Einasto and Saar 1976).

(b) Dynamical friction brakes the motion of companion galaxies.

This results in a decrease of the major semi-axis of the orbit until
the companion is destroyed by tidal forces (Tremaine 1976). The debris
of the companion falls onto the main galaxy. This process can increase
the mass and luminosity of the main galaxies and of the clusters of
galaxies by cosmologically significant amounts (Ostriker and Tremaine
1975, Gunn and Tinsley 1976). Dynamical friction predicts the absence
of companions of very low density near giant galaxies, which has been
confirmed by observations.

(c) Gas infall to main galaxies has been suggested by Oort (1970)
and Quirk and Tinsley (1973) as an important factor in the evolution of
galaxies. This infall may account for the very stable chemical compo-
sition of the disks of galaxies (Lynden-Bell 1975) or the high metal
content of the galactic gas (Ostriker and Thuan 1975).

The concentration of the hypergalactic gas towards a plane perpen-

dicular to the galactic plane may give rise to the formation of the
spiral structure (Jaaniste and Saar 1976, 1977a) and to the warping of
the distribution of galactic gas (Haud 1977).

8. GALACTIC AND HYPERGALACTIC POPULATIONS

Table 1 presents a summary of the principal galactic and hyper-

galactic populations according to our present knowledge. Galactic
populations are given according to Oort (1958) and Einasto, J6eveer and
Kaasik (1976). The hypergalactic populations of our own Hypergalaxy
are also given.
HYPERGALAXIES 57

Table 1
Galactic and hypergalactic populations

Populations of the Galaxy

ao M
Name E Z
(kpc) (M@)

Nucleus 0.5 0.005 10 8 0.04

Bulge 0.8 0.4 1010 0.02
Halo (s tars, 1010 10- 3
0.3 2.5
globular clusters)
Disk (stars,
0.1 7 6xl0 6 0.02
galactic clusters)
HI 0.02 6 3xl0 9 0.02

Populations of the Hypergalaxy

ao M
Name E Z
(kpc) (Mo)

Nucleus (Galaxy) 0.2 4 lOll 1O- L O.03

Core (Galaxy+LMC+SMC) 30: lOll 10- 2
E-disk (dwarf
0.1 100: 10 9 10- 4
ellipticals)
S-disk (dwarf
0.1 200: 10 9 10- 4
irregulars)
HI (Magellanic Stream,
0.1 60 10 9
Northern Streams)
Massive corona 1: 75 10 12
Hot gas 1: 75: 1011 :

We note that there exists a definite analogy between galactic and

hypergalactic popUlations. The main galaxy corresponds to the nucleus
of a galaxy, the central core - to the bulge of a galaxy. Galactic and
hypergalactic disks have also similar properties, both being quite flat
popUlations with an axial ratio of ~ 0.1. But there also exist import-
ant differences. A galactic disk is relatively homogeneous, a hyper-
galactic disk can be divided into two parts, E-disk and S-disk. An
E-disk contains elliptical galaxies which have no gas of their own; an
S-disk contains spiral and irregular galaxies containing their own gas.
Both populations are spatially segregated. A second difference is in
58 JAAN EINASTO

the spatial orientation of the planes of s~etry: the hypergalactic

disk is inclined to the galactic disk at 70 0 • The third difference lies
in the chemical composition: the metal content of objects in the hyper-
galactic disk is very low (Hartwick and McClure 1974, Canterna 1975,
Norris and Zinn 1975, Kunkel and Demers 1977), whereas in the galactic
disk the composition is close to solar composition. Of course, the
dimensions are also different.

Globular clusters seem to be a heterogeneous population. Most

globular clusters are strongly concentrated towards the galactic centre
and form a part of the galactic halo. Globulars with low central con-
centration (classes XI and XII) have a very low metal content, their
distances from the galactic centre are large, and many of them are con-
centrated towards the hypergalactic plane. Apparently these globulars
belong to the hypergalactic E-disk.

The next population is neutral hydrogen. This population is also

clearly divided into two parts - galactic and hypergalactic hydrogen.

The list of populations ends with the massive corona and hot gas.
Both the massive corona and the hot gas can be considered either as a
galactic or as an extragalactic population. Here the dual nature of
hypergalaxies is seen very clearly: as indicated in the introduction,
hypergalaxies can be equally well defined as giant galaxies with their
permanent environs or as groups of galaxies with one concentration
centre.

9. HYPERGALAXIES AS GALAXY COMMUNITIES

The bulk of available information suggests that hypergalaxies form

permanent aggregates which can be regarded as galaxy communities.
Different hypergalactic populations are bound into a single system by
the gravitation of the massive corona; these populations are in mutual
interaction due to both gravitational and gas dynamical effects.

The presence of a close dynamical link between systems of galaxies

and their main galaxies is very difficult to explain if galaxies in
hypergalaxies had been born independently. We conclude that galaxies
had already been born in hypergalaxies, i.e. galaxy formation is a
collective phenomenon. Statistical arguments indicating the collective
nature of galaxy formation have already been given by Ambartsumian
(1958) .

Galaxies can be divided into two classes: main galaxies and

companion galaxies.

The principal properties of hypergalaxies are determined by their

main galaxies. The density of the main galaxy determines its morpho-
logical type as well as the mass and the rnass-to-luminosity ratio of the
HYPERGALAXIES 59

whole hypergalaxy. Dense proto-hypergalaxies evolve into E-systems,

less dense proto-hypergalaxies to S-systems. Elliptical galaxies have
a smaller angular momentum than spirals. The dependence of the morpho-
logical type of the main galaxies on only one parameter - the initial
density - shows that in denser regions the momentum is also smaller.

The morphology of companion galaxies seems to be determined by the

initial conditions as well as by environment.

The evolution of the main and of the companion galaxies is differ-

ent because of environmental differences. The main galaxies possess
coronas; the companions move in these coronas and may be swept clean of
their own gas by the ram pressure of the coronal gas and destroyed by
tidal forces. Thus the main galaxies can grow at the expense of their
weaker companions. The main galaxies can also grow as a result of the
infall of gas. This process may be of importance for the chemical
evolution of galaxies, as well as for the formation of the spiral struc-
ture and the bending of the large scale distribution of galactic gas.

REFERENCES

Ambartsumian, V.A., 1958. Solvay Conf. Rep., Brussels, p. 241.

Canterna, R., 1975. Astrophys. J., 200, L63.
Chernin, A., Einasto, J. and Saar, E-.-,-1976. Astrophys. Space Sci.,
12., 53.
Christensen, C.G., 1975. Astro. J., 80, 282.
de Vaucouleurs, G., 1971. Publ. A.S.~, 83, 113.
Einasto, J., 1972. Tartu Astr. Obs. Teated 40 (Proc. First Europ. Astr.
Meet. 2, 29l).
Einasto, J.~ Haud, D., Joeveer, M. and Kaasik, A., 1976a. Mon. Not. R.
astr. Soc., 177, 357.
Einasto, J., Jaaniste, J., Joeveer, M., Kaasik, A., Kalamees, P.,
Saar, E., Tago, E., Traat, P., Vennik, J. and Chernin, A.D., 1974a.
Tartu Astr. Obs. Teated, 48, 3.
Einasto, J., Joeveer, M. and Kaasik, A., 1976. Tartu Astr. Obs. Teated,
54, 3.
Einasto, J., Joeveer, M., Kaasik, A., Kalamees, P. and Vennik, J.,
1977. Tartu Astr. Obs. Teated, 49, 3.
Einasto, J., Joeveer, M., Kaasik, A. and Vennik, J., 1976c, Proc. Third
Astr. Meet. Ed. E.K. Kharadze, Mezniereba, Tbilisi, p. 431.
Einasto, J., Joeveer, M., Kaasik, A. and Vennik, J., 1976b. Astr.
Astrophys., 53, 35.
Einasto, J., Kaasik, A., Kalamees, P. and Vennik, J., 1975. Astro.
Astrophys., 40, 161.
Einasto, J., Kaasik, A. and Saar, E., 1974. Nature, 250, 309.
Einasto, J., Saar, E., Kaasik, A. and Chernin, A.D., 1974b. Nature,
252, lll.
Fie1d~.B., 1975. Astrophys. Space Sci., 38, 167.
Golev, V.K. and Shcheg1ov, P.V., 1975. Astr. Circ. No. 874,4.
60 JAAN EINASTO

Gunn, J.E. and Tinsley, B.M., 1976. Astrophys. J., 210, 1.

Hartwick, F.D.A. and McClure, R.D., 1974. Astrophys:-J., 193, 321.
Haud, U., 1977. (in preparation).
Haud, U. and Einasto, J., 1977. Astr. Cirk. No. 958.
Jaaniste, J. and Saar, E., 1976. Tartu Astr. Obs. Teated, 54, 93.
Jaaniste, J. and Saar, E., 1977a. Astr. Zh. Letters, 3, 9.
Kiang, T., 1976. Mon. Not. R. astr. Soc., 143, 129.
Kunkel, W.E. and Demers, S., 1976. R.G.O. Bull. No. 182, 241.
Kunkel, W.E. and Demers, S., 1977. Astrophys. J., 214, 21.
Lynden-Bell, D., 1975. Vistas in Astr., 19, 299. ---
Lynden-Bell, D., 1976. Mon. Not. R. astr:-Soc., 174, 695.
Norris, J. and Zinn, R., 1975. Astrophys. J., 202, 335.
Oort, J.H., 1985. Ric. Astr. Specola Astr. Vatican, 5, 415.
Oort, J.H., 1970. Galactic Astronomy, I, 121, ed. H.Y. Chiu and
A. Muriel, Gordon and Breach. -
Ostriker, J.P. and Peebles, P.J.E., 1973. Astrophys. J., 186, 467.
Ostriker, J.P., Peebles, P.J.E. and Yahil, A., 1974. Astrophys. J.
Lett., 193, Ll.
Ostriker, J.~and Thuan, T.H., 1975. Astrophys. J., 202,353.
Ostriker, J.P. and Tremaine, S.D., 1975. Astrophys. J-.-,-202, Ll13.
Quirk, W.J. and Tinsley, B.M., 1973. Astrophys. J., l79,~.
Soneira, R.M. and Peebles, P.J.E., 1977. Astrophys. ~ 211, 1.
Tremaine, S.D., 1976. Astrophys. J., 203, 72.
van den Bergh, S., 1975. Nature, 257,-gI.
Vennik, J., 1977 (in press). ---

DISCUSSION

Ostriker: Do you find any correlation between the luminosity and the
separation of companion dwarf galaxies? In a recent study E. Turner and
I found an inverse correlation in his sample, that is, the more luminous
galaxies tend to be further away.

Einasto: In our analysis, we found the opposite result: bright

galaxies are strongly concentrated towards the main galaxy. To resolve
this discrepancy, original data should be compared.

Kiang: The Local Group has 2 hypergalaxies -

(1) How many has the Coma Cluster?
(2) How many hypergalaxies are there in Stephen's Quartet or Quintet?
(3) How many hypergalaxies have you discovered?

Einasto: (1) In rich clusters hypergalaxies have probably been

destroyed by close encounters.
(2) Stephen's Quartet is probably the core of a hypergalaxy.
(3) We have studied hypergalaxies in the northern hemisphere; our first
catalogue contains 60 objects.

Gursky: I am puzzled by your requirement that there should be an

enveloping corona around a hypergalaxy. How does your description
change if there is no such corona?
HYPERGALAXIES 61

Einasto: Available kinematic and morphological data strongly suggest

the presence of a massive corona. But formally a hypergalaxy can be
defined as a compact group of galaxies with one concentration centre.

Morton: What evidence do you have that other hypergalaxies besides our
own are distributed in a plane?

Einasto: In the NGC 4631 hypergalaxy, the main galaxy is seen edge-on
and has a flat rotation curve at a large distance, indicating the
presence of a massive corona. The optical companions have velocities
equal to those of the main galaxy. This is to be expected if the hyper-
galaxy is seen face-on.

Ekers: Westerbork HI observations of NGC 4631 (Weliachew and Sancisi,

Astron. Astrophys.~ in press) do not confirm the observations by Krum
and Salpeter that the flat rotation curve extends to a very great
distance. Further analysis of the Westerbork data by Sancisi shows that
there is no HI at the level claimed for the outer points. Consequently,
the flat rotation curve is only established to distances slightly
greater than the Holmberg radius.

Einasto: The Westerbork data show that the gas population has a smaller
extent. But over the whole observed range the rotation curve is flat
according to both the Arecibo and the Westerbork data.

Abell: In some respects, your work seems to me to parallel a similar

study by Holmberg of dwarf companions to galaxies. Do you find that
your data on the statistics of companion galaxies are in agreement with
those of Holmberg?

Einasto: Professor Holmberg studied companions up to a distance of

100 kpc from the main galaxy (for H = 50 km s-1 Mpc- 1 ). We study
companions up to a distance of ~ 1 Mpc. For this reason the results
are different. Our study is a natural continuation of Holmberg's work.

Holmberg: In my paper, I showed that one cannot find dwarf companions

at distances greater than 50 kpc because they get drowned in the back-
ground of unrelated distant galaxies.

Einasto: We do not agree with this conclusion.

Tifft: Is it correct (from your comments on NGC 4631 and the Local
Group diagram) that the plane of the hypergalaxy and that of the central
galaxy appear to be perpendicular?

Einasto: Yes, that is so.

N-BODY SIMUIATIONS AND THE VALUE OF n

J. Richard Gott, III

Princeton University Observatory

We have recently completed a series of N-body simulations of

galaxy clustering in an expanding universe (Aarseth, Gott and ~~rner
1977). The initial conditions and our results concerning galaxy
clustering will be summarized by Sverre Aarseth at this meeting. In
this paper I would like to tell about the implications of these models
for the value of n = 8wGp /3H02 (where fo is the present mean density
of the universe and H = go
kill s-l Mpc- is Hubble's constant). In the
standard Friedmann moaels with A = 0, n > 1 implies that the universe
will eventually recollapse while n < 1 implies the expansion will con-
tinue forever. As discussed in Gott, Gunn, Schramm, and Tinsley (1974),
there are a number of theoretical arguments to suggest that even the
unseen matter in the universe is clustered the way the galaxies are so
that virial mass determinations from groups and clusters and statistical
virial theorem methods can provide good estimates of the mean mass
density in the universe. We can utilize our N-body simulations to
check the accuracy of these techniques.

Our simulations contain 1000 equal point masses representing

galaxies contained in a spherical volume of present radius ~ 50 Mpc
(see Gott 1977 for more details). Two primary models have been analyzed,
an n = 1 Einstein - de Sitter model and an n = 0.095 open model with a
mean denSity compatible with a variety of arguments including the cos-
mological abundance of deuterium (Gott, Gunn, Schramm, and Tinsley 1974).
At the points in the simulations corresponding to the present epoch
both models have virtually identical power law covariance functions
str~tching over five decades in radius. Over the observed range
(10 > t;(r) > 1) both models have covariance functions which are best
fit by

(Gott, Turner, Aarseth 1977). This is in remarkable agreement with the

relation
s(r) Ct r-1. 8
63

M. S. Longair and J. Einasto leds.) , The Large Scale Structure of the Universe, 63-70. All Rights Reserved.
Copyright © 1978 by the IAU.
64 J. RICHARD GOTT. III

observed by Peebles (1974) over the same range. As Sverre Aarseth has
mentioned in his talk we believe the slope of the covariance function
in our models is influenced by relaxation effects. We find that the
slope of the covariance function is rather insensitive to initial con-
ditions. In particular for the models we have done it is impossible to
separate the n = 1 and the n = 0.1 models by looking at their covari-
ance functions. Thus it appears that there is little hope of determining
th~ value of n from studies of the covariance function over the range
10 - 1.

We have recently completed a preliminary analysis of the three

point correlation functions (Stark, Gott, and Aarseth 1977). The results
for both n = 1 and n = 0.1 models appear to be in good agreement with
the observational data of Peebles and Groth 1975.
While the n = 1 and n = 0.1 models have similar clustering prop-
erties they have rather different velocity distributions. The velocity
dispersions of galaxies relative to the Hubble flow and within clusters
in the 0 = 1 model are ~ 3 times as large as in the n = 0.1
model. This is simply because the galaxies in the 0 = 1 model, weigh
10 times as much as those in the n = 0.1 model. This large difference
in velocity dispersions makes it possible to easily distinguish between
the two models.

As Ed Turner described in his talk, we have used these N-body

simulations to check the group catalogue techniques of Gott and Turner
1977. We can see how well virial mass estimates from groups reflect
the true masses of the galaxies in the models. The N-body simulations
show that these techniques are accurate to about a factor of 2. The
simulations can be used to correct these methods for any systematic
errors. For the observational data this leads to corrected values of
n in the range
0.06 < n < 0.14
(Turner et al. 1977). This includes estimates using median M/L values
from all groups, and mean values from uncontaminated binaries, and
uncontaminated groups with 3 or more redshifts. It is interesting that
binaries give similar mass to light ratios (0 = 0.09) as do groups
(0 = 0.06) and clusters (n = 0.13).
Recently there has been renewed interest in statistical virial
theorem methods. Fall (1975) has pointed out that the excess potential
energy oW (per ga~xy) due to the clustering can be calculated by inte-
grating (s(r)/r) d r. Since the amplitude of the covariance function
is fixed by observation, the potential energy per unit mass oW is pro-
portional to n. Fall giv2s theore~i~al arguments suggesting that
o~(2/3)oW where oT = -¥p = ~ < (V-V )2 >. V is the root mean square
peculiar velocity of all galaxies in ~he sampl~ relative to the uniform
IIubble flow. The N-body simulations show that 5~(2/3)oW for all modelS
at the present epoch to an accuracy of 50% ~ott, Martin, Aarseth 1977).
N-BODY SIMULATIONS AND THE VALUE OF n 65

Fall adopted Vp ~ 300 km s-l and using the amplitude of the covariance
function found by Peebles deduced:

It = 0.05

Davis, Geller and Huchra (1977) have reanalyzed this problem using
a complete redshift sample of galaxies brighter than 13th magnitude.
If all galaxies had peculiar velocities Vp relative to the Hubble flow
then random pairs of galaxies should have line of sight velocity dif-
ferences of t:N r = 12' Vp/f'j. Such velocity differences between galaxies
can be measured for galaxies with separations of ~ 1 Mpc. The amplitude
of the covariance function is high enough that most such pairs seen in
the sky are real pairs and not background foreground projection effects.
The r.m.s. value of 6V f is computed using the method of Geller and
Peebles (1973): they f~nd 6V ~ 300 km s-l as compared with 6V r ~ 270
km s-l found previously by G~ller and Peebles with a smaller sample of
galaxies. This result is supported by Gott, Martin, and Aarseth (1977)
who find 6V r ~ 300 km s-l for an incomplete redshift sample in the
northern sky. (In all these studies the Virgo cluster is excluded
from the samples because with it removed the covariance functions of
these samples are equivalent to those obtained in deeper surveys and
have the appropriate power law shape. If Virgo is included it dominates
the covariance function and the extra potential energy due to it would
have to be included. Also Virgo may contain background foreground
contamination problems.)

Using V =..f3' 6V r /i2' , Davis et al. deduce It = 0.46 for the

northern gaiRctic cap and It = 0.23-ror-the southern galactic cap. They
also estimate the mean luminosity density in each region and find
1.0 x 10~ L® Mpc- 3 and 5.5 x 107 L® Mpc- 5 respectively. From deeper
surveys they deduce that 1he mean ~uminosity density for a fair sample
of the universe is 6 x 10 L® Mpc-. Thus it is no mystery why the
northern galactic cap yields a value of It that is higher by a factor of
two; that region simply contains twice as many galaxies as the average
for the universe. If the values are normalized to the average luminosity
densitY,both the north and south give similar estimates of It. The
average is It = 0.26.

A study of the velocity distributions in the N-body simulations by

Gatt and Aarseth (1977) indicates the pair velocity differences are
quite isotropic at all scales: thus if we pick any pair of galaxies
their peculiar velocity difference vector is uncorrelated with their
separation vector. This means when we sample close pairs in the sky
6Vr "'" 6Vtot/ -(3! Both It = 1 and It = 0.1 models show 6Vr to be independent
of radius for scales near 1 Mpc. This is in agreement with the observa-
tions. Gatt, Martin and Aarseth (1977) have shown that the statistical
method of Geller and Peebles does yield approximately correct estimates
of 6Vr at 1 Mpc. An interesting result found by Gatt and Aarseth (1977)
is that the true value of V1) is given by approximately Vp ~ 6V r where
6V r is the radial velocity uifference of pairs at 1 Mpc, (this relation
ho~ds for both the It = 1 and the It = 0.1 models) rather than the naive
66 J. RICHARD GOTT III

estimate Vp ~ {3' t::Nr/~. This is perhaps even more surprising when

one considers that motions on scales larger than 1 Mpc could in prin-
ciple boost Vp above the naive estimate. However, it is easy to see
how this comes about. V is the r.m.s. average value for all galaxies,
while 6V r is the averagePvalue for pairs. Consider the following
example: one cluster of 100 members and a velocity dispersion of
1000 km s-l, 10 small groups with 10 members each and velocity dis-
persions of 300 km s-l and 100 field galaxies with velocities of
100 km s-l relative to the Hubble flow. Say further that the clusters
and groups have sizes ~ 1 Mpc so that all have the same MIL ratio;
assume the field galaxies have no neighbors within 1 Mpc. For this
sample, V = 606 km s-l. Now the 100 galaxies in the cluster produce
4950 pair~, while the 100 galaxies in the groups produce only 450 pairs
and the 100 field galaxies produce no pairs at all, giving 6V r =
785 km s-l. So V = 0.8 6V r for this case. In principle one must know
the multiplicity ¥unction of galaxies (i.e. the distribution of group
sizes (cf. Gatt and Turner 1977) to correct 6V r for these statistical
effects and determine Vp' The multiplicity function may be determined
by making a group catalogue. With a proper treatment even the statis-
tical virial theorem methods require some knowledge of the groups
present. This brings us surprisingly close to the group catalogue
methods with which we started. Those methods do not throwaway the
additional information available as to which pairs actually go together
to form a group. By utilizing more of the available information group
catalogue methods may be even more accurate than the statistical virial
theorem methods. The N-body simulations show that only ~ ~ of the
galaxies have neighbors within 1 Mpc and that there are a large range
of cluster sizes. The N-body simulations have multiplicity functions
quite similar to those observed so we can regard the estimate
Vp ~ 6V r (1 Mpc) as reasonably reliable. This lowers the estimate of
Davis et al. by a factor of 3/2 to give:

n = 0.18
with an uncertainty of a factor of 2 due mainly to the uncertainty in
the amplitude £f the covariance function. If we used this value of
Vp ~ 300 km s- with the amplitude given by Peebles we would obtain
Fall's result n = 0.05jthe difference in n values is due to the fact
that Davis et al. find an amplitude of the covariance function that is
considerably lower than that found by Peebles.

Davis et al. also use solutions of the truncated BBGKY hierarchy

equations obtained by Davis and Peebles (1977) which give
Vp ~ ~ l:Nr(l Mpc) and yield values of n ~ 0.6. The N-body simulations
indicate that this BBGKY technique overestimates Vn by a factor of fr
and n by a factor of 3. In solving the truncated ~BGKY equations a
number of ad hoc approximations are made concerning both the evolution
of the two and three point correlation functions and the form of the
velocity distributions and the problem may lie in one or more of these
approximations.
N-BODY SIMULATIONS AND THE VALUE OF r2 67

Peebles (1976) has formulated a statistical virial theorem method

based on the observed amplitude of the three point correlati~ function
(peebles and Groth 1975). This predicts DV r (3 MPci ~ 830 Dl 2 kID s-l.
Since the observations show DV r (3 Mpc) ~ 300 km s- this gives

D = 0.13

This figure is based on Peebles original estimates of the amplitude of

the covariance function and should be compared with Falts value of
D = 0.05 for the same assumptions.

The Geller and Peebles (1973) statistical vi rial theorem gives

MIL ~ 140 and with the luminosity density found by Davis et al. yields
D = 0.12.

In conclusion/the different suitably corrected statistical virial

theorem methods yield values of D in the range

0.05 :(: D :(: 0.18

These results are consistent with those found by the group catalogue
methods and are inconsistent with D > 1 due to matter associated with
galaxies. The results are consistent with the value of D =_£.1 im~lied
from cosmological production of deuterium with Ho = 50 km s Mpc-
(Gott, Gunn, Schramm, and Tinsley 1974).

REFERENCES

1. Aarseth, S.J., Gott, J.R., and Turner, E.L.: 1977, in preparation.

2. Davis, M., Geller, M., and Huchra, J.: 1977, preprint.
3. Davis, M., and Peebles, P.J.E.: 1977, "Astrophys. J. Supp1."
34 (4), p. 425.
4. Fall, S.M.: 1975, "Monthly Notices Roy. Astron. Soc." 172, p. 23.
5. Geller, M., Peebles, P.J.E.: 1973, "Astrophys. J." 184, p. 329.
6. Gott, J.R.: 1977, Transactions of the I.A.U. (Grenoble) Vol. XVI A.
7. Gatt, J.R., and Aarseth, S.J.: 1977, in preparation.
8. Gatt, J.R., Gunn, J.E., Schramm, D.N., and Tinsley, B.M.: 1974,
"Astrophys. J." 194, p. 543.
9. Gott, J.R., Martin, E., and Aarseth, S.L.: 1977, in preparation.
10. Gatt, J.R., and Turner, E.L.: 1977, "Astrophys. J." 213, p. 309.
11. Gott, J.R., and Turner, E.L.: 1977, "Astrophys. J." 216, p. 357.
12. Gatt, J.R., Turner, E.L., and Aarseth, S.L.: 1977, in preparation.
13. Peebles, P.J.E.: 1974, "Astrophys. J. Letters" 189, p. L51.
14. Peebles, P.J.E.: 1976, ''Astrophys. Space Sci." 45, p. 3.
15. Peebles, P.J.E., and Groth, E.J.: 1975, "Astrophys. J. Supp1."
196, p. 1.
16. Stark, A., Gatt, J.R., and Aarseth, S.J.: 1977, in preparation.
17. Turner, E.L., Aarseth, S.L., Gatt, J.R., Blanchard, N.T., and
Mathieu, R.D.: 1977, in preparation.
68 J. RICHARD GOTT, III

DISCUSSION

Jones: If the peculiar velocities on large scales were as large as

indicated by the 24 hr microwave background anisotropy (say 650 km/sec),
how would this affect your estimate of n?

Gatt: Sandage and Tammann's studies indicate that perturbations of the

Hubble flow within the local supercluster are less than ~ 250 km s-l.
In any case if a peculiar velocity of the Earth were produced by
galaxies within ~ 20 Mpc, then the direction of the predicted motion
should be roughly in the direction of the Virgo cluster and should be
~ 250 km s-l. The recent microwave background studies, if correct,
suggest a velocity of ~ 600 km s-l in a different direction. This we
would have to ascribe to a bulk motion of the whole local supercluster
due to clustering on scales ~ 50 Mpc. The values of n, deduced above
from comparing the peculiar velocities of galaxies relative to the local
supercluster with the clustering within the supercluster would be
unaffected. A separate estimate of n can be obtained from the bulk
motion of the supercluster, if one knew the shape of the covariance
function from 50 Mpc out to the current Hubble radius. Unfortunately,
no observational data on this exists. Using the theory of Gott and
Rees for the covariance function at large scales, I have recently calcu-
lated that a motion of 600 km s-l for the local supercluster would imply
a value of n ~ 0.2.

Van der Laan: If in your simulations you were to introduce a mass

spectrum and a schematic form of tidal friction with its resulting mass
segregation, have you any idea of the effect on your results?

Gott: We have new simulations using 4000 bodies in which the masses of
galaxies are distributed according to a realistic Schechter type lumin-
osity function, but we have not analysed these yet. We have done
simulations where 2/3 of the galaxies have mass 1.0 and 1/3 of the
galaxies have mass 2.0. At the end the heavy galaxies have a covariance
function with approximately the same slope, but twice the amplitude of
the low mass galaxies, in accordance with theoretical expectations.
There is some evidence from studies of binaries and groups that E and SO
galaxies have M/L values ~ 2 that of spirals. This might explain why
Davis and Geller find that in a magnitude limited survey the covariance
function of E and SO galaxies is just twice the amplitude of that for
spirals.

Audouze: With the values for n which come out from your talk (n ~ 0.1)
it seems to me that according to Gott, Gunn, Schramm and Tinsley deuter-
ium may not be synthetized in sufficient quantities in a canonical model
of Big Bang nucleosynthesis.

Gatt: For Ho = 50 km s-l Mpc- 1 the value of n predicted by deuterium

synthesis is n = 0.1. I think that given the uncertainties, the esti-
mates of n from deuterium production and dynamical measurements are in
N-BODY SIMULATIONS AND THE VALUE OF .n 69

nice agreement. Of course, we think that it is quite suggestive that

these two completely different methods give similar values.

Ozernoy: Why did you not obtain by numerical stimulations a cutoff in

the covariance function due to the fact that gravitational instability
does not work at redshifts smaller than about n- 1?

Gott: The cutoffs at ~(r) ~ n- 3 , predicted by some theories for the low
n models due to exactly the effect you mention, have not shown up in the
N-body simulations. We have several lines of evidence to suggest that
non-linear relaxation effects are important in establishing the slope of
the covariance function over the observed range. Aarseth will talk
about this tomorrow.

Peebles: I hope it is accepted that the fact that richer groups contri-
bute more pairs than poorer ones causes no systematic error in the
esimate of ~Vr' if one does it right. In the form of the virial theorem
I like best at the moment, one uses ~Vr directly, with no attempt to
deduce Vr , and one relates this to an integral of the three-point corre-
lation function. This gives rather a higher n than Dr Gott mentioned.

Gott: The statistical virial theorem methods mentioned by Peebles do

calculate ~V in the proper way but they have implicit assumptions that
may bias therresults in large virialized clusters. In these clusters,
which contribute a significant fraction of the total pairs, the veloci-
ties of the individual pairs of galaxies are not due to their motion
about each other but to their random motion in the whole cluster. This
may well make the cosmic virial theorem estimates somewhat too high. If
one uses ~(r) z 68 r-l. 77 'as found by you and substitutes ~Vr ~ 300 km
s-l as found by Davis, Geller and Huchra into your cosmic virial theorem
using the 3-point correlation function it gives n = 0.13.

Davis: I would like to disagree slightly with the conclusions you

reached concerning my work with Geller and Huchra. We derive a lower
limit of n in the South of 0.26, which if translated to a fair sample
density would suggest n ~ 0.3. This estimate of n is a lower limit
because it does not include any peculiar motion on large scales and it
is not quite fair to conclude that large scale motion does not exist in
the Universe because it is not found in the N-Body simulations of the
Universe.

Gott: The lower limit you found used the naive estimate Vp = (1:3/1:2)
~Vr. The N-body simulations indicate that rather than a lower limit
this is in fact an overestimate. Large scale motions do boost Vp as you
suggest, but the statistical effects I mentioned have an even stronger
effect in decreasing Vp. The N-body simulations certainly do have large
scale peculiar motions as can be seen by inspection of redshift space
pictures. The simulations include both effects and give Vp ~ ~Vr. Thus
we would correct the n = 0.3 estimate you mention downward by a factor
of (3/2) to give n = 0.17.
70 J. RICHARD GOTT. III

Fall: In attempts to estimate the rms velocity of galaxies with respect

to the Hubble flow (V p ) by comparison with the relative velocities of
pairs (~Vr) it is important to recognize that in principle Vp and ~Vr
can have different scale dependences. Would you comment on the results
of your numerical experiments within this context?

Gatt: The N-body simulations indicate that ~V ~ const for pair separa-
tions 100 kpc < r < 3 Mpc as is found in the oEservations. The Vp we
are interested in measuring is the rms peculiar velocity of galaxLes
with respect to the mean Hubble flow defined for a large homogeneous
sample (r ~ 50 Mpc). Motions on scales 1 Mpc < r < 50 Mpc can boost Vp
relative to ~Vr measured at 1 Mpc. However the statistical effects I
mentioned in my talk make Vp lower with respect to ~Vr than one would
otherwise expect. The N-body simulations which produce reasonable
covariance functions include both these statistical effects and the
effects of large scale motions. They give the empirical result Vp~6Vr.

Tully: Implicit in your discussion is the assumption that most of the

mass in the Universe is distributed like the galaxies. This assumption
may well not be correct.

Gatt: These simulations do assume that the majority of the mass in the
Universe is clustered like galaxies. This includes any unseen matter
which falls into groups and clusters. While it is conceivable that most
of the mass is in some homogeneous component which does not participate
in the clustering, there are theoretical difficulties with this as out-
lined by Gott, Gunn, Schramm and Tinsley.

Fridman: Did you consider plane systems?

Gatt: No.
THE GALACTIC NEIGHBOURHOOD

G. A. Tammann 1) 2) and R. Kraan 1)

1) Astronomisches Institut der Universitat Basel
2) Hale Observatories, Pasadena, and
European Southern Observatory, Geneva

ABSTRACT
Several properties of the 131 galaxies known within 9. 1 Mpc are investi-
gated. 88 of these galaxies are concentrated into eight groups, leaving 33 percent of
true field galaxies. There are EI SO and SO galaxies among the field galaxies; their
types must be of cosmogonic origin. The groups have small velocity dispersion which
limits the mean mass-to-light ratio for the different types of group galaxies to 'JJl1 L <
20. Within the supergalactic plane the deviation from an ideal Hubble flow are small:
the changes of AHol <Ho> with distance and direction are not larger than ten percent;
the radial component of the peculiar motion of field galaxies is <25 km S-l. The
differential luminosity function of SlIm galaxies is well approximated by a Gaussian
with <M> = -15~7 and (J = 3~3. The luminosity function of EISO galaxies is much flatter
with a possible minimum, separating true E' s and'dwarf ellipticals (Reaves, 1977).
The sample galaxies are strongly concentrated toward the supergalactic plane; at a
distance of 4 Mpc of the plane the luminosity density drops to half its value. There is
also a pronounced luminosity density decrease with increasing distance from the Vir-
go cluster centre; at a distance of 30 Mpc the density has decreased by more than a
factor of 104 . The best estimate of the mean luminosity density within a sphere of 30
Mpc radius centered on the Virgo cluster is 1. 5 . 10 B L0 Mpc- 3 .

This is a first attempt to determine the mean properties of galaxies with-

in a dis tan c e -limited sample. &tch a sample should be as complete a s possible for
intrinsically faint galaxies, but it should also contain a statistically significant number
of galaxies. The best compromise seems to be a velocity limit of 500 km S-l,
corresponding to a distance limit of 9. 1 Mpc (a value of Ho = 55 km S-l Mpc- 1 is
assumed throughout; Sandage and Tammann, 1976).

1. THE SAMPLE DEFINITION

The catalogue of galaxies known with v 0 ~ 500 km S-l is given elsewhere
(Kraan and Tammann, 1978). It contains 184 entries. Galaxy types are from Sandage
and Tammann (1978) and other sources. The magnitudes are in general in the BT
system (de Vaucouleurs et al., 1976); for some fainter galaxies they are still provisio-
nal. The magnitudes are corrected for galactic absorption (Sandage, 1973) and for the
full amount of internal absorption (Holmberg 1958; 1964; with minor modifications for
71

M. S. Longair and J. Einasto (eds.); The Large Scale Structure of the Universe, 71-91. All Rights Reserved.
Copyright © 1978 by the fA U.
72 G. A. TAMMANN AND R. KRAAN

o ,roup' wlHun SOO km ,-I

Fig.I: The sample galaxies projected onto the sphere of supergalactic coordinates (the Local Group
members are excluded). The "zone of avoidance" Ubi < 15 0 ) is delineated with a heavy line The con-
centration of the galaxies toward the supergalactic equator is apparent. 46 galaxies lie within 498 of
the Virgo cluster center (the area is shown here too large); they are bona fide members of this cluster
(including the only known seven galaxies outside the Local Group with negative corrected velocities,
disqualifying the assumption that these galaxies could be foreground Objects, cf. Sandage and Tammann,
1977). The boundaries of seven nearby groups are shown with full lines; all sample galaxies within
these boundaries are considered as group members. Six galaxies with 400 < vo <. 500 km 8- 1 have been
proposed as members of more distant groups (de Vaucou1eurs, 1975); these groups are schematically
shown with broken lines.

the most edge-on galaxies). Ab solute magnitudes of Local Group and M 81 group mem-
bers are derived from individual distance determinations; for all other galaxies they
are calculated from the radial velocity v 0 (corrected to the Local Group; cf. Yahil et
al., 1977) and Ho. A justification of this procedure is given below (section IV).

It is believed that the present sample is essentially complete for absorp-

tion-corrected magnitudes of 12~ 0 for E/ SO galaxies and of 1 P 5 for all later types
(Sa-1m). At a distance of 9.1 Mpc this corresponds to a relatively bright completeness
limit in absolute magnitude of -18~ 3. But it should be noted that - whatever the lumi-
nosity function - galaxies with ~ -18~ 3 contain> 90 percent of all light.

II. THE DISTRIBUTION AT THE SPHERE: GROUP AND FIELD GALAXIES

The distribution in supergalactic coordinates of the galaxies in the pre-
sent sample is shown in Fig. 1. Three striking results emerge: (1) The galaxies are
strongly concentrated toward the supergalactic plane; (2) The galaxies are concentra-
ted to form groups; (3) 46 galaxies fall within 4~8 of the Virgo cluster centre. They
are certain members of this cluster and lie therefore outside the 9.1 Mpc limit. They
are excluded in the following discussion. Seven additional galaxies have also been ex-
cluded because they may belong to more distant groups (de Vaucouleurs, 1975). This
THE GALACTIC NEIGHBOURHOOD 73

Tab 1· The Distribution with Type

E SO Sa-Sd Sdm-Im Irr !:

< -18~ 3 2 2 26 4 1
34
6% 6% 76% 12% ( 3"/0)

> -18~ 3 15 5 11 63 1
94
16% 5% 12% 67% (1%)

leaves 131 galaxies which are known within a distance limit of 9.1 Mpc; they are re-
ferred to in the following as the" sample galaxies". Only 60 sample galaxies are con-
tained m the Shapley-Ames (1932) catalogue.

The distribution of the sample galaxies according to type is given in

Tab.l. The distribution is strongly luminosity dependent. Because it is also dependent
on position (there are clusters known with only E/ SO's) the relative type distribution
has no general meaning.

A. Groups of Galaxies.
Allowing for the projected position and to some degree for individual
distances and velocities it is possible to define seven groups containing 62 sample
galaxies. The following facts speak in favour of the reality of these groups: (1) Five
of the groups are historically well established; only two groups were formerly be-
lieved to be part of a larger complex (B4 and B5 in the CVn cloud); (2) With a mini-
mum of 5 and an average of 9 members with known red shift the groups are exceptio-
nally well defined; (3) All group members have Vo < 400 km S-l (the only exception
is the most distant M 101 group), it is therefore very unlikely that additional group
members could be found outside the 500 km S-l limit; (4) All galaxies within the
group boundaries are included; (5) The inclusion of any additional outlying galaxy
would more than double the total kinetic energy of that respective group; it is a priori
improbable that the galaxy, which is in projection the most outlying member, should
have the highest kinetic energy. (6) The groups are well separated in projection
and - according to several distance indicators - in space; (7) The group galaxies
are much more concentrated toward the super galactic plane than the field galaxies
(cf. Fig. 7).

There remain only a few ambiguities: Six dwarf galaxies between the
B3, B4, and B5 group have been assigned to the field; if they were treated as mem_
bers of any of these groups the following conclusion would not be altered because their
total light and mass is negligIble. The group B7 may actually consist of two separate
groups, a nearer southern group (containing NGC 55 and 300) with <vo> = 136, ( j =
21 km S-l, and another group with <v 0> = 248, CJ = 39 km S-l.

The seven groups, Bl to B7, and the Local Group (Yahil et aI., 1977)
are listed with some of their properties in Table 2. The groups contain all types of
galaxies, but true E/SO members are rather rare. Of the 18 E/SO group members
74 G. A. TAMMANN AND R. KRAAN

Tab.2: Groups within <vo> .: 500 km a- 1

Group Brightest n Gal E/SIJ Sa-Sd Sdm-Im Distance Mean radius Total Lum. Lum.Dens. <v> a(vo '
0
Member (Mpel (IOIOLElI Contrast
(Mpel

LG M31 26 13 3 10 - 1.5 11. 03 37 - 45

Bl Ie 342 5 1 3 1 4.1 0.5 7.41 865 224'1:23 50
B2 M81 15 1 4 8+ 2 Irr 3.3 0.6 4.36 229 24ot22 83
B3 MIDI 7 - 4 3 6.9 0.5 5.31 383 368t23 60
B4 NGC4449 12 1 2 9 4.7 0.6 I. 54 77 25H8 27
B5 NGC4736 10 - 4 6 6.3 1.0 4.98 49 347t8 25
B6 NGC 5128 7 2 2 3 6.5 1.0 13.87 179 232t21 56
B7 NGC55 6 - 5 1 3.3 0.5 5.54 570 1921:28 68

only four are true E's (Maffei 1 being the only bright one, the others - M 32, NGC 147
and 185 - being at the faint end), one is E/SO, and three are SO's. The remaining E's
have MB > -14" and should be classified as dwarf ellipticals (dE). The dE's are
known so far to occur only in groups and in the Virgo cluster (Reaves, 1956; 1977).
They may constitute a separate type of galaxies as further elaborated in section V. -
Among the group members are also the only two old-population irregulars, the irre-
gulars of type II (abbreviated here as Irr).

All sample galaxies together have a luminosity of 6.6 . 10 11 L 0 . Within

the sample volume of 3. 16 . 10 3 Mpc 3 this corresponds to a mean luminosity density
of 2. 1 . lOB L0 Mpc- 3 • The ratio of the luminosity density within each group and the
mean density defines what is given in Table 2 as luminosity density contrast. It can
be seen that the Local Group is not only the most extended group (its size is indepen-
dent of Ho) but also the group with the lowest contrast. This effect would be enhanced
if a larger value of Ho would have been chosen. This speaks against values of Ho
larger than 55. (It should be noted, however, that the Local Group shows a clear
luminosity concentration toward the centre, and the knowledge of faint dwarf members
in other groups may not be complete yet).

The last column of Table 1 contains the velocity dispersion within the
groups. These values are surprisingly small. It is clear that these small values can
only be found if exceptionally good red shift determinations are available. For
the present sample of nearby galaxies with many 21 cm-redshifts this is the case.
But since the typical mean errors of optical redshifts are in the order of 100 km S_l
(Sandage, 1978) it is in general quite difficult to find the true velocity dispersions of
groups.

B. Field Galaxies.
Fortythree sample galaxies, i. e. 33 percent, cannot be assigned to any
group. As mentioned above some of these galaxies may still be group members, but
the majority are certainly true field galaxies. The strongest evidence for this result
is the widely different concentration toward the supergalactic plane, the group
galaxies having a mean distance C from this plane of less than 1 Mpc and the field
galaxy one of more than 2 Mpc. Since the true galaxy density decreases with increa-
sing I;' the chance is considerably reduced to ever find group associations for the
field galaxies.
THE GALACTIC NEIGHBOURHOOD 75

Tab. 3: Field Galaxies

Type n Example C (Mpc) ntotal Fraction

E 0 - - 4+ 10 dE -
EISO 3 NGC 3115 -6.1 4 750/0
SO 3 NGC 404 0.5 6 50
Sab-Sb 2 NGC 2683 -3.5 6 33
Sbc-Sd 9 NGC 2903 -5.1 31 29
Sdm-Im 26 NGC 2188 -8.5 67 39

All types of galaxies are represented among the field galaxies with the
exception of pure E's, dwarf ellipticals (dE's) and old-population irregulars (Irr's).
The lack of E' sand Irr's is hardly significant because of their small number within
the present sample. The absence of field dE's is intriguing: it could be an observatio-
nal effect in view of the difficult detection (due to low surface brightness) and red-
shift determination of these galaxies, - but since there are not even candidates known
to be field dE's it is a reasonable working hypothesis that they may be confined to
groups and clusters.

Table 3 gives examples for field galaxies of different type. These ga-
laxies have exceptionally high C values. The one exception is the SO galaxy NGC 404,
but this galaxy is quite isolated. The galaxies in Table 3 are therefore most likely
true field galaxies.

In spite of the fact that no true E galaxy is known in the field, early-
type galaxies taken together (E to SO, excluding dE's) are locally at least as likely
to be field galaxies (43 percent) as later-type galaxies (36 percent).

The scarcity of early-type galaxies within the small groups of the pre-
sent sample. the occurrence of several EI SO galaxies in the rich Leo group. and the
preponderance of these galaxies in rich clusters suggest a positive correlation
between the relative number of early-type galaxies and the population size of the
galaxian aggregate. This could be taken as evidence that spirals evolve into SO andlor
E galaxies and that this transition is more pronounced in rich clusters with strong
interaction between the member galaxies. However. the fact that early-type galaxies
do occur as isolated field galaxies proves that not all early-type galaxies can be
formed by this process. Their origin must be cosmogonic.

III. THE MASS OF GROUP GALAXIES

An application of the vi rial theorem to the groups Bl to B7 gives virial
masses which differ by more than a factor 100. To reduce the scatter of individual
groups it is probably best to compare the total kinetic energy with the total potential
energy contained in these groups.
76 G. A. TAMMANN AND R. KRAAN

Adopting conventional mass-to-light ratios of 4, 20, and 30 (in solar

units) for Slim, SO, and E galaxies, respectively. one finds that the overall kinetic
energy is T = (1. 9 ± 0.3) . 10 62 J and that the overall potential energy isn=
1. 25 .
10 62 J.

If one assumes that the groups are relaxed one finds 2T/ 0.= 3.0, which
would suggest that the above mass-to-light ratios are too low by this factor. However,
due to the small velocity scatter in the groups the crossing times are very long
(1. 4 to 8· 1010 yr). It is therefore very que stionable whether the groups should be
assumed to be relaxed. All what one can then require is that the groups have negative
total energy. In this case the adopted mass-to-light ratios are too small by only a
factor of 1.5. This would roughly correspond to arne a n mass-to-light ratio of all
types of galaxies in the present groups of 'JJt/L = 17. But actually this is an upper limit
because the groups may very well have positive total energy (with crossing times
equal to or exceeding their age) and therefore may be in expansion. The result of
<'5Jl./L> ~ 20 (out to radii of the order of 1 Mpc) - and probably less for Slim galaxies
- can only be avoided for the present group members if one forcefully attributes
additional, outlying members to the groups. From the present material this would
seem to be quite artificial.

In a pioneering paper Kahn and Woltjer (1959) have shown, that the
apparent approach of M 31 and our Galaxy gives a handle to the mass determination
for these galaxies. The discussion of modern observations requires in fact that 'ill
(M 31 + Galaxy) > 2.8 . 1011 'ill0 (Yahil et al., 1977) or = 8· 1011 ~ (Lynden-Bell and
Lin, 1977). With a combined luminosity of these two galaxies of 10 1 L 0 this corres-
ponds to 'JJt/L > 2.8 or = 8. The result for these two spirals is in excellent agreement
with the above result for the groups B1 to B7. It is further strengthened by the fact
that there are other groups outside the present sample which lead to similar 'JJt/L
values (Materne and Tammann, 1976).

Other investigations of groups have occasionally led to higher mass-to-

light ratios. These results are based on more distant groups. Since the definition of
groups becomes more difficult with increasing distance and since the quality of the
observations (magnitudes and velocities) is typically inferior at larger distances, the
higher!lll/L-values should probably be given lower weight.

Pairs of galaxies also tend to yield relatively low mass-to-light ratios

in agreement with the present groups (Karachentsev, 1977). Higher masses have been
derived by Turner (1976) for a sample of pairs with an optically determined mean
velocity separation of 205 km S-l, but this sample raises still statistical problems
(van Albada and Freeman, 1977), and in addition with a mean external error of at
least 140 km S_l for a single velocity observation (as suggested by six new 21 cm-
velocities) the signal-to-noise ratio remains rather precarious.

In the case of spiral galaxies 'JJt/L;<: 10 would imply the presence of

massive halos. The best evidence for such halos came from the very extended, f 1 a t
rotation curves of some edge-on spirals (Krumm and Salpeter, 1977). However, the
large HI extent of these galaxies has not been confirmed by Sancisi (1977).

The conclusion that the sample galaxies have <'JJt/L> <20 leaves the
problem of the stability of clusters unsolved. The nature of the so-called missing
mass in aggregates with large velocity dispersion lies outside the scope of this
paper. It is well possible that the missing mass resides in a few (E) galaxies which
are, for some reason, not represented within the groups of the present sample.
THE GALACTIC NEIGHBOURHOOD 77

25

,~A
u
20
Leo
~~
r/~'"
~
Q.

~
15

10

5
"~
1000 1500

Fig. 2: The distance _ mean velocity relation for six nearby groups with particularly well determined
distances (Tammann, 1977) and for the Virgo cluster (Sandage and Tammann, 1976). Three relatively
nearby. rich groups or clusters are added (open circles) whose distances are known relative to the
Virgo cluster (Vis.vanathan and Sandage, 1977). The full drawn line corresponds to Ho = 55 km s-'Mpc-'.

IV. LIMITS ON THE DEVIATIONS FROM AN IDEAL HUBBLE FLOW

One can think of three types of deviations from an ideal Hubble flow: A.
A distance dependence of Ho; B. A directional dependence of Ho; and C. Peculiar
motions of groups and field galaxies. Their possible size shall be investigated in the
following.

A. The Distance Dependence of H o '

Good distance determinations from three or more independent methods
are available for some nearby groups and clusters. They define a Hubble diagramme
as shown in Fig. 2. The expansion is well represented by a single value for Ho of 55
out to a distance of ~ 25 Mpc (cf. also Tammann. 1977). Several distance indicators
beyond that limit - ScI galaxies (Sandage and Tammann. 1975). supernovae (Branch.
1977). and brightest group and cluster galaxies (Sandage. 1975; Sandage and Hardy.
1973) - exclude any significant changes of Ho beyond this limit.

B. The Directional Dependence of Ho'

The individual values Hi of the Hubble constant for several well- studied
groups and galaxies (within <vo> = 1600 km s-') are plotted versus the supergalactic
longitude in Fig. 3. Also shown is the suggested variation of Hi with longitude (de
Vaucouleurs, 1976). The two sets of data are in clear contradiction. This suggests
that the individual deviations from the mean value Ho = 55 are due to still remaining
problems of the distance determinations. An exact solution for the possible variations
depends very much on the weight attributed to a single value of Hi' In any case the
change of 6H/<Ho> with direction is smaller than 0.15 (Sandage and Tammann, 1976)
and probably still smaller (0.08) within an even larger volume (Sandage. 1975). Thls
78 G. A. TAMMANN AND R. KRAAN

o 60 120 180 240 300 360"

SGL
Fig. 3: The Hubble constant H. versus the supergalactic longitude. The points represent the same groups
as shown in Fig. 2. The some~hat more distant Grus group «vo> = 1580 km s-') is added as an open
circle. The dashed curve represents roughly the change of H. with longitude as suggested by de Vaucou-
leurs (1976). The almost perfect anticorrelation between the ~ints and the dashed curve strongly
suggests that the deviations from <Ho> • 55 km S .. 1 Mpc" l are not real, but due to remaining errors of
the di stance determinations.

conclusion is also compatible with the Rubin-Ford effect, even if its size is taken at
face value (Rubin, 1977).

It has been proposed that the local frame of inertia has large systematic
motions (~ 500 km s-') with respect to distant ScI galaxies (Rubin, 1977) and to the
3K-background radiation (Smoot et al., 1977). If these, mutually hardly compatible
motions, are real it is clear that they must involve very large volumes. The above
limit of AH/H o .... 0.1 for <vo> .$ 3000 km S-1 requires that the co-moving volume
has a radius of;;:: 50 Mpc.

C. The Peculiar Motions of Galaxies.

The groups within 20 Mpc in Fig. 2 deviate from the mean Hubble line by
(] = 93 km S-I. This value, however, is still strongly affected by errors of the
distance determinations. Allowing for these errors one finds that no group deviates
significantly from the mean Hubble line by more than 50 km S-1 (Tammann, 1977).
The true deviations could be still considerably smaller. As for the Local Group
Visvanathan and Sandage (1977) have shown that its peculiar motion in the direction
of the Virgo cluster is smaller than the error of the mean cluster velocity (t 60 km S-1).

The velocity distribution of the field galaxies within 9.1 Mpc is plotted in
Fig.4. No field galaxy is known with Vo < 100 km S-I! This limits the peculiar motions
of the Local Group and of the field galaxies to less than 100 km S-' (cf. Fisher and
Tully, 1975). The assumption that the field galaxies have constant space density leads
to a still more stringent limit: Monte Carlo calculations show that their observed
velocity distribution can be understood if their random motions are ~ 35 km S-1.
This value is in good agreement with an earlier, independent determination (Sandage
and Tammann, 1975a).

The lowest upper limit for the peculiar motions of field galaxies is set
by the expectation that they should have smaller random velocities than galaxies within
groups. The observed velocity dispersion in groups can be as low as 25 km S-1 (e. g.
B4 and B5); therefore the radial component of the peculiar motions of field galaxies
are probably smaller than this value.
THE GALACTIC NEIGHBOURHOOD 79

6
N
Fig. 4: The velocity distribution of fi e I d SlIm
galaxies of the present sample. The curved line is
calculated under the assumption of constant space
density and allowing for the discrimination of
intrinsically fainter galaxies at larger distances
2
(the luminosity function of Fig 5 is used to deter-
mine the effect), Note that there is no galaxy
with Vo < 100 km S-l
o
Vo

An important proviso should be made. The test galaxies for which the
above limits on the non-Hubble motions were derived lie all close to the supergalactic
plane. It can therefore not be excluded that the motions perpendicular to the plane
were larger. In addition the peculiar motions of field galaxies were derived from near-
by objects (vo < 500 km s-'); it can therefore not be excluded that the peculiar motions
are larger at larger distances.

In any case for the present sample there is observational evidence that
groups have random radial motions of less than 50 km s-' and field galaxies of less
than 25 km s-'. This is a justification for the above procedure to derive luminosities
for the sample galaxies from radial velocities and Ho. For even the nearest field
galaxies with Vo = 100 km s-' the error in distance becomes < 25 percent correspon-
ding to < O~ 5 in luminosity. The same upper limit is derived for the error of even the
nearest groups.

V. THE LUMINOSITY FUNCTIONS

The determination of the faint end of the galaxian luminosity function is
hampered by the fact that existing catalogues are very incomplete for intrinsically
faint galaxies. (The Shapley-Ames catalogue does not contain galaxies with M ;G
-15~ 5). Therefore the present sample is particularly well suited to define the faint
end of the luminosity function, especially if only nearby sub sample s are studied, for
which the completeness of dwarf galaxies is exceptionally good. In the following the
luminosity functions of SlIm and EI SO shall be investigated separately.

A. The Luminosity Function of SlIm Galaxies.

There are clear indications that the luminosity functions of Sa, Eh and Sc
galaxies differ systematically. The small present sample does not allow to investigate
these types separately. It should be stressed, however, that their combination is
forced. On the other hand the distinction between Sc and 1m galaxies is probably
ba sicly a function of luminosity. There is no spiral known with M > -16" and no 1m
galaxy with M < -20·. In the transition interval the ratio of spirals to irregulars de-
creases monotonically toward fainter luminosities. Also all other known parameters
change smoothly from Sc to 1m galaxies. The lumping of these two types seems there-
fore natural.

The subvolume within which the knowledge of SlIm galaxies is most com-
plete is represented by the Local Group and the M 81 group. Very few if any SlIm
80 G. A. TAMMANN AND R. KRAAN

SlIm I complele for EISO I complele for

N In LG and M8 I Group
M81 Group
+ MS - 14':"0 ,n LGand
Mel Group
, MS-I4~0

O~~-;-r----~--~----r---Tr---,----.---~-=~~--4---~----r-~~--~~--.---~----r-'
N SlIm I complele for
, mS+ll'!'5
Else
011
011
20

15 (M) "-I!f.'7 ,

,, 1
,,"3~3 ~' complete for
, mS+12'!'e
,,
to
,,
5 ,,
, ,- "
o~-~-~~~~~~~~~~-4~~~~~~~~~
-6 -10 -14 -18 -22 -10 -14 -22

Fig. 5: The differential luminosity function for SlIm galaxies (left side) and EISO galaxies (right side).
The upper panel IS for members of the Local Group and the M8l group only. It is likely that all galaxies
brighter than MB = -l4~ 0 are known in these two groups. The best fitting curve for the SlIm galaxies
IS a Gaussian with <MB> AS -15~ 7 and a (M) ttl 3~ 3. The lower panel shows all known galaxies within
9.1 Mpc (va .. 500 Ian S_l). The Sa-Sd galaxies (white histogramme) and Sdm-Im galaxies (hatched area)
are fully compatible with the same Gausman luminosity function as in the upper panel. The SlIm galaxies
are fully corrected for internal absorption; their completeness limit is therefore somewhat brIghter than
for EI SO galaxies.

galaxies can have remained undiscovered which are more luminous than IC 1613 and
HoI. The 0 n e additional candidate is the newly discovered, probable Local Group
dwarf in Sagittarius (Cesarsky et al., 1977). Considering that even HoIX with M =
-13~ 5 in the M 81 group is a relatively easy object, leads to the conclusion that essen-
tiallyall SlIm galaxies brighter than -14" are known in these two groups. And the
observation that the advent of IIlaJ plates and 21 cm-surveys have so far contributed
only very few additional 1m members of the Local Group and the M 81 group strongly
suggests that the number of undiscovered objects even considerably fainter than -14"
must be quite limited.

The differential luminosity function for this subsamp1e is shown in Fig. 5.

The distribution shows a maximum near -15~ '1 which, from the above remarks on
completeness, is almost certainly real. The distribution of galaxies brighter than
-14" can be well fitted by a Gaussian with <M> = -15~ 7 and (] = 3~ 3. The Gaussian
predicts that about three faint 1m galaxies remain to be discovered in the sub sample.
This small number may be somewhat too conservative because the error of the mean
magnitude is almost 1": if < M > = -14~ 7 then there could be still ~ 12 undiscovered
faint 1m's.

The rarity of very faint 1m galaxies is further supported by a recent

THE GALACTIC NEIGHBOURHOOD 81

SlIm VIRGO CLUSTER

30 ,,
N ,,,
,
20 ,,
I

I
I
10 ,
I
,._.J
,
I
I

E/SO
10

40 Fig. 6: The differentiallu_

minosity function of Virgo
cluster members (Sandage
30 and Tammann, 1978) The
same Gau ssian is fitted to
N the SlIm galaxies as in
Fig.4. The SlIm galaxies
20 (upper panel) and EI SO ga-
laxies (middle panel) are
combined in the lower pa-
nel. Note, however. the
10Ld~ striking difference between
the luminosity functions of
these two groups of types

21 em-survey of the M81 group (Sargent and Lo, 1977), which has provided only a
few additional members, and also by the absence of faint Im' s in the Virgo cluster
(Reaves, 1977a).

The differential luminosity function of all SlIm galaxies in the present

sample is also shown in Fig. 5. The Gaussian luminosity function of the nearby sub-
sample represents a very satisfactory fit for these extended data. The same
Gaussian fits also perfectly well the bright wing of the luminosity function of SlIm
galaxies in the Virgo cluster (Fig. 6).

The peaked luminosity function of Sand Im galaxies has been anticipated

to some degree by Holmberg (1969) who found that the luminosity function of spiral
galaxies only can be represented by a Gaussian. Incidentally it should be remarked
that the present luminosity function is reminescent of Hubble's (1936) Gaussian lumi-
nosity function with < M > = -14~ 2, but with much smaller dispersion of (] = O~ 84.
However, his function was based on the old distance scale and on only Shapley-
Ames galaxies. The partial agreement is therefore fortuitous.

It should be recalled that all absolute magnitudes of SlIm galaxies in the

82 G. A. TAMMANN AND R. KRAAN

present paper are corrected for the full amount of internal absorption. (The mean
correction amounts to O~ 42; note that the internal ab sorption not only decreases the
true luminosity of a galaxy, but that it also tends to flatten the luminosity function at
its bright end). It has first been pointed out by Kiang (1961) that some form of ab-
sorption correction (he choose a correction to face-on orientation) is necessary in
order to free the galaxy luminosities from random (?) orientation effects and to derive
meaningful luminosity functions. There are many problems which require an absorp-
tion-corrected luminosity function, e. g. the true mass-to-light ratios, the mean
energy spectrum and the total energy production of SlIm galaxies, as well as the re-
lative luminosities of SlIm and (absorption-free) E galaxies. It could be argued that
it would be preferable to have absorption-uncorrected luminosity functions to derive
the mean luminosity density in very large volumes and the brightness of the cosmic
sky light. However, the greatest uncertainty of these two parameters comes from
the unknown relative frequency of SlIm and E galaxies within very large volumes,
and there is therefore no reason to aim for an otherwise meaningless (uncorrected)
luminosity function.

There is a suggestion in the present sample that the SlIm galaxies in

groups have a wider luminosity dispersion than the field galaxies. This would mean
that very bright spirals and very faint irregulars occur preferably in groups. But in
view of the small sample and of the selection effects the difference is not significant.

One could imagine that the faintest irregulars cannot bind their hydrogen
and that this were the explanation for the scarcity of such systems. However, the
observed relative hydrogen richness of dwarf irregulars (cf. Huchtmeier et al., 1976)
contradicts this explanation. It seems therefore that the shape of the luminosity func-
tion of SlIm galaxies is tied to their formation process.

B. The Luminosity Function of E/ SO Galaxies.

The few known E, SO and dE galaxies within the Local Group and the M 81
group do not define a luminosity function (cf. Fig. 5). The same holds true for these
galaxies within the whole sample because of the incompleteness bias. There is how-
ever a strong suspicion that the luminosity function is at its faint end considerably
flatter than for SlIm galaxies. This suspicion is supported by the EI SO members of
the Virgo cluster, which are completely known to ~ -16~25 (Fig. 6).

The Virgo cluster EI SO members make it even possible that their

differential luminosity function has a broad maximum between ~ -18" and -20'. Evi-
dence for such a maximum has been given earlier by Abell (1975) for the luminosity
function of several clusters. In any case the Virgo cluster contains very few early-
type galaxies with ~ -14" to -15", which gives strong support for a minimum of
the luminosity function (Reaves, 1977a). Beyond this minimum, toward fainter mag-
nitudes, the number of galaxies increases again according to Reaves (1977). It was
proposed above to restrict the designation dE to these rather numerous, very faint
galaxies. Their lower luminosity limit is yet undetermined.

The luminosity functions derived here contradict only seemingly pre-

vious determinations (for a good review see Felten, 1977). The typical observational
limit for earlier investigations was .$ -15", and down to this limit the total number of
galaxies may increase indeed and may possibly be represented by an e-function. But
it seems now very unlikely that such an e-function should be extrapolated - as origi-
nally suggested by Zwicky (1957) - toward fainter magnitudes.

The present results indicate that the problem of "the" luminosity

THE GALACTIC NEIGHBOURHOOD 83

SlIm

40
30
......®
...J
30 o
"0
N 20~
...J

20

10
10

20 EISO 15

-7.0 -3.5 0 3.5 7.0 -7.0 -15 0 3.5 7.0

~ (Mpc)

Fig. 7: Left side: The number distribution of SlIm galaxies (upper panel) and E/&.J galaxies (lower pa-
nel) of the distance tfrom the supergalactic plane. Right side: The luminosity distribution in function
of C for SlIm galaxies (upper panel) and E/&.J galaxles (lower panel). Galaxies in groups are shown as
white histogrammes, field galaxies a8 hatched areas. (Note: the concentration toward the plane ap-
pears somewhat exaggerated because the sample subvolumes decrease with increasing ()

function is more complex than generally assumed. Different types of galaxies have
clearly different luminosity functions. The overall luminosity function depends there-
fore on the relative frequency of different galaxian types. Since this relative fre-
quency is strongly dependent on position (e. g. intra- and extra-cluster regions) the
overall luminosity function cannot have a general character. This dependence on
position is even more severe because SlIm galaxies may have different luminosity
functions inside and outside of groups, and because dE galaxies may be confined to
aggregates of galaxies.

VI. THE LUMINOSITY DISTRIBUTION WITHIN THE METAGALAXY

In this section systematic changes are investigated of the luminosity
density within the present sample volume. Beyond this it is attempted to describe
the luminosity distribution - as contributed by EISO and SlIm galaxies - within a
volume centered on the Virgo cluster and with a radius of - 30 Mpc. This volume
has frequently been designated as the Local Supercluster. This term is avoided here
as too programmatic; instead the name Metagalaxy shall be used. The reason is that
the total mass within the volume appears to be concentrated in a disc with a strong
radial density gradient, and that there is only one major density maximum in the
centre of the system, all other density fluctuations being of secondary and tertiary
84 G. A. TAMMANN AND R. KRAAN

Q.
2

Fig. 8: The luminosity density of SlIm galaxies (filled

o circles; full line) and of E/SO galaxies (open circles,
dashed line) in function of the distance C from the
supergalactic plane.
2 4 6 8 10
I~I (Mpc)

importance. This picture has little to do with superclusters, which are commonly
assumed to consist of two of more com par a b 1 e members (clusters). (For practi-
cal purposes the designation" supergalactic coordinates" are retained).

A. The Luminosity Distribution Perpendicular to the Supergalactic Plane.

The present sample volume was cut with equidistant planes parallel to
the supergalactic plane into ten subvolumes. The distribution of the sample galaxies
and their combined luminosities within these subvolumes is shown in Fig. 7. The
histogrammes do not reflect the true density run with distance C from the super-
galactic plane (because of the variable size of the subvolumes), but they do show (1)
a roughly symmetric distribution about the super galactic plane; (2) a strong concen-
tration toward this plane; (3) a more pronounced concentration of group galaxies than
of field galaxies; and possible (4) a stronger flattening of the system containing SlIm
galaxies than that of the E/SO galaxies.

The true luminosity densities of the subvolumes (combining volumes

with equal values of C) are plotted - separately for E/so's and SlIm's - in Fig. 8.
In spite of some scatter of the individual points the SlIm galaxies exhibit a clear
trend: the mean density of 4.2 . 10 8 L 0 Mpc- 3 within 2 Mpc from the plane decreases
to half its value at t = 4 Mpc and decreases by more than a factor of 10 at , '" 8 Mpc.
Only 15 percent of the light in the sample volume is carried by E/SO galaxies; most
of this fraction is contained in only a few galaxies. Therefore a well determined
density gradient cannot be expected for this type of galaxies. Indeed it is not much
more than a guess that the E/'E'IJ galaxies are also concentrated toward the superga-
lactic plane and that this concentration is possibly less pronounced than for SlIm's.
It would be. of course. of far-reaching significance if the lesser concentration of
E/so's could be substantiated.

The degree of concentration of the local galaxies is surprisingly high:

75 percent of their total light lies within ± 4 Mpc of the plane. Within C < 3.5 Mpc
almost aU light is contributed by group members. above this limit it comes from
field galaxie s. although the latter carry only 18 percent of the total light.
THE GALACTIC NEIGHBOURHOOD 85

as

04
.--. field galaxies only

02
~.",
~0-
~ 00
.--.
...~ ~~------~-----r------~-----+'
g 5 101
ct

2
Fig. 9: Lower panel: The luminosity density of SlIm
galaxies (filled circles, full line) and E/SO galaxies
(open circles, dashed line) in function of the distance
from the Virgo cluster centre. The local value, con-
taining the Local Group. is shown in parentheses.
Only galaxies wi!hin 4 Mpc of !he supergaIactic plane
o are considered. The corresponding relation for field
galaxies only is shown in the upper panel.
10 5 0 -5 -10
Virgo_ r IMpel _ AntlYlrgo

B. The Radial Luminosity Distribution Within the Metagalaxy.

Analogous to the previous paragraph the sample volume was cut into
five subvolumes by equidistant planes perpendicular to the supergalactic plane and
to the line of sight toward the Virgo cluster centre. The luminosity densities contri-
buted by EISO and SlIm galaxies within the different subsamples are plotted in Fig. 9.
Only galaxies with t < 3.5 are considered here in order to minimize the variable (! )
contribution of low-density regions at high I; -values. Due to the grouping of galaxies
the individual points have a large scatter, but they exhibit a clear trend: the density
decreases by roughly a factor of 10 from the subvolume nearest to the Virgo cluster
to the most distant subvolume. The trend is well confirmed if only the field galaxies
are considered; although their number is smaller they show less scatter because they
are free of clumping effects (cf. Fig. 8).

It has been known for a long time that the galaxy density is much higher
in the Virgo direction than in the Anti-Virgo direction (Shapley and Ames, 1932;
Reiz, 1941; Sandage et al., 1972), but the present result is surprising in as far as
the density gradient is still so steep in a sample volume 20 Mpc away from the Virgo
cluster.

The next aim shall be to derive the radial density distribution within the
whole Metagalaxy. A first attempt to do this comes from Jones (1976). The solution
is repeated here with independent data, combining the present sample (with I; < 3.5
Mpc) with observations of galaxies within 10· of the Virgo cluster centre (Sandage and
86 G. A. TAMMANN AND R. KRAAN

11

:~
--"""-'0

.\
'
Slim
\,- - .

\, \ ····· ...0""-.
./.101

eta \

\~
\
CI \
o
.9
EISO
7

o 0.5 1.0 15
log rYorgo IMpel

Flg.l0: The decrease of the luminosity density PL with the distance from the Virgo cluster centre
(logarithmic scale). The data for the outer regions are from Fig. 9; the data for the inner region.
(r" 3.5 Mpc) are from Sandage and Tarnmann (1978). The symbol. are the aarne a. in Fig. 9.

Tammann, 1978). The resulting density profile is shown in Fig. 10.

The luminosity density varies from a central volume of 1.5 Mpc radius
to a distance of 28 Mpc by about four decades. Almost everywhere the luminosity
density of S/(Im) galaxies is higher than of E/SO galaxies. The present data do not
define the densities between r = 3.5 and 10 Mpc. A density minimum may exist in this
range, as weakly hinted at by the EISO galaxies. Such a minimum is expected for a
bound cluster embedded in a freely expanding field. The eventual prove or disprove
for the existence of the minimum shall have important consequences.

If one assumes rotational symmetry of the Metagalaxy the density pro-

files of Fig.10 can be integrated. Using a linear interpolation (in log-log) for the range
r = 3.5 to 10 Mpc probably leads to an upper limit of the total metagalactic light. The
integration gives then 0.8 - 2.3 . 10 '3 L(,) with about 20 percent of the light in E/SO's.
The range of total luminosity is determined by two extreme assumptions: (1) the
flattening of the Metagalaxy is everywhere the same as determined in paragraph Ai
and (2) the true luminosity distribution is more or less spherical. The mean lumino-
sity density in a sphere of 28 Mpc radius, centered on the Virgo cluster, becomes
then - 1.5 . 10" L(') Mpc- 3 within a factor of two. With the above mean mass-to-light
ratio of < 20 this corresponds to a mean mass density of < 2 . 10- 31 g cm-S. If one
allows for enough "missing mass" to bind the inner 3 Mpc of the Virgo cluster ('JJl/L '"
140) the mean density is increased to only < 3. 9· 1O- s1 g em -3. This is less than O. 07
times the critical Einstein-de Sitter density.

The present investigation concerns a tiny volume and it is not meant to

have a bearing on the large-scale structure of the universe. But any cosmological
model must allow for at least one cell with the properties described above.

Acknowledgements. We owe to Dr. A. Sandage for unpublished data and

most stimulating discussions. Others - too numerous to be named here - have contri-
THE GALACTIC NEIGHBOURHOOD 87

buted with their suggestions and discussions. Dr. J. Materne has kindly performed the
vi rial calculations for the groups of galaxies. We thank Mrs. M. Saladin and Mr.
D. Cerrito for having made the manuscript ready for reproduction. &tpport of the
Swiss National Science Foundation is gratefully acknowledged.

REFERENCES
Abell, G. 0.1975, Galaxies and the Universe, ed. A. and M. Sandage and J. Kristian, Chicago: The
University of Chicago Press, p.616.
Albada, T. S. van, and Freeman, K. C .1977, private communication.
Branch, D. 1977, Supernovae, ed. D. N. Schramm, Dordrecht; Reidel, p.21.
Cesarsky, D. A .• Laustsen, S., Lequeux, J., Schuster, H. -E., and West, R. M.1977, preprint.
Felten, J.E. 1977, Goddard Space Flight Center Preprint, X-602-77-162.
Fisher, J. R., and Tully, R. B.1975, Astron. Astrophys. 44, 151.
Holmberg, E. 1958, Medd. Lund Obs. Ser.lI, Nr.136. -
Holmberg, E. 1964, Ark.f.Astron. 3, 387.
Holmberg, E. 1969, Ark. f. Astron.s;- 305.
Hubble, E. 1936, The Realm of Nebulae, New Haven: Yale Univ. Press, p.159.
Huchtmeier, W. K., Tammann. G. A. ~ and Wendker, H. J.1976, Astron. Astrophys.~ 381.
Jones, B.J.T.1976, M.N.174, 429.
Kahn, F.D., and Woltjer,L.1959, Ap.J.130, 705.
Karachentsev, 1. D.1977, this volume. --
Klang, T. 1961, M. N.122, 263.
Kraan, R., and Tammanrl,'"G. A.1978, in preparation.
Krumm, N., and Salpeter, E.E.1977, Astron.Astrophys.56, 465.
Lynden-Bell, D., and Lin, D.N.C.1977, M.N.181, 37.
Materne, J., and Tammann, G. A.1976, Proceedings of the Third European Meeting, ed. E. K.
Kharadze, Tbilisi, p.455.
Reaves, C. 1956, A. J. 61, 69.
Reaves. C. 1977, ProcYonference Evolution of Galaxies and Stellar Populations, New Haven:
Yale Univ. Press, in press.
Reaves, G. 1977a, private communication.
Reiz, A. 1941, Lund Cbs. Ann. No.9.
Rubin, V.C. 1977. I.A.U.Coll.37, li9.
Sanclsl, R. 1977, LA. U Symp. n. in press.
Sandage, A. 1973, Ap. J.183, 7ll.
Sandage, A. 1975, Ap. J.202, 563.
Sandage, A. 1978, in pre~
Sandage, A., and Hardy, E.1973, Ap. J.183~ 743.
Sandage, A., and Tammann, G.A.1975,~J.197, 265.
Sandage, A., and Tammann, G.A.1975a, Ap.J-:T97, 313.
Sandage, A., and Tammann, G. A.1976, Ap. J. 2~7.
Sandage, A., and Tammann, G. A.1976a, Ap. J-:207, Ll.
Sandage, A., and Tammann. G. A.197S. in preparation.
Sandage, A., Tammann, G.A.,. and Hardy, E.1972, Ap.J.172, 253.
Sargent, W. L. W., and Lo, K. Y.1977, Ann. Report Director Hale Obs.1976/77.
Shapley, H., and Ames, A.1932, Harvard Ann. 88, No.2.
Smoot, G.F., Gorenstein. M.V., and Muller. 1f.A...1977. preprint.
Tammann, G.A.1977, I.A.U.Coll.37, 43.
Turner, E.L. 1976, Ap.J.208, 30~
Vaucouleurs. G. de 1975, Galaxies and the Universe. ed. A. and M. Sandage and J. Kristian.
Chicago: The University of Chicago Press. p.557.
Vaucouleurs, G. de 1976, Ap. J. 205, 13.
Vaucouleurs, G. de, Vaucouleur5.A. de, and Corwin, H. G.1976. Second Reference Catalogue
of BrIght Galaxies, Austin: University of Texas Press.
Visvanathan, N .• and Sandage. A.1977, Ap. J. 216, 214.
Yahil, A., Tammann, G. A., and Sandage, A.1977. Ap. J. 217, 903.
ZWIcky. F. 1957, Morphological Astronomy. Berhn: Springer, p.224.
88 G. A. TAMMANN AND R. KRAAN

DISCUSSION

Peebles: This is a very elegant presentation. but I do think there are

some ambiguities. Your conclusion that the peculiar velocity must be
much less than 100 km s-l need not conflict with the idea that we are
moving at 300 km s-l or more peculiar velocity. All that is needed is
that we and the nearby galaxies are moving bodily (say, toward the Virgo
cluster). And I have the impression that the M/L values depend on the
detection of group members, so for example one can obtain many different
values of M/L for the M8l group depending on how one defines it.

Tammann: We are of course fully aware of your result that we and our
immediate neighbourhood may partake in a systematic journey toward the
Virgo cluster (1976, Ap. J., 205,318). But it seems to us that newer
observational evidence (as referenced in the text) tends to limit the
size of any peculiar motions and/or to increase the minimum volume which
could possibly move as one body.
The M8l group gives a relatively high M/L value, whereas other
groups (e.g. the IC 342 group) give very low values. This could mean
that the true M/L values change from group to group, or that the virial
solutions scatter considerably - for various reasons - about the true
value. It seemed to us reasonable to assume the latter and to determine
one mean M/L for all groups.

Davis: What, if any, galactic absorption corrections have been included

in your analysis?

Tammann: It has been believed for a long time that the galactic absorp-
tion could be derived from galaxy counts, until Noonan (197l.Ap. J., 76,
190) showed that faint galaxies are unsuited for this purpose because--
their number is affected by uncontrolled cosmological effects, and
brighter galaxies are too scarce to define the absorption at higher
latitudes. We have therefore relied on the cosec-law of colour excesses
(Sandage 1973; Visvanathan and Sandage 1978, in press) which implies
AB oc 0~13(cosec b - 1). We have, however, neglected the patchiness of
galactic absorption, because we feel that this effect is not yet sufff-
ciently controlled.

Huchra: (1) Do you correct the volume you use to derive the luminosity
density for the effect of galactic absorption?
(2) Your sample is very small, volumewise only a few thousand cubic
megaparsecs, so you have almost no information on the bright end of the
luminosity function where the mean galaxy density is ~ .001 or less -
there is a moderate chunk of luminosity there which you know nothing
about in this sample.

Tammann: As to your question: one sixth of our sample volume lies at

ibi<15°. You might therefore want to increase the mean luminosity
density by 17 percent. However, it is questionable whether a bulk correc-
THE GALACTIC NEIGHBOURHOOD 89

tion - even if of moderate size - should be applied, because the mean

luminosity density within this subvolume is not below average. We think
it is best to wait until we know how many low-latitude galaxies will
actually be added by future searches.
Your remark on the brightest galaxies is well taken. But do you
want us to extend our sample volume to contain the nearest quasars? It
is inherently impossible to gain maximum information on dwarf and
(super-) giant galaxies within the same sample volume.

Silk: One of your important conclusions was that the luminosity function
is very different for elliptical and spiral galaxies. Would you comment
on the various other studies in the literature that, while admittedly on
different samples, have found luminosity functions which are often
similar for both ellipticals and spirals?

Tammann: One of the few previous attempts to derive the luminosity

function separately for E and S galaxies comes from Holmberg (1969): he
also found widely different luminosity functions for the different
galaxy types.

Zeldovich: What 1S the distribution of galaxies perpendicular to the

supergalactic plane?

Tammann: The.concentration toward the supergalactic plane is (at least

locally) surprisingly strong: all groups lie within 3.5 Mpc of the
plane and 75 percent of the total light lies within 4 Mpc. We shall be
more specific on this problem in the written version of our talk.

van Woerden: I fear your census of galaxies with Vo < 500 km s-l may be
less complete than you think. (1) Several galaxies may be hidden in the
zone of avoidance; the recently discovered Circinus Galaxy shines
through a galactic window, but others may be hidden by heavy absorption.
(2) There are certainly many southern dwarf irregulars for which no
velocity is available yet.

Tammann: We fully agree that there are countless dwarf galaxies to be

discovered. However, the question for us is: (1) how many of them are
1m's? (2) are any of them members of the Local Group and the M8l group?
and (3) what is the number of the latter per absolute-magnitude interval?
After all our luminosity function for SlIm's galaxies predicts that half
of them are fainter than -15~7 and hence too dwarfish to enter the ----
Shapley-Ames catalogue. The next local 1m's to be discovered must help
us to fill up the low-luminosity end of the Gaussian luminosity function.
But it would take dozens and dozens of nearby(!) dwarf 1m's to make the
maximum near -15~7 disappear.
When I talked about completeness of the sample it was not with
reference to the number of galaxies, but to the total light within the
sample. Essentially all light is carried by the brightest galaxies, very
few of which should be lacking in the sample.
90 G. A. TAMMANN AND R. KRAAN

Turner: A comment on your carefully corrected (internal absorption,

inclination, etc.) magnitudes - these corrections certainly decrease the
quoted values of MIL; this is somewhat illusory, however, because for
the issue of mass discrepancies (e.g., rotation curve MIL vs. cluster
virial MIL) or for computing values of ~(=(M/L)L), the luminosities are
merely labels (bookkeeping devices) for the mass. It is only necessary
to measure the luminosities always in some (any) standard way. Put
differently, neither the virial mass discrepancies nor the value of ~
would be affected at all if every galaxy suddenly (and miraculously)
changed its luminosity by any constant factor. Converting your prefer-
red value of ~(O.04) to a MIL value in the usual, relatively uncorrected
luminosity system would give <MIL> ~30 M0/Lo' I think. For the compa-
rison of dynamical MIL values to stellar population values requires, as
you emphasized, careful evaluation of luminosities.

Tammann: I have always considered the mass-to-light ratio to be the

ratio of the true mass and the true luminosity. There may be applica-
tions where one wants to transform this physical quantity into a more
"label"-like parameter. But to avoid confusion one should clearly
distinguish between the two things, and for an observer the primary goal
should be - I think - to determine the first-mentioned quantity.

Kiang: (1) Faint galaxies are still being discovered in the Local Group.
Such galaxies would not have been discovered in any other group. Hence
I don't think that your sample is as complete at the low-luminosity end
as you say, nor that the general luminosity function is a peaked one.
(2) In calculating the luminosity density, the correction for
internal absorption should not be made. True, I was the only one of all
the authors who made this correction when deriving the luminosity func-
tion, but James Felten has now convinced me that, at least for the
purpose of getting the luminosity density, this should better not be
made. (Felten's point is that we must use the flux that eventually
emerges from the galaxy.)

Tammann: I am happy about your first comment, because it indicates that

our proposal of a peaked luminosity function for SlIm galaxies is at
least not trivial.
Your second comment, however, is a great surprise for me. I have
always been very impressed by your original proposal that internal
absorption corrections should be applied. We shall give arguments in
the written version of our paper why we believe that this is still
correct.

Ostriker: When you found the mean sizes, velocity dispersions and (MIL)
ratios, did you take straight number weighted averages or did you con-
sistently weigh each galaxy by its luminosity? The luminosity weighted
mass to light ratio is the one needed for cosmological discussions where
we mUltiply the mean luminosity density by <MIL>.
THE GALACTIC NEIGHBOURHOOD 91

Tammann: The virial masses are calculated according to the precepts set
out by Materne (1974, Astron. Astrophys.~ 33,451). They are therefore
strictly luminosity-weighted. --
THREE DIMENSIONAL ANALYSIS OF GROUPS OF GALAXIES

Jiirgen Materne
European Southern Observatory, c/o CERN, CH-12ll Geneva 23

A systematic and automatic classification of groups of galaxies has been

made only in two dimensions up to now (Turner and Gott, 1976). They in-
vestigated the projected density of galaxies at the sky. In this analy-
sis one can correct for foreground and background galaxies and superim-
posed groups only statistically. The properties of the groups and their
member galaxies will often not be recognized correctly. Therefore, a
three dimensional analysis is proposed.
A very powerfu11 method is taken from the cluster theory (for details
c.f. Bock, 1974). For example a hierarchy is built up for the groups,
keeping always the dispersion in the groups to a minimum. In this case
one has to scale the different coordinates in order to compare them. And
one has to define where to cut the hierarchy, i.e. where above the general
background groups are assumed to be real (Materne, 1977; in preparation).
The so found groups should be tested for reality: i) the crossing times
should be lower than the Hubble time; ii) the groups should be distinct;
iii) the groups should follow a luminosity-velocity relation. A final
test is to fit a distribution function to the groups and see if they are
stable against the fitting procedure.
This was done for all Shapley-Ames galaxies and DDO dwarf galaxies in a
region around the Leo group of de Vaucouleurs (1976). The hierarchical
clustering yielded s~x groups to which a Gaussian distribution function
was fitted: 1 (r_r)2 1 (v-v )2
n(r,v) -----2 exp(- c) exp(- c)
2n 0
r
20 2
r
Ifrr 0
v
20 2
v
It was found that the six groups follow the postulates i), ii), and iii).
In addition it was found that the groups seem to be dynamically bound
with conventional mass-to-1ight ratios.
Literature:
Bock, H.H. 1974, Automatische Klassifikation, Vandenhoeck and Ruprecht,
Gottingen.

93

M. s. Longair and J. Einasto (eds.); The Large Scale Strocture of the Universe, 93-94. All Rights Reserved.
Copyright © 1978 by the IAU.
94 JURGEN MATERNE

Turner, E.L. and Gott, J.R. 1976, Astrophys. J. Suppl. ~, 409.

de Vaucouleurs, G. 1976, in Galaxies and the Universe, ed. A. and M.
Sandage, J. Kristian, University of Chicago Press, Chicago and London,
p. 557.

DISCUSSION

de Vauaauleups: I would like to clear up a misconception about the

groups of galaxies which we have catalogued. We look not only at the
-distribution of galaxies on the sky but also their velocities, their
morphological types, the degree of resolution and so on. An experienced
observer takes into account many different parameters when he assigns
objects to a group.

Materne: I would agree but I was able to rediscover your groups in the
regions of sky analysed.

Zeldavich: Weighting by a Gaussian function is a very important improve-

ment over simply dividing space into cells, especially at the highest
levels of a hierarchy. This was noted in the review by Novikov and
myself in Advances of Astronomy (1965).

Gouguenheim: I wish to mention that Paturel has used the same method to
study the Virgo Cluster; he found several sub-groups and his general
conclusion is in agreement with the discussion of Tully and Fisher
(Astra. Astraphys., 1977, 54, 661).
SHORT COMMUNICATIONS 95

DETECTION OF AN OPTICAL HALO SURROUNDING

THE SPIRAL GALAXY NGC 4565
Dennis J. Hegyi

A halo surrounding the edge-on Sb giant spiral galaxy NGC 4565 has
been detected in the spectral band 3800-8600 A using a new photometric
instrument, an annular scanning photometer. The halo is brightest close
to the galactic nucleus and decreases in brightness until it reaches a
level of 1 part in 1000 of the sky at a galactic radius of 6.1 arc
minutes or 43 kpc from the galactic center. Because the scan path of
the ASP is circular, this point corresponds to a distance of 34 kpc from
the galactic plane. For comparison, the Holmberg radius of NGC 4565 is
equal to 7 arc minutes or 50 kpc. Preliminary V-I photometric data
indicate that the halo becomes redder with increasing galactic radius,
exceeding V-I = 1.9. Based on a variety of possible origins for the
light, it is concluded that the light is due to stars.

The rotation curve of NGC 4565 has been observed to be flat out to
100 kpc by Krumm and Salpeter (1977) from which the authors deduce a
galactic mass exceeding 10 12 Me' The combination of the optical and
radio data suggest that the mass necessary to explain the flat rotation
curve is contained in a massive stellar halo. Other authors (Einasto
et al. 1974, Ostriker, Peebles and Yahil 1974) have pieced together a
number of dynamical arguments which indicate that spherical halos may
make the dominant contribution to the cosmological mass density.

REFERENCES

Einasto, J., Kaasik, A. and Saar, E., 1974. Nature, 250, 309.
Krumm, N. and Salpeter, E.E., 1977. Astron. and Astrophys., 56, 465.
Ostriker, J.P., Peebles, P.J.E. and Yahil, A., 1974. Ap. J. (Letters)
193, L1.

DISCUSSION

Ginzburg: Is there a radio halo around NGC 4565.

Hegyi: The galaxy is a source of synchrotron radiation but the extra-

polation of its total radio flux density to optical wavelengths results
in an optical flux density only 10- 3 of our results.

Ostriker: The results you show appear to agree with the final results
H. Spinrad and I derived for the same galaxy (NGC 4565) by more or less
standard photographic and photoelectric methods.

Hegyi: Your results indicate a much brighter halo than we find by

perhaps as much as an order of magnitude.
96 SHORT COMMUNICATIONS

Huahra: What is the bolometric luminosity of the halo of NGC 4565 and
how much does it contribute to the total light?

Hegyi: The halo does not contribute a significant fraction of the total
light.

Tinsley: What is the luminosity of the halo and what is your estimate
of its MIL ratio?

Hegyi: All I can say is that the MIL value must be very large. It is
difficult to quote a single value for the luminosity because our obser-
vations are made in a very broad wavelength band.

Ekers: It is correct that NGC 4565 does not possess a radio halo but it
is also odd in that the disc emission is rather weak. I might also add
that recent Westerbork HI observations disagree with the results of
Krumm and Salpeter.

Zasov: Can you compare the luminosity distribution along the major and
minor axis of the galaxy?

Hegyi: The particular scan paths which we have chosen are relatively
insensitive to the eccentricity of halo brightness distribution. The
only conclusion that may be drawn is that our observations are consis-
tent with both a spherical halo and a slightly eccentric one.

Silk: What is the central surface brightness extrapolated from your

halo fit and how does it compare with that for elliptical galaxies?

Hegyi: An extrapolation of the de Vaucouleurs r! law to the galactic

centre would yield a higher surface brightness than that of most
elliptical galaxies.

Peebles: Have you tried fitting your data to a Hubble law?

Hegyi: We did n?t try very hard but my present impression is that the
de Vaucouleurs r 4 law gives a better fit.

THE SCATTER IN MASS-TO-LUMINOSITY RATIOS

Jaan Einasto

Mass-to-luminosity ratios of systems of galaxies as derived by

different authors have a large scatter from ~ 10 to ~ 200. The observa-
tional data used by different authors differ only slightly, so the
differences should lie in the treatment of the data. We have found that
in most cases < MIL > differences can be explained in the following
ways.
1. The relative motion of companion galaxies, located at a mean
distance R from the main galaxy, gives the inner mass M(R) of the
SHORT COMMUNICATIONS 97

system. In order to obtain the total mass the virial radius Ro. which
is approximately 6 times larger than the mean distance of bright
companions, should be used.
2. Double galaxies as well as groups of galaxies may form flat
systems (as our Hypergalaxy). If this system is seen face-on, the
observed velocity dispersion should be very small even if the true M/L
is large.
3. If systems of galaxies are picked out in redshift space, then
real systems can be split into subgroups with very small velocity
dispersions which do not correspond to the real velocity dispersion of
the system. On the other hand, foreground and background galaxies may
increase the true velocity dispersion.
4. If groups of galaxies are embedded in a massive corona then
most of the potential as well as kinetic energy of the system is concen-
trated in this corona. The conventional potential/kinetic energy test
is not sensitive to the presence of massive coronas.

DISCUSSION

Ostriker: I would like to make a remark about the use of the words
"virial discrepancy", "virial mass problem", "missing mass problem".
What one does is to measure the total light from a system and determine
its mass from dynamical arguments. The ratio of these numbers is the
M/L ratio. One then divides this M/L ratio by a hypothetical value
which may be thought appropriate and asks whether this ratio is unexpec-
ted or not. Since the "discrepancy" results from dividing a measurement
by an assumption, this is not a real discrepancy. Masses of systems can
only be determined by dynamical arguments.

Tinsley: I would like to strengthen Ostriker's comment that it is

inappropriate to talk about "luminous mass" with an M/L '" 10 (or 5, 20,
etc.) solar units. Even in ordinary populations with M/L '" 10, nearly
all of the mass is invisible. Spectroscopic studies of galaxies show
that the light is dominated by giant stars (especially in the red-
infrared), whose own M/L is much less than 1; most of the mass in a
system with M/L '" 10 is in faint dwarfs that hardly contribute at all to
the integrated light. Therefore, it is quite inappropriate to refer to
the quantity "luminosity x 10" as "luminous mass". I strongly agree
with Ostriker's remark.

Rood: The value M/L ~ 7 or 10 for individual galaxies is not an assump-

tion taken out of the hat. It is based on studies of the internal
motions of galaxies - stellar velocity dispersions and rotation curves.

Tremaine: One of the assumptions everyone has made is that all galaxies
of a given type really do have a universal mass-to-light ratio. There
seems to be no reason why the M/L values of galaxies or groups of a
given type should not vary by a factor of 10. Having said this I hasten
to add that I don't believe it but the possibility should be kept in
98 SHORT COMMUNICATIONS

mind, since it would resolve many of the inconsistencies in different

people's results.

Jaakkola: In their standard context, the low values of M/L and n,

obtained recently in some well-known investigations, imply an "open"
Universe with hardly any influence of galaxy concentrations on the
Hubble flow. The picture of steady recessional motion of galaxies
resembles conspicuously the mechanistic world picture of 200 years ago
and, therefore, it possesses in a certain sense a metaphysical content.
One can even notice a feed-back to the ancient Ptolemaic idea of
"perfect motion". The usual inference about a smooth redshift-distance
relation as an argument in favour of the expansion hypotheses is
incorrect. The case is actually the opposite - Hubble himself consider-
ed this result as a natural consequence of a photon interaction
mechanism for redshifts and he may be taken as a fair authority as
regards the Hubble relation! Taking into account the fundamental role
of gravitation and the absurdity of dynamics without matter, a low
value of n(= Pobs/Pcrit) should be considered as a conflict between
theory (Pcrit) and observations (Pobs).

STABILIZATION OF SYSTEMS OF GALAXIES BY SUBCLUSTERING

L. M. Ozernoy and M. Reinhardt

Subc1ustering might help to solve the vi rial theorem paradox for

systems of galaxies by hiding a major part of the potential energy in
gravitationally bound subsystems. We have shown (Ozernoy and Reinhardt
1976, Astr. Astrophys.~ 52, 31) that even in groups of galaxies there is
mass segregation, in the-Sense that bright group members tend to be con-
centrated towards the centre. Recently Wesson and Lermann (1977,
Astrophys. Sp. Sci., 46, 327),rea1izing the importance of subc1ustering,
proposed a quantitative method for estimating its effect on the stabil-
ity of systems of galaxies. However, their assumption about the
frequency of subsystems of multiplicity n is not in accord with
Holmberg's (1962) result. The mean frequency of galaxies in pairs is
0.37 for the Turner and Gott groups (1976) and 0.23 for the de Vauceu-
lours groups (1976), in good agreement with the value of 0.25 required
by Holmberg's distribution. Assuming Holmberg's frequency of gravita-
tionally bound subsystems and that they are homogeneously distributed
throughout the system, we have for the ratio of the total potential
energy of a system of N equal masses n to the potential energy calcula-
ted in the usual way neglecting subclustering ns ' n/n s ~ 1+(Rc )/«r2>N),
if the velocity dispersion <or 2 (n» = constant. Here Rc is the effec-
tive radius of the system and <r2> the mean distance of binaries. The
assumption 0r 2 (n) = const is reasonable for n ~ 7, when Holmberg's
distribution holds, since or 2 (2) = 203 km s-1 according to Karachentsev
(1974), and increases to only ~ 1000 km s-1 for rich clusters. Since
Karachentsev's data give an <r2> = 33 kpc for Ho = 55 km s-1 Mpc- 1 , we
have n/ rt.s ~ 4 for groups of galaxies wi th Rc ~ 1 Mpc and N = 10. Thus
it seems that subc1ustering cannot remove the mass discrepancy for rich
clusters and for groups only in moderate cases.
SHORT COMMUNICATIONS 99

However, so far we have considered only systems of galaxies of

equal mass, which is clearly unrealistic. Particularly in rich
clusters, we often find two very bright galaxies in the central region,
which might form a physical pair. For example, if NGC 4874/4889 in the
Coma cluster are gravitationally bound, the binding energy of this pair
alone would exceed ns for the whole cluster, derived by the usual method,
by a factor of 11.5 for Rc ~ 2 Mpc and N ~ 800. Thus this single pair
could stabilize the whole Coma cluster. It is interesting that irregu-
lar clusters of galaxies, which on the average show a particularly high
mass discrepancy, have also a high number of multiple subsystems.

For groups we give as an example Nos 48 and 49 in de Vaucouleur's

catalogue (1976). The ratio of the binding energy of the pairs NGC
3504/3512 and NGC 3607/3608 to ns is np/ns = 15.4 and 10.4, respectively,
as compared to the observed mass discrepancy MVT/M = 9.1 and 10.9 (Rood
et al. 1970). From this we draw the tentative conclusion that subclust-
ering, especially binary galaxies, must be taken into account in the
correct calculation of the potential energy of systems of galaxies and
that this might remove the virial mass paradox, where it still seems to
exist. This would also be important for the determination of the
cosmological density parameter n.

DISCUSSION

Turner: You suggest the substitution of a hierarchical mass model for

the usual continuous one in the evaluation of the potential energy.
Neither approximation is necessary, of course; the potential energy may
be evaluated directly from the observed positions of the individual
galaxies. This is the standard procedure for small groups and has been
used also on rich clusters. For the Coma cluster such direct evaluation
of the potential energy agrees fairly well with that derived from the
continuous mass model.

Reinhardt: Calculating the potential energy of a given system of

galaxies from the observed positions, you do not know, because of pro-
jection effects, whether some galaxies form gravitationally bound sub-
systems or not. On the other hand there is a convincing statistical
evidence of the existence of such substructures, as we have pointed out.
Therefore, it is necessary to take into account the binding energy of
subsystems in the calculation of the potential energy. Neglecting sub-
clustering will in general give a wrong result.

Turner: I agree that the two bright galaxies at the centre of Coma may
well be a binary; I do not agree that this binary dominates the cluster
potential energy.

Reinhardt: We are happy that you agree with us in that the two bright
galaxies in Coma might form a bound pair. However, in this case it
follows directly from the data that its binding energy greatly exceeds
that of the whole cluster. There is no way out of it.
100 SHORT COMMUNICATIONS

Peebles: If you want to assume the mass is mainly in the two brightest
galaxies near the centre of the Coma cluster then the potential varies
as r- 2 , and to match the observed counts of galaxies as a function of
distance from the centre of the cluster, number density N(r) ~ r-~, you
must assume the velocity dispersion varies as v 2 ~ r- 1 , which seems not
to be observed.

Reinhardt: I think you have misunderstood me. We assume that in the

central pair of galaxies there is concentrated only a small part of the
total mass of the system, but a large part of the binding energy. Of
course, this is not the same and our assumption does not require any
change of the velocity distribution.

MORPHOLOGICAL INVESTIGATION OF PAIRS CONTAINING

MARKARIAN GALAXIES*
c. Casini and J. Heidmann

We have obtained large scale photographs or electronographs for 40

pairs containing Markarian galaxies: 6 Markarian-Markarian pairs and 34
Markarian-normal pairs, most of which are physical close pairs with two
similar components (here "normal" means with no UV excess).

We have studied the distribution of types for the Markarian and for
the normal components; the most frequent type is spiral, with more
ordinary spirals than barred ones; next are compact for the Markarian
and elliptical for the normal galaxies; the only one irregular is of
clumpy type.

A comparison was made between isolated Markarian galaxies and those

in pairs; for both, spirals are dominant with ordinary ones more frequent
than barred ones. There are more compacts, fewer irregulars and much
fewer lenticulars among pairs of Markarian galaxies than among isolated
ones.

Morphological peculiarities have been investigated. In the

Markarian-normal pairs, the Markarian is more often peculiar; thus
peculiarity is associated with UV excess; there is no tendency for
peculiar Markarian galaxies to pair with peculiar normal galaxies;
peculiar Markarians have more often a diffuse spectrum but it cannot be
said that morphological peculiarities are located at the same positions
as the UV excess.

Six noteworthy peculiar galaxies are described which contain double

strings of condensations, double or triple nuclei, clumpy structure or
whorl shape.

*Submitted to Astr. Astrophys. Suppl. Series.

SHORT COMMUNICATIONS 101

DISCUSSION

Tinsley: On the correlation of morphological peculiarity with ultra-

violet excess, I would like to mention a study that R. B. Larson and I
made earlier this year. We studied the distribution of galaxies in
Arp's "Atlas of Peculiar Galaxies" in the (U-B, B-V) two-colour plane.
Interacting galaxies have an unusual colour distribution, in the sense
that many have a strong ultraviolet excess. We interpreted the colour
in terms of unusually high star formation rates, which are presumably a
consequence of the interaction. Maybe Dr Heidmann is seeing a similar
effect.

CLUSTER MEMBERSHIP OF SEYFERT GALAXIES

K.-H. Schmidt

I would like to comment on very preliminary results concerning the

cluster membership of Seyfert galaxies. Two years ago van den Bergh
stated in a discussion of the bright classical Seyfert galaxies that
these objects were mainly field galaxies. We extend this discussion to
about seventy objects in the list of Drs Khachikian and Weedman
(Astrophys. J. 3 192, 581, 1974). By a cross correlation of this list
with Dr Abell's catalogue of rich clusters taking account both of the
positions in the sky and the distances of the objects, we find only two
certain cluster members - the well-known case NGC 1275 in the Perseus
cluster and Markarian 298 in the Hercules cluster - and three probable
or possible other cluster members. Taking into account the fact that
only some fifty objects of Khachikian and Weedman fall within the
distance range of the Abell catalogue, we find that only a few per cent
of the Seyfert galaxies are members of Abell clusters. Within a factor
of two, this percentage is the same as the percentage of cluster
galaxies relative to all galaxies. Therefore, this result does not
point to a pronounced field membership of the Seyfert galaxies.

Further, there seems to be a hint that Seyfert galaxies are found

in clusters. Four out of five certain or probable members of Abell
clusters are class 2 Seyferts. Although the statistics are as yet poor,
this result is unexpected if one takes into account the fact that 53 of
the 70 Seyferts in the list of Khachikian and Weedman belong to class 1
and only 17 to type 2.

DISCUSSION

Abell: David Jenner and I found a Seyfert galaxy in the Virgo cluster.
It is a normal-appearing spiral, whose NGC number I have forgotten (off
hand). We shall publish a note on it.

Huehra: When you compute probabilities from the cluster and Seyfert
catalogue cross correlation, do you correct for the very different sky
coverage of the catalogues? The majority of the Seyferts come from
Markarian's first lists which cover only a small region of the sky.
102 SHORT COMMUNICATIONS

Schmidt: Yes, I correct for this effect.

Ambartsumian: Have the clusters to which Seyfert galaxies belong any

special peculiarities?

Schmidt: No, they are normal clusters.

THE RESULTS OF OBSERVATIONS OF DOUBLE

GALAXIES IN THE UBV SYSTEM
A1eksander Tomov

In the Astronomical Observatory of the town of Be10gradchik a one-

channel e1ectrophotometer in the UBV system with pulse counting has been
installed at the 60-cm Cassegrain telescope. The photoamp1ifier used is
of the type EMI 6256B (1). With this telescope the author has carried
out observations of 80 double galaxies during the period 1974 to the
middle of 1977, using the differential method of observation with 3
diaphragms (24", 54" and 135") depending on its diameter.

The most important results are the correlation between the colour
indices of individual components of double galaxies, Holmberg having
obtained the correlation coefficient R(B-V) = 0.80 ± 0.06 (2).

The author set himself the aim of checking the degree of correlation
for the two colours (B-V) and (U-B) for individual classes of double
galaxies; E-E and S-S - 60 double galaxies and E-S - 20 double galaxies.

The correlation coefficients R for the 60 double galaxies (E-E and

S-S) are R(B-V) = 0.885 ± 0.046 and Rl(U-B) = 0.824 ± 0.067.

The correlation coefficients for 20 double galaxies (E-S) are

R(B-V) = 0.620 ± 0.085 and R1(U-B) = 0.512 ± 0.098.

In addition the author selected individual galaxies taken from

reference (3) randomly together as double galaxies. It turned out that
in this case, too, there is a correlation between the colour indices.

The correlation coefficients R' for 55 randomly selected pairs of

E-E and S-S galaxies are
0.580 ± 0.072
0.382 ± 0.085.

For E and S galaxies selected randomly, the 27 pairs have correlation

coefficients
0.125 ± 0.090
0.084 ± 0.098.
SHORT COMMUNICATIONS 103

(1) Tomov, A., 1973. Bulletin of the Astronomical Observatory of

Varna, Bulgaria.
(2) Holmberg, E., 1958. Medd. Lund Astr. Obs., Ser II, No.136.
(3) de Vaucouleurs, G., 1972. "Integrated magnitudes and colours for
bright galaxies in the UBV system".

ON THE ORIGIN AND EVOLUTION OF PAIRS OF GALAXIES THAT

HAVE DIFFERENT PHYSICAL AND KINEMATIC CHARACTERISTICS
I. Pronik and L. Metik

We have gathered data on U,B,V,K magnitudes, radial velocities,

spectra, morphological types, radio-emission, dimensions and other
characteristics for 47 pairs of galaxies (Metik and Pronik 1978). The
following results were obtained:

1. Colour indices of members of pairs are given in figure 1. It

shows that the pairs divide into three separate sequences. Red galaxies
with (U-B) > + O~5 co-exist with galaxies of any colour, whereas blue
galaxies co-exist only with blue or very red ones. Moreover in the
pairs of mixed colour the smaller the colour index of the blue galaxy
the redder is its component. Pairs of blue galaxies form a separate
group (sequence I). Both
(U-BI, members of this sequence
have either negative or
small positive colour indices
(U-B). The rest of the pairs
of galaxies form sequences
II and III. They lie along
two lines corresponding to
• (U-B) ~ + O~5. Such colour
index belongs to the brighter
-0.5
component (subscript 1) of
n the pairs of sequence II and
to the fainter one (subscript
2) of sequence III. Members
-1.0
n of pairs change their colours
0.0 along each sequence: in I -
both members, in II - the
Figure 1. Correlation of colour indices fainter component, in III -
of members of pairs. Crosses - pairs of the brighter one.
sequence I, dots - II, and circles - III.
2. Sequences of pairs selected according to their colours show
also differences in radial velocities (V r ). Pairs of sequences I and II
have on average smaller differences in radial velocities than those of
se~uence III. Moreover sequences II and III show a dependence of
19j6Vri on 61 (U-B) I: 19l6V r i decreases with decreasing 61 (U-B) I in
sequence II whereas in sequence III we have the opposite case.
104 SHORT COMMUNICATIONS

3. Members of sequence I are spiral or irregular galaxies. Ep,

Seyfert and Markarian galaxies are also present. There are E-E and E-S
pairs in sequence II and no Markarian and Seyfert galaxies. E-E and
S-E pairs are present in sequence III; Markarian and Seyfert galaxies
are also present.

4. Absolute magnitudes and spectral types. There is a connection

between these characteristics, but different for each sequence. The
brighter members of the pairs of sequences I and II have more later
spectra whereas III sequence have earlier ones.

5. Infrared data at 2.2~m. (V-K) colours for 10 galaxies of

sequence II and for 10 galaxies of sequence III are available. On
average galaxies of sequence II are redder than those of sequence III.
If confirmed, this means that the former on average are more "metallic"
than the latter.

6. We did not find differences in the radio luminosities of the

members belonging to the I, II and III sequences.

7. The dimensions of the pairs are in good agreement with the

data of Karachentsev (1970): pairs of spiral galaxies have larger sizes
than those of mixed ones.

8. The origin and evolution of pairs of galaxies were considered

by Heidman (1976). He pointed out two main phenomena: both galaxies of
the pair were born at the same time or the smaller component is the
result of ejection activity of the main one. The character of the
19!6Vrl and 6(U-B) relation for sequences II and III shows that the
velocities and colours of fainter pairs are not accidental. The ejec-
tion origin of the fainter component cannot ensure this. If this result
is supported by future observations it will mean that members of pairs
of galaxies were born at the same time.

9. The data suggest that the sequences I, II and III are different
in their origin and evolution. The three sequences of pairs show differ-
ences in morphology and velocities too. It is well known that kinemati-
cal properties of systems of galaxies are formed before the time of star
formation. This suggests that sequences I, II and III separated before
the matter began to condense into galaxies. The members of these
sequences had different primary conditions of formation and so evolved
in different ways. Main galaxies of pairs of sequence II with high MB
have later spectral types than those of sequence III. We conclude
that the latter contain more massive stars than the former.

REFERENCES
Heidman, J., 1976. Proc. Third Europ. Astron. Meeting, Tbilisi, p.267.
Karachentsev, I., 1970. Vestnik Kiev University, 12, 13.
Metik, L. and Pronik, I., 1978. Astron. Zh., in press.
SHORT COMMUNICATIONS 105

MARKARIAN GALAXIES WITH DOUBLE AND MULTIPLE NUCLEI

E. Khachikian

Saakian, Petrosian and I have studied the central regions of

Markarian galaxies using observations made with the SAO 6-metre telescope
and the 2.6 m and 0.5 m telescopes of the Burakan Observatory. In a
paper soon to be published, we show that in those galaxies which have
strong ultraviolet excesses, i.e. Markarian Galaxies, double or mUltiple
nuclei are found with a significantly greater frequency than they are in
ordinary galaxies. This result suggests that these are among the most
active nuclei. The photographs in our paper show the 60 examples of
double or multiple nuclei found in the first 8 lists of Markarian
g·alaxies.

DISCUSSION

A. V. Zasov: How often do the wide emission lines occur in such galaxies?

Khachikian: Usually about ~ 85% of all Markarian galaxies have emission

lines.

EQUIDENSITOMETRIC DETERMINATION OF ANGULAR DIAMETERS

AND MEAN SURFACE MAGNITUDES OF COMPACT GALAXIES
W. Hagner and N. Richter

In a previous publication (1975) we showed that it is possible to

determine the maximum diameters of galaxies on photographic plates by
means of the photographic equidensity method. This method was applied
especially to 235 members of the Coma cluster. On Tautenburg plates it
was possible to determine for these objects equidensitometric angular
diameters down to 10 arc sec.

The next step of our investigation was to apply this method to more
than 700 compact galaxies selected in 4 fields around the globular
cluster M3 near the galactic North pole. It was found that it is possi-
ble to determine real angular diameters down to 5 arc sec to 17 and
partly 18 magnitude on Tautenburg plates in the V system if the condi-
tions of seeing are 2 arc sec or better. Figure 1 shows the diameter-
magnitude diagrams for one of these fields both for stars of the
standard sequence around M3 and for the compact galaxies. The diameters
of the compact galaxies clearly lie above the straight line, which
represents the diameter-magnitude relation for the standard stars. From
this result we conclude that these equidensitometric diameters of compact
galaxies are real angular diameters and not produced by scattering
effects due to seeing or diffusion in the photographic emulsion.

Using these diameters it was possible to determine the mean surface

brightness of the compact galaxies from the integral brightness already
determined by iris photometry.
106 SHORT COMMUNICATIONS

constructing a diagram for the relation between integrated magni-

tudes IDv and the mean surface magnitudes IDv/o" we find a strong and
nearly linear dependence. Another result is that galaxies with a mean
surface brightness of 2l~5/o" must be considered to be compact galaxies,
which is an extension of Zwicky's result, who postulated for these
objects a mean surface brightness of 20m/O" and brighter. Therefore
this photometric criterion for optical compactness can be extended to
2l~5/D". In Figure 2 this relation for field IV is given.

In Figure 3 we can see the same relation for all types of galaxies
in the Coma cluster. We find that there exists only a very weak
dependence between the integrated brightness and mean surface brightness.
The latter we find in the mean about 23m/o" with a scatter of about 2
magnitudes.

Figure 4 shows once more for comparison the least squares straight
lines for normal and for compact galaxies derived from Figures 2 to 4.

REFERENCE
Richter, N. and Hagner, W, 1975. Astron. Nachr. 296, 221.

..
11

16 ,.""

to

Q'

..., 06

0.4
12 14 11 20-
11-

Figure 1
SHORT COMMUNICATIONS 107

22
field. N
21 C."r~.t 6~l....1.~ 21

20

1, 1,
.,

I:
~I

.q

lt~
f
-vro'L-~ __-L__L-_~I~I~-L__~~~-L__~~ .... /0.
40 ~1 ~2. 41 "" 1S ~, ~1 -11 1,
"'v_
Figure 2

" .... '

COWla.
..Ll tJr'~
Ctu~ter
- l1

- ZO
."
I
.... /a·
11 ... 1'\ 1" ~s ., 11 11 .,

Figure 3

l~

Z4
"
U U

za Z2

Z4 11

21 ZO

I
..,
...
1=
..... /rl' .... ... AI.Al.fJt "1''''4 <f'''1
At ~"/a·

..."
Figure 4
INTERACTING SYSTEMS

Alar Toomre
Massachusetts Institute of Technology

I will confine this brief review to just two examples of galaxies

that seem to have interacted recently with impressive consequences:
(1) the well-known Messier 51 system NGC 5194/5 ~ deservedly VV 1, and
(2) the so-called Cartwheel, a ring within a ring, separated by vaguely
spokelike features and accompanied by two smaller galaxies.

These two situations deserve special emphasis, not only because of

the presence in Tallinn of Vorontsov-Vel'yaminov who at least two years
ago still regarded both as "enigmatic", but perhaps especially because
they long fascinated also that other grand pioneer of our subject, the
one whom we abbreviate as Zw. Indeed the Cartwheel was discovered by
Zwicky (1941) himself - - from Palomar with a 46-cm Schmidt telescope,
despite its _34 0 declination - - who then thought it "one of the most
complicated structures awaiting its explanation on the basis of stellar
dynamics". And in the very next paragraph of that old article, Zwicky
went on to remark about M5l that "tidal actions operative during this
encounter may have caused the two spiral arms in the big nebula" and,
more significantly, "notice also an indication of a tidal effect in the
small nebula, which has the aspect of a closed barred spiral".

Even I do not pretend that these two rich examples have yet shed
all of their mystery. However, I do believe that modern observations
offer some unusually clear and instructive signs of recent tidal damage
in both cases - - and also that the explanation of the Cartwheel as a
tidal remnant has turned out to be almost embarrassingly simple.

1. THE NGC 5194/5 PAIR

The main evidence for tidal interplay between the two partners in
this classic system has now grown to four items: two still refer to
their out~ ~hapeh, and the other two to some newly-measured veloc{tie~.
Those shapes and their orientations are shown beautifully in Figure 1,
where a fairly standard (but already integrated) photograph has been
superposed by M.S. Burkhead upon the photographic sum of five (!) deeply
109

M. S. Longair and J. Eznasto (eds.), The Large Scale Strncture of the Universe, 109-116. All Rights Reserved.
Copyright © 1978 by the fA U.
 110 ALAR TOOMRE

Figure 1. Recent composite photograph of NGC 5194/5, by Burkhead

exposed IIIaJ negatives, all taken by himself using the Palomar l22-cm
Schmidt telescope. Very distinctive here -- as in the older IIIaJ photo-
graph by van den Bergh (1969; reprinted slightly darker by Toomre 1974)
and in others which Schweizer and especially Lynds have kindly shown me
in the past few years -- is the almost comma-like lower outline of 5194,
including its abrupt western edge that curves smoothly all the way from
the northwest to the broad extremity in the south. And even more tell-
tale, as I have stressed before, are the faint long plumes or streamers
which seem to extend from 5195 toward the 2 and 8 o'clock directions.

I doubt that these long streamers could have been known to Zwicky
(1941), since his sketches of various faint details even in the 1959
Handbueh article seem rather fanciful when compared with Figure 1. His
"indication of a tidal effect in the small nebula" must have referred
more to the strange impression of three spikes of a crown immediately to
the north of 5195 that can be had from photographs not quite as burnt-
out as the one above. On the other hand, I am afraid that Vorontsov-
Vel'yaminov (1975a,b, 1977) was quite seriously mistaken when he wrote
repeatedly that 5195 "does not exhibit the slightest evidence of the
'tail' and 'bridge' shown in the model" constructed by Toomre and Toomre
(1972 = TT; see esp. Fig. 21), and that "none of the filaments •.• pre-
~icted by the theory are observed, at least on the photographs published
$0 far". If from nothing else, this last remark suffers from the over-
~ight that, far from p~edicting any tidal plumes from 5195, my brother
INTERACTING SYSTEMS 111

-1

2 N

Sun~
Figure 2. An improved model of the tidal encounter between NGG 5194/5.
The left scenes are as if viewed from the Sun, the right set as if from
the west. The main revisions from Figure 21 of TT are that the final
5
(= present) time here is only 2.0 instead of 2.4, gnd that thg sate lite
angles have been changed to is = _75 0 and Ws = +10 (from -60 , -15 ).
Less important, the line of intersection of the 5194 spin plane with the
"sky" has been turned to a more realistic PA = _100 , and the main orbit
angles have become i4 = _7S o and W4 = _200 • The tilt S4 = -22?S now.
The gravity was also softened moderately at close range.
112 ALAR TOOMRE

and I were frankly "peeking at the answer" as we then sought to .£mLta.te.

those striking features published already by van den Bergh.

This is not to imply that I remain very happy with the naive model
concocted by TT. In retrospect, we were surely too greedy in presuming
that the counterarm of 5194 had already narrowed itself kinematically,
by mere lapse of time, into a curving feature more photogenic than the
broad southern extension seen in Figure 1 and, incidentally, also in the
2l-cm radio data of Shane and Bajaja (1975). And those tidal plumes
from our make-believe 5195 were themselves perhaps unduly droopy, since
we had not bothered to "optimize" them via adjustments of the angle ws.
Fortunately, as Figure 2 illustrates using still only the massless par-
ticles without any original random motions, it was easy to lessen those
flaws considerably via modest changes of parameters.

Enough about the outer shapes. Now what about the velocities?
Obviously models such as shown in Figure 2 also imply something about
those. For instance, as TT noted already, the purported tidal plumes in
Figures 1 and 2 make sense only if it is the "east or southeast side of
the main surviving body of NGC 5195 which most nearly approaches us".
If that side were to recede, the tidal model would be dead. Schweizer
(1977) checked: the east side approaches.

+
r. 470 490

/~')

...' .': '. ~ /cJr "

,~ • \~! I ,~
.... ... ') , ' ~
(>'..... \. """~,~' '->;--"v' :"
\)

Figure 3. Comparison of four "channel maps" of neutral hydrogen in NGC

5194, obtained by Shane and Bajaja (1975; see also Shane 1975) with the
WSRT and filters of ~27 km/s width centered upon heliocentric speeds of
430,450 ... krr./s, against four similar maps implied by Figure 2 after
adopting systemic velocities of 464 and 594 km/s for 5194/5.
INTERACTING SYSTEMS 113

Another interesting check is provided by Figure 3. As its lower

row indicates, there is something odd about the 2l-cm narrow-filter maps
of neutral hydrogen produced with the Westerbork radio array and kindly
made available by W.W. Shane. Instead of the familiar sectors symmetric
about the major axis that one expects for a flat disk in normal rotation,
these maps convey more the impression of an S shape, particularly in the
two channels closest to the systemic velocity. A natural suspicion,
voiced in part already by Weliachew and Gottesman (1973) in discussing
their Owens Valley results with 4x coarser angular resolution, is that
such abnormalities are further evidence of tidal damage. The four theo-
retical maps in the top row hardly discourage that notion.

Before leaving the M5l system, however, two ironies or conceivable

difficulties also deserve to be aired. One irony of course concerns the
spiral structure of 5194, situated mostly in deeper regions that suffered
little in this simple tidal picture. As one who has also wondered aloud
(TT, p.664) whether even that splendid structure might nevertheless have
resulted indirectly from the encounter, I must here repeat sadly that
such hopes seem to be fading. A truer story may well be that a fine
pre-existing spiral got rather dented lately in its exterior.

A second and very different irony is that suddenly even the plumes
in Figure 1 seem short in comparison with the 15+ arc min long region to
the northwest in which Burkhead (1977) discovered and traced a distinct
excess of faint light at about B = 25 or 26 magi sec 2 level. And to com-
pound that surprise, Giovanelli, Haynes and Burkhead (1977) reported
that they in turn had detected, among various 2l-cm residue in the sur-
roundings of M5l known in part already to other observers like Shane
(1975), also some with speeds as large as 650 to 700 kmls and roughly
coincident with Burkhead's optical extension! Whether tidal debris (from
yet farther out in 5l95?) or something more primordial, here surely is a
reminder that even old friends may not be entirely what they appear.

2. THE CARTWHEEL

Figure 4 shows the clearest photograph of the Cartwheel that I have

yet seen. As noted already, the explanation of ring galaxies like this
one has lately turned out to be astonishingly simple. According to
Theys and Spiegel (1977), that accomplishment was almost entirely theirs
- - though I have recently been told also that Eneev, Kozlov and Sunyaev
noticed the basic phenomenon in their own experiments three years ago.

To repeat this explanation, Figure 5 shows the transient outcomes

of six different vertical bombardments of a given disk of randomly set
test particles by another imagined galaxy which penetrates without any
influence except its own inward pull of gravity. I used to think, in
all the work leading to the TT paper and even for some time afterwards,
that intrusions deeper than the one shown at the top of Figure 5 - - and
still yielding shapes reminiscent of M5l - - would leave the target disk
much too splattered. This impression remains valid for orbits rather
114 ALAR TOOMRE

Figure 4. IIIaJ photograph of the Cartwheel, taken by Blanco with the

4-m telescope at Cerro Tolo10 Inter-American Observatory. North is up,
and east is again to the left.

inclined in the ~ect sense relative to spin, but evidently the vertical
orbits (with W = 0) already contradict my old intuition. In fact, al-
though not shown here, I now know conversely that shapes with off-center
"nuclei" like those resulting from impacts 2 or 3 in Figure 5 can them-
selves be improved bZ the arrival of the intruder in a ~~og~ade orbit
tilted as much as 45 from the vertical.

The Cartwheel itself appears to have required nothing even as fancy

as tilting the orbit. We see that something like the next-to-bottom row
in Figure 5 already provides a fair imitation. In reality, as Fosbury
and Hawarden (1977) discussed, the blame probably belongs to the more
easterly neighbor: unlike the other, its velocity increment is plausibly
small, and so is its gas content. On a dark print of Blanco's negative,
this galaxy even shows a semblance of a tidal hook. Most important,
Fosbury and Hawarden also estimated that the intense HII regions whicr
mark the ring have a net expaYL6..toYl (assuming the "spokes" trail) equal
to about one-third of their rotational velocity -- which indeed jibes
with the "expansion that is perhaps smaller by a factor of 2 or 3" con-
jectured by Lynds and Toomre (1976) in a similar context.

As one final corroboration, notice that the simulations near the

bottom of Figure 5 -- just like the movie which Bob Reynolds from Ohio
State University kindly prepared and lent me for this Symposium -- even
INTERACTING SYSTEMS 115

o
t=6 12 o 18 o 24

o o

o o o

o o o

o o o

(~
o o
t=6 12 18 o 24

Figure 5. Six vertical penetrations of a disk of 2000 test particles.

Long ago, these merely circled an unmarked central body 2x as massive as
the parabolic intruder. Viewed from 45 0 latitude, their rotation appears
counterclockwise. Gravity was softened again near both massive bodies.
116 ALAR TOOMRE

Figure 6. Radial locations of 40 particles from the

symmetric encounter shown at the bottom of Figure 5.

manage to imitate the observed ~nnet ring. Very reminiscent of later

ripples that develop after a pebble has been dropped into water, this
new zone of crowding here results (at least in theory) from the falling
back and second rebound of the many independent "oscillators" that were
yanked inward by the intruder. Figure 6 tells the same in more detail.

I am very grateful to V.M. Blanco, M.S. Burkhead, R. Giovanelli, B.

Reynolds and W.W. Shane for entrusting me with unpublished material, to
the NSF for its support, and to J. Einasto for his gracious hospitality.

REFERENCES

Burkhead, M.S.: 1977, A~tnophy~. J., submitted.

Fosbury, R.A.E., and Hawarden, T.G.: 1977, Monthly Notiee6 Roy. A4tnon.
Soc.. 178, 473.
Giovanelli, R., Haynes, M.P., and Burkhead, M.S.: 1977, P~oc.. IAU Symp.
77, in press.
Lynds, R., and Toomre, A.: 1976, A4tnophy~. J. 209, 382.
Schweizer, F.: 1977, A4tnophY4. J. 211, 324.
Shane, W.W.: 1975, in La Vyrtami..Que de6 Galaue6 SpVta1u, CNRS, Paris
(ed. L. Weliachew), p. 217.
Shane, W.W., and Bajaja, E.: 1975, private communication.
Theys, J.C., and Spiegel, E.A.: 1977, A4tnophY4. J. 212, 616.
Toomre, A.: 1974, P~oc.. IAU Syrnp. 58, 347.
Toomre, A., and Toomre, J.: 1972, A4tnophY4. J. 178, 623.
van den Bergh, S.: 1969, A4tnophY4. Lett~ 4, 117.
Vorontsov-Ve1'yaminov, B.A.: 1975a, P~'ma A4tnon. Zh. 1, 3.
Vorontsov-Vel'yaminov, B.A.: 1975b, A4tnon. Zh. 52, 491.
Vorontsov-Ve1'yaminov, B.A.: 1977, A4tnon. A4tnophy~. Suppl. 28, 1.
Weliachew, L., and Gottesman, S.T.: 1973, A4tnon. A~tnophy~. 24, 59.
Zwicky, F.: 1941, App~ed Mec.hani~ (von Karman volume), p. 137.
Zwicky, F.: 1959, Handbuc.h det PhY4~k 53, 373.
SHORT COMMUNICATIONS 117

ON THE TIDAL ORIGIN OF M5l-TYPE SYSTEMS

B. Vorontsov-Velyaminov

We established MSl systems as a class of interacting pairs of

galaxies in 1959. In 1975 I showed that among 160 systems with this
appearance the ratio of luminosities (and hence masses) of the compon-
ents varies from 1:1 to 1:0.01. In the latter case, the companion is
comparable in mass and dimension to an isolated H II region. The number
of such MSl-type systems increases as the luminosity of the companion
decreases. Small companions cannot draw spiral arms from the primary
companions by means of tides. Yet, they are observed at the tips of
spiral arms. Thus, they must originate at the same time as the spiral
or form within them. In some cases, these companions emanate from the
spiral arm, as is shown by comparison of observations and calculations.

There are cases, such as VV 20, 21, 244, 247 and others, in which
the components have bridges or tails, possibly of tidal origin, much
fainter than the original spiral arms. They form large angles with real
spiral arms and probably lie in another plane.

There are some "twice M5l-type" galaxies with two similar compon-
ents at the ends of opposite arms. It is impossible to believe that
this resulted from their simultaneous arrival from infinity or from very
elongated closed orbits to symmetrical, very close positions relative to
the primary. In VV 470 the relative dimensions of the components are
equal to those in the case of M5l.

It is noteworthy that there are cases (M5l, VV 19, VV 20 and others)

where, besides a curved filament, really or only apparently connecting
the components, there is also a straight and more massive filament
connecting them which is not predicted by the tidal theory.

The fine tidal arms ingeniously obtained by this theory disappear

in a more realistic treatment in which there is self-gravitation and
small dispersion of velocities (modelled by F. Hohl). The hypotheses of
collisions to explain ring galaxies and "Mergers" fails on the grounds
of statistics and because of the existence of tight nests of galaxies.

THE FREQUENCY OF RING GALAXIES AND THE PROBABILITY

OF THEIR FORMATION BY COLLISIONS
V. Dostal and V. Metlov

A survey of ring galaxies which were discovered in 1960 by B.

Vorontsov-Velyaminov has been carried out using the Morphological
Catalogue of Galaxies. Clusters of galaxies were also included. Such
galaxies without spiral arms constitute 0.7% of all galaxies down to
l5~, and are 100 times less frequent than spiral galaxies. This per-
centage is the same for clusters, but there are fields where ring
118 SHORT COMMUNICATIONS

galaxies are 3-5 times more frequent than the average. The probability
of nearly central chance collisions of galaxies as proposed by some
theories has been calculated. It was assumed that a ring galaxy is
formed from the encounter of a spiral galaxy with another galaxy if its
mass or luminosity is not less than 1% of that of the principal galaxy.
The minimum separation of their centres was taken to be less than 2.7
kpc, but the angles between the vectors of the relative velocity and the
planes of galaxies were not restricted. The lifetime of the rings once
they are formed was assumed to be 10 9 years and the distribution of the
galactic velocities was calculated according to a Maxwellian distribu-
tion. The number of galaxies of given absolute magnitude per unit
volume was taken from Holmberg (Stars and Stellar Systems, 9, 123). The
curvature of the trajectories was taken into account. So conditions
were chosen to be most favourable for nearly central collisions.

The mathematical expectation of the ratio of the number of rings

with apparent magnitude less than a given value to the corresponding
number of spiral galaxies is only 10- 5 percent. This figure cannot be
changed more than one order of magnitude depending on the adopted para-
meters of the velocity distribution which were changed within its perm-
issible limits. The calculated value of Nr/N s thus is 10 3 times smaller
than the observed ratio of rings to the number of spirals.

Collisions in double and mUltiple galaxies which have been dynami-

cally connected since their formation are also of very small probability.
In the case in which the deceleration of one of the galaxies occurs, the
eccentricity of its orbit becomes smaller. The study of globular
clusters has shown that increase in the ellipticity of orbits only takes
place in some cases for orbits in the rather dense parts of the Galaxy.

If the companion had a radial orbit from the beginning, then

according to Toomre (IAU SymposiuD No. 58, 360, 1974), after 2 or 3
oscillations it must coalesce with the primary. If we observe the first
collision of this system, the components must have been formed with very
small relative tangential velocity (less than 1 km s-l for a mass of
lOll M@ and an orbital period of 10 10 years).

The fact that all these types of collision are of very small proba-
bility suggests that the mechanism of formation of the ring must be
inherent in the galaxy. Perhaps the frequent presence of a companion
plays some role in its development.

DISCUSSION

Tovmassian: The mov~e and slides shown by Prof. Toomre were very
impressive but I would still like to warn that one must be careful when
discussing interacting systems. As an example I would like to mention
the galaxy NGC 520 which has a very curious shape and which was consid-
ered an interacting pair by Toomre and Toomre. Radio observations of
SHORT COMMUNICATIONS 119

this galaxy made by myself and R. Sramek with the Green Bank interfero-
meter revealed an unresolved radio source in the very centre of this
object. In my opinion, this rules out the hypothesis of interacring
galaxies. The results of our observations, and also a photograph of
NGC 520, were published in the first issue of Astrofisika in 1976.

Toomre: You may be a little too pessimistic. Why couldn't that compact
radio source be the nucleus of one of two almost overlapping systems?
And maybe it was even "turned on" by some recent tidal accretion of gas
from its neighbour? Anyhow, I do know that Stockton in Hawaii has
recently measured the rotations of those two seemingly edge-on hulks in
NGC 520. He finds both of their southeast ends to be receding, as
indeed they should be on a simple tidal picture.

Ambartsumian: In the ring film by Reynolds you showed some results

calculated for very close encounters. But since distant passages are
much more probable, what are their effects?

Toomre: As several workers have found, passages of similar masses at

distances greater than one or two disk diameters rapidly yield nothing
remarkable or even very noticeable. Slightly deeper encounters where
the outer parts of two systems indeed graze or mildly penetrate each
other are the ones that tend to produce the nicest bridges or tails.
Still deeper intrusions at, say, half a disk radius often result only in
a great deal of tidal splatter - although the rings arising from the yet
more accurate (but rarer) central hits are, of course, a delightful
exception.

Sunyaev: Dr Toomre presented beautiful slides and films of the rings

which appear in close encounters of galaxies. T. Eneev, N. Koylov and
I have also made films of our simulations of encounters of galaxies
which also demonstrate that ring features appear due to tidal interac-
tions. I want to mention one interesting case. In some simulations, a
point mass M passes by the "galactic" disk of massless particles in
Keplerian orbits about a central mass of mass 2M. A significant frac-
tion of the particles is captured by the point mass. A disk is formed
around the point mass. The most interesting thing is that the plane of
the newly-formed disk is perpendicular to the plane of the first galaxy.
A smaller fraction of the particles is evaporated due to tidal effects.

Kiang: Your picture shows that the ring is a bransient feature. It

dissolves even as the satellite leaves the scene. Do you, then, find
objects in the vicinity of the Cartwheel (and the like) that can plaus-
ibly be identified with the projectile?

Toomre: Yes indeed. Not only do we find those two candidates quite
close to the Cartwheel (see Fig. 4), but it has been stressed by several
observers since the late 1960's that good rings almost invariably have
close companions. In his 1973 thesis, Theys went one step further: he
noted that those companions often tend to lie near the minor axis.
120 SHORT COMMUNICATlONS

Morton: How do the spokes form in the Cartwheel?

Toomre: I am not sure, but I presume it is some intense shear-instabil-

ity rather than just simple gravitational clumping that tends to make
the interstellar gas extra uneven and lumpy in that outward-rushing zone
of strong crowding. In any case, like Fosbury and Hawarden, I regard
those spokes as gaseous and young stellar debris that has already been
left behind by the main circular shock wave, and that is now falling
back down toward that inner ring (like the intermediate trajectories in
Figure 6).

Richter: A question about shock waves. I understand that the ring in

the Cartwheel nebula may contain a shock front producing population I
stars. However, from your movies it seemed to me that at the time when
the ring is most prominent, most of the population II stars (except from
near the nucleus) have themselves been swept into the ring. Hence the
disk should contain nearly pure population I, whereas the population in
the ring should be mixed. Therefore the disk spiral arms (or the
"spokes") would need to be bluer than the ring. Is this right?

Toomre: No, I don't think so. All random motions were omitted in these
simple simulations. Had they been included for population II stars,
their "ring" would undoubtedly be much broader and less impressive.

Komberg: It was found by Smirnov and myself in a sample of nearby

spirals that HII regions and DB associations are distributed in ringlike
zones superposed on the spiral structure. In these ring structures the
younger objects seem to be further from the nucleus. Also there are no
visible companions. What does this mean from your point of view?

Toomre: I think the strong and sharp rings are a very special and rare
class of galaxies. They are not at all to be confused with the faint
and broad ringlike distributions of extra light seen fairly often in the
exteriors of spiral galaxies with at least some hint of barred or oval
structure. And of course the neutral hydrogen itself often tends to be
depleted (used up?) near the centres of disks, as Roberts first stressed
about a decade ago. I suspect your effect is related to the hydrogen.

Heidmann: Yesterday I showed a photograph of Markarian 271 where there

appears to be a single one-armed spiral feature, like a whorl, connected
to a small barred spiral. Do you think such a model could fit into your
models for rings, with some special parameters?

Toomre: I won't know until I try! I doubt it will fit easily, though,
since any deep-set spiral structure must need help from the self-gravity
which my test particles simply omit.
ENCOUNTERS OF SPHERICAL GALAXIES :
N-BODY SIMULATIONS AND COMPARISON WITH THEORETICAL PREDICTIONS

Peter Biermann and Roland Wielen

Astronomische Institute Astronomisches Rechen-Institut

der Universitat Heidelberg
Bonn, F.R. Germany F.R. Germany

We have studied hyperbolic encounters of spherical galaxies by self-

consistent N-body simulations. Each galaxy is represented by a Plummer
model. A galaxy contains up to 250 'particles'. The force between the
particles is properly softened in order to avoid an unrealistic internal
evolution due to close encounters of the particles. The results of these
numerical experiments, carried out by R.W., are compared with theoretical
predictions based on the impulsive approximation, made by P.B . . In order
to facilitate the assessment of such a comparison, the passing galaxy is
here represented by a rigid gravitational field, moving on a straight
line. In the table given below, we compare the relative mass 10ss,6M/M,
and the relative change in the total internal energy of a galaxy, 6E/E,
for two cases. Both galaxies are of equal mass and size ; the impact
parameter is equal to the median radius of a galaxy (containing 50% of
its mass in projection) ; the initial relative velocity would correspond
to the orbital excentricity e as given, if the galaxies were mass points.
While the agreement is fair with respect to the energy transfer 6E, the
theory predicts a mass loss of more than 10% (or 25 particles) in cases
where no mass loss is observed in the N-body simulations The probable
reason for this discrepancy is the failure of the impulsive approximation
in these cases. A mass loss of the order of 1% or less cannot be ruled
out in the N-body calculations, because of the limited number of parti-
cles. The significant increase of the internal energy of a galaxy during
an encounter leads to an expansion of the outer parts of the galaxy. This
expansion may finally lead to a mass loss as soon as the outer parts have
expanded beyond the tidal radius of the galaxy caused by its cluster
environment. The inelasticity of galactic encounters has implications for
the dynamical evolution of groups and clusters of galaxies. The groups
and clusters will shrink in radius because of the loss of orbital energy
which is used to expand the individual galaxies. The final amount of
collapse of the group or cluster is governed by the relative amount of
the internal binding energies stored initially in the individual galaxies
and released during the galactic encounters, in comparison to the bing-
ing energy of the group or cluster.
e=2.61: N-Body Imp. Approx. e=9.68 : N-Body Imp.Approx.
o % 10.6 % o % 2~1 %
23.3 % 20.4 % 6.9 % 9.5 %
121
M. S. Longair and J. Einasw (eds.), The Large Scale Strncture of the Universe, 121-122. All RIghts Reserved.
Copyright © 1978 by the fA U.
122 PETER BIERMANN AND ROLAND WIELEN

DISCUSSION

Ostriker: Two comments on your interesting results may be appropriate:

(1) I quite agree that the relative change in energy per collision
is usually large compared to the relative loss of mass. The energy
change is typically concentrated near the half mass point, but the mass
loss occurs only in the outermost regions. Thus, it is difficult for
N-body integrations with a small N to have enough particles in these
regions (~m/m ~ 5 x 10- 3) to give an accurate result.
(2) For galaxies like known galaxies (cluster or field objects),
Vrms -, 200 km s-l, but great clusters have Vrms .~ 1000 km s-l. Thus the
maximum contraction would be 5% and quite generally one can show that a
catastrophe will not, cannot, occur. The maximum density of the cluster
is 1/N 2 of the density of the initial galaxies, where N is the initial
number of galaxies in the cluster.

Wielen: (1) My point of view is that the mass loss from galaxies in
clusters occurs mainly "indirectly": galactic encounters 'excite' some
stars in the galaxies to such energies that they can escape by crossing
the tidal radius of the galaxy. Stars are usually not directly 'kicked
out' of the galaxies by galactic encounters, i.e. they do not acquire
positive energies from the encounter. Therefore, the (indirect) mass
loss will strongly depend on the tidal radius and hence on the cluster
environment of the galaxies.
(2) I have not claimed that the collapse of a cluster will be
complete. The collapse will be stopped as soon as the binding energy,
originally stored in the galaxies, is transferred to the cluster. This
corresponds exactly to your density limitation. I agree that the Coma
cluster will not contract significantly in the future. However, one can
turn around that argument: Perhaps the inelasticity of galactic
encounters was so effective in the past that most of the binding energy
has already been transferred from the galaxies to the cluster, thus
leading to the presently observed distribution of binding energies
between the galaxies and the Coma cluster. In many groups and small
clusters, the velocity dispersion in the cluster is smaller or equal to
the velocity dispersion in the galaxies. These groups and clusters
should be able to contract significantly in the future due to the
inelasticity of galactic encounters.
ON THE MAGELLANIC STREAM, THE MASS OF THE GALAXY AND THE AGE OF
THE UNIVERSE

D. Lynden-Bell
Institute of Astronomy,
The Observatories,
Cambridge CB3 OHA, England.

The Magellanic stream has been fitted with high accuracy in both
position and velocity by the tidal tearing of a Magellanic Cloud. To
get the good fit to the high velocity at the stream's tip at a suitable
distance from the Galaxy we need either a large mass for the Galaxy, or
a large circular velocity for the Sun, or both. An extragalactic
method of determining the circular velocity yields the high value of
V = 294 ~ 42 km/sec and an orbit of poor accuracy for the relative
m6tion of the Galaxy and the Andromeda nebula. Very large masses are
needed if Andromeda and the Galaxy were formed together. A new direct
determination of Hubble's constant from the "superluminal" expansion
observed in VLB radio sources gives an age of the Universe of 9 billion
years. Either larger masses still or smaller distances within the
local group are necessary to bring Andromeda back towards us in so short
a time.

Discovery of optical object~ltying in the direction of the Magellanic

Stream and other hydrogen clouds stimula~21 my interest in explaining
the stream. \"ork by Hartwick and Sargent has since shown that the
apparent coincidences are caused by projection since the velocities of
the optical objects differ from those of the hydrogen. Nevertheless
over the intervening years D.N.C. Lin and I have run some forty thousand
test particle orbits comprising(290 passages of a Magellanic Cloud
clothed with 200 test particles 3. Typically we start our Magellanic
Cloud at apo-galacticon with its rotation inclined at some'specified
angle to the plane of the Magellanic Cloud's orbit. As our Magellanic
Cloud SWEepS past perigalacticon and on, it is tidally torn making a
bridge towards the Galaxy and a tail away from it. We have discovered
orbits which give a tail and no bridge and others that give bridges
without tails. We have been at pains to fit the rather straight run
of velocity with angle around the sky which is observed in the Magellanic
Stream. l.Je find this straightness to be a characteristic of orbits
that are somewhat retrograde in the sense that the spin of the Magellanic
Cloud has an inclination of 1100 to 130 0 to its orbital angular momentum
about the Galaxy. A description of the orbits searched over can be
found in our paper. We search in dimensionless units and when we
123

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 123-130. All Rights Reserved.
COPYright © 1978 by the [AU
 ~
-

!l
Figure 1. Our best fitting orbit for the Magellanic Cloud about the Galaxy. The small orbit of
The Milky Way is seen near the focus. Picture is taken from an angle 300 above the orbit plane and
the distance of poin~of the stream from that orbit plane are indicated. The orbit has e = 0.7, ~
t""
i = 130°, ~M = 1/50, P = 110°. -<
Z
o
t'r1

~
t'r1
t""
t""
THE MAGELLANIC STREAM, THE MASS OF THE GALAXY AND AGE OF TilE UNIVERSE 125

finally find an eccentricity e, a mass ratio llM' a Hagellanic Cloud

size r, an inclination i and a direction of spLn at apogalacticon P
that can scale exactly onto the observations, we determine the length
and velocity scales. Using Newton's gravitational constant G, we can
then determine the mass of the Galaxy. We found that such a fit was
not unique unless a circular velocity for the Sun's motion around the
Galaxy was first specified; however a single best fit configuration did
well in all cases if the velocity scale (and therefore mass scale) was
changed for each choice of circular velocity assumed. (See Figure 1.)

To explain the high velocity of approach towards the Sun of the tip
of the Hagellanic stream we need either a large velocity for the Sun
around the Galaxy so that there is less to explain in terms of infall,
or a large infall velocity. A large infall velocity can arise either
because the tip of the stream has been pulled inwards by a very heavy
mass, or because the tip has fallen very deep into the Galaxy's potential.
We found that the almost great circle shape of the stream was spoiled by
parallax if the tip of the stream was too close to the Galaxy. We
could not fit our best model to the positions of the stream on the sky
if the tip (w?s closer than about 16 kpc from the galactic centre .(IBvies
and Wright 4J have assumed that the stream's tip is a factor of two
closer.) For the following circular velocities we get the following
masses for the Galaxy:

300 290 250

3.8.10 11 4.3.10 11 7.6.10 11

It is evident that unless the circular velocity is very large, then the
Galaxy must have a heavy halo. This result turned our interest towards
the value of the circular velocity.

The radial velocities of the galaxies in the local group reflect

the motion of the observer which is the compound of the motion of our
galaxy with the Sun's motion within it. It has always been assumed
that these two velocities cannot be separated by extra-galactic measure-
ments so that the speed of the Sun's circular motion about the galaxy
V cannot be determined this way. However this is false. First
d~termine the Sun's motion, U, with respect to the galaxies of the local
group excluding the Andromed~ subgroup. Assuming these minor galaxies
define the same zero of velocity as the centre of mass of the Galaxy
and Andromeda, the momenta of the Galaxy and Andromeda must be equal
and opposite. Thus if Q is the galaxy's velocity, Andromeda's will be
-llQ, where II is the mass ratio MG/MA. The observed radial velocity of
Andromeda will therefore be

v = (-ll G-U)~ = (-ll(U - V ~) -Y).~A

Ar - - -A - c- - -
where £
is the direction of the circular motion and ~ is unit vector
-A
towards Andromeda, from which we may deduce
126 D. LYNDEN-BELL

v
c
Thus knowledge of the mass ratio ~ yields V. Applying this method to
10 independent members of the local group yields V = 294 ± 42 km/sec
for th t5 yelocity of free circular motion at the Su~'s distance from the
centre . The depen~5~ce on ~ has been removed by using the relation-
ship of Tully & Fisher which gives the luminosity and thus the mass
of a spiral galaxy proportional to the 5/2 power of its maximum circle
velocity. With Andromeda's circular velocity at 265 km/sec the least
squares result is insensitive to ~ in the range ~ to 1.25 and the most
probable result is always close to 290 km/sec.

This method yields not only V but also U and hence the Galaxy's
velocity G = U - V t
and Andromed~'s velocity -~G. With the aid of
these we ;an ~alcurate orbi(5)for the galaxies ab;ut one another. The
total ~BsS MA + MG required t~2bring them close to one another
2 x 10 years ago is > 1.3 x 10 M. A smaller circular velocity will
yield a yet larger mass. My recen~ interpretation of "superluminal"
expansion in VLB gives directly a Hubb~e constant of 110 ± 10 km/sec/Mpc,
and a mat}mgr timescale of only 9 x 10 years for the age of the
Universe' Unless the distance to the Andromeda nebula has been
seriously over-estimated (by a factor of 1.5 or more) then the mass
needed to bring the two galaxies togrzher in so short a time as 9 x 10 9
years will be of the order of 4 x 10 M. However there is weak
evidence that other galaxies, Maffei I ~n particular, intervened to
complicate our naive picture of binary dynamics.

The Age of the Universe

·
F 19ure 2 sows
h VLB 0 b servat10ns
. (9) 0 f t h e separat10n
. 0 f components
in the radio galaxy 3C120. They were originally interpreted (using a
Hubble constant H = 55 km/sec/Mpc) as a burst expanding at 5c followed
by another expand~ng at 8c. However, the rediscussion that follows
shows that a single expansion along the dotted line at 4.6c on that
scale, accompanied by a rebrightening of a "central" object is a better
interpretation. We picture the geometry involved as a flash of very
strong waves in a tube formed by an accretion vortex about a black hole.
This flash propagates both ways along the vortex tube and escapes as two
equal pulses of strong waves, each travelling outwards at the velocity
of light. The electric fields of each strong wave pulse accelerates
the electrons it encounters and causes them to radiate synchro-compton
radiation in the radio region. It is the regions currently exciting
electrons that are seen as radio bright. Although the material of the
sources is not moving rapidly outwards the excited patches move out with
the velocity of light. If the two pulses set out an an angle 8 to the
line of sight, then the one moving towards the observer Slwill move with
apparent lateral speed c sin 8 (.)with respect to central black ho~e 0,
while the other source S2 - wi~r ~ove with apparent lateral speed ~ ~~~o~ l2)
The apparent lateral separation velocity of Sl and S2 will thus be
Vs 2~ (~while the ratio of the separations will Ee 1 - cos 8. (4)
S1n 8 1 + cos 8
THE MAGELLANIC STREAM, THE MASS OF THE GALAXY AND AGE OF THE UNIVERSE 127

16
I
II I
I ' I
E 14 I I
U II
co
I
M I
1/
-< 12
I[
Vl I
>- I I
~ I I II
Vl 10 I I II I I
Z
...,<t I
II

I
II I ~
8 I I
I

71 72 73 75 76 77
3C120 I
I
6 • 2·8 em I

o 3·8 em
• 6 ·0 em
5
If)
o
z
o
u 4
W
If)
U
0:::
« I
3

(0)

{ oj

O~--~~--~ __ ~~ ____- L____ ~ ____~____L -__~

70 71 72 73 74 75 76 77 78
Figure 2.
EPOCH
128 D. LYNDEN-BELL

Application to 3C120, z = .033

Figure 2 shows the observations given in reference 10. The 3.Scm
flux shows an outburst in 1973 which caused a double source. This was
seen as a simple double expanding at about v = 5c (solid line) until
1974.2 when the double model failed to fit w~ll, consequent upon a small
increase of flux. In 1975 there followed a violent increase of flux
and immediately a two source model fitted well again, but at about the
same separation as it was at 1974.0. The source began to dim in late
1975 and in 1976.1 it was again a well-determined double although the
flux at this time is not given. The authors of paper 10 have inter-
preted these events as the two superluminal expansions of 5c and Sc
given by the heavy straight lines. However on my picture this would
involve the vortex tube up the rotation axis of the radio galaxy to turn
between the two events and I consider this most unlikely, especially as
the observed position angle of separation on the sky are the same for
both doubles. The following interpretation of events is more natural
to me. 1972.5 - 1973 a double expanding source 51' 52 was born at 0
and expanded throughout 1973 as its flux diminished. In early 1974
the object at 0 brightens, confusing the neat double interpretation of
the data. In 1975.2 0 has increased in brightness so much that it forms
a double with 51' the source moving towards us and 52 is outshone and
ignored. 0 dims in late 1975 leading to a confused picture with 51'
o and 52 all seen. Finally in 1976 0 dies away leaving the old expand-
ing source 51 52. It is noted that this latest point fits on the dotted
expansion line ~4.6c) of the old source, which is a perfectly good alter-
native to the full line (5c) if one ignores confused data. The peculiar
interest of this new interpretation arises because it not only removes
the discrepancy of two different superluminal velocities observed from
the same object, and therefore probably at the same 8 but, because we
see 0 at times, we can measure the ratio of the separations 05 1 to 05 2
and hence determine 8. With 8 determined, the apparent transverse
separation velocity is known from (3) so the observed angular velocity
gives the distance and hence Hubble's constant. The small z = 0.033
of 3C120 makes the cosmological corrections unimportant for this source,
so the Hubble constant derived in this case is independent of q .
However the same method applied to 3C345 should yield a reasonagly good
value of qo too.

If 8 were 90 0 the clear double seen in 1975.4 would have been half
way up to the dotted line. Its fractional offset [051-45152)/5152
corresponds to the offset of 0 from the centre of 51 and 52 and ~s
therefore just 4 cos 8. From this we find 8 = 6So ± 10°. It is
particularly interesting that this angle is close to the inclination of
the apparent disk of 3C120 measured from Arp's picture on which it appears
to be a disturbed spiral. Measuring his picture we get i = 55 0 - 60°
but this may well be in error by 10 0 due to the fact that the disturbed
disk does not have a circular outline. Furthermore the observed separa-
tion of the radio double, although fairly close to the apparent mirror
axis of the optical image is not along it. It seems reasonable to
assume that ejection is taking place <llong the rotation axis, but the
THE MAGELLANIC STREAM. THE MASS OF THE GALAXY AND AGE OF THE UNIVERSE 129

galaxy has imperfect circular symmetry. As in such a case i should

equal e and both determinations have similar errors we shall adopt
e = 64 ± BO • The light echo theory then gives us
v 2c/sin 64 0 = 2.22c ~ 0.19c
s
The authors of reference S adopt H = SSkm/sec/Mpc, and obtained Sc for
this velocity, but our correction ~o the dotted line interpretation of
Figure 2 yields 4.6c on that scale. To get the answer deduced from the
light echo theory we must change the Hubble constant by a factor
4.6 ± .2
= 2.1 ± 0.2. We deduce a Hubble constant of ltS±12km/sec/
2.22±.19 Mpc.
Similar work on She 3C273 observations yields e = 79 0 and a Hubble
constant of l13±10 km/sec/Mpc. The statistics of the superluminal
expansions of Sl S2 for 3C120, 3C273 and 3C34S yields Ho=110±20km/sec/Mpc.
In summary this physical method of measuring H yiel§s Ho=llO±lOkm/sec/Mpc
and a maximum age of the Universe H -1 =(9 ± 1) x 10 years provided the
• •• 0
1nterpretat10n 1S correct.

References

1. Lynden-Bell, D.: 1976, Monthly Notices Roy.Astron.Soc. 174, p.69S.

2. Hartwick, D., and Sargent, W.L.W.: 1977, preprint Astrophys.J.
3. Lind, D.N.C., and Lynden-Bell, D.: 1977, Monthly Notices Roy.
Astron.Soc.
4. Davies, R.D., and Wright, A.E.: 1977, Monthly Notices Roy.
Astron.Soc. lBO, p.7l.
S. Lynden-Bell, D., and Lin, D.N.C.: 1977, Monthly Notices Roy.
Astron. Soc. lBl, p.
6. Tully, R.B., and Fisher, J.R.: 1977, Astron.Astrophys.S4, p.66l.
7. Lynden-Bell, D.: 1977, Nature in press.
B. Lynden-Bell, D.: 1977, Proc. of the Cambridge NATO Summer School
(editors S. Mitton and C. Hazard) ,on Quasars.
9. Cohen, M.H.: 1977, Nature, 26B, p.40S.
10. Cohen, M.H., Linfield, R.P., Moffet, A.T., Romney, J.D, Seielstad,
G.A., Kellermann, K.I., Shaffer, D.B., Pauliny-Toth, I.I.K.,
Preuss, E., Witzel, A., Schilizzi, R.T., Geldzahler, B.J.
1977, Nature, 26B, p.40S.

DISCUSSION

de Vaueouleurs: Do you believe that the spin axis of the accretion disk
in the black hole model coincides with the spin axis of the galactic
disk?

Lynden-Bell: If the system has been left alone without interaction for
a long time, then yes.

de Vaueouleurs: If so, does e 6B o agree with the inclination derived

from the shape of 3C 120?
130 D. LYNDEN-BELL

Lynden-Bell: i55 0 - 60 0 ± 10 0
=
e68 0 ± 10 0
=
I think this is good agreement. I took e 64 0 •

THE STELLAR COMPONENT OF THE MAGELLANIC STREAM

William E. Kunkel

The problem of the Magellanic Stream (MS) is re-examined consider-

ing also stellar systems (globular clusters and dwarf spheroidal
galaxies) lying between 25 and 250 kpc from the Galactic center (Kunkel
and Demers 1976). The distribution of these systems is shown to be non-
random, with a planar component whose inclination to the Local Group is
43 0 , and to the local supergalaxy exceeds 60 0 • The velocity distribu-
tion of members of the group shows a symmetry typical of debris associ-
ated with a disruptive event at perigalacticon of a victim moving in an
elliptical orbit.

These data are examined in terms of tidal disruption affecting the

Magellanic Cloud system, with a view to constraining the multidimension-
al character of admissible solutions from modelling efforts using
computer simulation. The observed angle between the plane of the MS and
that of the stellar or Magellanic Plane Group (MPG) is explained in
terms of one being residue from a bridge and the other from a tail. The
angle depends to a first approximation on only the mass ratio of the
victim to the perturber, which is found to be 0.028, consistent with
classical estimates for the masses of the LMC and the Galaxy.

Color-magnitude studies of the stellar systems, as well as the

marked composition difference between the MS and the Magellanic Clouds
point to a considerable age for the epoch of the most disruptive event,
possibly just less than the age of the Galactic globular cluster system.

The sense of orbital motion (which also determines the status of

bridge or tail) is determined independently from the age of the oldest
(and most numerous) stellar component in the Wing of the Small Cloud,
from the dominance of HI in the MS and of stars in the MPG, and from the
uniformity (or its lack) of debris material in position along the debris
orbits.

The model most favored by these arguments moves in a sense opposite

to that of Davies and Wright (1977), and of Lin and Lynden-Bell (1977).
The first encounter producing the most pronounced disruption (and much
of the MPG and MS) occurred some time ago; the current position of the
Magellanic Clouds near perigalacticon correspond to a second or possibly
subsequent encounter with relatively mild interaction; the orbital
period is of the order of 5 x 10 9 years. The total Galactic mass lying
inside perigalacticon consistent with this model is a factor of 3 or 4
greater than the classical values in current use.
SHORT COMMUNICATIONS 131

A more complete account of this work will appear elsewhere.

REFERENCES

Davies, R.D., and Wright, A.E., 1977. Mon. Not. R. astr. Soc., 180, 71.
Lin, D.N.C., and Lynden-Bell, D., 1977. Preprint.
Kunkel, W.E., and Demers, S., 1976. R.O.B., no.182.

DISCUSSION

de Vaucouleurs: May I remind you that the mass of the LMC is about
1.5 - 2.0 x 10 10 Mo?

Kunkel: Such a high mass would satisfy interest in a more maSSLve

Galaxy. I had used the lower classical value (of 6 x 10 9 based on
McGee and Milton 1966) merely to show that no serious conflict exists
for even the older mass estimates.

FINE STRUCTURE IN THE MAGELLANIC STREAM

A. G. D. Philip and A. G. Davis

Dr Erkes and I, of Dudley Observatory, and Dr Ken Turner, of the

Carnegie Institution of Washington, used the 300 foot radio telescope
at the National Radio Astronomical Observatory, Greenbank, West Virginia
in March of 1977 to measure neutral hydrogen at 21 cm in three 20 strips
across positions indicated by Mathewson, Cleary, and Murray (1974,
~strophys. J.~ 190, 291) as being part of the Magellanic Stream. In
--- h m
region "A" (centred at a = 22 44 , 0 = +24 .5) no hydrogen was found
above the background (3 x 10 19 H atoms cm- 2 ). In a revised map
Mathewson and Schwarz (1976, Mon. Not. R. astr. Soc.~ 176, 47P) show
that this region is not occupied by the stfieam. In regions "B" and "c"
(centred at a = 23 h lOm, 0 = +11 0 and a = 0 OOm, 8 = -17 0 ) hydrogen was
detected at the velocities found by Mathewson and Schwarz. Maps of the
structure found in these regions are shown Ln Figure 1.

The contour intervals are 10 K km s-l. The maps cover 20 in decli-

nation and about 30 minutes in right ascension. The location of major
features of the stream agree with a less detailed map of Mathewson,
Murray and Schwarz (1977, preprint). Fine structure can be seen in each
map; the stream is broken up into two main components running approxi-
mately parallel to each other. The half-power beam width was 10 arcmin.
132 SHORT COMMUNICATIONS

Figure 1. Hydrogen maps of regions "B" and "e" ~n

the Mage11anic Stream.
II

CLUSTERS OF GALAXIES
PHOTOELECTRIC SURFACE PHOTOMETRY OF THE COMA CLUSTER*

Jorge Melnick 1 , Simon White 2 and John Hoesse1 3

I European Southern Observatory, Geneva, Switzerland.
2 Institute of Astronomy, Cambridge, England.
3 California Institute of Technology, Pasadena, USA.

A photoelectric search is presented for light from a smooth inter-

galactic background in the Coma cluster of galaxies.

A fully orientable 2-inch remotely operated sky monitor was mounted

on the optical axis of the Palomar 60-inch reflector in front of its
secondary mirror. This small telescope was used to continuously measure
a fixed patch of sky, while the main telescope carried out drift scans
across the cluster covering an over-all area of 78'(RA) by 14'(DEC).
This procedure reduced the effect of sky brightness fluctuations to an
r.m.s. level of less than I percent of the sky in the individual scans.
The observations were carried out in two intermediate-width bands cen-
tred at 4930 A
(g) and 6550 A
(r). A complete description of the ob-
serving and data-reduction procedures has been given elsewhere (Melnick,
White and Hoessel, 1977).

Surface brightness distribution in r after removal of sky and foreground

star contamination, and smoothing with a 3!5 FWHM Gaussian. Contour
levels are in 20 ct/sec steps equivalent to 26.6 mag/sec 2 •

The over-all light distribution in the cluster is seen in the

figure to follow very closely the distribution of galaxies in the clus-
ter. The two supergiant galaxies in the cluster centre are seen to
*The pape~ was read for the authors by J. Materne.
135
M S. Longair and J. Einasto reds.), The Large Scale Structure of the Universe, 135-136. All Rights Reserved.
Copyright © 1978 by the fA U.
136 JORGE MELNICK ET AL.

have a common envelope, extending more than 750 kpc in the EW direction
(H = 50). More quantitative results are obtained by separating the
data into seven semi-annular groups. The results are shown in the
table, where the outer radius of each annular region is indicated.
Also included are the visual magnitude of each region obtained from the
present g and r data, V, and the total contribution to the light in
each region by galaxies brighter than VZ5 = l8~0, VZ5. A comparison
between the V and VZ5 magnitudes suggests that between 15 and 20 per-
cent of the light of the cluster comes from galactic halos, faint
galaxies, and the intergalactic medium (IGM).

R g r V VZ5

0-6 10I!l38 ± 0.09 91!197 ± 0.09 101!l20 10~54

6-12 11.24 ± 0.09 10.62 ± 0.08 10.98 11.09
12-18 11.45 ± 0.15 10.94 ± 0.14 11.23 11.37
18-24 11.59 ± 0.13 11.07 ± 0.12 11.37 11.36
24-30 13.04 ± 0.32 12.17 ± 0.22 12.69 12.27
30-36 14.53 ± 0.90 13.46 ± 0.58 14.11 13.52
36-42 14.23 ± 0.49 13.74 ± 0.45 14.03 14.51
0-24 9.55 ± 0.07 9.05 ± 0.07 9.34 9.53
0-42 9.48 ± 0.09 8.97 ± 0.08 9.26 9.41

An estimate of the true intergalactic luminosity can be obtained

from our data alone by considering only the counts coming from regions
well away from galaxies brighter than VZ5 = 18.0. The mean of both
bands averaged over the seven regions considered above gives
LIGM/LGAL = 0.18 ± 0.05 which must be considered as an upper limit since
much of the observed intergalactic light may still come from material
which is bound to individual galaxies. It is concluded, therefore,
that 25 percent is a conservative upper limit to the amount of light
that might come from a true intergalactic background. If the cluster
is bound by intergalactic material, this medium cannot consist of any
usual stellar population.

REFERENCE

~felnick, J., White. S.D.H., and Hoessel, J.: 1977

M.N.R.A.S. 180, pp 207-218.
DISCUSSION

Ostriker: Do you have any information on the colour of the background

light in comparison with the colour of the galaxies?

Materne: B-V is about 2 but with a very large error.

AbeLL: Is it consistent with the colour of galaxies, say B-V~ I?

Materne: Within the errors, yes.

RADIO PROPERTIES OF ABELL CLUSTERS

C. Lari
Laboratorio di Radioastronomia, Bologna
G.C. Perola
Istituto di Scienze Fisiche dell'Universita, Milano

Since the early 60's the association of some radio sources

with rich clusters of galaxies was noted (Mills, 1960; van den Bergh,
1961) and subsequent investigations concentrated mostly on the radio pro-
perties of clusters in the Abell catalogue. This paper is confined to
this subject, although one should bear in mind the importance of associa-
tions with other scales of clustering which do not appear as entries of
. -1
that catalogue. In thlS paper we shall use Ho= 100 kID sec and the
radio power P in WHz-l at 408 MHz.

1. RADIO LUMINOSITY FUNCTION OF GALAXIES IN ABELL CLUSTERS

The first item that attracted general attention is whether ga-

laxies inside Abell clusters have a probability of being radio sources
different from that of the galaxies outside.This item is best studied by
constructing a radio luminosity function (RLF). Recent determinations
based on large samples have been obtained by Owen(1975), Riley(1975),
Mills and Hoskins(1977). The first two are in the form of spatial density
or fraction of clusters with radioemission, the third, based on an attempt
Lo identify each source specifically with a cluster member, can be more
directly compared with the bivariate RLF of the general population of
galaxies, which expresses quantitatively the dependence of the probabi-
lity of radioemission on the optical absolute magnitude M of the galaxies.
A recent study at 1415 MHz with the WSRT (WCI05 project, D.Harris, C.Lari,
J.Vallee, A.Wilson) of B2 sources located within one Abell radius, RA ,
of Abell clusters with D~5, combines a good positional accuracy with a
reasonably good resolution even for D=5 clusters, and therefore improves
over Mills and Hoskins (MH) in the identification process. We shall refer
extensively to the preliminary results we have extracted from it. WCI05
contains all Abell clusters, surveyed in the B2.1-5 catalogues, with
D~5 (21 o<Dec<42°, tot. of 279). MH studied all clusters south of +18 0
with D~4, plus a random selection of D=5 and 6 (tot. of 247). The two
samples are completely independent, while there is a large overlap
between WCI05 and the samples used by Owen and by Riley.

137

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 137-147. All Rights Reserved.
Copyright © 1978 by the fAll.
138 C. LARI AND G. C. PEROLA

• FRG from WelD5

Log RPl
log F(P)
• from WC105 o IIII.. S-20 not fAG from WelDS
• Not fRG within o.3RA , from Mill, .nd Hoaklns (1917)
X MilI5 .nd Hoskl"s (197n

-2
-2

230 23A 238 242 2.(.6 250 25A 258

logP(408MHz) (WHZ- 1)

234 23.8 242 :250 2M 258

LogP(40B MHz) (W Hz-1)

Figg. I and 2. Radio luminosity function of the first ranked galaxies

and of the other member galaxies in Abell clusters.

In WCI05 we find 76 sources which can be associated with 63 clusters.

About 90% of the "gOOd" associations can be unambiguously identified with
a single galaxy. All the identifications are with E type galaxies, except
that a few objects are too faint to classify on the PSSP. This allows a
comparison to be made with the local (z~O.I) bivariate RLF constructed
by Auriemma et al(1977) for the E+SO type galaxies. These authors find
that, at logP>24.5 and up to at least 26, the probability of radioemis-
sion depends on M as 10-0.6M, irrespective of P. This dependence weakens
as logP decreases below 24.5.

l.i. RLF of the first ranked galaxies (FRG).

Fig.l gives the fractional RLF of the FRG's from MH and from WCI05,
along with an expectation based on the bivariate RLF in Auriemma et aI,
and on the distribution in Mv of the 82 FRG's in Sandage(1972). The MH
points (from which we have subtracted the D=6 clusters to avoid an evolu-
tionary bias relative to WCI05) are in excess over our estimate at logP
greater than 25. This may be due to a statistical fluctuation, but also
to the stricter criteria adopted by us in the identification process.
It is very important to solve this dilemma, because WCI05 gives only a
slight excess (1.50) over the expectation at logP>25, while the excess
becomes 3.50 if the two samples are combined. The latter result, if con-
firmed by a study of the MH clusters technically homogeneous to WCI05,
would be difficult to evaluate as a real excess, because of the uncerta-
inty on the 10-0.6M dependence and the probable presence of systematic
biases in the RLF used for comparison. One of these is the incidence of
the SO galaxies in the normalization adopted by Auriemma et al: their
exclusion as galaxies with radio properties different from those of the
typical morphologies of the FRG's would reduce the discrepancy. Moreover
evolutionary effects cannot be excluded at logP>25, although a recent
study (Ulrich et aI, in prep.) shows that in a range of redshifts compa-
RADIO PROPERTIES OF ABELL CLUSTERS 139

rable to that of D=4,5 clusters, the general RLF is still the same as the
"local" one up to logP=26. A possibility is a difference of the evolutio-
nary rate in the clusters. A real excess would imply in any case that the
cumulative lifetime of FRG's as strong radio sources is longer than for
equal magnitude galaxies outside of clusters.

l.ii. RLF of the other cluster members.

In Fig.2 we give from WCI05 the number of sources per cluster iden-
tified with galaxies other than the FRG, which, if members, would be bri-
ghter than Mv =-20 (to avoid background contamination). The same function,
but for sources within 0.3RA' derived from MH is also given. The agree-
ment is good, even if 30% of the associations in WCI05 lie outside 0.3RA.
It appears that: a)the forms of the RLF for the FRG's and for the other
galaxies are rather similar, but the proportion of FRG's fainter than
logP=24.8 is significantly smaller than that of the other members.
b) The form of the RLF for the members other than the FRG's is fairly si-
milar to that of the general RLF for galaxies brighter than My=-20. Com-
parison in absolute value (that is, per galaxy) with the latter requires
knowledge of the average number of members brighter than -20 in the
clusters surveyed. From counts available for a restricted number of near-
by clusters we estimate the normalization factor to be between 10 and 15;
with such a value, the agreement with the expectation would be reasonably
good. We note that WCI05 does not confirm the presence of a turn down
in the cluster RLF found by Owen (1975) below logP=24.8, but agrees with
the finding by Auriemma et al, based on 5 nearby clusters, that from
logP=22 to logP=24 the RLF in clusters does not differ significantly
from the general one.

l.iii. Correlation with Bautz-Morgan type.

Guthrie(1974), McHardy(1974), Tovmasyan and Shirbakyan(1974) found

that powerful sources occur more frequently in BM class I clusters than
in any other class. The results from WCI05 are given in Table 1.

Table 1. Distribution of sources per Bautz-Morgan class

B-M class I I-II II II-III III

(a) 7( .28) 5(.20) 2(.08) 3( .12) 8( .32)
(b) 2(.06) 3(.08) 7(.18) lOt .26) 15(.42)
(c) lOt .13) 7(.09) lOt .13) 16(.21) 33(.44)
(d) .22 .ll .13 .21 .33

The key of the table is: (a) number (and fraction) of clusters with the
FRG detected (we find no correlation between radio power and EM class);
(b) clusters with a galaxy detected other than the FRG. It is clear that
EMI are more frequent in (a) than in (b). (c) is the distribution among
140 C. LARI AND G. C. PEROLA

76 nearby clusters (McHardy,1974). Although there can be doubts that (c)

is an adequate representation of clusters in D=4,5 it is noteworthy that
(b) and (c) are consistent with each other, so that the effect found by
the previous authors has to be attributed to the FRG's. The last line (d)
is the distribution predicted from (c) using the 10-0.6M dependence in
the RLF and the estimate by Sandage and Hardy(1973) of the deviation of
<M > of the FRG's from the overall mean, per BM class. Comparison of (d)
wiIh (a) suggests that the excess of FRG's in BMI with radioemission
can be attributed to a fair extent to the magnitude effect in the RLF.

We have seen however in l.i that this effect may be insufficient to

account for the fraction of radioemitting FRG's. The absolute magnitude
is conceivably not the only "good" parameter on which the probability of
radioemission may depend. Other parameters, like the colours and the
morphological properties, can be important. Tovrnasyan and Shirbakyan
(1974) have made an analysis on the PSSP of 355 clusters in D=5 surveyed
by them and by Owen. They find that outstandingly bright D galaxies,
dumbells and peculiar ellipticals appear to have a higher probability of
radioemission than the ordinary giant ellipticals. For the WCI05 sample
a similar study is in progress, but we mention that, out of 25 FRG's
detected, 11 and 5 are of D and dB type respectively.

l.iv. Correlation with richness class R.

Several authors (van den Bergh,1961; Rogstad and Ekers,1969;

Tovrnasyan and Shirbakyan,1974; Owen,1975; Riley,1975) found that the
percentage of radioemitting clusters is at best only weakly correlated
with R. This is not surprising because the probability of radioemission
in a cluster depends basically on the bright end of the optical luminosi-
ty function, and, for instance, Sandage(1972) found no correlation between
R and the average My of the FRG's. From WCI05 we have constructed the
RLF for clusters with different R, and we find (above 10gP=24.2, a range
of P where all richness classes in our sample are surveyed at best):
a)the fraction of FRG with radioemission is independent of R, except
that an excess of low statistical significance is found for R=2+3 rela-
tive to R=O+l. b)The fraction of sources per cluster not associated with
the FRG is 0.13 (R=O), 0.15 (R=l) , 0.45 (R=2+3), that is about three times
as many sources per cluster are found in R=2+3 than in R=O+l. Moreover
the percentage of clusters with more than one source increases with R,
confirming a finding by MH. These results indicate that the number of
sources found in clusters is proportional to the number of individual
candidate galaxies.

l.v. RLF of Spiral and Irregular galaxies.

These galaxies are weak radioemitters and cluster spirals have been
detected only in deep radio surveys of nearby clusters. A systematic
study of their properties is in progress (see Jaffe and Perola,1976).
Slight differences are found between cluster and field galaxies, and
between cluster and cluster. The latter appears to correlate with the
presence of optical emission lines (Jaffe, Perola and Valentijn, 1976).
RADIO PROPERTIES OF ABELL CLUSTERS 141

D FRG_I •• I••

~ Otherp •••I •• Mv S-20

--0- Coma Clu•. pi Mv 5

(norfMIllzed)

Fig.3. Radial distribution of radiogalaxies in Abell clusters.

Fig.4. Distribution of spectral index between 408 and 1415 MHz.

2. RADIAL DISTRIBUTION OF RADIOGALAXIES IN ABELL CLUSTERS

The radiogalaxies brighter than My=-20 in WCI05 have the radial

distribution (with respect to the centres in the Abell catalogue) shown
in Fig.3. For comparison is given also the fit to the distribution of
galaxies brighter than Mv=-19.2 in the Coma cluster (Bahcall,1973), nor-
malized to the same number of objects. Clearly the "composite radio clu-
ster" is more centrally condensed than the optical cluster. The excess
peak within 0.2RA is due to the FRG's and follows from the combination
of two effects, the first that the FRG's have a more centrally peaked
distribution than the other galaxies, the second that, from the lO-o.6M
dependence, the brightest galaxy has the largest probability in a cluster
to be a strong source. Since the FRG in the first two BM classes tend to
be more centrally located than in the others, on the basis of the results
in l.iii, we expect the radioemitting FRG's to have a narrower distribu-
tion than the general one. This effect has been noted already (Leir,1976).
After subtracting the FRG's, the radiogalaxies follow remarkably well the
optical distribution out to 0.5RA' but are proportionally far less nume-
rous than the galaxies beyond that point. Radiogalaxies brighter than
-20.5 have an even narrower distribution. These results we tentatively
explain as due to a progressive decrease with increasing radial distance
of the fraction of very bright galaxies in the "average" cluster. This
segregation in magnitude would be strongly enhanced by the M dependence
in the radioemission probability. This explanation need however to be
supported by specific counts on a representative number of clusters.
For the moment we cannot exclude that Fig.3 is evidence that a galaxy
is more likely to be a radio source the closer it is to the cluster
centre.
142 C'. LAR! AND G. C. PEROLA

3. SPECTRAL INDEXES OF RADIO SOURCES IN ABELL CLUSTERS

Several authors (e.g. Baldwin and Scott,1973; Slingo,1974; Colla et

al,1975; Roland et al,1976) have shown that steep spectrum sources are
found more frequently in clusters than outside. This result has been
interpreted as evidence of enhanced efficiency in the radio source confi-
nement by the intergalactic gas inside clusters, which allows the effects
of radiative losses to show up as a spectral steepening. In Fig.4 we
give the distribution of the spectral index between 408 and 1415 MHz for
sources with 10gP>24.2 in the WCI05 and in a complete sample of galaxies
outside Abell clusters (from Colla et al,1975, and Fanti et al,1977b).
Sources between 0.2 and one RA have a distribution similar in form to
that of the outside sources, although the median, indicated by an arrow,
is larger by about 0.15. Sources within 0.2RA have a definitely broader
distribution, and the median is larger by about 0.2. So the above results
are confirmed and it is found that the spectral index distribution in
clusters correlates with the distance from the centre, the one of the
sources in the innermost regions being the more dissimilar from that of
the outside sources. This strengthens the validity of the interpretation
mentioned. Roland et al(1976) find that the distribution is broader for
sources in BMI to II than in the other classes. This reflects the fact
that radiogalaxies other than the FRG are more centrally located in the
BMI to II classes than in the others: we find 60% against 32% within .2RA.

4. RADIO SOURCE STRUCTURES IN ABELL CLUSTERS

In the last 9 years it has become progressively more clear that the
content of sources per morphological type is different inside and outside
clusters, and a great deal of observational and theoretical work has been
devoted to the "peCUliar" radio structures found in clusters. We shall
not review this work here, but limit ourselves to statistical properties.
A careful comparison in/out need to be based on samples chosen from com-
plete radio surveys and having a similar distribution in P, because radio
morphology and size are to some extent a function of P (see Fanaroff and
Riley,1974; Gavazzi and Perola,1977).

4.i. Distribution of largest physical size.

The great majority of the WCl05 sources are within 24.2 and 25.2 in
10gP. For comparison we therefore use sources in the same power interval
from two samples of B2 sources identified with galaxies outside Abell
clusters brighter than ffip=15. 7 (Fanti et al,1977a) and with 14<mv < 17
(Fanti et al,1977b). These sources have also been mapped at 1415 MHz with
the WSRT. The distribution of the largest physical size (LPS) for 58
inside and 41 outside Abell clusters is given in Fig.5. The two distri-
butions are practically identical, with the median value of the LPS
(indicated by an arrow) within 15 kpc, and a maximum value in both cases
of 300-400 kpc.
RADIO PROPERTIES OF ABELL CLUSTERS 143

outside Abell Clusters inside Abell Clusters

24.2 < Log P < 25.2 24.2 < LogP < 25.2
15

10

Fig.5. Distribution of the largest physical Slze.

4.ii. Morphological types.

The percentage distribution of morphological types in the two sam-

ples is given in Table 2. Symbols: D=aligned double; V=V-shape or highly
misaligned double; T=tail or head-tail; CX=complex structure; H=halo;
PW=source extended on opposite sides of peak (this category probably
contains some unresolved doubles); NC=slightly resolved not classified;
NR=not resolved.

Table 2. Distribution of morphological types

D v T CX H PW NC NR

outside (%) 54 2 5 2 7 15 15

inside (%) 12 7 3 17 12

FRG's (%) 24 12 4 12 12 8 12 16
Note two striking differences: a) outside clusters more than half
of the sources have a D structure, while inside less than 20% show such
a structure. On the other hand the size distribution of the D's in the
two samples appear rather similar, except that, if the PW sources are
all considered unresolved doubles, there would be proportionally more
doubles with LPS less than 100 kpc inside the clusters. We recall that
previous analyses of the size of D sources in the 3CR catalogue (which
are generally brighter than those in our samples) also found no statisti-
cally significant differences between in and out (Hooley,1974; Burns and
Owen,19n) .

b)The V+T+CX type sources amount to at most 10% outside, but make
up 35% inside. It is remarkable that the percentage of these.types toge-
144 C. LARI AND G. C. PEROLA

ther with the D sources inside (52%) and outside (63%) are rather close
values. This result supports the opinion that the D "missing" from the
cluster sample are sources of the type V, T and CX, whose morphology is
dramatically affected by the physical conditions prevailing in the intra-
cluster medium, in particular the dynamical action of a denser than ave-
rage intergalactic gas. Its effects can be of various kinds, like (1)
drag on radiocomponents associated with galaxies moving through the medium
(Miley et al,1972); (2) buoyancy of radiocomponents (Gull and Northover,
1973); (3) asymmetric ram pressure on moving components (W.Christiansen
in Rudnick and Owen,1977); (4) bulk motions of the gas. In view of these
effects, it would be of considerable interest to study the radial distri-
bution of the various types. Unfortunately the statistics in the WCI05
sample is too poor for telling significant radial dependences. It seems
however that the FRG's (see Tab.3) differ from the rest. In particular
the H type sources in the sample are associated with an FRG, while only
one T source is. On the other hand, the percentage of V sources is the
same for the FRG's and the other galaxies. This result, along with the
finding by Owen and Rudnick(1976), and confirmed in WCI05, that T sources
are on average associated with less bright galaxies than the V's, supports
their view that while effect (1) is mostly responsible for the tail stru-
cture, effects (2,3) are likely to be more important in determining the
V shapes, being these sources associated with galaxies which can be
suspected to move at a relatively lower speed with respect to the medium.

Thanks to a higher resolution, the work by Rudnick and Owen(1977)

on several sources in cluster reveals the structure of small sources. In
their statistical sample of 37 sources, 11% are small doubles «30kpc)
and 16% are V sources with size less than 50kpc. It is remarkable that
a high percentage of doubles be distorted into a V when their size is
less or comparable to the main body of the optical galaxy. Moreover,
Rudnick and Owen note the curious effect that these V sources tend to
point away from the cluster centres, and interpret it as due to pressure
gradients or buoyancy in the general cluster field.

On the contrary, T type sources appears randomly directed (e.g.

Harris,1977).This supports the idea that their elongation is generally
in the direction of the galaxy velocity. The rather sharp bendings in
some of the longest tails can be due to buoyancy (Cowie and McKee,1975),
but there is no clear evidence that the tails are affected by bulk motions
in the gas, like a general outflow or inflow (Lea,1976). It has been
suggested that galaxies with radio tails may be endowed with peculiarly
high velocities. An analysis for 12 well known tails (Baggio,Perola and
Tarenghi, in prep.) shows the contrary, that is the Quantity x= ~v/a
(ratio of radial velocity difference with respect to cluster average,
to cluster velocity dispersion) appears normally distributed.

5. CORRELATION BETWEEN RADIO AND X-RAY EMISSION IN CLUSTERS

To check the existence of a correlation, we have used the 15 Abell

clusters with D$3 in the Ariel catalogue, all of which have been surveyed
RADIO PROPERTIES OF ABELL CLUSTERS 145

in radio. We find that 10/15 (67%) contain at least a radio source with
logP>24.2. The fraction of clusters in general with a radiogalaxy more
powerful than logP=24.2 is instead 25%. However, the 15 X-ray clusters
have a peculiar richness distribution: 1 with R=3, 7 with R=2, 6 with
R=l and 1 with R=O. If the prediction is corrected accordingly, the
expectation becomes 45% that is 7 out of 15. It seems to us that, on the
statistically limited basis of the sample used, there is little evidence
for the existence of such a correlation.

References

Auriemma, C., Perola,G.C., Ekers,R., Fanti,R., Lari,C., Jaffe,W.J.,

Ulrich,M.H., 1977, Astron.Astrophys.57, 41
Bahcall,N.A., 1973, Astrophys.J. 183, 783--
Baldwin,J.E., Scott,P.F., 1973, Mon.Not.Roy.Astron.Soc. 165, 259
Burns,J.O., Owen,F.N., 1977, Astrophys.J. 217, 34
Colla,G., Fanti,C., Fanti,R., Gioia,I., Lari,C., Le~ueux,J., Lucas,R.,
Ulrich,M.H., 1975, Astron.Astrophys. 38, 209
Cowie,L.L., McKee,C.F., 1975, Astron.Astrophys. 43,337
Fanaroff,B.L., Riley,J.M., 1974, Mon.Not.Roy.Astron.Soc. 167, P31
Fanti,C., Fanti,R., Gioia,I.M., Lari,C., Parma,P., Ulrich,M.H., 1977a,
Astron.Astrophys.Suppl. 29, 279
Fanti,R., Gioia,I.M., Lari,C.~Ulrich,M.H., 1977b, Astron.Astrophys.
in press
Gavazzi,G., Perola,G.C., 1977, Astron.Astrophys., in press
Gull,S.F., Northover,K.J.E., 1973, Nature, 244, 80
Guthrie,B.N.G., 1974, Mon.Not.Roy.Astron.So~168, 15
Harris,D.E., 1977, Highlights of Astronomy, VOl.4, part 1.IV
Hooley,T., 1974, Mon.Not.Roy.Astron.Soc. 166, 259
Jaffe,W.J., Perola,G.C.,1976, Astron.Astrophys. 46,275
Jaffe,W.J., Perola,G.C., Valentijn,E.A., 1976,Astron.Astrophys. 49,179
Lea,S.M., 1976, Astrophys.J. 203, 569
Leir,A.A.; 1976, Toronto M.Sc~hesis
McHardy,I.M., 1974, Mon.Not.Roy.Astron.Soc. 169, 527
Miley,G.K., Perola,G.C., Kruit,P.C.van der, Laan,H.van der, 1972,
Nature 237,269
Mills,B.Y., 1960, Austr.J.Phys.13, 550
Mills,B.Y., Hoskins,D.G., 1977,:preprint
Owen,F.N. ,1975, Astrophys.J. 195, 593
Owen,F.N., Rudnick,L., 1976, Astrophys.J., 205, Ll
Riley,J.M., 1975, Thesis
Rogstad, D.H., Ekers,R.D., 1969, Astrophys.J. 157,481
Roland,J., Veron,P., Pauliny-Toth,I.I.K., Preuss,E., Witzel,A., 1976,
Astron.Astrophys. 50, 165
Rudnick,L., Owen,F.N., 1977, Astron.J., 82,1
Sandage ,A. , 1972, Astrophys.J. 178, 1 --
Sandage ,A. , Hardy,E., 1973, Astrophys.J., 183, 743
Slingo,A., 1974, Mon.Not.Roy.Astron.Soc. 168, 307
Tovrnasyan,G.M., Shirbakyan,M.S., 1974, Astrofizica 10, 29
Van den Bergh,S., 1961, Astrophys.J. 134, 970
146 C. LARI AND G. C. PEROLA

DISCUSSION

Abell: Is there a single known case of a tail source that is not ~n

some kind of cluster or group that can be recognized, say, on the sky
surveys? What about V-shaped sources?

Perala: I know of no examples of head-tail radio sources which are

outside clusters or groups of galaxies. There is an example of a V-
shape source (B2 source 0034+25), which is associated with a fairly
isolated galaxy, which, however, (from its redshift) may just be in an
underpopulated area of the Perseus supercluster.

Langair: In Cambridge, Ian McHardy has completed a survey of a statis-

tical sample of 60 Abell clusters which are 4C radio sources. His
conclusions agree substantially with those described by Drs Perola and
Ekers. One correlation which McHardy finds is that the radio trail
sources are almost exclusively associated with Bautz-Morgan class III
clusters. Few of them are associated with Bautz-Morgan class I. This
provides a neat picture in which in the BM III clusters, all the
brightest galaxies have roughly the same absolute magnitudes and hence
they must all be in motion with respect to the dynamical centre of the
clusters.

Ostriker: At a given optical luminosity, is there a significant differ-

ence between the probability that a first ranked galaxy or any other
galaxy will be a radio source? That is, is there a correlation between
rank and radio properties independent of the correlation between optical
luminosity and radio properties?

Tinsley: Further to Dr Ostriker's question, Dr Perola has found that

the probability of a first-ranked galaxy being a radio source increases
with its optical luminosity. Gunn has shown that this leads to an
important bias in the Hubble diagram for qQ' if the galaxies are select-
ed by their radio emission because the opt~cally brightest galaxies are
thereby selected and the value of qo obtained in such a sample is too
large if the selection effect increases with redshift. Perola's result
suggests that samples of galaxies used in the Hubble diagram should not
be obtained from radio catalogues, as it has been done in some studies.

Perala: It seems to me that the correct statement is as follows: the

addition to the Hubble diagram for first ranked cluster galaxies of
objects in very distant clusters detected in radio surveys may lead to
a bias, much in the same way as the Bautz-Morgan class effect, which in
principle at least can be corrected for. A similar statement, however,
is not correct when the Hubble diagram is constructed only with galaxies
selected entirely from radio samples, irrespective of their being members
or not of rich clusters, provided that only objects with P(408 MHz)
> l0 25 WHz- 1 are used, or that the radio limit of the sample is properly
taken into account. The latter statement is explained in Auriemma et
al. (1977) and also, but rather telegraphically, in Fanti and Perola
(1976) at the Cambridge Symposium on radio sources.
RADIO PROPERTIES OF ABELL CLUSTERS 147

Longair: In response to Dr Tinsley's remarks and to amplify Dr Perola's

answer, it should be noted that the correlations between radio and
optical luminosity refer only to low luminosity radio sources. The
Abell cluster radio sources from the B2 Catalogue are at the very low
end of what one normally calls a radio galaxy. If one restricts atten-
tion to the radio sources, which are classical doubles, there is little
or no correlation between radio and optical luminosity for these cD
galaxies.

Komberg: Is there a correlation between the properties of elongated

radio structures of radio galaxies in clusters of galaxies and the
values of velocities of these galaxies, relative to the centre of the
cluster?

PeroZa: At present we do not have sufficient data on the redshifts of

galaxies in the clusters.
RADIO OBSERVATIONS OF CLUSTERS OF GALAXIES: THE TAIL SOURCES

R.D. Ekers
Kapteyn Astronomical Institute
Groningen, The Netherlands

The "head-tail" class of radio galaxies (e.g. Ryle and lJindram, 1968
and Miley et al., 1972) are of particular interest in the context of this
symposium since they provide some of the clearest evidence for a diffuse
medium.

Models to explain the morphology of these radio galaxies (e.g. Jaffe

and Perola 1973, Pacholczyk and Scott 1976, Cowie and McKee 1975) gene-
rally involve some or all of the following three concepts:
i) a dynamic pressure cr PIG VG 2 to bend the tail back away from
the galaxy,
ii) a static pressure cr PIG T IG to confine the back part of the tail,
and I
ii) a Mach number cr TiG to give the desired geometry. The minimum
pressure in the radio em~tting region can be determined rather reliably
from equipartition agreements and the velocity of the galaxy, VG, from
measurements of galaxy redshifts. Hence we can then use the arguments
(i) - (iii) to determine the density, PIG, and temperature, TIG, of the
intergalactic medium.

I do not intend to dwell on details of these models here but to

discuss two new observational results concerning the velocities and
environment of the head-tail radio galaxies.

At first it was thought that radio galaxies with tails had except-
ional high velocities with respect to the intergalactic medium, e.g.
NGC 1265 has a radial velocity component of +2 200 km/sec with respect
to the mean velocity of the Perseus Cluster. However, in a recent analysis
of 14 head-tail radio galaxies Ulrich (preprint) has shown that the dis-
tribution of velocities is in fact consistent with that expected if the
galaxies were selected at random from their clusters. This result has
important implications for the deduced density of the intercluster medium
since by relation i) we need a higher value of PIC to compensate for a
lower average value of VG, and when combining relations ii) and iii) we
need a lower temperature to keep the galaxy supersonic and hence again
a higher value of PIC to still have static confinement.
149

M S. Longair and J. Einasto (eds), The Large Scale Structure of the Universe, 149-151. All Rights Reserved.
Copyright © 1978 by the IAU.
150 R. D. EKERS

It has also been generally assumed that the head-tail phenomenon is

confined to rich clusters of galaxies and that the relatively high inter-
galactic densities required only pertained in these regions of space. If
we take a sample of 21 head-tail sources now known, most of these are in
fact in Abell clusters. However, in many cases these tail sources have
only been found because of the observations of Abell clusters. If we form
an unbiased sample using radio catalogues we then have only 8 objects of
which 2 are in Abell clusters, 3 are in the much poorer Zwicky clusters
and the others are obscured or are in the southern hemisphere, and we
have inadequate information for a quantitative classification.

Hence it may still be the case that the head-tail sources are always
found in some type of clusters but they are certainly not confined to
the rich clusters. On? interesting example of a head-tail source in a
poor cluster is B2 101J+35 (Ekers, Fanti, Lari and Ulrich, preprint).
This is a long tail (700 kpc) in a "medium compact" Zwicky cluster with
a line-of-sight velocity dispersion of 584 km sec-I. Model calculations
suggest a number density of about 10- 3 cm- 3 for the intergalactic medium,
similar to the value found for rich clusters. Since clusters of this type
are 100 times more numerous than the rich Abell clusters it is clear that
their contribution to the density of the Universe can be much greater
([6 '\, 0.2).

Finally, there is another interesting result in the region of

B2 1615+35. This Zwicky cluster, together with the Abell clusters 2197,
2198 and 2162 and another Zwicky cluster 1611.6+3717 all have the same
average redshift and form a clear supercluster complex 30 Mpc in extent.
If we assume that the intergalactic density we find in the region of the
tail source pertains to the entire supercluster then the volume involved
is much greater and such an intergalactic medium in the superclusters
could make a dominant contribution (~ '\, 0.2) to the density of theUni-
verse.

References
Cowie, L.L. and McKee, C.F.: 1975, Astron. Astrophys. 43, 337
Jaffe, H.J. and Perola, G.C.: 1973, Astron. Astrophys.26, 423
Miley, G.K., Perola, G.C., Van der Kruit, P.C., Van der~aan, H.: 1972,
Nature 237, 269
Pacholczyk, A.G. and Scott, J.: 1976, Astrophys. J. 203,313
Ryle, M. and Hindram, N.D.: 1968, Mon. Not. Roy. Astron. Soc. 138,

DISCUSSION

de Vaucouleurs: I refer to the last two lines of your last slide show-
ing both superclusters and groups, each contributing [6 = 0.2. This
seems to be redundant because small groups are the components (the sub-
units) of superclusters.

Ekers: Yes, if all groups are in superclusters.

RADIO OBSERV ATIONS or CLUSTERS or GALAXIES: THE TAIL SOURCES 151

Chernin: The mean velocities of galaxies within a cluster are close to

the sound velocity of the gas which is smoothly distributed and is in
virial equilibrium. So the strongly supersonic motion you refer to is
hardly the case. Perhaps, you mean with respect to low temperature
clouds?

Ekers: I included the supersonic velocity case because this is required

for some tail radio source models. However, especially since we nm.
have evidence that the tail sources have the same velocity distribution
as the rest of the cluster galaxies, I agree that the supersonic case is
less likely. Low temperature clouds do not seem very likely because of
the smooth form of many of the tail sources.
EXTENDED RADIO EMISSION IN CLUSTERS OF GALAXIES:
RECENT WESTERBORK OBSERVATIONS

E.A. Valentijn and H. van der Laan

Sterrewacht,
Leiden, The Netherlands.

Whether galaxy clusters, in addition to a number of radio sources

associated with individual cluster members, have an extended radio
emitting region designated as cluster halo is in an interesting way re-
lated to several questions concerning cluster characteristics:
i) do relativistic electrons leak out of active cluster galaxies and
accumulate; how do they propagate in the cluster;
ii) what is the strength and topology of the intracluster magnetic
fields;
iii) is there an association between radio-synchrotron and either in-
verse Compton or free-free X-ray emission;
iv) what static thermal pressure helps to confine the discrete extended
(head-tail) radio sources in clusters.

The observations are problematic, especially for the nearby clusters

for which the expected angular sizes of the cluster haloes are of the
order of 30' arc:
i) single dish telescopes have the best brightness sensitivity but
blend discrete sources due to lack of resolution;
ii) aperture synthesis telescopes are insensitive to extended emission
due to missing short spacings (for the Westerbork Synthesis Radio
Telescope at 610 MHz 8 % 30');
iii) the galactic foregroun~a~ontinuum and its polarization is difficult
to separate from suspected cluster emission.

We present results for five clusters studied with the WSRT:

I) In the Coma cluster a radio halo (Coma C) has been detected (Jaffe
et al. 1976; Valentijn, 1977). Assuming a Gaussian brightness distribu-
tion for the halo at 610 MHzafluxof 1.2 ±.5 Jy and a F.W.H.M. dia-
meter of 33' was found. Additional single dish WSRT observations proved
the strong polarized signal from the direction of Coma to be galactic
foreground radiation. If the X-ray emission from Coma is at least 75%
thermal, as indicated by Fe-line emission, then the intracluster magne-
tic field must be at least 10- 7 Gauss to keen the inverse Compton X-rays
below the permitted level. Compare this value with H = 4.10- 7 Gauss.
eq
153

M. S. Longair and J. Einasto (eds.), The Large Scale Strncture of the Universe, 153-155. All Rights Reserved.
Copyright© 1978 by the1AU.
154 E. A. V ALENTlJN AND H. V AN DER LAAN

2) In the Perseus cluster Gisler and Miley (1978) found no evidence at

610 MHz for whole cluster emission. NGC 1275 does have its own radio
halo on a 5' scale (see also Miley and Perola, 1977), whose morphology
is similar to, possibly identical with that of the extended X-rays source
centred on NGC 1275 (Gorenstein et al. 1977). According to an inverse
Compton interpretation of the X-rays, the X-ray to radio surface bright-
ness ratio would imply a magnetic energy density ~ 500 x below the radio-
deduced equipartition value, so this X-ray component is most probably
thermal.

3) In the Hercules cluster no extended radio halo has been detected. Two
wide angle radio trails (NGC 6061 and NGC 6034) have been mapped (Valen-
tijn and Perola, 1977; van Breugel and Valentijn, 1978). In order to
confine the expansion of these tails by a de~3e an9 hot intracluster
medium a lower limit on the product pT 7 > 10 (10 K cm- 3 ) is derived.

4) In A2256 there appear four to six radio trails in the 610 and 1415
MHz maps and in addition there is eccentrically located cluster emission
not identifiable with individual galaxies (Bridle and Fomalont, 1976;
Bridle et al. 1978). The galaxies in A2256 represent a normal radio
luminosity function·_ 4 7 -3
Values of pT 7 > 5.10 (10 K em ) are derived from static thermal con-
finement of the tailed radio sources. This set of quantities and espe-
cially the relatively low spectral index a(1415, 610) = 0.7 for the very
extended component implies particle reacceleration over extensive volumes
and may indicate the presence of a strong intracluster magnetic field.

5) AI314 contains several radio trails (Vallee and Wilson, 1976; Wilson
and Vallee, 1977; Vallee, 1977). There is no evidence for a cluster halo.

The limited range of surface brightness sensitivity of the ~~RT

defines the actual range of pT 7 derived from observables from 10 to
10- 3 (10 7 K cm- 3 ). Uncertainty in the electron to nuclear energy ratio
n adds an uncertainty n 4 /7. The density and pressure estimates in indi-
vidual cases are therefore only indicative and uncertain by a factor of
ten.

Breugel, W.J.M. van and Valentijn, E.A.: 1978, in preparation.

Bridle, A.H. and Fomalont, E.B.: 1976, "Astron. and Astrophys." 52, 107.
Bridle, A.H., Fomalont, E.B., Miley, G.K.: 1978, in preparation.
Gisler, G.R. and Miley, G.K.: in preparation.
Jaffe, W.J., Perola, G.C. and Valentijn, E.A.: 1976, "Astron. and Astro-
phys." 49, 179.
Miley, G.K. andPerola, G.C.: 1977, "Astron. and Astrophys." in press.
Valentijn, E.A. and Perola, G.C.: 1977, "Astron. and Astrophys." in press.
Valentijn, E.A., in preparation.
Vallee, J.P. and Wilson, A.S.: 1976, "Nature" 259,451.
Vallee, J.P.: 1977, in press.
Wilson, A.S. and Vallee, J.P.: 1977, "Astron. and Astrophys." 58, 79.
RECENT WESTERBORK OBSERV ATIONS 155

DISCUSSION

Ginzburg: If you observe high radio brightness in the central regions

of a cluster of galaxies, you can estimate, making a number of assump-
tions, what the intensity of the radio halo in the cluster should be.
What is your view on this question?

van der Laan: My views on this question are not related to the observa-
tions which I have presented. It is my view that extended "halo"
sources in clusters may be detected in low frequency radio surveys. The
cosmic ray electron component is expected to suffer severe synchrotron,
inverse Compton and adiabatic losses which conspire to steepen the
spectrum of the halo emission. These sources would not be observable at
the wavelengths at which we have made our observations. There is no
contradiction between the observation of intense radiation from the
central regions and the absence of a cluster halo.
CLUSTERS WITH EXTENDED RADIO EMISSION AT HIGH FREQUENCIES

R. Wielebinski
Max-Planck-Institut fUr Radioastronomie, Bonn, F.R.G.

The existence of 'haloes' in clusters of galaxies was deduced by

Ryle and Windram (1968) for the Perseus cluster and by ~Jillson (1970)
for the Coma cluster at 408 MHz by comparing total flux measured by a
single dish with the sum of fluxes of radio sources found in the field.
A direct measurement of the extended source Coma C was made by Jaffe et
al. (1976) at 610 MHz. The failure to detect the halo of Coma at
higher frequencies is attributed by all authors to a steep spectrum of
this extended component.
The 100-m radio telescope of the MPIfR was used successfu11y tomap
normal galaxies with 3 mK r.m.s. noise at 2.7 GHz and 1.5mK at 4.8GHz.
These noise levels are still slightly higher than the confusion limit
of the telescope and are ideally suited to map low-luminosity extended
regions. To overcome problems of dynamic range limitation a sidelobe
cleaning procedure was developed which allows the recognition of fea-
tures 30 dB below a strong source.
The survey of clusters of galaxies at 2.7 GHz contains some 30
clusters, all suspected to have extended emission on the scale> 5'.
Positive detections of extended emission could so far be reported for
A754, A2142, A2256 and A2319 (Haslam et al. 1977). The extended emis-
sion does not permeate the whole cluster volume, but is usually present
near the dominant galaxies. All these clusters are known to be X-ray
sources. On the other hand, most of the clusters reported to be ex-
tended by Owen (1974) have been resolved into individual radio sources.
Clusters for which our 4~6 beam at 2.7 GHz gives not sufficient angular
resolution are being mapped at 4.8 GHz with a 2~6 beam.

Haslam, C.G.T., Kronberg, P.P., Waldthausen, H., Wielebinski, R.,

Schallwich, D. 1977, Astron. Astrophys. Suppl. (in press)
Jaffe, W.J., Perola, G.C., Valentijn, E.A. 1976. Astron. Astrophys.
49, 179
Ryle-;-M., Windram, M.D. 1968, Monthly Notices Roy. Astron. Soc. 138,1
Willson, M.A.G. 1970, Monthly Notices Roy. Astron. Soc. 151, 1
157

M. S. Longair and J. Einasto (eds.}, The Large Scale Structure of the Universe. 157-159. All Rights Reserved.
Copyright © 1978 by the JAU
158 R. WIELEBINSKI

ABEll1656 tllMA n1 tH

An 11.1 cm map of COMA (Abell 1656). The contours are 0,5, 10, ...
30, 40, ... 210 mJy/beam area. Beamwidth is 4.6 arc min to 3 dB points.
Note 5 mJy/beam area = 10 mK Tb' Note extended emission is seen to
surround the sources 5C4.85 and 5C4.81 which are merged due to the
beam and steep spectrum of the tailed source. The extent of "the halo"
is some 15 arc min, which corresponds to a linear size of ~ 500 kpc
assuming the distance of Coma to be ~ 100 Mpc.
CLUSTERS WITH EXTENDED RADIO EMISSION AT HIGH FREQUENCIES 159

DISCUSSION

Silk: Is there any tendency for X-ray clusters to possess extended

radio emission generally?

Wielebinski: So far all the extended radio sources detected are also
X-ray sources. But the statistics are still poor and we must await
correlations between various parameters, e.g. richness, to see which
correlation is unique.

Tarenghi: Is the bridge between NGC 1265 and NGC 1275 real?

Wielebinski: No. This slide shows a simple addition of several maps.

Side lobes give the effect of a bridge. Using sidelobe removal a new
map of Perseus cluster is now being made. Still the conclusion about
the absence of a large Perseus halo is suggested by the present map.
RADIO HALOES AROUND GALAXIES AND IN CLUSTERS

V. L. Ginzburg
P. N. Lebedev Institute of Physics
USSR Academy of Sciences
Moscow, USSR

The question of whether or not our and other normal galaxies have
some sort of halo - an extended region containing, in particular, cosmic
rays - has been discussed for no less than 25 years. Such a "cosmic ray
halo" (CRR) appears as a radio-halo, although the absence of the latter
is not evidence against the presence of CRR. The point is that the
relativistic electrons responsible for the radio emission from the
radio-halo undergo synchrotron and Compton losses which are practically
absent in the case of the cosmic-ray proton-nuclear component. Possibly
because the discussion concerning the existence of the radio-halo in the
Galaxy has lasted for years it has acquired a particular character. The
latter is clearly reflected in the report by Baldwin (1976) who empha-
sized that: "In this discussion so far I have avoided the use of the
phrase "radio-halo". It arouses antagonism in otherwise placid astrono-
mers and many sought to deny its existence ... " Such a situation
evidently reflects the difficulties that arise in detecting the radio-
halo of our own Galaxy when account is taken of other confusing galactic
sources as well as of the metagalactic background.

I have always considered the existence of some CRR and, probably,

radio-halo to be most probable if not inevitable. This point of view
was based on dynamical considerations as well as on radio-data while the
objections to the existence of a radio-halo, to say nothing of a CRR,
seemed quite unconvincing (Ginzburg and Syrovatskii 1964, Ginzburg 1967,
also literature cited in Baldwin 1976, Ginzburg and Syrovatskii 1964,
Ginzburg 1967, and Ginzburg and Ptuskin 1976). Clear proof based on
observations and convincing interpretations of the data were needed. In
my opinion such proofs have been obtained and to show this is the aim of
the present remark. I mean, first of all, discussion of radio-astronom-
ical data for our Galaxy which gives evidence in favour of the existence
of a radio-halo with characteristic dimension R ~ 10 kpc and with a
rather high luminosity (Bulanov et al. 1976 and Ginzburg & Ptuskin 1976)~
*Note also that in the radio-disk model used the half-thickness of
the radio-disk is assumed in Baldwin (1976) to be equal to 750 pc which
exceeds considerably the value accepted before, to say nothing of the
half-thickness of the gaseous disk.
161
M. S. Longair and J. Einasto (eds), The Large Scale Structure of the Universe, 161-163. All Rights Reserved.
Copyright © 1978 by the IAU.
162 v. L. GINZBURG

More obvious and in a certain sense more convincing are the radio-data
for the "on edge" galaxies NGC 4631 (Ekers and Sancisi 1977) and NGC 891
(Allen et al. 1977; some data about this galaxy were presented already
in Baldwin 1976). Since NGC 891 resembles our Galaxy, the fact that it
has a radio-halo with a characteristic dimension of the order of 10 kpc
(or if we take the distance from the galactic plane at which the radio-
brightness at a wavelength of about 50 cm is half, then with a dimension
of 4-5 kpc) is rather weighty additional proof in favour of the presence
of approximately the same type of radio-halo around our Galaxy. Note
that not a single case is known in which a galaxy with a rather bright
radio-disk component was lacking a radio-halo. If the disk has low
brightness in the radio band, this simply indicates the absence of
sufficiently powerful sources of cosmic rays (or, more precisely, of
their electron component). What may account for such a situation is
another question not connected directly with the halo problem (the most
natural explanation is that in a radio-quiet galaxy supernovae are
anomalously rare, which in turn requires explanation).

The data on cosmic rays and in particular on the existence of the

radioactive isotope lOBe does not in the least contradict the existence
of a galactic CRR; the opposite opinion expressed in the literature
appeared to be due to a misunderstanding (Ginzburg and Ptuskin 1976,
Ginzburg 1977). Moreover, recent data on the amount of lOBe indicates a
rather rapid cosmic-ray outflow from the region of the gaseous disk,
where the main sources are concentrated. Further developments of the
techniques of cosmic-ray i-sotopic analysis as well as the use of gamma-
ray data (Ginzburg 1977) may sup?lement the radio method for studies of
the Galactic CRR in important ways.

Thus, there is now every reason to believe that cosmic rays are not
trapped near the galactic plane but escape from the gaseous disk and
form a CRR which results naturally in the appearance of a radio-halo.
Further observations and the corresponding theoretical analysis of
various relevant problems must be aimed at,in particular a clarification
of the character of CRR transition into intergalactic space. One may
think that this transition, at least quantitatively, is not universal
but depends on the type of the cluster to which a given galaxy belongs.
For sufficiently dense and large clusters it is quite possible that the
CRR and even radio-halo of some galaxies lose their individuality and
cosmic rays fill all the cluster with an increased intensity. Then,
naturally, a cosmic-ray halo (CRR) and, in principle, a radio-halo must
be formed for the whole cluster and not for each galaxy.

REFERENCES
Baldwin, J.E., 1976. "The structure and content of the Galaxy and
galactic gamma-rays", p.206. Proc. Intern. Symposium, Greenbelt,
Maryland, USA.
Ginzburg, V.L. and Syrovatskii, S.L, 1964. "Origin of Cosmic Rays",
Pergamon Press, London and New York.
Ginzburg, V.L., 1967. lAU Symposium No.3l (ed. R. von Woerden),
Academic Press, London.
RADIO HALOES AROUND GALAXIES AND IN CLUSTERS 163

Ginzburg, G.L. and Ptuskin, V.S., 1976. Rev. Mod. Phys., 48, 161.
Bulanov, S.V., Dogel, V.A. and Syrovatskii, S.I., 1976. Astrophys. and
Space Sci., 44, 255.
Ekers, R.D. and S~cisi, R., 1977. Astron. and Astrophys., 54, 973.
Allen, R.J., Baldwin, J.E. and Sancisi, R., 1977. Astr. Astrophys.,
(preprint) •
Ginzburg, V.L., 1977. 15th Intern. Cosmic Ray Conference: Invited and
Reported Lectures, Plovdiv, Bulgaria. (For an extended version of
this report see Uspekhi Fisicheskih Nauk, 1978.)

DISCUSSION

Valentijn: Is there according to your calculations in situ particle

acceleration needed to explain the observed spectral index distribution
of galactic haloes?

Ginzburg: Those computations which I mentioned have been carried out

under the assumption that the sources of relativistic electrons are
within the disk. Electrons will diffuse into the halo and will lose
their energy (in particular, through radio emission). Evidence about
the conditions existing in the halo of the Galaxy and of other normal
galaxies give us no foundation for expecting any effective acceleration
in these regions. Nevertheless, it is quite possible to imagine condi-
tions in which it would be necessary to consider the acceleration of
electrons in haloes or in clusters of galaxies.

Ostriker: In rich spiral clusters, electron cosmic rays will diffuse

from the disks of galaxies into their radio haloes, and from these into
the general intergalactic cluster space thereby producing a cluster
radio halo. Has anyone in your group computed the expected properties
of such cluster haloes?

Ginsburg: No. Nobody in our group has so far done it. However these
computations need to be done and I hope that they will be done.
Certainly it will be necessary to specify a great number of parameters
(characteristics such as the halo of clusters, sizes of haloes, radio
luminosity, spectral index, etc., which obviously will depend on the
type of the cluster, the magnetic field between galaxies, etc.).
X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES

J. L. Culhane
University College London
Mullard Space Science Laboratory
Holmbury St. Mary
Dorking, Surrey, UK.

X-ray astronomy has, in the past year, seen the publication of the
second Ariel (2A) and fourth Uhuru (4U) catalogues of X-ray sources. A
number of new X-ray cluster identifications and the confirmation of sev-
eral others has resulted. In this review I will briefly summarise the
situation regarding identifications and, for the 2A clusters, discuss
the luminosity function and the possible relationships between a number
of cluster X-ray and optical properties. Superclusters have been tenta-
tively proposed as a class of X-ray sources and I will comment briefly
on recent observations of these objects. Cluster structure has been
studied by the Copernicus and SAS-3 spacecraft and by a number of rocket
observations with imaging X-ray telescopes undertaken by the Harvard
Centre for Astrophysics. I will review the current situation regarding
structural measurements. Finally I will discuss the present status of
Iron line observations at 6.7 keV in cluster spectra and the estimates
of Fe abundance that result from these data.

X-RAY CLUSTER IDENTIFICATIONS AND CORRELATIONS

Although as many as 62 clusters have been proposed as X-ray sources

by a variety of workers (see Culhane (1977», the number of reasonably
secure identifications as of July 1977 is probably between 32 and 40
(Culhane (1977) Jones and Forman (1977), McHardy et al. (1977». The
2A* cluster sample of 38 objects includes 25 Abell clusters and from the
well known 2A sky coverage, McHardy et al. have deduced the differential
luminosity function shown in Figure 1 using the maximum volume method.
The high luminosity upper limit corresponds to the absence of sources of
power greater than 3.10 38 watts at a flux above 1 Ariel count s-l (~3
Uhuru counts s-l) over 90% of the sky. The total density of X-ray
clusters approaches that of all Abell clusters at Lx ~ 10 36 watts and
thus all clusters should emit X-rays at luminosities above this value.
Luminosity functions for richness class 0, 1 and 2 are shown in Figure 2.
*Compiled from the 2A catalogue of Cooke et al (1977). Observations
carried out with the Leicester University Sky Survey instrument on ArielV.
165

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 165-177. All Rights Reserved.
Copyright © 1978 by the fA U.
166 J. L. CULHANE

o
-6 APPROX. DENSITY 2
- - - - - OF ALL ABEU CLUSTERS
i
NORMALISED

L
N..JvIBER 3
DENSITY 5
-0·5
-7

No OF SOURCES OF
Lx IN BINS FROM
Lx 10 -0,25 TO Lx 10 025
2
-10
-6
3
R =1
2
R ~O
-1·5
-9

36·75 37'75 38'75

36·25 3725 3825

po(o) 04.9 10-7 "PC- 3
Loo Lx (WATTS, 2 - 10 KeV)
po(1) = 3.8 10-7 MPC- 3
po(2)o 1'1 10-7 "PC- 3

Figure 1. The differential Figure 2. Luminosity function for

luminosity function for 2A X-ray richness class 0, 1 and 2 Abell
sources associated with Abell clusters normalised to the space
clusters (McHardy et al. (1977)). density of the appropriate rich-
The number of sources in each ness class (McHardy et al. (1977)).
luminosity interval is indicated.

The function for each class has been normalised to the total space den-
sity of clusters of that richness. It is clear that Lx is an increas-
ing function of richness with each richness class increasing the prob-
ability of finding an X-ray source of given luminosity by a factor three.

From the overall luminosity function, the contribution of clusters

to the diffuse X-ray background can be set at 14% for the 2 - 10 keV
range. X-ray emitting Seyfert galaxies could be responsible for a further
6% and hence these two source classes could account for 20% of the flux
assuming no source evolution.

For the 2A Abell clusters, McHardy et al. have computed the prob-
ability of a random coincidence between a 2A X-ray source and an Abell
cluster as less than 0.7%. Although the southern clusters associated
with 2A sources do not form a complete sample, it is clear that no gross
inconsistencies exist between the two hemispheres.
X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES 167

For those 2A clusters which have measured values of velocity dis-

persion (ov), Lx is plotted against 0v in Figure 3. The dashed line
indicates the Lx ~ ov 4 relation proposed by Solinger and Tucker (1972)
which is clearly not a good fit to the data although there is an indic-
ation that clusters with large Lx will also tend to have large 0v. The
relation between kTx ": and ffa; is shown in figure 4 (Mitchell, Ives and
Culhane (1977». Although the sample of clusters is smalle~, the data
indicate that kT x ~ ov 2 which is consistent with a hot gas origin for
the X-rays.

2319 <,()!
,
1-1029

38
75<'-
,)426

I- 2256
1656 - 1
1
154 1
19<00
t ~
1
1

37-5 LOGlOLx
2147/2152/
2199:,
168 I -+
'" 1
1
-576
1
37
SOLl~GER ______ J
TUCKER 1 .C~rus
1
1
1 -1367
194 1
36-5
J,f21 vor
I .
1
LJ.~I '1060

lOG 10 (RADIAL Ov)-)o

1+-----~----I
2-5 3-0 3-5 1000

Figure 3. X-ray luminosity (Lx) Figure 4. Temperature (Tx)

plotted against cluster velocity against 0v for 8 clusters (Mit-
dispersion (ov)' The original chell, Ives and Culhane (1977».
Solinger and Tucker (1972) rel-
ationship is also shown (McHardy
et al. (1977).

It has been apparent for some time that clusters with high Lx freq-
uently contain a dominant central cD galaxy. Since this type of morphol-
ogy is well described by the Bautz-Morgan class I, McHardy et al. have
*kT x values are from X-ray spectral observations with the University
College London proportional counter spectrometer on Ariel V.
168 J. L. CULHANE

plotted the distribution of both X-ray and general clusters with Bautz-
Morgan class. From their results (Figure 5), it is clear that the prob-
ability of a cluster producing an X-ray source of Lx ~ 10 37 watts is
three times greater for Bautz-Morgan class I clusters than for clusters
in general. This conclusion is also supported by the work of Bahcall
(1977a) who has, in addition shown that the relation between Lx and
central galaxy density (Figure 6) is consistent with a hot gas origin
for the X-ray emission.

e- X-RAYCW5TERS
Lx ). 10 37 WATTS

.• .. . ...
6-

~
0
::!
::; ~~~i~
0
4- ~e ON
2-
,.
00-
8M CLASS
40- 40_

I
GENERAL CLUS TERS

30- ,,- ,0 .0-

",,""J(
20- ~ ~ ..! ........... t': • 0S 3
20-

10
---_ ... ----_ ... --- o 0> 3 10-
10-
~~4'·--~~~~~IO~4T4--~~~-U~,~04UO--~~-L~~
][ ][-1II III: Lx( 2·10 KeVl erljjs 5- 1
BM CLASS

Figure 5. Bautz-Morgan distrib- Figure 6. Central galaxy density

utions for X-ray clusters and for plotted against Lx for the clusters
clusters in general. All clusters indicated. Relationships between
of distance class ~ 4 and Lx ~ Lx and No are also shown (Bahcall
10 37 watts. (1977a)).

A relation between spiral galaxy content and Lx is proposed by

Bahcall (1976b) and McHardy et al. Using the 2A cluster sample for Lx
values and spiral content results from Oemler (1976) and Melnick and
Sargent (1977), both groups have demonstrated that clusters with the high-
est values of Lx have the lowest percentages of spiral galaxies. The
correlations observed are consistent with thermal emission from a hot
intracluster gas which strips the spiral galaxies by ram pressure.

The 2A clusters discussed above were identified as a result of scans

of the detector fields of view around the sky and the observation of
X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES 169

significant departures from background counting rates when detectable

X-ray sources were in view. Ricketts et al. (1977) have searched the
Ariel V survey data for clusters using a point summation technique (PST)
in which the position of known clusters are used as arguments of entry to
the Ariel V data base and all data referring to individual source, posi tions
are summed and checked for significance. A total of 360 cluster posit-
ions were searched in this way. In addition to detecting the cluster
sources mentioned above, a further 13 Abell clusters were observed to
have significant X-ray emission while upper limits were assigned to 22
further clusters. The list of positive detections includes Abell 2218,
a rich distance class 6 cluster for which Gull and Northover (1976) have
measured a drop in microwave background temperature due to scattering by
hot intracluster gas (Sunyaev and Zel'dovich (1972)).

THE RELATION BETWEEN X-RAY SOURCES AND SUPERCLUSTERS

Following the circulation of the 4U catalogue, Murray et al. (1977)

have suggested the possibility of an association between unidentified
high latitude X-ray sources and superclusters. Using a definition based
in part on Abell's (1961) discussion of second rank clusters, they find
12 such objects within the area of sky covered by the 4U catalogue. Five
X-ray sources are candidates for association with the superclusters and
three of these have their error box centres within less than 1.5 0 of a
supercluster centre. Murray et al. point out that with an expected number
of accidental coincidences of less than 0.17, the probability of observing
three coincidences is less than 0.003 «3a). However the proposed ident-
ifications are based purely on the association since the sources are too
weak to permit measurement of their X-ray extent.

The five supercluster locations have been the subject of an Ariel V

PST search by Ricketts et al. The results of this search are given in
table 1. Only one of the five candidates is detected at the 4U level
(4U1203-06). Upper limits are set to the X-ray emission fromtheiremaining
four Uhuru sources that are between two and four times less than the 4U
fl ux values.
Table 1 - X-ray supercluster candidates
4U source 4U intensity 2A intensity or 3a
(4U CPS) upper limit (2A CPS)
0134-11'" 2.7 + 0.6 :;::0.43 (1. 3U)t
0443-09 1.8 + 0.3 :;::0.30 (0.9U)t
1203-06 2.2 + 0.4 0.6 (4.9a) ( 1.8U)t
-
1456+22 3.0 + 0.5 :;::0.32 (0.7U)t
2259+16 3.0 + 0.6 :;::0.47 (l.5U)t

* GSFC OSO-8 result gives a 99% confidence upper limit of ~ 0.4 Uhuru
counts s-l.
t 1 Ariel count s-l ~3 Uhuru counts s-l.
170 J. L. CULHANE

In addition, OSO-8 observations (Pravdo et al. (1977» place an upper

limit on the X-ray flux from 4U0134-11 that is a factor 6.5 below the 4U
value. Thus the association of X-ray emission with superclusters rather
than with individual clusters which may be members of a supercluster
is perhaps premature.

THE X-RAY STRUCTURE OF CLUSTERS

Uhuru observations have explicitly demonstrated the extent of cluster

X-ray sources in six cases (Kellogg and Murray (1974». A variety of
observations of structure have been made in the case of the brighter
cluster and a number of these have been summarised elsewhere (Culhane
(1977, 1978». Unfortunately the X-ray structural data available do not
yet discriminate between the various models.

In the case of the Perseus cluster, it is clear that up to 20% of

the emission is associated with the active galaxy NGC1275 while the
remainder of the flux originates in a diffuse source of about 10 in
angular extent (Wolff et al. (1976), Gorenstein et al. (1976». The
extended emission may be explained by an isothermal gas sphere or by
various adiabatic models. The origin of the X-ray emission from around
NGC1275 is not yet clear but the similarity between X-ray and radio
(Miley and Perola (1975» brightness distribution is striking.

The structure of the Virgo cluster has been studied by Gorenstein

et al. (1977) whose 0.5 - 1.5 keV map of this source is shown in Figure 7.
The diffuse emission is centred on M87 and not on the optical centre of
the cluster. The measured surface brightness distribution and two poss-
ible models are illustrated in Figure 8. The solid line results from an
isothermal gas sphere and requires a source associated with M87 to contrib-
ute about 10% of the total cluster flux. The dotted line is obtained
from a power law (in radius) surface brightness distribution and could
result from gas infall onto M87 where accretion is regulated by cooling
at the centre (Cowie and Binney, (1977». The observed increase of diffuse
source size with energy also favours this model.

The other clusters are less well studied at present. Mitchell et

al. (1975) have observed the central 10' of the Centaurus cluster using
the Copernicus X-ray reflectors and find that the flux (0.5 - 1.5 keV)
is four times larger than is predicted by an isothermal gas sphere model.
This may indicate emission from an active galaxy (NGC4696) or an accretion
process dominated by the central galaxy. The effect of a central galaxy
on cluster emission has received further support from the recent work of
Schnopper et al. (1978) who find a point like source coincident with a cD
galaxy at the centre of Abell 478. The X-ray luminosity of this source
is approximately 30% of that of the cluster as a whole.
X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES 171

01.

012

006

o OOOJ;------;-5--;,O~~15----,2;;;1o
: 1-3 13-40 CORE RADIUS (cremin)
u
::: 4-12 .41-52 ~ 10- 1
."7
'"c
.. .::: ... . . . "1'
... ...
~
,

..·.'. •.[.• "il1tIJ'liiE•E' • i

o
u

. :::: :;::lj::;:::!:::iW:=:!::::.
N

. .

J
VIRGO CLUSTER
I 05-15keV

~'_',"m_m_ _ _ _:::_::_:;:_:_ _
10-''--_----'-_ _--'--_ _.1...._ _' - - _ - '
o 5 10 15 20 25
RADIAL DISTANCE (aremini

Figure 7. An X-ray map (0.5 - Figure S. Relative surface bright-

1.5 keY) of the Virgo cluster ness deduced from the map in figure
(Gorenstein et al (1977». The 7 plotted against radial distance.
bright central region is coincid- The solid line represents an iso-
ent with MS7. thermal gas sphere. the dotted line
is a power law in radial distance.

THE DETECTION OF IRON LINE EMISSION IN CLUSTER X-RAY SPECTRA

Studies of X-ray spectra over a limited energy range (e.g. 2 - 10

keY) can not discriminate between thermal and non thermal emission pro-
cesses. Although some evidence in favour of thermal emission has been
obtained from observations over more extended energy intervals (Ricker
et al. (1976). Davidsen et al. (1975). Malina et al. (1978» the detect-
ion of an emission feature due to highly ionised iron in the Ariel V
spectrum of the Perseus cluster by Mitchell et al. (1976) (Figure 9) has
established the presence of hot (T ~ 10SK) gas in clusters of galaxies.
The Perseus observation was confirmed by Serlemitsos et al (1977) who
also detected similar features in the spectra of the Virgo and Coma
clusters (Figure 10). The Coma observation has been confirmed by Ariel V
with increased significance (Culhane 1975) while Mitchell and Culhane
(1977) have detected an iron feature in the spectrum of the Centaurus
cluster.
172 J. L. CULHANE

PERSEUS CLUSTER
0001
+HIGH GAIN
o LOW GAIN

PERSEUS

~ 0001

~
Ne 0

.
~, ~

., ... u '"z>-
:>
10 0

e
.
1
u

> 00005 COMA

'"

0 6 8 10 12 16
ENERGY (keV)

Figure 9. The X-ray spectrum of Figure 10. Fe emission features

the Perseus cluster with the Fe for the Perseus, Virgo and Coma
emission feature indicated {Mitchell clusters {8erlemitsos et al.
eta1. (1976)). (1977))

The observations made by Ariel V and 080-8 are summarised in table 2.

It is apparent that in cases where both instruments have observed the same
cluster, there is good agreement between the measured fluxes although the
uncertainties are necessarily', large, due to the,low fluxes and the, relatively
poor energy resolution of the detectors. There is no evidence for intrin-
sic broadening in the lines. Furthermore all the observations indicate

Table 2 - Cluster iron line fluxes

Cluster Feature Flux Feature equivalent width (eV)
(10- 3 photons cm- 2 s-l)
Ariel V 080-8 Ariel V 080-8
Perseus 3.4+0.4 4.4+0.8 360+50 490+90
- - - -
Coma 0.7+0.1 1.2+0.7 400+60 280+170
- - - -
Virgo - 2.1+0.6
- - 850+250
-
Centaurus 1.0+0.2 - 890+170 -
X-RAY OBSERV ATIONS OF CLUSTERS OF GALAXIES 173

that peak feature emission occurs at around 6.8 keV with an uncertainty
of 0.2 keV. Since the underlying continuous spectra indicate temperatures
in the range 30 - 100. 106 K, the emission features must be due mainly
to transitions in Fe XXV and Fe XXVI. However a large number of lines
from these ions are blended into the observed emission features.

In order to estimate the iron abundance in the gas, it is necessary

to calculate a feature to continuum ratio as a function of temperature
and compare it with the observed values. This may be done by assuming
that the hot intra cluster gas is a coronal plasma in ionisation equil-
ibrium. Transitions contributing to the feature include the Fe XXV and
XXVI resonance lines (k), the Fe XXV 3p and 3S transitions (l,F) and a
variety of satellites of the resonance lines which arise due todielectron-
ic recombination and inner shell excitation (Bhalla et al. (1975». The
total feature flux is shown plotted against temperature in Figure 11
together with the percentage contributions of the transitions mentioned
above. The ion balance calculations used are those of Jacobs et al.(1977)
which are currently the best available. The calculated feature to
100·0'r--,---,--,--,--,--,--,--,----,
'''7Satelllte lines

"-

".
.---
"-

/
/,y
,,--
/-
;..--
--
----=-----~

/Jordan/
~(1970)/

/ L-summ.,s (1974)

! I
+-- F.XXYI(R)
i
a Total Fe ___
c
a feature flux

.
;;

10-52
1·0
~
if
,~

.,
~
'"

'"
~ 10- 53
50 60 70 eo 90 100 16~~0-~2~0-~30~~4~0-~5~0-~6~0-~7~0-~e~0-~~-~loo
T ell; 106 ) OK T (x 10') 'K

Figure 11. Calculated Fe Figure 12. Calculated line

feature flux plotted against to continuum ratio plotted
temperature. The fractional against temperature for three
contributions of groups of tran- different ionisation balance
sitions are also shown (R-resonance, calculations.
I-intersystem(3p), F-forbidden (3S),
dielectronics and inner shell satellites).
174 J. L. CULHANE

continuum ratio is shown in figure 12 for three different ion balance

assumptions. Although the latter lead to gross disagreement below 30.10 6 K,
little difference is apparent in the temperature range appropriate to
the clvster spectra.

The iron abundances deduced are given in table 3. Although there

are significant differences between the values derived by the Ariel V
and OSO-8 groups, data are only available for both the Perseus and Coma
clusters and the errors are large.

Table 3 - Iron abundance

NFe
Cluster T(106 K) Nre- (Cosmic)'"
Ariel V OSO-8
Perseus 66 0.23+0.06 0.45+0.08
- -
Coma 70 0.13+0.04 0.38+0.23
- -
Virgo 43 - O.46±O.20
Centaurus 32 O.40±O.12 -

* Assumed Cosmic Iron Abundance NFe

NH
=

CONCLUSIONS

In the past year considerable progress has been made in our under-
standing of the extended X-ray sources in clusters of galaxies. Between
30 and 40 clusters are known X-ray sources. The cluster luminosity
function has been determined and indicates that essentially all clusters
of galaxies will be X-ray sources. There is a good correlation between
Lx and cluster richness, central galaxy density and the percentage of
spiral galaxies present. The latter point is strongly suggestive of the
presence of hot gas whose role is also indicated by the relation between
kT x and avo Searches for further X-ray identifications are in progress
but the proposed association between X-ray emission and supercluster
gas may be premature.

Studies of cluster structure are at present confined to the brighter

sources. Although these data are as yet unable to throw much light on
the mechanism of heating the intracluster gas, there are indications
that massive central galaxies playa significant role in the process.
Both isothermal and a variety of polytropic gas sphere models are still
permitted by the available data.

The discovery of line emission in the cluster X-ray spectra demon-

X-RA Y OBSERVATIONS OF CLUSTERS OF GALAXIES 175

strates the presence of hot gas and indicates a a model dependant iron
abundance of between 10% and 40% of the cosmic value.

The subject of cluster X-ray emission is now firmly established as

a major source of data on both the evolution of galaxies and the large
scale structure of the universe.

REFERENCES

Abell, G. : 1961, Astrophys. J., 66, 607.

Bahcall, N.: 1977a, Princeton University preprint.
Bahcall, N.: 1977b, Princeton University preprint.
Bhalla, C.P., Gabriel, A.H., and Presnyakov, L.P.: 1975, Mon. Not. R.
astr. Soc., 172, 359.
Cooke, B.A., Ricketts, M.J., Maccacaro, T., Pye, J.P., Elvis, M., Watson,
M.G., Griffiths, R.E., Pounds, K.A., McHardy, I., Maccagni, P.,
Seward, F.D., Page, C.G., Turner, M.J.L.: 1977, Mon. Not. R. astr.
Soc., in press.
Cowie, L.L. and Binney, J.: 1977, Astrophys. J., 215, 723.
Culhane, J.L.: 1977, Highlights in Astronomy, Proc. IAU XVI General
Assembly, Grenoble, 1976.
Culhane, J.L.: 1978, To be published in Q.J. Roy. Astr. Soc.
Davidsen, A., Bowyer, C.S., Lampton, M., Cruddace, R.: 1975, Astrophys.
J., 198, l.
Gorenstein:-p., Fabricant, D., Topka, K., Tucker, W.H., Harnden, F.R.:
1977, Astrophys. J., 216, L95.
Gorenstein, P. and Harnden:-F.R.: 1976, Bull. Am. Astr. Soc., 8, 445.
Gull, S.F. and Northover, K.J.: 1976, Nature, 263,572. -
Jacobs, V.L., Davis, J., Kepple, P.C., Blaha, M::-1977, Astrophys. J.,
211, 605.
Jones:-C. and Forman, W.: 1977, Centre for Astrophysics preprint.
Kellogg, E.M. and Murray, S.: 1974, Astrophys. J., 193, L57.
Malina, R., Lea, Susan, M., Lampton, M., Bowyer, C.S::- 1978, Astrophys.
J., in press.
McHardy, I., et al.: 1977, in preparation.
Melnick, J. and Sargent, W.L.: 1977, Astrophys. J., in press.
Miley, G.K., and Perola, G.C.: 1975, Astron. Astrophys., 45, 223.
Mitchell, R.J., Charles, P.A., Culhane, J.L., Davison, P.~N., Fabian,
A.C.: 1975, Astrophys. J., 200, L5.
Mitchell, R.J., and Culhane, J.L.-:--1977, Mon. Not. R. astr. Soc.
Mitchell, R.J., Culhane, J.L., Davison, P.J.N., and Ives, J.C.: 1976,
Mon. Not. R. astr. Soc., 175, 29P.
Mitchell, R.J., Ives, J.C. and Culhane, J.L.: 1977, Mon. Not. R. astr.
Soc., in press.
Murray, S.S., Forman, W., Jones, C., Giacconi, R.: 1977, Centre for
Astrophysics preprint.
Oemler, A.: 1974, Astrophys. J., 194, 1.
Pravda, S.H., Mushotzky, R.F., Becker, R.H., Boldt, E.A., Holt, S.S.,
Serlemitsos, P.J., Swank, J.H.: 1977, Nature, in press.
176 j, L CULHANE

Ricker, G.R., Scheepmaker, A., Mrecher, K., Ryckman, S.G., Ballintine,

J.E., Boty, J.P., Downey, P.M., 1ewin, W.H.G.: 1976, Astrophys. J.,
205, 165.
Ricketts, M.J., et al.: 1977, in preparation.
Schnopper, H.W., Delvaille, J.P., Epstein, A., Helmken, H., Harris,
D.E., Strom, R.G., Clark, G.W., Jernigan, J.G.: 1978, Astrophys.
J., in press.
Serlemitsos, P.J., Smith, B.W., Boldt, E.A., Holt, S.S., and Swank, J.H.:
1977, Astrophys. J., 211, 163.
Solinger, A. and Tucker, W.H.: 1972, Astrophys. J., 175, 1107.
Sunyaev, R.A., and Zel'dovich, Y.B.: 1972, Com. Astrophys. Space Sci.,
4, 173.
Wolff: R.S., Mitchell, R.J., Charles, P.A., Culhane, J.1.: 1976,
Astrophys, J., 208, 1.

DISCUSSION

Chinearini: How many random associations would you expect between

superclusters and X-ray sources? How was the Supercluster sample
selected?

Culhane: I would refer you to the preprint of Murray and co-workers for
a detailed discussion but, so far as I can remember, the probability of
the three X-ray sources accidentally coinciding with superclusters is
less than 0.003. The selected superclusters were broadly Abell class II
clusters with a number of additional criteria. However, you will recall
that Ariel V and OSO-8 place 30 upper limits on the flux from four of
the five candidate Uhuru X-ray sources that are between two and six
times lower than the reported Uhuru values.

Sunyaev: Can you distinguish experimentally the Ka line of weakly

ionized iron from the La line of hydrogen or helium-like iron ions?
This is very important because we are all very interested in the
chemical composition of the intergalactic gas in clusters. If there is
a lot of iron, a large fraction of the gas must be secondary and not
primaeval. In this case explosions of supernovae are an extremely
important source of IGG. More speculative, but still possible, situa-
tions may exist in which active galactic nuclei give a significant part
(but not all) of the radiation from clusters. The Ka line may be formed
in the cold gas in the vicinity of nuclei. In this case hot gas
(kT ~ 10 8 K) inside the cluster may be of a primaeval chemical composi-
tion.

Culhane: Yes, we can distinguish between these lines at 6.4 and 6.8 keV.

Audouze: In relation to the last part of your talk, I would like to

call attention to the work of one of my colleagues - Laurent Vigroux
(1977, Astr. and Astroph. Letters). He used the iron abundance obtained
by the Goddard group and assumed that the iron distribution is homogen-
X-RAY OBSERVATIONS OF CLUSTERS OF GALAXIES 177

eous in clusters such as Coma and Perseus. He concluded from his model
that the production of heavy elements should occur at the beginning of
the evolution of the galaxies in the cluster (i.e. prompt initial
enrichment).

Culhane: A problem with this model arises from the present lack of
spatial information on the distribution of Fe emission in the cluster to
which I drew attention during my talk. But in any case the abundance
estimates refer only to the X-ray emitting gas which is ~ 8 x 10 12 M0 •
The X-ray abundance estimates then tell us that the detected mass of Fe
is < 5 x 10 9 Me' which is much less than the value of ~ lOll Me deduced
by Vigroux who assumes that the Fe/H ratio is constant throughuut the
cluster gas.

Chernin: Is there any room for clumpiness to reduce the Fe abundance?

Culhane: At the present time the only condensations in the X-ray

emission are seen in the case of NGC 1275 in Perseus « 20% of total
flux) and M87 in Virgo « 10% of total flux) but the available X-ray
maps of these and other clusters have a low resolution, so data of a
better angular resolution are required.

Ostriker: I cannot resist remarking that several years ago anyone who
built models of galactic evolution (such as Thuan and myself),and
estimated the ejecta from stars which are an extrapolation of those
which we see now, would have found exactly the amount of intracluster
iron which is now observed.

Tinsley: Models for the early evolution of galaxies allow a large

amount of chemically enriched gas to be ejected. In a 1975 paper in
P.A.S.P., Larson and Dinerstein predicted that clusters of galaxies
should contain a significant mass of gas with approximately the solar
metal abundance, as a result of ejection from young galaxies, and this
gas has subsequently been observed.
GAS IN GALAXY CLUSTERS

Joseph Silk
Department of Astronomy, University of California
Berkeley, California 94720

Gaseous matter almost certainly cannot account for a significant

amount of the binding mass in the cores of rich clusters. Implications
of the variety of upper limits are well known (Tarter and Silk 1974),
and will not be described here. However intracluster gas provides
important clues to the evolution of galaxy clusters, and the present
review is devoted to elucidating its role.

I. X-RAY OBSERVATIONS

X-ray data have provided the principal source of evidence for the
existence of a hot intracluster gas. For some time, spectral fits to
the continuum in the energy range 0.2 - 30 keV have marginally favored
thermal as opposed to non-thermal (or power-law) intrinsic source
spectra. Discovery of an emission feature at 7 keV, identified with
Fe XXV and Fe XXVI line emission, has confirmed the presence of hot
intracluster gas in four cluster sources (Mitchell et aZ. 1976;
Serlemitsos et aZ. 1976; Mitchell and Culhane 1977). An unexpected find-
ing is that the iron abundance is similar in at least three of the
sources, and is generally within a factor of seven of the solar iron
abundance. This latter conclusion appears to be moderately independent
of various models for the gas distribution (Bahcall and Sarazin 1977).
The inferred masses of gas are comparable to those within the luminous
regions of the cluster galaxies, and only amount to a fraction (typically
10% in a hydrostatic model) of the cluster dynamical mass.

Recent observations with high spatial resolution (Gorenstein et aZ.

1977) indicate that the extended Virgo source is centered on M87 and is
small (core radius approximately 100 kpc), and may therefore be confined
by the gravitational potential well of M87 itself (Mathews 1977). Sev-
eral other clusters are known to be extended sources of X-ray emission,
typical core-radii being several hundred kpc (Kellogg and Murray 1974),
usually larger than, or comparable to, the characteristic scale of the
galaxy distribution. Gas temperatures based on X-ray spectral data are
now available for some fifteen cluster sources (Jones and Forman 1977);
179

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 179-188. All RllJhts Reserved.
Copyright © 1978 by the IAU,
180 JOSEPH SILK

there is a weak correlation with X-ray luminosity, the less luminous

source tending to have lower temperatures.

II. RADIO OBSERVATIONS

The morphology of tailed radio galaxies (TRG's) has provided dra-

matic evidence for the presence of intracluster gas. Most TRG's are as-
sociated with clusters; the few that are not (see below) may be associ-
ated with superclusters. There seems little doubt that the long (often
curved and typically several hundred kpc) tailed sources are produced
and maintained by interaction with the intracluster gas (Jaffe and Perola
1973; JP). Rudnick and Owen (1974) find that two-thirds of cluster radio
sources show significant amounts of distortion or misalignment, as com-
pared with the classical double-lobe structure. Common to the various
plasmoid ejection models is the confinement of the head of the TRG by
ram pressure:

(1)

where B is the internal magnetic field, Pcr denotes the cosmic ray density
in the tail, p is the intracluster gas density. and v is the plasmoid
velocity relative to the intracluster gas. Use of the equipartition
field in the ex~anded tail region and conservation of flux to infer B(JP)
yields p = 10- 2 g cm- 3 for 3C 129. This does not differ significantly
if the ejection is subsonic (Cowie and McKee 1975), but a substantially
lower value of p was inferred in a model where reacceleration of electrons
occurred in the tail (Pacholczyk and Scott 1976). In fact, this latter
conclusion is erroneous, as cosmic ray pre'ssure was neglected in applying
(1). If turbulence - driven reacceleration of electrons is important and
results in fields below the equipartition value in the tail, the left
hand side of (1) is actually increased, and p is larger than in the JP
model. Ram pressure confinement generally leads to minimum densities
p ~ 10- 27 g cm- 3 • This density is similar to the mean gas density found
in the case of X-ray clusters. The thermal pressure of the intracluster
gas is rather more model-dependent. If local acceleration occurs, no
constraint can be set on the gas temperature T. On the other hand, JP
argue from the constancy of the tail width that the thermal pressure
must balance the internal pressure in the tail, thereby enabling minimum
pressures of ~lO-ll dynes cm- 2 to be inferred for the intracluster gas.
Tail curvature may reflect the galaxy trajectory, systematic gas flows
(such as cluster winds or infall), or buoyancy in the cluster potential
well.

Are there any TRG's outside rich clusters? Jaffe (1976) cites
several examples. However all are in or very near well-known super-
clusters, apart from "a clear cut example of a tailed source in a sparse
group of galaxies, that associated with NGC 7385". In fact, Murray et
al. (1977) find a nearby X-ray source (4U2259 + 16) in a region of "un-
usually high cluster density". The known TRG's outside rich clusters
might be in regions, such as superclusters, where diffuse hot gas is
GAS IN GALAXY CLUSTERS 181

present, and thermal confinement of the tailed source is therefore pos-

sible. The equipartition pressure in the case of NGC 7385 is a factor
~100 lower than found for typical TRG's in rich clusters, amounting to
~10-13 dynes cm- 2 (Schilizzi and Ekers 1975). This is consistent with
confinement by an ambient gas of density ~10-5 cm- 3 at T ~ 10 8 K (as
would be implied were the X-ray source identification correct).
The correlation between decametric spectral index and presence of a
radio source in an Abell cluster is well known, steeper spectral indices
being found in clusters. There are indications that the 26 MHz spectral
index steepens with increasing Lx (Erickson et al. 1977). The low fre-
quency data also indicate a correlation between the 26 MHz luminosity
L26 and Lx: for 14 cluster sources studied by Erickson et al., one finds
d log Lx/d log L26 ~ 0.5 ± 0.3 (20). What perhaps is not as generally
appreciated is that the higher radio frequency data show an equally good
correlation. Utilizing the 1400 MHz survey of Owen (1975), one finds
that d log Lx/d log L1400 ~ 0.7 ± 0.3 (20) for 24 cluster sources.

III. OPTICAL DATA

The apparent correlation between X-ray luminosity and cluster mor-

phological type may largely be due to a correlation between Lx and cluster
richness (R) (Jones and Forman 1977). The data on some 36 X-ray cluster
sources can be fitted by d log Lx/dR ~ 0.23 ± 0.13 (20). Bahcall (1977a)
finds a similar correlation between Lx and central counts of galaxies.

Melnick and Sargent (1977) find that the radial distribution of

spirals often differs from that of ellipticals and SO's in X-ray
clusters. The ratio of spiral to disk systems decreases, and the ellip-
tical fraction increases, with increasing Lx. The spiral fraction also
decreases with increasing velocity dispersion ~v. In a larger sample of
14 clusters, Bahcall (1977b) confirms that the fraction of spiral
galaxies decreases with Lx.

The weak correlation between Lx and ~v has persisted (but not

strengthened) as more data has become available (Faber and Dressler 1976).
One presently has, for 13 X-ray clusters with measured ~v, d log ~v/d
log Lx ~ 0.13 ± 0.08 (20).

Another form of optical evidence for diffuse intracluster matter

is due to intergalactic bridges and filaments. The diffuse light in the
Coma cluster is the best studied example of this sort, but appears to
largely consist of blue stars (Thuan and Kormendy 1977, and references
cited therein). The interpretation of quasar absorption lines still
arouses sufficient controversy that they will not be cited here as pro-
viding any unambiguous evidence for the existence of intracluster gas,
theoretical arguments notwithstanding (e.g. Silk 1970).
182 JOSEPH SILK

IV. IMPLICATIONS

The near-solar abundances of iron in at least four X-ray cluster

sources appears to have eliminated pure infall models of cluster evolu-
tion; detailed spatial mapping is still needed to confirm this conclu-
sion. One aspect of these early models has survived: infall through the
gravitational potential well of the galaxy cluster remains the best
candidate for supplying the thermal energy of the gas (Silk 1973). The
near coincidence between cooling and Hubble times in the cluster cores
can be understood if cooling regulates the amount of gas present, the
gas being accreted onto the most massive, slowly moving central galaxies
(and possibly initiating the formation of these systems) (Silk 1976;
Cowie and Binney 1977). This could account for the frequency of occur-
rence of cD galaxies in luminous X-ray clusters (where more gas may have
initially been present). If this infall occurred sufficiently long ago,
no evidence of recent star formation may remain in the galaxy colors.
The dominant galaxies might actually end up being reddened if preferen-
tially enriched material has been accreted. One might expect that the
colors of cD galaxies should not fallon a simple linear extrapolation
of the galaxy color-magnitude relation.

The gas density in the cluster core ~ Tl/2, whence we expect that
at photon energy E, dLx/dE ~ Tl/2 exp(-E/kT). This relation is consist-
ent with current data.

The correlation both of low (26 MHz) and high (1400 MHz) frequency
radio fluxes with Lx can be simply understood if the magnetic fields in
the extended radio sources (that may be largely contributing to LR : Aizu
1977) originated together with the enriched intracluster gas. Assume
that the relativistic electron density is maintained in equipartition
with the intracluster field. This may not be implausible since local
acceleration mechanisms are evidently required. Suppose that the intra-
cluster magnetic fields and relativistic particles are initially produced
by galactic supernova remnants. Since the rate at which enriched gas is
produced by evolving stars should be proportional to the supernova rate,
one would expect that the total cosmic ray electron energy Eel is ap-
proximately proportional to the gas mass M . Consequently one obtains
(assuming similar volumes for the X-ray an3a~adio sources) LR ~ Lx (E el )2
(Mgas)-2(~v)-I. This suggests a proportionality similar to that observed
(slnce Lx ~ (~v)4).

If equipartition is attained, the magnetic field strength should be

proportional to E1i2 r- 3/ 2 (where r is the source dimension). Ignoring
any variation in e r, the synchrotron lifetime ts (at a fixed frequency)
varies as E-3/4, or Mglfs4 in the protogalactic infall model. Since
Lx ~ M~a~' a shorter electron lifetime ts ~ L~3/ 10 (and therefore steeper
radio spectrum) is indicated in high luminosity X-ray clusters.

The decrease of spiral content with Lx and ~v supports ram pressure

stripping (Gunn and Gott 1972) as the mechanism for converting spirals
to SO's. A quantitative estimate of its significance can be obtained in
GAS IN GALAXY CLUSTERS 183

individual clusters where spiral galaxies that pass within a critical

radius rs of the cluster centre (core radius rc) are stripped (Tarter

*
X2 ~DF ) xmIn
2 (DF*) tw
mln (km s-l)
Cluster r /r (rs model) (King r /r
s c x c
model)

Abell 262 2.4 0.34 (7) 8.8(7) 760 (6.0)

326 2.0 0.19(9) 12.5 (9)
426 5.3 14.5(21) 42.7(21) 2420 0.91
576 1.2 1. 2 (12) 6.5(12)
1060 2.2 13.6(12) 25.9(12) 1335 > 0.85
1656 5.3 0.16(4) 6.9 (4) 1560 2.5
2666 0.6 0.47(8) 0.57(8)
Virgo 1.4 0.99(15) 7.2(15) 1070 0.38

*
Deg. freedom.

1977). The absence of spirals within the central regions satisfactorily

fits the data of Melnick and Sargent (1977), whereas the fit to the
radial distribution of spirals with a uniform (King) distribution is
significantly worse (except for Abell 2666, where rs < r c ). X-ray core
radii (rx) generally appear to be ~0.5 rs (except for Abell 262, where
rx is poorly determined), indicating that ram pressure stripping is ef-
fective until the gas density has dropped to about ~O.l of its central
value. The ratio rs/rc is low for three clusters with low velocity
dispersion (~v), and larger for the two clusters with high ~v. The ram
pressure stripping model predicts such correlations: rs should scale
both with r x ' and as ~v4/ 3 (for rs > r c ). More data on rx and ~v would
help establish whether these indications of correlations are correct.
The decrease of spiral content with ~v and Lx provides further support
for ram pressure stripping.

Richness is a logarithmic measure of counts of luminous galaxies

within an Abell radius, and may be taken to give a measure of cluster
optical luminosity Lc£ (since cluster core radii show small variations):
specifically, R ~ 2.1 £n (Lc£/Lo). If we assume a relation of the form
d log Lx/dR = 8 (8 = 0.23 ± 0.13; § III), then Lx cr Lc £5.1 8 . With Mgas cr
cluster mass Mc£' one finds that Mc£ ~ L~2/ 5 ~ Lc£28 ~ Lc £O.46. Other
functional fits to Lx(R) also yield a weak dependence of Mc£ on Lc£.
Hence Mc£/Lc£ decreases with increasing richness, and ~v ~ (1.6)8K.
Measurement of cluster velocity dispersions (and masses) would enable us
to verify whether Mgas cr Mc£ for X-ray clusters where Mgas cannot be
determined directly.

The available data (for five sources) is consistent with ~(~v)2/3x

kT (where ~ is the mean molecular weight), expected in most thermal
models (Silk 1973). The weak (Lx, ~v) correlation (Solinger -and Tucker
184 JOSEPH SILK

1972) is presently incapable of distinguishing between different theoret-

ical models; this situation may change when improved X-ray source size
data becomes available, and can be incorporated into a search for a more
general correlation.

Inverse Compton X-radiation, produced by scattering of the microwave

background radiation by relativistic electrons, seems to account only for
a small fraction of the observed fluxes, since the iron-line data and
limits on hard X-ray fluxes indicate that most of the emission must be
thermal. One would also expect that the correlation between Lx and LR
should improve towards lower radio frequencies (Harriss and Romanishin
1974). A correlation between Lx and n (Gould and Raphaeli 1977), utiliz-
ing the predicted decrease in the microwave background radiation towards
rich clusters (Sunyaevand Zel'dovich 1972), should be potentially cap-
able of directly discriminating between and non-thermal models; however
it is presently inconclusive, recent results (Lake and Partridge 1977;
Rudnick 1977) having failed to confirm earlier indications of a positive
effect (Gull and Northover 1976) in known X-ray clusters. There are
indications, however,of the "cooling" effect in very rich clusters.

Nevertheless, inverse Compton X-rays are bound to be important in

nearby clusters at photon energies ~ 30 keV, and also in distant clusters
at z : 1 (e.g. Rees 1967). Assuming equipartition fi~ld strengths, the
ratio of inverse Compton to thermal X-ray fluxes is Lic/Lih ~ 0.01 (1+z)4
[Mgas(z)/Mgas(O)]-l. If the gas mass does not change with time, non-
thermal X-ra~will dominate at z ~ 1.5. The gas mass cannot exceed the
dynamical cluster mass, typically ~10 Mgas(O): hence inverse Compton X-
rays must dominate at z ~ 4.

v. ORIGIN AND EVOLUTION

The gradual increase of the elliptical to disk galaxy content of

clusters with increasing Lx and decreasing Bautz-Morgan type suggests
that the morphological distribution of galaxy types may be determined
when the cluster forms. This is not a unique interpretation, but a
possible explanation may be that galaxies undergo strong mutual inter-
actions when clusters form. This can only occur to any significant extent
if the galaxies are themselves newly formed protogalaxies. Thus we come
to one of the crucial unresolved issues of galaxy formation theories:
do galaxies form long before, or more or less coeval1y with, clusters?

The gas content of clusters leads support to the notion of coeval

formation. The amount of enriched intracluster gas attests to its
protoga1actic origin: current mass loss rates fail to account for it.
If galaxies formed in isolation, one might expect the stellar content
to rapidly dominate over the gas content as star formation proceeds,
just as happens when the disks of spiral and SO galaxies are formed. In
rich clusters, this process cannot have occurred, for some lOll Me per
galaxy is present in the form of enriched intrac1uster gas. This amount
of matter is comparable to a typical disk mass. While ordinary spirals
GAS IN GALAXY CLUSTERS 185

develop in isolation, it seems plausible that cluster galaxies must be

inhibited from making enriched stellar disks by interacting with neigh-
boring galaxies (cf. Ostriker 1977). One can account for intracluster
gas if the enriched matter produced by ordinary stellar evolution in
protogalaxies stays in gaseous form (De Young 1977): only if coeval
formation of galaxies and clusters occurs does this seem easy to arrange.

There are several mechanisms by which proto galactic interactions

can affect galactic evolution. The most important may be due to col-
lisions, which will be frequent between pro to galaxies in newly formed
clusters (Silk 1977). Initially, relative encounter velocities may be
low, and coalescence can occur. Once protogalaxies have fallen through
the cluster potential well, after roughly a crossing time (~109 yr) has
elapsed, the encounter velocities in rich clusters will be large compared
to the internal motions within the protogalaxies. While the proto galaxies
are largely gaseous, subsequent collisions will be highly disruptive.
Diffuse protogalactic gas will be shock heated, and escape from the proto-
galaxies, but dense clouds will cool effectively, and remain bound to
the protogalaxy. If the clouds are initially near the Jeans mass, the
shock passage will trigger star formation.

Star formation will result in the production of enriched material,

which will be ejected by supernova-driven wind flows from the less mas-
sive galaxies into the intracluster medium. Evaporation (Cowie and
Songaila 1977) and ram pressure may also playa role. More massive
galaxies will be capable of retaining supernova e~ecta, and make succes-
sive generations of stars. Galaxies of mass ~ 10 1 Me may provide most
of the intracluster gas, and should be relatively unenriched (and bluer)
compared to much more massive systems. The metal abundance of gas in
the cluster core is expected to approach that of the most massive cluster
galaxies.

How can these speculations be verified? X-ray cluster galaxies

should have undergone more recent star formation than other clusters,
and this should be reflected in their colors, particularly at z > 1,
where the cluster gas fraction (and therefore infall and ram pre~sure
effects) may be enhanced. The color changes should exceed those expected
on the basis of theories of the evolution of isolated ~alaxies. Lx
should be larger by a factor ~ (present gas fraction)-. Measurements
of the spatial distribution of line emission will help elucidate the
role of galaxy interactions. To distinguish between theoretical models,
a multiparameter analysis is needed that incorporates the principal
X-ray, optical, and radio parameters. This should become feasible with
the new generation of X-ray satellites.

I am indebted to Drs. J. Binney, J. Hutchins, C. F. McKee, and J.

Tarter for valuable discussions, to many colleagues for providing pre-
prints of recent results, and to NASA and NSF for research support.
186 JOSEPH SILK

REFERENCES

Aizu, K. 1977, Pub. Astron. Soc. Japan, 29, 33.

Bahca11, J. N. and Sarazin, C. L. 1977, Astrophys. J. Letters, 213, L99.
Bahca11, N. A. 1977a, Astrophys. J. Letters (submitted).
Bahca11, N. A. 1977b, Astrophys. J. Letters (submitted).
Cowie, L. L. and Binney, J. 1977, Astrophys. J., 215, 723.
Cowie, L. L. and McKee, C. F. 1975, Astron. and Astrophys., 63, 337.
Cowie, L. L. and Songaila, A. 1977, Nature, 266, 501.
DeYoung, D. S. 1977 (preprint).
Erickson, W. C., Mathews, T. A., and Viner, M. R. 1977 (preprint).
Faber, S. M. and Dressler, A. 1976, Astrophys J. Letters, 210, L65.
Gorenstein, P., Fabricant, D., Topka, K., Tucker, W., and Harden, F. R.
1977, Astrophys. J. Letters (submitted).
Gould, R. G. and Rephaeli, Y. 1977 (preprint).
Gull, S. R. and Northover, K. J. E. 1976, Nature, 263, 572.
Gunn, J. E. and Gott, J. R. 1972, Astrophys. J., 176, 1.
Harriss, D. E. and Romanishin, W. 1974, Astrophys~., 188, 209.
Jaffe, W. 1976, I. A. U. Symposium No. 74 (in press). ---
Jaffe, W. and Pero1a, G. C., Astron. and Astrophys., 26, 423.
Jones, C. and Forman, W. 1977 (preprint).
Kellogg, E. and Murray, S. 1976, Astrophys. J. Letters, 193, L37.
Lake, G. and Partridge, R. B. 1977 (preprint).
Mathews, W. G. 1977 (preprint).
Melnick, J. and Sargent, W. L. W. 1977, Astrophys. J., 215, 601.
Mitchell, R. J. and Culhane, J. L. 1977, Monthly Notices-Koy. Astron.
Soc., 178, 75P.
Mitchell, R:-J., Culhane, J. L., Davison, P. J. N., and Ives, J. C.
1976, Monthly Notices Roy. Astron. Soc., 175, 29P.
Murray, S., Forman, W., Jones, C., and Giacconi, R. 1977, Astrophys.
J. Letters (submitted).
Ostriker, J. P. 1977, in The Evolution of Galaxies and Stellar Popula-
tions, ed. B. M. Tinsley and R. B. Larson (Yale Univ. Press), 369.
Owen, F. N. 1975, Astrophys. J., 195, 593.
Pacho1czyk, A. G. and Scott, J. S~976, Astrophys. J., 203, 313.
Rees, M. J. 1967, Monthly Notices Roy. Astron. Soc., l3~29.
Rudnick, L. 1977 (private communication).
Rudnick, L. and Owen, F. N. 1977, Astron. J., 82, 1.
Schilizzi, R. T. and Ekers, R. D. 1975, Astron~and Astrophys., 60, 221.
Serlemitsos, P. J., Smith, B. W., Boldt, E. A., Holt, S. 5., and Swank,
J. H. 1977, Astrophys. J. Letters, 211, L63.
Silk, J. 1972, Astrophys. J., 172,563. ---
Silk, J. 1973, Ann. Revs. Astron.- and Astrophys., 11, 269.
Silk, J. 1976, Astrophys. J., 208, 646.
Silk, J. 1977, Astrophys. J. (submitted).
Solinger, A. B. and Tucker, W. H. 1972, Astrophys. J. Letters 175, Ll07.
Sunyaev, R. A. and Zel'dovich, Ya. B. 1970, Astrophys. Sp. Sci-.-,-7, 3.
Tarter, J. 1977 (private commumication). -
Tarter, J. and Silk, J. 1974, Quart. Jour. Roy. Astron. Soc., 15, 122.
Thuan, T. X. and Kormendy, J. 1977 (preprint).
GAS IN GALAXY CLUSTERS 187

DISCUSSION

Gisler: Further evidence with regard to the ram pressure stripping of

cluster galaxies may be obtained by looking at the gas content of
cluster galaxies. I have just completed a study of emission line fre-
quencies in a large sample of optical galaxy spectra, obtained from a
literature survey covering the last twenty years (approximately 1300
galaxies), and I have found that the frequency of the emission line
objects is indeed much lower in the Zwicky compact clusters than in less
compact associations (Gisler 1977, submitted to Mon. Not. R. astr. Soc.).

Abell: What is the basis of the 10 14 Mo of evolved matter that had to

be ejected?

Silk: X-ray observations, in particular of the Perseus and Coma

clusters, can be fitted satisfactorily by a range of thermal models that
determine the total mass of X-ray emitting gas to within a factor or
roughly 3. In the simplest isothermal models, the central density is
typically ~ 3 x 10- 3 cm- 3 , and the X-ray core radius is ~ 0.5 Mpc.

Ostriker: Incidentally, in the galactic evolution models, you get about

10 M@ ejected per unit solar luminosity and hence if the luminosity of
the cluster is 10 13 Lo ' you get 10 14 Mo ejected.

Tinsley: There is plenty of evidence for blue galaxies in clusters ~n

the redshift interval 0.5 to 1.0 which will be discussed in my talk on
Thursday.

Parijskij: I would like to mention our new results which we have

obtained concerning the Sunyaev-Zeldovich effect in the Coma cluster.
We first observed a small dip in the microwave background radiation in
the direction of the Coma cluster in 1970 and several later observations
give similar positive results but with large errors. Our latest obser-
vations were made with the 600-metre RATAN radio telescope at a wave-
length of 4 cm. We have again observed a dip in the direction of the
Coma cluster having ~T/T = - 2 x 10- 4 ± 0.7 x 10- 4 K and we have an
estimate of its angular size. It is somewhat less than the size of the
X-ray source in the cluster but we are not certain if the difference is
significcnt or, if real, whether it is due to a temperature or a density
effect.

Boynton: This result for the Coma cluster is consistent with zero at
the two-sigma level and is therefore consistent with the very recent
measurement of G. Lake and R. B. Partridge at A = 9 rom:
~T/T < 0.05 ± 0.05.
However, this latter limit implies no significant indication of the
Sunyaev-Zeldovich effect at an even lower level of significance for this
particular cluster.
188 JOSEPH SILK

Zeldovich: Observers should also try to see the effect beyond the maxi-
mum of the spectrum of the microwave background radiation where a
corresponding small increase in the temperature of the radiation in the
direction of the Coma cluster should be observed. It is, however, a
very difficult observation but we need to be sure about it!
 COMPUTER SIMULATIONS OF GALAXY CLUSTERING

Sverre J. Aarseth
Institute of Astronomy
University of Cambridge, England

1. INTRODUCTION AND INITIAL CONDITIONS

The aim of the present work (performed In collaboration with J.R.

Gott and E.L. Turner) is to account for the observed distribution of
galaxies in terms of the gravitational instability picture. Specific-
ally we assume that all the matter is contained in galaxies. The
evolution of such a system can then be studied by N-body simulations
once the initial conditions are specified. Our approach is essentially
experimental; a variety of models are computed and the results are
compared with observations.

The models are characterized by the following set of parameters:

N, n, m, S, ~o' €, Ro' Here N denotes the total number of galaxies;
th~ ind~x n is used to describe the initial density fluctuations op/p cr
M-2 -nib (n = 0 corresponds to a random distribution, whereas n = - 1
gives rise to a 'flat' fluctuation spectrum favoured by Gott and Rees
(1975)). The galaxy masses m are usually taken to be equal; alterna-
tively we select two mass groups or a general mass function. A random-
ized component of kinetic energy, 'rp =: ST H, may be superimposed on the
initial Hubble velocity flow ~i = H-~i with total kinetic energy TH
(S = 0 denotes a 'cold' universe). We adopt the standard Friedmann
cosmology with A = 0 and concentrate on two values of the final density
parameter ~o =: 81TGpo/3Ho2, where Po is the present mean density of the
universe: (i) a closed universe with ~o = 1 (parabolic case) and (ii)
an open universe with ~o = 0.1. The calculations are assumed to start
at a red-shift Zst when the primordial density fluctuations have
reached a value op/p ~ 1 on galactic mass scales. Associating this
epoch with a galaxy collapse time ~ 10 9 yrs then gives an expansion
parameter at the present epoch of Ro = 1 + Zst, whereupon the initial
value of ~ may be evaluated if ~o ~ 1. The scale of the system is set
from considerations of the total luminosity; we adopt a final radius
Ro = 50 Mpc for a Hubble constant Ho = 50 kID sec- l Mpc- l .

Two separate computer codes are used, depending on whether the

galaxies are assumed to be mass-points or extended distributions of
189

M. S. Longair and J. Einasto (cds.), The Large Scale Structure of the Universe, 189-196. All Rights Reserved.
Copyright <i) 1978 by the IAU.
190 SVERRE J. AARSETH

characteristic size E. In the latter case the 1 galaxies are represented

by a 'soft' potential of the form Gm/(r2 + E2)2. Treating close mass-
point pairs by a two-body regularization technique, we have computed
several models with N = 1000 galaxies. A soft potential permits more
particles to be studied; here we report on two models with N = 4000
galaxies and a softening parameter clRo = 1 x 10- 3 (i.e. E = 50 kpc).
The equations of motion for each galaxy are advanced according to the
Ahmad-Cohen (1973) integration scheme which treats neighbours and dis-
tant particles on different time-scales. We consider a spherical
region of space in isolation and therefore neglect the effect of
external galaxies. Any galaxy which approaches the expanding boundary
is reflected with decreased radial velocity. This procedure preserves
the mean density in the co-moving frame and has the effect of cooling
the peculiar motions near the boundary. The boundary itself is assumed
to obey an appropriate Newtonian equation of motion throughout.

The calculations give rise to an extensive data bank which will be

used for a variety of tests and comparisons. In this paper we report
briefly on the covariance function and discuss some properties of the
clustering process. More detailed considerations will be published
elsewhere (Aarseth, Gott and Turner (1977), henceforth referred to as
AGT) .

2. THE COVARIANCE FUNCTION

The covariance function ~ provides a precise tool for comparing

the final models with observations. It is defined in terms of the
probability dP of finding a galaxy within a volume dV at the distance r
from a randomly chosen galaxy: dP = n (1 + ~(r)) dV, where ng is the
number density of galaxies. Peebles f1974) finds ~(r) ~ 68 r- l . 8 for
0.06 ~ r ~ 60, r being measured in Mpc, where the amplitude is uncertain
by a factor of ~ 2. The 64 rouble question is then whether any of the
models can reproduce a satisfactory power-law covariance function over
the observed range.

Analysis of models I and II yields covariance functions ::;(r) ~ 110

r- l . 9 and ::;(r) ~ 190 r- l . 9 , respectively, over the range ::;(r) ~ 10 4 to
~(r) ~ 1 for the scaling Ro = 50 Mpc. Here Model I is a closed universe
with parameters no = 1, n = 0, B = 0.027, 1 + Zst = 13.8, whereas the
open Model II is characterized by ~o = 0.1, n = - 1, S = 0, 1 + Zst =
30.9. Both models contain N = 1000 galaxies of equal mass which have
been treated as mass-points (E = 0). The amplitudes which are somewhat
high may be reduced by ~sing smaller values of Zst, whereas the exponent
is very close to the observed value for both models.

It is a striking feature of the mass-point models that the covari-

ance function continues to increase (~ith a slightly st~eper slope) at
smaller distances, reaching ::;(r) ~ 10 at rlRo ~ 2 x 10 5 in both models.
This property is accounted for by a number of extremely close binaries.
No such small scales are present in the initial distributions where the
COMPUTER SIMULATIONS OF GALAXY CLUSTERING 191

closest binaries are ~ 10- 2 of the corresponding scale factor R. The

net shrinkage is mainly due to relaxation effects between binaries and
single galaxies which are well understood from theoretical considerations
(Heggie 1915). In the case of soft potentials the form of ~(r) is
similar to the mass-point models over the observational range but the
maximum amplitude is considerably smaller.

Models with two mass groups ml and m2 have also been considered.
In one model with m2 = 2m we find that the amplitude of ~(r) for the
heavy mass component lS atout twice that for the light component, whereas
the respective slopes are 2.1 and 1.9. We may associate the heavy com-
ponent with E and SO galaxies which show some evidence of having twice
the M/L value of spirals (Turner 191 6) but similar luminosity functions
(Shapiro 1911). Our result is then in qualitative agreement with
covariance functions determined separately for ellipticals, lenticulars
and spirals (Davis and Gel~er 19 16). The mass effect in the simulated
models is another manifestation that significant two-body relaxation
has taken place. A further indication of relaxation is provided by an
initial feature in the covariance function at large r which disappears
during the evolution.

To conclude this discussion, it appears that different cosmological

models (i.e. closed vs open) cannot be distinguished from the analysis
of ~(r) for the present range of parameters. However, the N-body
simulations do demonstrate that the gravitational instability theory
can account for the observed galaxy clustering.

3. THE GROWTH OF CLUSTERING

Pictures of the final models reveal a significant amount of galaxy

clustering (cf AGT) and it is therefore of interest to understand the
clustering process. The growth of clustering is illustrated by a 16 mm
time sequence movie which shows the projected xy-distribution of N =
4000 galaxies in co-moving coordinates. This is an open model with
basic parameters n = - 1, S = 0, ~o = 0.1, E = 50 kpc, 1 + Zst = 32.
We employ a general mass spectrum derived from the luminosity function
of galaxies in small groups (Turner and Gott 19 16), where the initial
density fluctuations 6p/o are comparable to a similar distribution of N=
1000 equal galaxy masses. Although the movie only displays the two-
dimensional development, the apparent clustering is quite pronounced.
Two snap-shots of the evolution are displayed in Figure 1; there is some
indication of the clusters grouping together into super-clusters as
suggested by the observations. One cluster forms at the boundary and
survives throughout the calculation; hence the boundary effects do not
appear to suppress the clustering unduly.

A proper analysis of the clustering process lS best carried out in

three dimensions. Some preliminary results have been obtained uSlng a
simple approach based on spherical symmetry. For this purpose we employ
an operational cluster definition as follows. The integration scheme
192 SVERRE J. AARSETH

R= 4·3

Figure 1 Projected distribution in the xy-plane for the

model Qo=O.l, N=4000 at expansion factors R=4.3 and R=15.4.
COMPUTER SIMULATIONS OF GALAXY CLUSTERING 193

provides information about the number density of neighbours (i.e. neigh-

bour sphere radius and corresponding membership). For simplicity we
assume that the galaxy with the highest associated density is located at
a cluster centre. We then include the mass Ms within spheres of increas-
ing radius Rs until the corresponding density contrast C = R3Ms/NRs3
falls below a prescribed value Ccl (N is also the total mass). A density
contrast of Ccl = 5.5 represents turn-around in an Einstein-de Sitter
universe, but alternative values may also be considered. The properties
of such a cluster can then be analysed as for an isolated system,
whereupon the next cluster is identified in a similar manner subject to
the condition that no galaxy is counted more than once.

Table I gives the fre~uency distribution of identified clusters

with more than 12 members for Ccl = 5.5. The first line refers to the
~o = 0.1 model discussed above and the second line is for a similar
closed model with 1 + Zst = 10.7.

Table I

Model > 400 400 - 200 200 - 100 100 - 50 50 - 25 25 - 13

["10 0.1 1 o 4 6 9 31
rio =1 1 1 3 4 8 31

The table contains 51 and 48 clusters with respective total populations

of 2422 and 2748 galaxies. Adopting Ccl = 22 instead we find similarly
29 and 21 clusters with a corresponding total membership of 1731 and
1804 galaxies. The average mass of the cluster galaxies in Table I
tends to be somewhat higher than the total average (typically c, 20 -
30 per cent), but only in one case does the excess reach a factor of 2.
On the other hand, a few clusters show a small deficiency of heavy mem-
bers; this is perhaps not surprisi~g considering that only about 26 per
cent of all the galaxies exceed the average mass. In any case it should
be emphasized that the relatively small mass differences between cluster
galaxies and field galaxies in these models do not allow for the pos-
sibility of mergers, which would be more likely in clusters.

Further analysis may show whether some of the richest clusters

contain independent sub-clusters not identified by the present procedure.
In principle at least the stability of hierarchical clustering would
provide a test for distinguishing between closed and open models. All
clusters in Table I have significant negative binding energies and are
therefore ~uite stable. This is still the case (with one exception)
after subtraction of the negative energy stored in close pairs, e~uiva­
lent to evaluating the total energy in the two-body approximation. Most
of the members are located inside half the estimated cluster radius,
indicating a significant core-halo structure. The associated crossing
times are defined by tcr = Ms5/2/(2iEcli)3/2, where Ecl is the corrected
cluster energy. Typical values of the dynamical age t/tcr range from
~ 2 to ~ 10, hence there has been time for significant internal relaxa-
tion.
194 SVERRE J. AARSETH

Tidal interactions between the clusters may also be of dynamical

importance, with shear motions generating angular momentum of opposite
sign. The effectiveness of such tidal torques may be tested by evaluat-
ing the dimensionless quantity y = J2/2TI, where J, T and I are the
total angular momentum, kinetic energy and moment-of inertia with respect
to the cluster centre. For most clusters in Table I, y ~ 0.01 and hence
the externally induced rotational energy is very small. This result may
also have implications for the tidal origin of rotation inside galaxies;
to be effective this process requires the galaxies to be relatively
closer to each other than is the case for the clusters.

In conclusion, the present dynamical models can account quite well

for the observed covariance function and are also capable of producing
clusters with a wide range of membership. The insensitivity of these
quantities to the initial conditions (particularly Q) is somewhat sur-
prising and needs to be understood better. However, the velocity dis-
tribution provides more sensitive tests for distinguishing between
cosmological models, as discussed by Gott elsewhere in this volume.

REFERENCES

Aarseth, S.J., Gott, J.R., and Turner, E.L. 1977, in preparation.

Ahmad, A., and Cohen, L. 1973, J. Compo Phys. 1£, 389.
Davis, M., and Geller, M.J. 1976, Astrophys. J. 208, 13.
Gott, J.R., and Rees, M.J. 1975, Astron. & Astrophys. ~, 365.
Heggie, D.C. 1975, Monthly Notices Roy. Astron. Soc. 173, 729.
Peebles, P.J.E. 1974, Astrophys. J. Letters, 189, L51.
Shapiro, S.L. 1971, Astron. J. 76, 291. ---
Turner, E.L. 1976, Astrophys. J:-208, 304.
Turner, E.L., and Gott, J.R. 1976, Astrophys. J. 207, 6.

DISCUSSION

Peebles: I have produced similar films, but have always called them
"propaganda films" because it seemed to me that it is very dangerous to
compare them too closely to the real Universe. There are two important
lengths in the model, the radius R, and R/N 1 /3, where N is the number of
particles, the latter being the radius at which the initial fluctuations
are non-linear. Since we do not know how to model the non-linear part
of the mass distribution, I felt that one should only examine the struc-
ture that develops on physical scales ~ R/N 1 /3. Unhappily, since N is
limited to ~ 1000, this leaves only a very restricted range of scales.
You have taken a bold step in going to much smaller scales, and you
might be right, but I think it is a little dangerous.

de Vaucouleurs: Could you offer some comments - possible reasons - for

the facts that (1) your calculations match satisfactorily the observed
two point covariance function, but (2) the galaxy and cluster distribu-
COMPUTER SIMULATIONS OF GALAXY CLUSTERING 195

tions shown in your film does not resemble the real Universe as depicted
by Peebles from the Lick counts?

Aarseth: In the film each galaxy is shown with the same intensity,
whereas the masses are in fact selected from a broad Schechter function.
A proper representation would certainly improve the visual impression of
the simulated cluster picture. In any case this question cannot be
settled by a subjective inspection. What is needed, is a quantitative
comparison using methods described at this meeting.

Petrovskaya: What is the time unit in the flim and what may one say
about the time scale of the clustering process?

Aarseth: The initial epoch is taken to be 10 9 years and the final epoch
is the present time, corresponding to an expansion factor of 32. A
significant amount of clustering can already be seen after four or five
expansion times.

Fall: My question is related to Prof. Peebles' comment about character-

istic scales and their possible effect on the reliability of your calcu-
lations. As he has pointed out, one might worry that the point-mass
calculation would not accurately reflect the clustering of real galaxies
on scales smaller than the initial mean intergalaxy separation or,
equivalently, the characteristic size of a galaxy. Also, one might
worry that the real distribution of galaxies was already clustered and
had significant velocities at epochs, corresponding to the beginning of
your calculations. If the final distribution of galaxies in your calcu-
lations depended sensitively on galaxy sizes, velocities etc., then the
interpretation would be complicated. My question is simply, do you find
any evidence that the final distribution depends sensitively on these
effects?

Aarseth: We have done calculations with mass-points and soft potentials

as well as different initial position and velocity distributions, and
find no significant differences in the final correlation functions.

Zeldovich: The initial peculiar velocities if they are different in

different places are equivalent to temperature. The Jeans mass grows
and small clusters are prevented from forming. But various peculiar
velocities increase the amplitude of perturbations, each one on its own
mass scale. They can even lead to clustering in the absence of initial
density fluctuations.

Davis: Could you comment on the growth rate of the cosmic potential
energy T or the cosmic potential energy U in your simulations? In the
self-similar ~ ~ 1 models, T and U will scale as R(1-n)/(3+n), where R LS
the cosmic factor and n is the initial spectral index of perturbations.
If this is not so in your simulations, then non self-similar effects are
dominating the solution.
196 SVERREJ.AARSETH

Aarseth: We plan to check the growth-rates of the peculiar energies,

but have not done so yet.

Ostriker: One of the most interesting aspects of Aarseth's simulations

is that the amplitude of the correlation functions for masses M = 2MI
are twice as large as those with M = MI. Can I ask all those who have
made these numerical simulations whether they would have expected this
result or does it indicate that the calculations are wrong?

PeebLes: I have made N-body calculations to test this point. The

simulations start with a clustering hierarchy with points distributed
according to the empirical 2, 3 and 4-point correlation functions. The
velocities are chosen so that on taking statistical averages the distri-
bution will be kept in equilibrium, and the model is then run forward in
time. It is found that the 2 and 3 point correlation functions do not
change appreciably with time. If there is a mass distribution, the
cross-correlation function for different masses does not change with
time, i.e. the light particles do not float up to the dense spots.

Ostriker: Have you answered my question?

PeebLes: If the simulations had resulted in the clustering distribution

we observe in the Universe, I would not have expected to see mass
segregation.

Gott: We were not at all surprised that our simulations with two mass
groups (M2 = 2MI) gave the results that the galaxies of mass 2MI had a
covariance function of just twice the amplitude of the galaxies of mass
MI. The three point correlation results of Peebles and Groth show that
a tight binary galaxy has just twice the covariance function of a normal
galaxy, essentially because each of the two galaxies in the binary
galaxy brings the expected average number of companions with it. Of
course, one galaxy with mass 2M is dynamically indistinguishable from a
tight binary of mass (MI+MI), and should, therefore, have a similar
covariance function.
SHORT COMMUNICATIONS 197

AN ANALOGY BETWEEN SIMULATED AND ACTUAL

CLUSTERS OF DIFFERENT KINDS
G. Paal

A study of clusters is presented by means of the function Nm(9) ,

i.e. the total number N of members heavier than a limiting mass m, as a
function of the limiting projected angular distance from the centre 9.
As a rule such a function is found to possess a well-defined maximum
change of slope - an apparent "deflection" - if the limiting mass is
large enough (see Figure 1 a-g). Some possible definitions of this
deflection and the corresponding special "core radius" of the cluster
are described else-
AarSE'th's simLiated model Whitp's modPi and obsprvations where (Astrofizika,
N 700 bodlr.;
7, 435, 1971). The
80 500 bodIes
V2!J<1546 typical occurrence and
60 rather regular behav-
40 iour (systematic
temporal variations)
of this deflection in
a sample of 17 config-
urations simulated by
60 S.J. Aarseth and S.D.M.
White for different
conditions prove this
feature to be a natur-
N o. @1.0 al consequence of the
80 500 bodIes
gravitational inter-
60
action of many bodies
: ~t~""::~ in a statistical
sense; it is apparent-
/to.32 ly a characteristic of
NO.5 @1. core formation and the
80 500 bodIes
evolution of clusters.
60
/ m;:t.S In the case of compu-
40 ~-_m.2.D ter data the existence
20 ~ cluster of galaxlf!'s
of such deflections
.Jfo.l9 cannot possibly be due
0.5 1.0 mm on PSS plaIr.; ~
to any observational
Furtt..r clQ!JUms of observotd clust.,-., Astroliziioa 7,pp.438, 445. error or uncertainty
Figur~ 1. or erroneous personal
judgement. The corres-
ponding diagrams, Nm(9) for clusters of galaxies (see also loco cit.)
are found to be practically indistinguishable from those of star clusters
and simulated systems. They show the very same type of deflection, the
same remarkably small dependence on the choice of the limiting mass,
that of the "centre", the field correction etc. As simulations give just
the same diagrams as have been obtained observationally it is plausible
198 SHORT COMMUNICAnONS

to regard also these observational findings as dynamically necessary and

not chance coincidences, unless the actual clusters of galaxies are
stabilized by some continuously distributed intergalactic "hidden matter"
instead of the masses of observable discrete bodies themselves. However
one can hardly accept a radically different explanation for the appar-
ently identical phenomenon in the case of systems of galaxies on the one
hand and those of stars on the other. Estimates of the crossing time tc
of the systems together with morphological arguments and numerical
experiments indicate that actual "rich" clusters of galaxies might
typically be near a particular epoch of very rapid systematic cosmologi-
cal evolution. A considerable decrease of the aforementioned "core
radii" (a transition between two nearly constant values) with the other
characteristic sizes of the systems almost unchanged (Figure 1 a-d).
The corresponding phenomena were detected earlier observationally and
consequently these empirical data, together with their consequences, now
seem to be once again corroborated (see e.g. Ast~on. Naeh~iehten, 297,
311, 1976). These remarkable analogies suggest that a similar mechanism
is responsible for the formation and evolution of conspicuous cores of
clusters and a similar solution of the somewhat controversial situation
with regard to mass segregation both for clusters of galaxies and open
star clusters (e.g. by properly choosing the initial conditions).

STRUCTURES AND NUMBER-DENSITY DISTRIBUTIONS

IN CLUSTERS OF GALAXIES
F. W. Baier

I would like to present some preliminary results of an investigation

of 50 clusters of galaxies. For studying the clusters, Schmidt plates
taken with the 2-metre universal telescope at Tautenburg and the prints
of the Palomar Sky Survey were used. For all the clusters we determined
isopleths to obtain detailed information about the distribution of
galaxies. We have found well isolated clusters, double clusters, multi-
ple systems and rather complicated structures.

Because we cannot say anything about physical connections between

the various groupings in the cluster regions under investigation, we
certainly have to expect a high percentage of random projections of
independent clusters. But we cannot exclude the possibility of physical
connections between the various groupings in those cases in which we
observe an irregular but well isolated structure.

For the isolated clusters we calculated projected radial number-

density distributions and radial cumulative distributions. From the
latter we found the total populations and dimensions of the clusters.
But these quantities have to be considered upper limits for those
clusters which are not well isolated because neighbouring clusters
produce secondary maxima in the radial number-density distribution and
increase both the total number of galaxies and the distance from the
centre to the apparent border of the cluster.
SHORT COMMUNICATIONS 199

We find total cluster radii to be between 2 and 4 Mpc and total

cluster populations between 30 to some hundreds of galaxies to the
limiting magnitude of our counts which is about one and a half magnitudes
brighter than the plate limit.

From the projected number-density distribution we determined the

radius of the sphere around the cluster centre which contains 20 galaxies.
We found that this radius is about 0.3 Mpc independent of the total popu-
lation of the cluster if it contains more than 170 galaxies. In any case
a value of 0.2 Mpc is a lower limit for the radius of the sphere defined
in this way. This corresponds to an upper limit for the mean number
density in this sphere for all clusters.

Of course we have counted galaxies to different absolute magnitudes

because of the different distances of the clusters, all of which are
detected to the same limiting apparent magnitudes. This effect has
little influence on the radii of the spheres containing 20 galaxies
because of luminosity segregation in the centres of clusters of galaxies.
But the total cluster characteristics will depend on these magnitude
limits.

We will discuss this effect in detail in connection with a more

extensive sample of clusters in a future paper. Our results obtained up
to the present time are published in a series of papers in Astronomische
Nachrichten.

MATHEMATICAL APPROACH TO THE PROBLEM OF CLUSTERING

J. Burczyk and A. Zieba

The mathematical theory of the description of the structure of

point-sets in a plane or in space can be applied to the investigation of
the distribution of different extragalactic objects. Mathematically we
can define rigorously such ideas as cluster, nucleus of cluster, mono-
or polynuclear structure, background and so on. First we choose a
critical distance that determines the scale on which we investigate the
distribution of objects. The critical distance plays a similar role to
the wavelength of radiation by means of which we investigate the internal
structure of matter. It is obvious that the description is different for
different critical distances. To obtain a complete picture we have to
analyse the evolution of this description as the critical distance
changes.

The mathematical theory of the description of the structure of point

sets can be applied to investigations of individual clusters as well as
to the statistical description of a family of clusters. We can charact-
erize strictly the given distribution as well as compare different
distributions.

If the point-set under under investigation is very numerous the

200 SHORT COMMUNICATIONS

description can be very complicated. In this case it is worth applying

the method of statistical reduction. The main idea of the method of
statistical reduction is the following: we replace the real population
by a new, reduced population which is distributed according to the same
rules as the real population. The process of reduction is a mathematical
one based on the theory of probability. As a result of the reduction
method we lose information but we gain on simplicity of description. We
can lose much information but gain much simplicity and it depends on our
choice. In every particular case we are able to regulate this choice in
an optimal way. As we know, it is very difficult to compare the
structures of two sets with different numbers of points. An additional
important advantage of the method of statistical reduction is the fact
that different sets after reduction can possess the same number of
objects.

As the starting-point in describing the distribution of galaxies we

must know the coordinates of galaxies or their numbers in squares by
rectangular division of the map. In the first case the method is not
only more accurate but also applicable to investigations of individual
clusters. Applications to distribution of different extragalactic
objects are in progress. In the second case we have already some
statistical results. Professor Rudnicki will describe some of them.

RADIO EHISSION OF ABELL CLUSTERS IN THE GB AND GB2 REGIONS

Adam Hichalec and Jerzy Machalski

The observational material is based upon pencil-beam surveys made

with the NRAO 300-ft radio telescope at 1400 MHz. The GB survey
(Maslowski 1972) covers 0.1586 sr in the zone 7h 15m ~ a ~ l6 h 20m,
45~9 ~ 0 < 5l~8. The GB2 survey made by Hachalski in 1975 was carried
out in the region defined by 7h 08m ~ a ~ l6 h 58m, 3l~8 ~ a ~ 39%
(0.2828 sr).

The aim of this work is as follows:

(1) to identify the sources in Abell clusters,
(2) to estimate the luminosity distribution at 1400 MHz and
compare it wi th that of Owen (1975) ,
(3) to search for a dependence of spectra on luminosity.

~ve have obtained the following results: GB GB2

Number of Abell clusters in the region 102 181
Number of detections within 0.3 rc
(cL Wills 1966) with 5 1400 ;;;. 0.09 Jy; 31* 39
with 5 1400 < 0.09Jy no data 12
*cf. Hichalec (1977) .
SHORT COMMUNICATIONS 201

Distances to the clusters were estimated from the relation

log ze = -4.697 + 0.224 m10 (Corwin 1974).
The luminosity distribution we have obtained is presented in Figure 1,
while Figure 2 shows a spread of spectral indices against luminosity •

. t... . .
11

..
.
II

~ 11
" ~(l4DG. 211S)

.
-

-
II

t '"
. . .. ..,-,,-... '..
11'1 UII.CDI}
1--'
-31 -!4 -II - o.
I\.~~·
ll,I,Is.4Do)
I
II
+
Figure 1. Observed luminosity
(z~ S1400) distribution of GB -3! -24 -11
and GB2 radio sources stronger LI.( z~ s,., I
than 0.09 Jy which coincide
with Abell clusters within Figure 2. The spread of spectral-
0.3 rc (full line). The dashed index against luminosity for GB
line indicates an extension of sources (upper diagram) and GB2
the distribution to lower flux sources (lower diagram). Vertical
densities taken from the GB2 and horizontal bars indicate
survey data. typical rms errors.

Our analysis leads to the following conclusions:

(1) The observed luminosity distribution (based on observations of
weaker clusters) is distinctly much more asymmetric in comparison with
that given by Owen (1975). This suggests that strong radio galaxies are
less numerous in Abell clusters than was estimated by Owen.
(2) There is no evidence for a correlation between spectral index and
luminosity (see Figure 2) from the above observations. The spread for
the GB data is considerably greater, mainly because of the use of a high
frequency spectral index which is more sensitive to cut-offs in this
spectral region.
(3) About 10% of radio sources coinciding with Abell clusters seem to
have flat or inverted spectra. For some of them radio structural data
are available. These are:
A 0951 (compact radio source, Owen and Rudnick (1976»,
A 1314 (weaker radio source IC 711; Webber (1974), Vallee and
Wilson (1976».
202 SHORT COMMUNICAnONS

REFERENCES

Corwin, H.G., 1974. Astron. J., 79, 1356.

Maslowski, J., 1972. Acta Astron:: ~, 227.
Michalec, A., 1977. Acta Cosmo1ogica, 6, 47.
Owen, F.N., 1975. Astrophys. J., 195,593.
Owen, F.N. and Rudnick, L., 1976. Astrophys. J., 203, 307.
Va11~e, J.P. and Wilson, A.S., 1976. Nature, 259, 451.
Webber, J.C., 1974. Pub1. Astron. Soc. PacifiC:-86, 223.
Wills, D., 1966. Observatory, 86, 140.
III

LARGE SCALE SYSTEMS

THE LOCAL SUPERCLUSTER

G. de Vaucouleurs
The University of Texas
Austin, Texas 78712

1. INTRODUCTION

The first quantitative evidence for a large scale density excess or

"metagalactic cloud" including the Local Group was obtained by Holmberg
(1937) and confirmed by Reiz (1941). The present concept of the Local
Supercluster (LSC) as a flattened aggregate of field galaxies, small
groups and larger clouds centered at the Virgo cluster was formulated 25
years ago (de Vaucouleurs 1953) even before large-scale superclustering
was recognized as a general phenomenon. See review papers in V;';"-taJ., ~n
A~~onomy (1956), Sov~et ~~onomy (1960), SeienQe (1970), Publ. A~tAon.
SOQ. PaQ~6~Q (1971), and TAU Symp. No. 63 (Abell 1974).
Although the concentration of the brighter galaxies (m ~ 15) toward
the supergalactic plane could conceivably be fortuitous (Bahcall and Joss
1976), statistical tests d~monstrate the very low probability of the ran-
dom clumping hypothesis (1976a) and confirm the reality of the local su-
percluster as a physical association (1975a, b, c; de Vaucouleurs ana
Corwin 1975).

The first attempts to detect an anisotropy -- and in particular a

rotational component -- in the velocity field of nearby galaxies were
made by V. Rubin (1951) and Ogorodnikov (1952) with conflicting results.
A specific disk model of the supercluster in differential rotation and
expansion proposed by the author (1958, 1964, 1966, 1976b) predicts de-
partures from a linear-isotropic Hubble flow in good agreement with ob-
servations. A spherical non-rotating model has been discussed by Peebles
(1976) and a rotating-expanding spheroidal model by the author and Peters
(1972). Both models indicate a measurable slowing down of the Hubble ex-
pansion by the local density excess, but a non-rotating model cannot re-
produce details of the observed velocity field. The characteristic "dou-
ble wave" first-order effects of differential supergalactic rotation have
been detected in the velocity residuals of solar motion solutions for the
nearer groups (Stewart and Sciama 1967) and, perhaps, even within the Lo-
cal group (de Vaucouleurs et ai. 1977a, b).

205

M. S. l.ongair and J. Einasto (ed.<.), The Large Scale Structure of the Universe, 205-213. All Rights Reserved.
Copyright © 1978 by the IAU
206 G.DEVAUCOULEURS

A number of possible objections to these results (van Albada 1962;

Sandage and Tammann 1975) have been examined in detail and found to be
invalid or contradicted by concrete evidence (1964, 1976b). Further
tests are presented in §§ 4, 5.

The effects of a local inhomogeneity imbedded in an isotropic back-

ground have been discussed by Silk (1974), Mavrides (1976, 1977) and
others. More complex hierarchical relativistic cosmological models have
been explored by Wesson (1975). Theoretical models for the origin and
evolution of superclustering in general, and of the Local Supercluster
in particular, have been discussed within the framework of the gravita-
tional instability picture (Peebles 1974, Doroshkevich et at. 1974, 1976)
or, alternatively, the primeval turbulence concept (Ozernoy 1969, 1974).

2. SUPERGALACTIC LATITUDE EFFECTS

A. Supergalactic concentration. The strong concentration of gal-

axies toward the supergalactic equator in the northern galactic hemi-
sphere (NGH) is well documented at all magnitudes m < 14 (1960, 1975b)
and is probably detectable down to m ~ 16 (Carpenter-196l). In the
southern galactic hemisphere the supergalactic concentration is signifi-
cant only for the brighter galaxies (m < 11), an indication of the out-
lying location of the Local Group in the LSC.

The Local Group and nearby groups are

concentrated toward the equator of a Local
Cloud inclined 14° to the equatorial plane
of the LSC (1975a); this local plane also
controls the distribution of the nearest
1
intergalactic H I clouds (de Vaucouleurs
and Corwin 1975), including the Magellanic
Stream and the presumed orbital plane of
the Magellanic Clouds (Einasto et al. 1976;
Mathewson et at. 1977; Davies and Wright
1977). More generally large galaxy clouds
within the LSC tend to be flattened or elon-
gated in directions inclined less than 35°
to the equatorial plane of the LSC (1975c). Figure 1

The nearby dwarf elliptical galaxies of the Sculptor type are also
strongly concentrated toward the supergalactic equator especially in the
NGH (Karachentseva 1969), as are the DDO dwarfs, and more generally the
low-velocity galaxies of all types (1965, 1975a; Tully and Fisher 1977a,
b). (Figure 1).

B. Orientation of galactic planes. An unpublished search by the

author in 1953-54 did not detect any strong tendency of the planes of
galaxies toward parallelism to the supergalactic plane either in the
position angles of the major axes of galaxies close to the SG equator in
the Reinmuth catalogue (1926) nor in the distribution of the poles of 202
THE LOCAL SUPERCLUSTER 207

large spirals in the Danver catalogue (1942).

However, according to Reinhardt and Roberts (1972), the ellipticity

of galaxies in the (first) Re6ehen~e Catalogue (RCl) is greater in low
galactic latitudes. An unpublished analysis of the mean ellipticity of
lenticular and spiral galaxies in the Se~ond Re6ehen~e CiUalogue (RC2)
by type and SG latitude confirms this finding: galaxies of a given type
tend to have greater than average ellipticity near the SG equator; how-
ever, the effect is indicated in both galactic hemispheres which detracts
from its significance.

C. Color excess. A tendency for galaxies of a given type to appear

redder than average in low supergalactic latitudes was detected by Takase
(1972) from an analysis of the corrected colors Co of 510 galaxies in RCI.
This result was confirmed by an independent analysis of color residuals
of 262 galaxies from new McDonald data (de Vaucouleurs et al. 1972).
This effect was interpreted as evidence for intergalactic extinction and
reddening (see also Schmidt 1975), or, alternatively, as an effect of
local density on the composition and intrinsic colors of galaxies (Abadi
and Edmunds 1976).

However, because the galactic and supergalactic equators are nearly

orthogonal, there is a loose negative correlation between galactic and
supergalactic latitudes; as Gula et al. (1975) have noted this makes the
results sensitive to errors in the galactic extinction correction. A
new, detailed analysis of the fully corrected total colors (U - V)r of
468 galaxies in RC2 fails to detect any significant color excess near the
supergalactic plane in either hemisphere. It is not clear why the two
previous studies showed a positive effect, but the nil result from the
new data which are far superior in quality and quantity has greater weight.

D. Intergalactic extinction. There is some evidence for an anti-

correlation between counts of faint galaxies and clusters and supergal-
actic latitude B. An unpublished study by the author in 1954 indicated
a negative correlation between the apparent surface density of bright
galaxies (m < 13, Shapley-Ames) and faint galaxies (m < 21, Hubble
counts). Zwicky (1962) has commented on the marked deficiencies of VD
and ED clusters in the areas covered by the Virgo cluster and Ursa Major
Cloud along the SG equator. This conclusion is confirmed by Holmberg's
(1974) analysis of the distribution of Zwicky's D, VD and ED clusters;
the observed deficiency in the Virgo area could be explained by an intra-
cluster extinction of 0.25 mag. This needs to be confirmed by a study
of colors.

3. SUPERGALACTIC LONGITUDE EFFECTS

The apparent distribution as a function of SG longitude L of gal-

axies in the equatorial belt (IBI< 30°) is an important indicator of the
structure of the LSC; in particular the direction of the SG center is con-
sistently determined to be that of the Virgo cluster (L = 104°) whether
208 G. DE V AUCOULEURS

low latitudes (IBI < 10°) are included or not. This result holds at all
magnitudes 11 < m < 14 (1975b) within statistical fluctuations caused by
the cloud structure of the LSC. Only the nearest galaxies and groups
(~ < 10 Mpc) show pronounced departures; in particular, several nearby
groups and clouds are concentrated in the UMa-CVn area (L < 90°) (1965,
Figure 3).

4. SUPERGALACTIC KINEMATICAL EFFECTS

Systematic departures from the ideal linear-isotropic Hubble expan-

sion are an important corollary of the supergalactic hypothesis. A rota-
ting-expanding model of the LSC accounts for the observed departures with
a minimum number of free parameters (1958, 1964, 1966, 1972, 1976b). Sev-
eral possible objections or alternative interpretations have been examined
and tested as follows:

(a) The velocity anisotropy does not exist or is not significant

(Bahyl 1974; Sandage and Tammann 1975 = ST V). The data on Sc I and nearby
groups presented in ST V have been examined in detail (1976b). They sbow
the same typical velocity anisotropy with the same phase and amplitude
(at given modulus ~o) as had been previously derived from the totality of
available (m,z) data (loc. cit., Figs. 5-7). Tully and Fisher (1977a, b)
have confirmed the center-antic enter anisotropy in their extensive sample
of nearby spirals and irregulars.

(b) The anisotropy of the (m,z) relation could conceivably be pro-

duced by an anisotropy of the distribution of absolute magnitudes at
given m (van Albada 1962), but detailed examination of the supergalactic
distribution of luminosity classes shows that no such effect is present
(1964) .

(c) Another possible cause of anisotropy is the dependence on space

density of the effective mean distance of a magnitude-limited sample in
the presence of a significant dispersion oM of absolute magnitudes (Teeri-
korpi 1975). While this bias can account for apparent departures from
linearity in z(m) in some di-
rections and in some data, it
does not follow that the ani so- vA
(kms-')
IBI < 30"
tropy of the relation is in
fact and quantitatively the
result of the supergalactic
anisotropy of the space den-
sity of galaxies. Two tests
are possible:
1000
(I) Compare the ampli- • m"11 0-129.uN " I 0(19581
tudes of the velocity aniso- 01'0"31.5. uN "04(1977)
tropy derived from several
o 180" SGL 360"
magnitude-limited samples
having different dispersions Figure 2
THE LOCAL SUPERCLUSTER 209

TABLE 1. NORTH-SOUTH MODULUS ANISOTROPY AT z = constant t

Data log Vo lI].1o(B - E) NB,N E a (].10)
ST V, Sc I 2.8-3.3 +0.82 ± 0.22 19, 9 0.51, 0.63
T > 2, 0.5 < II. < 1.4 2.8-3.2 +0.46 ± 0.10 92, 27 0.43, 0.60
t area B (60° < L < 120°), area E (240° < L < 300°), both IBI < 30°.

OM; since any statisEical bias is proportional to O~ the anisotropy

should decrease as OM 2 if it is a statistical artifact, but should be
independent of OM if it is a real property of the velocity field. Fig. 2
shows the dependence of < Vo > on SGL for 2 corresponding samples, one of
old data having a large dispersion oM ~ 1.0 mag (m = 11.0-12.9, de Vaucou-
leurs 1958, Table 1) and one of the latest data for which oM ~ 0.4 mag
«].10> = 31. 5, see § 5). The amplitude is only slightly less in the new
data.

(II) Compare the absolute magnitudes (or distance moduli ].10) derived
from the m(z) relations in a bias-free range of z. Table 1 shows two re-
cent estimates for the north and south galactic polar caps (60° x 60°
areas Band E in de Vaucouleurs 1976b, Fig. 1); the first is from 28 Sc I
galaxies with ].10 derived from H II regions (Sandage and Tammann 1975; de
Vaucouleurs 1976b, Tables 4 and 7); the second is from 119 spirals (T> 2,
Sb or later) having a luminosity index II. = (L+T)/IO in the range 0.5 < II.
< 1.4. Both sets indicate that at the mean redshift < log Vo > ~ 3.0-
where the statistical bias is negligible (Sandage and Tammann 1975;
Teerikorpi 1975), galaxies are about 0.6 mag (or 30 percent) more distant
in area B. Conversely, at a given mean distance < log 1I > ~ 10 Mpc, red-
shifts are greater by about 30 percent in area E as all previous studies
had indicated (1976b, Table 6).

5. MOTION OF LOCAL GROUP AND HUBBLE CONSTANT

New solutions from RC2 data are presented in Table 2 for the motion
of the Local Group with respect to some 300 galaxies having distance mo-
duli 27 ::: ].10 .::: 33. The distance moduli were derived from fully corrected

TABLE 2. LOCAL GROUP MOTION IN SUPERGALACTIC COORDINATES

< ].10 > N <1I> <z > V L B <H>

(A) 29.70 94 8.71 0.0030 343 64° _ 2° 103

± 65 ± 10 ±19 ± 5
(B) 31.20 98 17.38 0.0052 348 102 - 9 90
± 80 ± 23 ±20 ± 3
(C) 32.17 96 27.26 0.0083 727 100 +27 92
±135 ± 18 ±13 ± 3
210 G,DEVAUCOULEURS

total magnitudes BT and, independently, from isophotal diameters Do as

a function of luminosity index A through the formulae

M; = -19.25 - 1.4(A 2 - 1), log Do = 4.23 - 0.25(A 2 - 1).

The zero points are fixed by 19 Local Group and nearby galaxies having
revised distances 6 < 4 Mpc derived from primary and secondary indicators
(1977a, b). Only galaxies of revised types Sb and later (T > 2) and
0.45 < A < 1.55 were considered. The standard error of the mean distance
moduli derived from both B~ and Do is o(Vo ) ~ 0.4 mag (exclusive of zero
point 0 0 ~ 0.15 mag). The solar motion relative to the extended Local
Group (V S = 336 km s-l toward £S = 107°, b S = -16°, solution B from de
Vaucouleurs et at. 1977, Table 1) was subtracted to calculate the Local
Group apices A, B, C. The main conclusions are
(a) H is substantially constant (outside Local Group) and indepen-
dent of ~ with <H> = 95 ± 3 (internal m.e., exclusive of zero points);
the luminosity selection bias which results in an apparent increase of
V/~ with V or ~ (Teerikorpi 1975) is not in evidence;
(b) redshifts at ~o = const. are systematically larger in the SGH
and, if this anisotropy is attributed only to a motion of the Local Group,
it implies a velocity Vc - 350 ± 50 km s-l toward SGL ~ 80° ± 10°, SGB ~
_5° ± 15° relative to an all-sky sample of some 200 galaxies in the dis-
tance range 5 < ~ < 20 Mpc (A, B shells);
(c) the Local Group apex (Fig. 3) is still in the same general di-
rection indicated by previous
studies (de Vaucouleurs and
Peters 1968) and in good agree-
ment with directions calculated
from a kinematical model of the
Local Supercluster (1972);
(d) at greater redshifts
(C shell) the effect of solar
motion is less well determined
and may include a component of
the systemic motion of the LSC
relative to more distant gal-
axies. The C shell overlaps
with the inner shell (1600 <
Vo < 3500, N = 22) of the Rubin
et at. (1976a, b) sample of dis- Figure 3
tant Sc I galaxies (LG velocity
Vc = 406 ± 164 toward SGL = 2So, SGB = -24°,'D i~ Fig. 3); the velocity
vectors Vc agree in amplitude, but differ by - 90° in apex directions.

6. VELOCITY ANISOTROPY AT ~o = const.

Classical cos A-term solutions have limited significance when higher-

order harmonics are present, i.e., when the velocity field is not merely
a reflection of solar motion. Solutions of the form log Vo - <log Vo >
S(~o - W
o ) were made for 262 galaxies of type T > 2 and in the range
THE LOCAL SUPERCLUSTER 211

TABLE 3. VELOCITY-DISTANCE RATIOS AT ~o = const.

Areas <~o> 29.0 30.5 31.5
L IBI from BT
O
log Do BT
O log Do BT
O log Do < Hr >

Vr 675 631 1028 1054 1720 1455

A,B,C
Hr 107 100 82 87 87 73 89
< 180° < 30°
n 16 16 55 44 61 57
Vr 853 793 1267 1408 2378 1959
D,E,F
Hr 135 126 101 ll2 ll9 98 ll5
> 180° < 30 0
n 7 7 18 7 16 26
Vr 380 419 1348 1075 1752 1668
G, J
Hr 60 66 107 85 88 84 82
< 180 0 > 30 0
n 5 4 6 6 14 12
Vr 635 685 1349 1428 2069 1777
H, K
Hr 101 108 107 ll3 104 89 104
> 180 0> 30 0
n 5 5 9 7 18 13

<Hr> 101 100 99 99 99 86 97

0.5 < ~ < 1.5 for different sectors of the SG equatorial belt. The char-
acteristic SG anisotropy is still in evidence in both < H > and Seven
when < ~o > '" const.

Another approach consists in applying differential corrections

~log V = +0.2(~o - ~o) to the corrected velocities Vo in small intervals
of ~o ~nd calculate < log Vo~o) > at ~o = const. The corresponding mean
reduced velocities Vr and Vr/~ = Hr are given by Table 3 for 4 large sec-
tors. The mean Hubble ratio is systematically higher by - 30 percent in
the SGH (L > 180 0 ) at both low and high SG latitudes; further, the all
sky average < H > '" const., independent of ~ or Vo ' which confirms that
this sample is substantially free of statistical bias.

Dr. W.L. Peters and Messrs. G. Bollinger, H.G. Corwin, D. Monyak

and W.L. Pence contributed to several phases of this work which was sup-
ported in part by grants from the U.S. National Science Foundation.

REFERENCES

Abadi, H.I. and Edmunds, M.G. i976, A.&A. 45, 319.

Abell, G.O. 1974, IAU Symp. 63, 79.
Bahcall, J.N. and Joss, P.C. 1976, Ap.J. 203, 23.
Bahyl, V. 1974, B.A.C. 25, 115.
Carpenter, R.L. 1961, P.A.S.P. 73, 224 (see also A.J. 66, 607).
Danver, C.G. 1942, Lund Ob~. Ann. No. 10.
Davies, R.D. and Wright, A.E. 1977, M.N.R.A.S. 180, 71.
de Vaucouleurs, G. 1953, A.J. 58, 30; 1956, V~~ in A~~on. 2, 1584;
1958, A.J. 63, 253; 1959, ACTP. nH. 36, 977; 1960, Sov. A~~. 3, 897;
212 G. DE V AUCOULEURS

A.J. 69, 737; 1965, "Nearby Groups of Galaxies" in Vol. 9 of

1964,
S~ and Stellan Sy~tem~, 557; 1966, AttL Conv. Co~mologia (Padova,
1964), 37; 1970, Se~enee 167, 1203; 1971, P.A.S.P. 83, 113; 1972, in
IAU Symp. No. 44, 353; 1975a, Ap.J. 202, 319; 1975b, Ap.J. 202, 610;
1975c, Ap.J. 202, 616; 1976a, Ap.J. 203, 33; 1976b, Ap.J. 205, 13;
1977a, in IAU-CNRS Coll. No. 37 and 263, CNRS, Paris, 301; 1977b,
Compt. Rend. P~ 284(B), 227.
de Vaucou1eurs, G. and Corwin, H.G. 1975, Ap.J. 202, 327.
de Vaucou1eurs, G. and Peters, W.L. 1968, Nat~e 220, 868.
de Vaucou1eurs, G., Peters, W.L., and Corwin, H.G. 1977a, Ap.J. 211, 319;
1977b, IAU-CNRS Call. No. 37 and 263, CNRS, Paris, 149.
Doroshkevich, A.G., Sunyaev, R.A., and Ze1dovich, Ya. B. 1974, in IAU
Symp. No. 63, 213.
Doroshkevich, A.G. and Shandarin, S.F. 1976, M.N.R.A.S. 175, 15P.
Einasto, J., Hand, U., Joeveer, M., and Kaasik, A. 1976, Tantu P~ep~. 10.
Gu1a, R., Rudnicki, K., and Tarraro, I. 1975, Aeta Co~mol. 315, 3, 39.
Holmberg, E. 1937, Lund Ob~. Ann. No.6, 52; 1974, A.&A. 35, 121.
Karachentseva, V.E. 1969, Ve~t. K~ev U~v., ~~on. S~. 11, 114.
Kiang, T. and Sas1aw, W.C. 1969, M.N.R.A.S. 143, 129.
Mavrides, S. 1976, M.N.R.A.S. 177, 709; 1977, IAU-CNRS Call. No. 37 and
263, CNRS, Paris, 549.
Mathewson, D.S., Schwarz, M.P., and Murray, J.P. 1977, Ap.J.Lett., in
press.
Ogorodnikov, K.R. 1952, P~oblem~ 06 Co~mogony (Mo~eow) 1, 150.
Ozernoy, L.M. 1969, J.E.T.P. Lette~ 10, 251; 1974, in IAU Symp. No. 58,
85.
Peebles, P.J.E. 1974, Ap.J. Lett. 189, L51; 1976, Ap.J. 205, 318.
Reinhardt, M. and Roberts, M.S. 1972, Ap. Lett~ 12, 201.
Reinmuth, K. 1926, "Die Herschel Nebel," V~. St~nW. /ludelb~g, Bd. 9.
Reiz, A. 1941, Lund Ob~. Ann. No.9.
Rubin, V.C. 1951, A.J. 56,47.
Rubin, V.C., Ford, W.K., Thonnard, N., Roberts, M.S., and Graham, J.A.
1976a, A.J. 81, 687; 1976b, A.J. 81, 719.
Sandage, A.R. and Tammann, G.A. 1975, Ap.J. 196, 313.
Schmidt, K.H. 1975, Ap. &Spaee Se~. 34, 23.
Silk, J. 1974, Ap.J. 193, 525.
Stewart, J.M. and Sciama, D.W. 1967, Nat~e 216, 742.
Takase, B. 1972, P.A.S. Japan 24, 295.
Teerikorpi, P. 1975, A.& A. 45, 117.
Tully, R.B. and Fisher, R. 1977a, "A Picture of the Local Supercluster",
(preprint); 1977b, IAU-CNRS Call. No. 37 and 263, CNRS, Paris, 95.
van Albada, G.B. 1962, in IAU Symp. No. 15, 411.
Wesson, P.S. 1975, Ap. & Spaee S~. 32, pp. 273, 305, 315.
Zwicky, F. 1962, in IAU Symp. No. 15, 347.
THE LOCAL SUPlcRCLUSTER 213

DISCUSSION

Zeldovich: What is the effective flattening of the supercluster ~n

terms of the ratio of axes of distribution of galaxies?

de Vaucouleurs: About 0.4 to 0.5 for the outer regions, see Vistas in
Astronomy, vol. 2 (1956).

Zeldovich: Is it due to rotation?

de Vaucouleurs: No, not in the sense of a Newtonian spheroid ~n centri-

fugal equilibrium.

Zeldovich: The cosmological Hubble constant should be measured at

distances where density uniformity is established - not in superclusters.

de Vaucouleurs: Of course, but we can measure directly only small

distances, say < 50 Mpc, so a compromise must be accepted. Except in
the heavily populated supergalactic equatorial belt in the north galac-
tic hemisphere, H is approximately constant and independent of direc-
tions (see Table 3). Ho ~ 100 km s-1 Mpc- l should be a good enough
approximation to the asymptotic, low-density value of the Hubble constant.

Van der Laan: Referring to Figure 1, can you say what this diagram
means? Have the space densities of galaxies been cor~ected for the cos 2 e
factor at high latitudes?

de Vaucouleurs: All the effect occurs very close to the supergalactic

equator, e = 0, and hence these correction factors are very small in
comparison with the rapid fall-off in the number density of galaxies.
214 SHORT COMMUNICATION

A TOUR OF THE LOCAL SUPERCLUSTER

R. Brent Tully and J. Richard Fisher

To complement our movie, we offer three steroscopic pictures of a

region enclosing the Local Supercluster. A coordinate system is defined.
x axis: in a line from the Virgo cluster to the Galaxy.
Y axis: orthogonal to x in the plane of the supercluster;
the supercluster lies in the x-y plane.
Z axis: perpendicular to the supercluster; the plane of the
sky lies in the y-z plane.
The distance scale is established assuming a uniform Hubble flow. The
Virgo cluster is at the origin (0, 0, 0). If Ho = 75 km s-l Mpc- 1 is
taken, then the Galaxy is at (14.7, 0, 0). The boundaries of the volume
displayed are then:
-5 < X < 14.7 Mpc
-14<Y<17 "
-10 < Z < 10 "
The pictures may be studied with a standard stero viewer or, with diffi-
culty, by splitting the images with a card and viewing from a distance
of about 15 em. The three perspectives are:
(a) a view essentially of the y-z plane (in fact, there is
a 13 0 rotation about the z axis),
(b) after a 30 0 counterclockwise rotation about the y axis
from (a),
(c) after a 90 0 counterclockwise rotation about the y axis
from (a), i.e. a view of the x-y plane.
The viewer at vantage point (a) is edge-on to the disk of the Local
Supercluster and the viewer at (c) is looking down from the pole.
SHORT COMMUNICATION 215

.... :.;:,-

-" •...
..
.. _. wi .. .. ~ .1 ....

.. -::..

c
216 SHORT COMMUNICATION

DISCUSSION

de Vaucouleurs: Is the velocity range sufficient to cover the whole

supercluster?

Tully: The Virgo cluster has a redshift of 1100 km s-l while our outer
boundary is at 3000 km s-l. The number density of galaxies beyond Virgo
appears to falloff significantly faster than the falloff which would
be expected from incompleteness in the sample. However, we could get
into semantic difficulties about just what we mean by the term super-
cluster.

Abell: In seeing this extremely interesting film, I had the impression,

perhaps incorrect, that many of the groups and clusters of points seemed
elongated towards us (i.e., the Galaxy). Can this not be an effect of
velocity dispersion within the groups?

Tully: There are quite a number of elongated features but I am not

aware of a tendency for them to point toward us. Of course, there is
the very prominent region in Ursa Major - Canus Venatici which extends
away from us, but in this region we know from non-redshift distance
indicators that the low redshift objects are nearer and the high red-
shift objects are farther away.
LARGE SCALE CLUSTERING IN THE UNIVERSE

P. J. E. Peebles
Joseph Henry Laboratories
Princeton University, Princeton, N. J., U.S.A.

1. INTRODUCTION

In 1924 Hubble presented the first generally convincing evidence,

from the identification of variable stars, that some of the brightest
"spiral nebulae" are galaxies of stars well outside the bounds of our
own Galaxy. This led him to reconsider the idea that the faint
"spiral nebulae," which were known to be much more a1undant than
bright ones, might be similar objects at greater distances. If the
galaxies were uniformly distributed through space the number brighter
than apparent magnitude m would vary as

E
-z
V 10 3

10

8
m

Fig. 1. Hubble's test of the distribution of galaxies.

217
M. S. LongQlr and J Einasto (eds.), The Large Scale Structure of the Universe, 217-226. All Rights Reserved.
Copyright © 1978 by the 1AU.
218 P. J. E. PEEBLES

log N« m) = 0.6 m + A.

Hubble (1926) tested this relation with the data shown in figure 1.
The point at the extreme right hand side is based on Fath's (1914)
counts of nebulae with diameters greater than 'V 4". The straight line
shows the slope expected under the homogeneity assumption. The agree-
ment is very encouraging, though, as we now recognize, it must be in
part fortuitous because the counts at m ~ 12 are influenced by the
concentration of galaxies in and around the Virgo cluster, that is,
the Local Supercluster, and also in the Local Group. Shapley (1938
and earlier references therein) persistently emphasized that the
homogeneity assumption surely is not the whole truth for the galaxy
distribution is strongly clustered on scales at least as large as that
of the Local Supercluster, and he even ventured to ask whether there
really is an ultimate scale on which the universe appears close to
homogeneous. However, most people were quick to accept Hubble's con-
clusion that the galaxy distribution is fairly uniform. Hubble (1926)
pointed out that this agrees with Einstein's (1917) homogeneous world
model. Milne (1933) suggested that homogeneity might be a logical
consequence of what we mean by the universe. Dingle (1933) was more
cautious in arguing that, at least within the framework of general
relativity theory, homogeneity is not a necessary property of the
world model but rather a particularly simple mathematical model subject
to empirical test. However, he seemed not to be inclined to question
the observational situation.

As is discussed at this conference two recent developments

indirectly confirm Hubble's intuition at an accuracy much better than
anything previously available. First, the microwave back~round is
isotropic about us, on angular scales ~ 10' to oi!i ~ 10-. This does
not directly measure the isotropy of the matter distribution because
the coupling of the microwave radiation to matter in the present
universe is very weak. It does imply some strong indirect constraints:
redshift, whether interpreted as cosmological (expansion) or
gravitational, must be constant to tenth percent accuracy in different
directions, as must the "initial values" at the horizon. The X-ray
background appears to be isotropic to 'V 1 percent accuracy on scales
larger than a few degrees. Since an appreciable fraction of this
radiation does come from known sources - galaxies and clusters of
galaxies - this implies a similar limit on fluctuations in the matter
density integrated to the horizon, ct 'V 3000 h- 1 Mpc (H = 100 h km 3- 1
Mpc- 1 ) •

Despite the remarkable preclslon of these two tests there remains

Shapley's point - the galaxy distribution is strongly clustered, and the
nature of this clustering ought to be of considerable interest to
cosmology. In the first place, of course, it tells us something about
what the universe is like. More directly, galaxy clustering is
interesting because the relevant dynamics seems to be simple enough
that we might hope to deal with it quantitatively and in some detail.
Perhaps if we can see how to measure the galaxy distribution in the
LARGE SCALE CLUSTERING IN THE UNNERSE 219

right way it will prove to be the "Rosetta Stone" by which we learn the
underlying significance of the clustering of matter!

The data on hand permit fairly detailed studies of the nature of

the galaxy distribution (within "local samples") on scales ~ 10 h- 1 Mpc.
Some aspects of this are summarized in the next section. The distribu-
tion on scales ~ 30 h- 1 Mpc is much less well understood. Mandelbrot
(1977) has given a particularly attractive discussion of the point that
the usual description of the distribution of mass in the universe by a
classical continuous and differentiable function p(x, t) [or, more
generally, Tij(xk)] is in a sense a fiction because p fluctuates on
whatever scale one chooses to measure it, from the limit of classical
physics, perhaps 10- 13 cm, to at least 30 h- 1 Mpc ~ 10 26 cm. This is a
range of some 39 decades. The standard Friedman-Lemaitre cosmological
model can be a reasonable approximation only because (or if!) this
progression of clustering ends within the next two decades, so the
density is close to uniform on the scale ct ~ 3000 h- 1 Mpc.
De Vaucouleurs (1971 and earlier references therein) has emphasized
that this nominal two decade hiatus in the clustering is small compared
to the 39 decades that came before, so it is well to check very
carefully whether the clustering really does terminate. Some aspects
of the observational situation are discussed in section 3.

2. CLUSTERING ON SCALES { 10 h- 1 Mpc

The best sample of the galaxy distribution on large scales is the

Lick catalog (Shane and Wirtanen 1967), which lists counts of galaxies
brighter than m = 19 in 10' by 10' cells. The effective depth of the
catalog is about 250 h- 1 Mpc. A more detailed but shallower sample is
the Zwicky catalog (Zwicky et. al. 1961-68), with an effective depth
~ 50 h- 1 Mpc.

One very natural choice of statistics to use in analyzing the

galaxy distribution is the n-point correlation functions used in the
theory of non-ideal gases. (We know this is a natural choice because
in the original analyses of Neyman, Scott and Shane 1953, Limber 1954,
and Rubin 1954 all, apparently independently, chose statistics closely
allied to the two-point correlation function.) Recent work along this
line has been reported at length in the literature (Groth and Peebles
1977 and earlier references therein) so only a few of the main results
will be summarized here.

The two-point correlation function, ~(r), is the generalization to

a distribution of points of the usual auto-correlation function for a
continuous density function,

The galaxy two-point correlation function at small scales is given to

good accuracy by the simple power law model
220 P. J. E. PEEBLES

s (ro/r)Y, ro:t 5 h- 1 Mpc,

Y 1.77 ~ 0.04, 100 h- 1 kpc ~ r ~ 10 h- 1 Mpc.

We can be fairly sure the estimate of s has not been seriously affected
by systematic errors in the catalogs because the results from the
independent Lick and Zwicky samples are quite similar. The galaxy
three-point correlation function is known to accuracy comparable to that
of S. The galaxy four-point function is only just detected above the
noise, but again the results from the two samples are consistent. All
these statistics agree with the picture that the galaxies are arranged
in a hierarchical clustering pattern, with the characteristic density
within clusters of size r scaling as r- Y. This also agrees very well
with the conclusion of de Vaucouleurs (1971 and earlier references
therein) that the most dense clusters of size r have density that scale
as a power of r, with index very close to y.

3. CLUSTERING ON SCALES ~ 10 h- 1 Mpc

A first interesting question is whether the power law hierarchical

clustering pattern observed at r ~ 10 h- 1 Mpc might simply be
extrapolated to arbitrarily large scales. A cosmology of this sort has
been discussed by Wertz (1971). This extrapolation would imply that
the map of angular positions of galaxies brighter than m should appear
equally "rough," in the ensemble average, at all m. That is, suppose
galaxies are counted in cells of fixed size and shape randomly placed in
the sky. Then the expected rms fluctuation in the counts of galaxies
brighter than m would be independent of m (Mandelbrot 1975). This is
not what is observed: the ratio of rms fluctuations in the Zwicky and
Lick samples is quite constant at a value of about 3, 10' ~ 8 ~ 20.
This value is consistent with what is expected if the Zwicky catalog
is a "fair sample" of the clustering in the Lick catalog.

A related aspect is the size of ro (eq. [lJ). This is a

characteristic clustering scale on which the density fluctuates by a
factor of about 2. If clustering extended over the full depth of the
sample ro would be expected to be comparable to the sample depth. The
fact that it is much smaller in the Lick survey, ~ 250 h- 1 Mpc,
suggests that ro has not been affected by the sample size.

In the Lick sample there is in addition to the small-scale

clustering a definite large-scale variation over angular distances of
perhaps 40 0 (Shane 1976, Groth and Peebles 1977). It is difficult to
know what part of this might be due to local effects such as variable
obscuration in the Galaxy, what part might be a true large-scale
component in the galaxy distribution. A rough upper limit to the
latter at 10 ro is (Peebles and Hauser 1974)

s(50 h- 1 Mpc) ~ 0.025. (2)

LARGE SCALE CLUSTERING IN THE UNIVERSE 221

That is, the large-scale component in any case seems to be a relatively

small perturbation from homogeneity in the sample.

The clustering length ro represents a mean over the distribution,

and one certainly can find some spots in the Lick sample where the
density remains higher than twice the overall mean over distances
substantially larger than roo An example is given by the cross-
correlation between the Lick counts and the positions of Abell clusters.
One simply "stacks" the galaxy counts around all clusters in a chosen
distance and richness class to get the mean number density as a
function of angular distance from the cluster center. By using the
galaxy luminosity function (which is itself constrained by the variation
of this "stacked" densi ty with cluster distance class) one can find from
the angular cross-correlation the mean spatial density of galaxies as a
function of linear distance from a cluster center. This mean space
density is twice the overall mean in the Lick sample at the distance
(SeIdner and Peebles 1977)

A second example is the clustering in the positions of the Abell

clusters. The auto-correlation function ~cc(r) describing the spatial
distribution of Abell clusters reaches unity at the clustering length
(Hauser and Peebles 1973)

rs ~ 30 h- 1 Mpc, ~cc(rs) = 1. (4)

This can be compared to the depth of the Abell (1958) catalog,

640 h- 1 Mpc.

The three lengths r o ' r a , and rs measure an interesting

progression in the scales of galaxy clustering in general, in the
concentration around rich clusters, and in the clustering of the rich
clusters themselves. Only in the second case can one trace the
clustering effect to distances substantially larger than the
characteristic clustering length given here. (This is because
systematic errors in the catalogs cause systematic errors in the
auto-correlation function estimates but not in the cross-correlation
function estimates if the errors are not common to both catalogs.)
However, it does appear that all three lengths are comfortably smaller
than the samples sizes, which suggests that the data now available are
adequate to measure the large scale clustering in the universe. Of
course, one would very much like to test this against deeper samples.

A deep sample that is available is the angular distribution of

radio sources, most of which are at distances ~ ct = 3000 h- 1 Mpc.
SeIdner (1977) has found very weak but apparently significant clustering
in the 4c radio catalog. This is illustrated in figure 2, which shows
the angular power spectrum for the distribution of sources in the full
catalog, S ~ 2 Jy. This statistic is a measure of the mean number of
sources in excess of that expected for a random distribution and within
222 P. J. E. PEEBLES

0.4

o --- -- --- -- - - - --" -- ----- ----- ------ -. ---- -- - --- ---- -- ----- --

-0.4

o 160 200
j

Fig. 2. Test of clustering in the 4c catalog.

angular distance ~ 1800/~ of a randomly chosen source in the catalog.

(This is the Legendre transform of the angular correlation function.
The advantages of this power spectrum are that estimates at successive
~ values are nearly statistically independent and_ the effects of
clustering on different scales are displayed at different ~-values.
For details see Peebles 1973. )

The power spectrum is negative. averaging about - 0.15. which says

there are on the average about 0.15 fewer sources around a randomly
chosen one than is expected for a un~ random distribution. This is
known to be due to confusion: if two sources happen to be closer than
~ 0?5 in the sky they will be counted as one (Webster 1976). The
spectrlm reaches a minimum at ~ ~ 100. It rises at larger ~ because it
is starting to "resolve" the anti-correlation at e :;, 0'?5 caused by
confusion. Much more interesting is the fact that the spectrum at
~ ~ 30 (8 ~ 6 0 ) is higher than at ~ ~ 100 (8 ~ 2 0 ) by ~ 0.1. That is.
the mean number of neighbors in excess of random increases by about
0.1 in going from 2 0 to 60 from a source, as if the sources were
clustered on scales of perhaps 40 , The effect is larger if the sample
is limited to a higher flux level.

Now this apparent clustering, if real. can be interpreted in two

rather different ways. Most of the 4c sources are at distances
~ 3000 h- 1 Mpc. and so one might well suppose that the clustering is
,among soyrces at about this distance. An angle of 40 subtends
~ 200 h- Mpc at this distance. To produce the observed degree of
LARGE SCALE CLUSTERING IN THE UNIVERSE 223

clustering the spatial clustering of the sources would have to be

sl.lbstantial because the effect is washed out by the many clusters seen
in projection. A rough estimate of the wanted spatial two-point
correlation function for the sources is

This goes well beyond the clustering scales mentioned above (cf. eq.
[2]). The second and possibly more conservative interpretation is that
we are seeing clustering in the small fraction of sources that are at
distances much closer than 3000 h- 1 Mpc. Some evidence that this is so
is the fact that there is quite a substantial cross-correlation between
the radio source positions and the Lick galaxy counts (at distances
~ 250 h- 1 Mpc). Thus a significant number of the 4c sources are fairly
close, and, since these sources evidently tend to be near galaxies, and
galaxies tend to cluster, these sources should tend to cluster.
SeIdner (1977) has been able to find a self-consistent model along
these lines that does reproduce the observed clustering of sources
around galaxies and among themselves for the range of flux levels
2 < S < 9. The radio luminosity function that is wanted is close to
standard estimates except that it is larger at the low power end, but
not by a factor that seems unreasonable. In this interpretation the
radio source clustering lengths are in line with equations (1), (3)
and (4) rather than equation (5).

4. CONCLUSIONS
My impression is that we may understand at least the broad outlines
of the large scale distribution of galaxies. The main question is. have
we "fair samples" of the distribution? I think the two major and direct
pieces of evidence that we do are (1) the n-point correlation functions
derived from the Zwicky and Lick samples are related to each other as
expected if both are fair samples, and (2) the characteristic clustering
lengths (eqs. [1], [3J and [4]) are well within the sample sizes. I t is
possible to find contrary indications, as in equation (5), and clearly
it will be a major task to decide whether such indications can be
"explained away" as systematic errors in data or interpretation.
All the data discussed here were obtained "by hand," so to speak,
and the more I have studied the data the more I have been impressed
with the enormous effort acd the scrupulous attention to detail that
was devoted to each catalog. This means it will be no easy matter to
improve the observational situation, even with automatic scanning
devices, though surely that time eventually will come. There are two
extensions of the data that would be very important and are technically
feasible now. First is a very deep survey in limited selected areas,
following Hubble (1936), for the purpose of extending the test of
Hubble's count-magnitude relation (duly corrected for cosmology),
extending the test of equation (1) to smaller r, and testing whether
the apparent clustering in angular positions scales with depth as
224 P. J. E. PEEBLES

predicted from the clustering measures taken from the Lick and Zwicky
samples. Some preliminary results on this subject have been reported
by Phillips et. al. (1977). A more difficult project is to improve the
estimates of the shape of the galaxy correlation function ~ at
r ~ 10 h- 1 Mpc. This could be done in a survey about twice the depth
of the Lick catalog, in sample areas spread over some 10 to 20 degrees,
with galaxy magnitude standards controlled and consistent in all
samples to about Om.03 (to assure that systematic errors in magnitudes
do not introduce appreciable spurious clustering).

Finally, it .night be mentioned that the situation will be very

different when we have adequate samples of galaxy redshifts at
D ~ 50 h- 1 Mpc. In the present analysis, where we have only very crude
measures of individual galaxy distances, a major problem is that the
angular distribution of a sample at great depth appears very close to
random because we are seeing many clusters in projection. As a result
a small systematic error in the angular distribution can be translated
into a very large error in the estimate of the spatial clustering.
Redshift data will allow us to avoid this problem and will present us
with the great opportunity of studying the large scale kinematics as
well as the large scale distribution of galaxies in the universe. Some
aspects of this great project are discussed at this conference by
M. Davis.

ACKNOWLEDGMENT

This research was supported in part by the National Science

Foundation.

REFERENCES

Abell, G. 0.: 1958, Ao~aphy~. J. Suppt. 3, P 211.

de Vaucouleurs, G.: 1971, Pubt. Ao~an. Soe. Paci6~e 83, p 113.
Dingle, H.: 1933, Monthly Nat{e~ Roy. Aotnan. Soe. 94, p 134.
Einstein, A.: 1917, S. B. Plte.uM. Akad. OJ~~. p 142.
Fath, E. A.: 1914, Ao~ophy~. J. 28, p 84.
Groth, E. J. and Peebles, P. J. E.: 1977, Ao~ophy~. J. 216.
Hauser, M. G. and Peebles, P. J. E.: 1973, Ao~ophy~. J. 185, p 757.
Hubble, E. P.: 1926, Ao~ophy~. J. 64, p 321.
Hubble, E. P.: 1936, Aotnophy~. J. 84, p 517.
Limber, D. N.: 1954, Ao~ophy~. J. 119, p 655.
Mandelbrot, E. E.: 1975, C.R. 280A, p 1551.
Mandelbrot, E. E.: 1977, Fltact~, Freeman, San Francisco.
Milne, E. A.: 1933, Z. Ao~ophyJ... 6, p 1.
Neyman, J., Scott, E. L., and Shane, C. D.: 1953, AotnophyJ... J. 117,
P 92.
Peebles, P. J. E.: 1973, AJ..~ophyJ... J. 185, P 413.
Peebles, P. J. E. and Hauser, M. G.: 1974, AotnophyJ... J. Suppt. 28,
p 19.
LARGE SCALE CLUSTERING IN THE UNNERSE 225

Phillipps, S., Fong, R., Ellis, R. S., Fall, S. M., and Mac Gillivray,
H. T.: 1977, preprint.
Rubin, V. C.: 1954, P~oc. Nat. Acad. Sci. 40, p 541.
SeIdner, M.: 1977, Dissertation, Princeton University.
SeIdner, M. and Peebles, P. J. E.: 1977, A6~ophy~. J. 215, p 703.
Shane, C. D.: 1976, private communication.
Shane, C. D. and Wirtanen, C. A.: 1967, Pub!. Lick Ob~. 22, part 1.
Shapley, H.: 1938, P~oc. Nat. Acad. Sci. 24, P 282.
Webster, A.: 1976, Monthly No~ceh Roy. A6~on. Soc. 175, p 71.
Wertz, J. R.: 1971, A6~ophy~. J. 164. p 227.
Zwicky, F., Herzog, E., Wild, P., Karpowicz, M. and Kowal, C. T.:
1961-1968, Catalogue 0 -6 GaiaUeh and C!U6.teM 0 -6 Gaia.Ueh. in
6 vols., California Institute of Technology, Pasadena.

DISCUSSION

Kiang: When I made the statement "galaxies are clustered on all scales"
ten years ago (Monthly Notices 1967), I was looked at askance by most
astronomers. I am glad that this idea is now getting generally accepted,
thanks largely to the work of Dr Peebles and co-workers. But the impres-
sion should not be formed that this idea originated with Peebles, nor
with his results on the covariance function. In fact, I arrived at this
idea by first following Neyman and Scott's method of definite clusters
but finding that the size of clusters obtained increased with the size of
cells used in the analysis.

Longair: I will discuss in my lecture tomorrow the evidence on the

isotropy of the distribution of radio sources. In brief, one may say
that a large number of surveys have now been analysed using the method
of power spectrum analysis by Webster and no evidence of anisotropy has
been found. It is important that this result is found in large surveys
such as the Molonglo surveys in which the effects of confusion are very
small.
In interpreting the results of the cross-correlation analysis
between Shane-Wirtanen counts and 4C radio sources, it should be rememb-
ered that the luminosity distribution of radio sources is very broad.
Roughly, one would expect about 20% of the 4C sources to have redshifts
less than ~ 0.1-0.2. These sources will be correlated with the overall
galaxy distribution and hence will certainly contribute to the effect
found by Peebles.

Peebles: In Webster's analysis he only sets upper limits to the aniso-

tropy. Our effect is at the upper limits he sets to the anisotropy. It
should be noted that one can distinguish confusion and clustering using
the power spectrum method.

Bolton: The relationship between radio sources and Shane-Wirtanen

galaxies can be shown to be due to the identified galaxies alone if the
radio sample is divided into three classes - identified galaxies,
quasars and unidentified sources.
226 P. J. E. PEEBLES

Peebles: That is just the point I was trying to make: one can argue
that the apparent clustering in the 4C sources is due to the relatively
small fraction of relatively close radio galaxies. We find the effect
you mentioned in the 3C catalogue. There does not seem to be enough
data to repeat this test in the 4C catalogue.

THE ISOTROPY OF THE UNIVERSE ON SCALES EXCEEDING THE HORIZON

L. Grishchuk

I would like to describe in a few words work which was done by

Zeldovich and myself. It gives some restrictions on the amplitude of
possible very large-scale density fluctuations mentioned by Peebles.
The main question investigated in this work is the following. What can
be said, using known observational data and some general hypotheses,
about structure of the Universe beyond the region accessible for obser-
vation at the present epoch? In fact we consider density fluctuations
(as well as rotational perturbations and gravitational waves) with
wavelengths larger than the horizon. We use the observational fact that
the quadrupole-type anisotropy of the microwave background radiation is
absent at the level of oT/T < 10- 4 . It is interesting to know if it may
happen that, at the present epoch, there exists a significant density of
perturbations (say, with the dimensionless amplitude of the order of
10- 1) which we do not even suspect because the corresponding wavelength
is very long and therefore direct observation of the entire perturbation
is not possible. Such a direct observation will be possible only in the
remote future when the horizon becomes equal to the corresponding wave-
length. To answer the question we make a natural but very important
assumption. Namely, we assume that the harmonic perturbations of differ-
ent wavelengths are not correlated in any particular way. Otherwise,
they might fit together in such a way that all perturbations (and,
therefore, oT/T) would be especially small within the horizon while sig-
nificant perturbations could take place just beyond the horizon. A
situation of this kind would imply that an observer at the Earth occupies
a unique position in the Universe. We assume, on the contrary, that all
observers are equivalent. All of them, even causally unconnected obser-
vers, could detect similar restrictions on the anisotropy of the micro-
wave background, oT/T < 10- 4 . Nevertheless, the question still exists
whether small perturbations unnoticeable by every observer within his
horizon can represent different parts of a significant long wavelength
limit. The main result of this investigation can be formulated in the
following way. The observational data on oT/T in combination with the
natural hypothesis on the statistical independence of different harmonics
leads to the conclusion that in the Universe there are no significant
(i.e. with the amplitude exceeding oT/T) density fluctuations on any
spatial scale larger than the horizon. (The paper will be published in
Astr. Zh. U.S.S.R., November-December, 1977.)
SHORT COMMUNICATION 227

DISCUSSION

Zeldovich: First, a historical point: Einstein assumed the Universe

was uniform but it was Friedman who proved that uniformity is incompat-
ible with a static model of the Universe. Second, Peebles used nT/T to
exclude the very crude, very inhomogeneous island model of the Universe.
But Sachs and Wolfe have shown that since nT/T of relic radiation is
less than 10- 3 , this proves that perturbations of density of amplitude
1% on scales, say, 500 Mpc are absent. Using rather natural statistical
assumptions, Grishchuk and myself have shown that one can extrapolate
this result - density perturbations are small even on scales greater
than the horizon. This is curious because we are dealing with regions
which in principle cannot be observed!
RESULTS ON THE LARGE SCALE DISTRIBUTION OF EXTRAGALACTIC OBJECTS
OBTAINED BY THE METHOD OF STATISTICAL REDUCTION

Konrad Rudnicki and Stanislaw Zieba

Jagiellonian University Observatory
Cracow, Poland

1. INTRODUCTION

The phenomenon of clustering of galaxies was known earlier than

the nature of the so called extragalactic nebulae. It was discovered as
early as the end of XVIII century by W. Herschel and confirmed by his
followers. After the introduction of Schmidt cameras into astronomy,
Zwicky suggested that all galaxies participate in clustering. After the
discovery of extragalactic radio sources some astronomers have been
trying, in different ways, to apply the concept of clustering to them
also.

The concept of clustering was, until the last two decades, taken to
mean that all the objects under consideration belong to some individual
systems called clusters. Those individuals mayor may not belong to
individual "clusters" of the second order, and those to the clusters of
third and higher orders. Kiang (1967) drew attention to the fact that
the overall existence of individual clusters is not the only possible
model of clustering. He proposed another picture called continuous
clustering. This picture is based, in principle, on the concept of
irregularities or non random fluctuations in the distribution of objects
on all possible scales. Configurations of regions of higher or lower
density (number of objects per unit volume) can in some places manifest
themselves as well defined individual clusters, but these are rather
exceptions. In general, the distribution of extragalactic objects in
Kiang's picture can be described by the statistical parameters of the
fluctuations, and not by the characteristic parameters of individual
clusters, clusters of clusters, etc., as was attempted in the classical
papers of Abell, Neyman, Zwicky and their collaborators.

It is no wonder that the idea of the overall existence of individ-

ual clusters was considered first historically. Early investigators
were fascinated by such systems as the Coma, Hydra I or Perseus
clusters. These were the first to be elaborated and described in detail.
Hundreds of others, similar to them, were discovered in farther regions
of space. Therefore, general catalogues of galaxies such as those of
229
M. S. tongair and 1. Einasto (eds.), The Large Scale Structure of the Universe, 229-239. All Rights Reserved.
Copyright © 1978 by the 1AU.
230 KONRAD RUDNICKI AND STANISLAW ZIEBA

Abell (1958) and Zwicky (Zwicky et al. 1961-1968) were made assuming
that clustering means the existence of individual clusters. Even in
the last few years single, field galaxies were considered sometimes to
be clusters with one member. This is rather a psychological problem
and its practical consequences were discussed in depth by Rudnicki
(1976). Only after first attempting to describe whole regions of the
sky by listing individual clusters and possible superclusters and find-
ing it impossible to see any notion of "a cluster as an individual"
which would be adequate to the real distribution was our attention
drawn (Flin et al. 1974, see also Rudnicki 1976) to Kiang's concept of
continuous clustering which had been overlooked.

One cannot say that today the controversy between the concept of
the overall existence of individual clusters (Zwicky-Neyman proposition)
and of continuous clustering (Kiang-Flin proposition) is in any sense
resolved. One of the strongest observational arguments in favour of
the existence, in general, of individual clusters is the peculiarity of
first rank galaxies in almost all the strong condensations of galaxies.
This is explained theoretically by the concept of the multiplication of
galaxies described by Arp and Ambartsumian. This postulates that all
the galaxies are offsprings of protogalaxies, ejected from them directly
or from their descendants. A region consisting of offsprings of a
single protogalaxy is certainly an individual cluster, at least at the
beginning but subsequently irregular ejections and individual motions
may blur the picture. There is another philosophical argument, namely
we are placed inside a well defined system, our Supergalaxy (the Virgo
cluster, in Zwicky's terminology). One would rather dislike the idea
that we inhabit a very particular part of the Universe. On the other
hand, in addition to the difficulties already mentioned with the
division of the space between individual cluster cells, the strongest
argument in favour of continuous clustering is the spectrum of charac-
teristic dimensions of irregularities in the distribution of extragalac-
tic objects. In the course of more detailed investigations it gets more
and more complicated. Also the controversy about the existence of
higher order clusters, especially the fact that there is a smooth trans-
ition between what some investigators call a superclusters and a single
regular cluster of the same size (Rudnicki 1967) can be easily resolved
by the Kiang-Flin proposition.

There are two ways of seeking a solution to this fundamental

question of which proposal fits the real Universe better. One is to
search for a better notion and a better practical criterion for picking
out individual clusters in the sky; the other - to develop more sensi-
tive statistical methods. We shall limit ourselves here to the second
approach only.

2. SHORTCOMINGS OF FORMER METHODS

Excluding some methods of only historical value and a few others of

limited applicability, the most widespread methods today are the
RESULTS ON THE LARGE SCALE DISTRIBUTION OF EXTRAGALACTIC OBJECTS 231

following four:

(i) Correlation method (called also autocorrelation method) is

very simple and permits one to obtain the characteristic dimensions of
clustering provided that they are very distinct.

(ii) Dividing the investigated field into cells and comparing the
standard deviations of numbers of objects in these cells with a statis-
tical distribution which is given a priori. By changing the sizes of
cells, characteristic dimensions of the fluctuations can be obtained
(the method of subsequent subdivisions, cf Zwicky 1957). This method
has various names and various mathematical realisations and gives also
one parameter of distribution (dimension). There arise difficulties of
interpretation when comparing the results based on different numbers of
objects or of cells.

(iii) Analysis of the distribution of distances between the nearest

(or second, third, etc. nearest) objects (Pilkington and Scott 1965).
This, when applied to the nearest neighbours only, gives rather primitive
information. When extended to more distant neighbours it brings diffi-
culties of interpretation of the results. Besides it is very sensitive
to all possible selection effects which bring serious difficulties ln
comparing the results based on different sources of observational
material.

(iv) Power spectrum (cf Webster 1976a,b) is certainly one of the

best methods used today. Its shortcomings lie in the complicated
mathematics which makes practical programming and preparation of
observational material difficult. It needs much computer time. In
addition, in this method some effects of the general gradient of the
distribution over the surveyed field may be mixed with the effects of
clustering.

3. METHOD OF STATISTICAL REDUCTION

The method of statistical reduction, also called the reductional

method, was elaborated by Andrzej Zieba and his collaborators and has
already been published (A. Zieba 1975, S. Zieba 1977, Garbaj 1976) and
there is no need to present it here in detail. We want, however, to
present its simplest and most frequently applied variant in extragalac-
tic astronomy for those who have not had the opportunity to get
acquainted with it earlier.

An area of celestial sphere, which mayor may not be rectangular,

is divided into quadratic cells called elementary domains. An area so
divided is called a chart of population. We form pairs of these cells.
Each pair, called a fundamental domain, consists of two numbered cells
(elementary domains). Those, in every fundamental domain, lie at the
same distance and the same direction from each other (Figure 1).
232 KONRAD RUDNICKI AND STANISLAW ZIEBA

We consider now all the

possible pairs of objects distri-
i buted in the surveyed field and
l-
12 calculate the probabilities that in
a fundamental area chosen at random

" t
the pair of objects drawn at random
is distributed in everyone out of
six possible ways: 0,0; 0,1; 1,0;
1,1; 0,2; 2,0. The formulae,
I
worked out by A. Zieba, for the
probabilities of such configura-
tions for a given chart and funda-
Figure 1. Scheme of elementary mental domains are unexpectedly
and fundamental domains. simple. He call a set of such
probabilities the disposition of the
population reduced to two objects. Only four out of these six probabil-
ities are independent. Then, we calculate the same probabilities for the
so called randomized distribution which is the disposition of a chart
where the numbers of objects in individual elementary domains are the
same as in the real chart of the population, but the elementary domains
are shuffled at random. In addition we calculate a standard disposition
of objects distributed at random over the elementary domains. So we
have three sets of parameters:

distribution of two objects chosen at random

in a fundamental domain chosen at random

0,0 0,1 1,0 1,1 0,2 2,0

probabilities of the above distributions

for the real
chart
(disposition
of a chart)
for a shuffled
chart
(randomized
disposition)
for a chart
populated at
random
(standard
disposition)

Four independent parameters can be formed out of P11 for the

description of the real chart. The following parameters have simple
physical interpretations.
RESULTS ON THE LARGE SCALE DISTRIBUTION OF EXTRAGALACTIC OBJECTS 233

(a) Concentration index c. It is defined as:

c
2

where k is the number of elementary domains. When there are more

highly populated elementary domains on the real chart than the random
distribution gives, c is larger than 1. On the other hand, c smaller
than 1 indicates that there is a deficiency of highly populated cells
in comparison with objects distributed at random.

(b) Index of grouping:

P ll k-l Pll
g k2
Sl1 k-c 2

When highly populated areas avoid each other the value of g is less
than 1. When they show a tendency to gather together it is larger
than 1.

(c) Index of weak anisotropy:

k2 POI - P IO

2 k - (l+c)

When scantily populated areas ("holes") in fundamental domains appear

more frequently in the first elementary domains, al has a negative
value. In the opposite case - the value is pos~t~ve. This index
indicates the degree of overall anisotropy (i.e. gradients) in the
general background of the surveyed objects.

(d) Index of strong anisotropy:

P02 - P2Q P02 - P20
az k2
R02 + R20 2

The physical sense of this index is similar to al but shows the general
anisotropy in the distribution of "condensations" and not of "holes",
It shows the anisotropy (gradient) in the distribution of dense fluctua-
tions. The last two indices are, of course, direction dependent.

In addition, the so-called structural index s = c/g was introduced

by Stanislaw Zieba. When the sizes of elementary domains are of the
same order as the sizes of the fluctuations, the concentration index
tends toward its maximum, the grouping index toward its minimum. Thus
the structural index is very sensitive to characteristic sizes of
fluctuations and is a powerful tool for studying the distribution of
these sizes.
234 KONRAD RUDNICKI AND STANISLAW ZIEBA

4. ADVANTAGES OF THE METHOD OF STATISTICAL REDUCTION

The main advantages of the reductional method are as follows:

(i) Because every real population is reduced to two "mean" objects

only, the results for populations of various numbers of objects can be
compared immediately to each other without special discussion.

(ii) There exist easy formulae for calculating the standard devia-
tions of each of the defined parameters. Therefore any discussion of
confidence level is simple.

(iii) The effect of anisotropy of the background and anisotropy of

clustering (the existence of which is still under discussion) is auto-
matically separated from the effects of clustering as such.

(iv) Application of the concentration index gives results of the

same kind as the method of subsequent subdivision.

(v) In application to the clustering problem, the grouping index

calculated for different distances of elementary cells gives equivalent
information (not identical in mathematical sense!) to the correlation
method.

(vi) The structural index permits one to obtain the same results as
in the power spectrum method but, as it appeared in practical applica-
tion, is more sensitive, i.e. shows more individual maxima at a suffi-
ciently high confidence level.

(vii) Since the observational material is reduced to two objects,

any observational selection effect is meaningless for the results,
provided this effect is independent of the coordinates.

(viii) The formulae of the method are rather simple, easy for numeri-
cal calculation, and do not require much computing time. If needed,
calculation for division into cells of intermediate sizes, additional
configurations of elementary domains within the fundamental domains,
etc., may be performed at once.

(ix) When the four independent parameters are not sufficient for
describing the real distribution, additional parameters may be easily
formed by applying the reduction to a number of objects greater than
two and/or by using fundamental domains containing a larger number of
elementary domains (cf Zieba 1975 and Garbaj 1976).

The method described here has all the advantages of the other
methods and in addition some features which are characteristic of it
alone. We are sure it deserves more attention among astronomers working
on the distribution of extragalactic objects, not to mention other
applications.
RESULTS ON THE LARGE SCALE DISTRIBUTION OF EXTRAGALACTIC OBJECTS 235

5. RESULTS

In recent years, a group of about a dozen astronomers, working at

the Jagiellonian University Observatory or associat~d scientifically
with it, have been working on the theoretical development and practical
application of the reductional method. Some of these results have been
published already, some others are to be presented here separately as
short contributions. We want to confine ourselves here to some other
results obtained by Miss Jolanta Burczyk, Stanislaw Zieba, Marek Urbanik
and Konrad Rudnicki.

(a) Subclustering in well defined individual clusters

The phenomenon of subclustering inside a large, even fairly regular

cluster is well known. This problem was studied in Cracow by means of
the concentration index. A field of 52 square degrees was covered by
overlapping plates of two colours and different exposure times taken
with the Palomar Schmidt Telescope. The field contains two large
clusters: 1105.3 + 2835 and 1115.2 + 3013 (clusters Nos 5 and 14 in
field 156 of Zwicky's Catalogue). It was divided into squares 75 x 75
(arcmin)2 which corresponds roughly to the size of the nuclear areas of
these clusters. Smaller squares 15 x 15 (arcmin)2 were also studied
inside every large square separately. By comparing plates with differ-
ent exposures, galaxies in different magnitude ranges were studied
separately (Burczyk and Rudnicki 1978). The results are as follows:

magni tude range concentration index

IS' x 15' 75' x 75'
(mean value)
blue magnitudes
m < 15.8 2.9 1.6
15.8 < m < 18.0 1.5 1.3
18.0 < m < 19.4 1.3 1.1

yellow magnitudes
m < 15.0 1.5 1.5
15.0 < m < 16.3 1.7 1.6
16.3 < m < 19.0 1.6 1.3

The formalism of the reductional method allows one to compare

immediately the results of large and small scale clustering. In terms
of the hierarchical clustering model one can interpret these numbers to
mean that subclustering inside clusters is stronger than the "main"
clustering itself. In a continuous clustering model the same numbers
say that the larger the structures are the less prominent they are. I
have to confess, however, that we have found a region of the sky where
236 KONRAD RUDNICKI AND STANISLAW ZIEBA

concentration indices for small cells are smaller than for larger ones.

(b) Distribution of objects in large regions of the sky

A large set of statistical reductions was performed for a quadratic

region of the sky centered on the north galactic pole which included the
entire area down to b = +66 (S. Zieba 1977). Galaxies from the Uppsala
Catalogue (divided into 7 magnitude classes and into 7 morphological
types), clusters of galaxies from the Abell (divided into 3 distance
classes) and Zwicky (divided into 4 classes) catalogues, as well as the
radio sources from catalogue 4C (4 intensity classes) have been studied.

For individual galaxies, the spectrum of characteristic dimensions

is continuous with some maxima and no distinct new results have been
obtained except that, for distances larger than those of large individ-
ual clusters, spiral galaxies show a distribution tending to random,
whereas elliptical, SO and irregular galaxies show complicated struc-
tures in all dimensions investigated. The fact that irregular and
elliptical galaxies show a similar distribution on the celestial sphere
is surprising and has certainly some cosmogonic implications.

The distribution of radio sources was found to be random. This

confirms what was previously generally accepted. However, it should be
pointed out that for smaller sizes the structural index is systematical-
ly less than unity, and for larger sizes it tends to one. This means
that for small, and only for small distances, the distribution of radio
sources tends to uniformity (Figure 2).

$4
Figure 2. Structural index for

,I, I II + ~uul ~
2~--~~~-T-r~--~r~2 radio sources with S ~ 2.0 x 10- 26
Jy. Sections of the line which do
.: I· !'Bu,1 not differ significantly (more
than 3a) from neighbouring
3S' 1° 20 ';0 ,6 ..". 15' 20· sections are marked with vertical
strokes.
An additional analysis has been made for the PKS and GB radio
sources lying outside the north galactic cap (Urbanik 1977). It was
shown that the trend toward uniformity in distribution cannot be
explained by a confusion effect alone. On the other hand, individual
classes of radio sources studied separately, exhibited a tendency toward
clustering in small isolated regions. However this property is not
common to any of these individual classes over the whole sky nor to all
types of sources in any of the selected regions.

When selecting clusters of galaxies from existing catalogues, the

assumption is implicitly made that the individual cluster model is
accepted. The picture obtained here is rather complicated. A continu-
ous distribution with many maxima is visible for every class of cluster.
Figure 3 shows a review of the most distinct maxima. Besides the 6
RESULTS ON THE LARGE SCALE DISTRIBUTION OF EXTRAGALACTIC OBJECTS 237

Size of e.le""l.e.lI'l.t~l\'"~
do 'I¥l<\~'l'\S

Figure 3. Significant (more than 30) maxima

(characteristic sizes of clustering) in the
distribution of clusters of galaxies. Circles-
maxima known from other authors. Crosses-
maxima newly obtained.

maxima already known earlier (Kiang 1967, Kiang and Saslaw 1969, Bogart
and Wagoner 1973, Kalinkov 1974) marked by circles, 8 new maxima were
found. The multiplicity of maxima seems to speak in favour of the con-
tinuous clustering picture. The anisotropy indices for all the investi-
gated objects and regions show no significant deviations from zero.

Thus, using the statistical reduction formalism, all the main

features known already in the distribution of objects are found in one
set of calculations, and besides some new effects appear visible, which
is consistent with the picture of continuous clustering rather than with
the hierarchical clustering of individual systems.

The problem still remains open, but if we insist on remaining with

the hierarchical picture. a very complicated model for structures of
different order has to be created.
238 KONRAD RUDNICKI AND STANISLAW ZIEBA

The author gratefully acknowledges the assistance of Prof. Andrzej

Zieba, Mr Wal ter Murawski and Mr Marek Urbanik in preparing this review.

REFERENCES

Abell, G.O., 1958. Astrophys. J. Suppl. Ser. 3, 211.

Bogart, R.S. and Wagoner, R.V., 1973. AstrophYs. J., 181, 609.
Burczyk, J. and Rudnicki, K., 1978. To be published in Acta Cosmologica.
Flin, P., Machalski, J., Maslowski, J., Urbanik, M., Zieba, A. and
Zieba, S., 1974. "Confrontation of Cosmological Theories with
Observational Data" (editor M. S. Longair), p .121.
Garbaj, T., 1976. Acta Cosmologica, 4, 7.
Kalinkov, M., 1974. "Galaxies and Relativistic Astrophysics" (Springer
Verlag), p.142.
Kiang, D., 1967. Monthly Notices Roy. Astro. Soc., 135, 1.
Kiang, D. and Saslaw, W.C., 1969. Monthly Notices Roy. Astro. Soc.,
143, 129.
Pilkington, J.D.H. and Scott, P.F., 1965. Mem. Roy. Astr. Soc.,69, 183.
Rudnicki, K., 1967. Astronomicheskii Zhurnal (USSR), 44, 77.
Rudnicki, K., 1967. Astron. Nachr., 297,317. --
Urbanik, M., 1977. Phys.D. Thesis - Jagiellonian University.
Webster, A.S., 1976a. Monthly Notices Roy. Astr. Soc., 175, 61.
Webster, A.S., 1976b. Monthly Notices Roy. Astr. Soc., 175, 71.
Zieba, A., 1975. Acta Cosmologica, 3, 75. ---
Zieba, S., 1977. To be published: Acta Cosmologica, 6, 100.
Zwicky, F., 1957. "Morphological Astronomy" (Springer-Verlag).
Zwicky, F., Herzog, E., Wild, P., Karpowicz, M. and Kowal, T.C., 1961-
1968. "Catalogue of Galaxies and Clusters of Galaxies",
California Institute of Technology, Pasadena. 6 vols.

DISCUSSION

Ostriker: It may be that the characteristics of the brightest cluster

galaxies will not be helpful in distinguishing between different models
of clustering. This is due to the fact that in a local extreme concen-
tration of galaxies (whatever the origin of the maximum is), interactions
between the galaxies can produce objects like cD systems. Tidal stripp-
ing, cannibalism and equipartition combine to produce a centrally located,
low surface brightness supergiant system.

Rudnicki: I am very glad to hear it because I considered the first rank

galaxies to be the strongest argument against my personal opinion on the
clustering.

Holmberg: Your analysis does not take into account local variations ~n
the galactic absorption, which may seriously affect the numbers of
identified clusters. On account of the limited time, I refer to my
paper of 1974, published in Astronomy and Astrophysics.
RESULTS ON THE LARGE SCALF DISTRIBUTION OF EXTRAGALACTIC OBJECTS 239

de Vaucouleurs: I should like to remind statisticians that the study of

irregularities in the distribution of galaxies should take into account
the patchiness of the obscuring interstellar medium in our Galaxy follow-
ing the methods first developed by Arnbartsumian some 40 years ago and
first applied by Agekian and a few others in this country.

Fessenko: There are indications that the ~ajority of the clusters of

Zwicky and Abell are a result of the chance projection of independent
galaxy systems with smaller dimensions and at different distances from
us. Consequently, superclusters have been discovered among objects,
which are not real clusters. Variability of the observational conditions
can explain not only the dependence of the angular diameters of the
apparent false clusters on distance, but also explains the analogous
dependence for false superclusters.

Rudnicki: All the three last remarks are made on the same topic. We
have investigated the observed distribution of objects. The final inter-
pretation of the results has, of course, to take into account the
galactic extinction also. I know Fessenko's paper and I am sure he is
right to some extent, but I hope not completely. If he is, then every-
thing we are doing is without any meaning.
HAS THE UNIVERSE THE CELL STRUCTURE?

Mihkel Jdeveer and Jaan Einasto

Tartu Astrophysical Observatory
Toravere 20244
Estonia, U.S.S.R.

1. INTRODUCTION

As demonstrated by de Vaucouleurs (1956, 1975a,b, 1976), Abell

(1961), Karachentsev (1966), Kiang and Saslaw (1969) and others,
clusters of galaxies as well as groups of galaxies have a tendency to
form second-order clusters of galaxies, often called superclusters.
Recent statistical studies by the Princeton group (Davis, Groth and
Peebles 1977, SeIdner and Peebles 1977a) have confirmed this result.
Statistical studies, however, give little information about the internal
structure of second-order clusters.

To get a clearer picture of the distribution of galaxies in space,

we have studied the three-dimensional distribution of galaxies and
clusters' of galaxies. We have used recession velocities of galaxies and
mean velocities of clusters as distance indicators supposing, following
Sandage and Tarnrnann (1975), that the expansion of the Universe is
uniform.

2. OBSERVATIONAL DATA

The distribution of the following objects has been studied: Abell

clusters of galaxies (Abell 1958), Zwicky clusters of galaxies (Zwicky
et al. 1961-68), groups and pairs of galaxies, and single galaxies.

The second Reference Catalogue of Bright Galaxies (de Vaucouleurs,

de Vaucouleurs and Corwin 1976) has served as the source of the radial
velocities for galaxies. Radial velocities of the Abell clusters have
been taken from a compilation by Corwin (1974) supplemented by some new
determinations (Faber and Dressler 1976). Only nearby objects have been
studied: Zwicky clusters belonging to the "near" distance class and---
galaxies having redshifts not in excess of 15 000 km s-l have been used.

241

M. S. Longair and I. Einasto (eds.), The Large Scale Structure of the Universe, 241-251. All Rights Reserved.
Copyright © 1978 by the IAU.
242 MIHKEL JOEVEER AND JAAN EINASTO

3. SPATIAL DISTRIBUTION OF NEARBY GALAXIES AND CLUSTERS OF GALAXIES

To study the form of large-scale clusters, we compare the apparent

distribution of these objects on the sky with successive cross-sections
in space. To do this, the sky has been divided by parallel small
circles into zones. The cross-sections of the space along the zones
represent conic volumes with the Galaxy at the tip of the cones. This
particular representation, used also by Fall and Jones (1976) in their
study of the Rubin-Ford effect, has proved extremely useful in studying
the spatial distribution of galaxies.

The overall large-scale distribution of galaxies was studied by

dividing the sky into six cross-sections parallel to the celestial
equator, each zone being 15° in width. The distribution of galaxies and
clusters of galaxies in these sections exhibits the following features.

Clusters of galaxies are not randomly distributed but form long

chains. Small groups of galaxies and galaxies have a tendency to form
disk-like aggregates. These densely populated formations are separated
from each other by big holes: regions of low space density of galaxies.
The cross-sections between 0 = -45° and 0 = +45° show the presence of
chains of clusters of galaxies in the Perseus, Coma, Hercules and Fornax
regions. Big holes can be seen between the Local Supercluster and the
Perseus chain of clusters, between the Coma and Hercules supercluster
and in some other regions (Figure 1).

We have studied in detail the Perseus supercluster which is very

favourably situated just behind a big hole and there is therefore prob-
ably very little confusion with foreground objects. Galaxies in the
magnitude interval l2~ - l4~5 in the Perseus region of the sky have
been plotted in Figure 2. Giant galaxies in this apparent magnitude
interval are at a distance of about 100 Mpc, if we adopt a Hubble
constant H = 50 km s-l Mpc- 1 •

The most prominent feature in this figure is a chain of clusters of

galaxies, which contains the Perseus cluster Abell 426, clusters A 347,
A 262, NGC 507, NGC 383, and a number of groups of galaxies. We shall
call this chain the Perseus chain of clusters. On the sky it covers a
strip about 45° long and has half width about 1?5. Practically all
galaxies of the chain are located in a strip 4° wide (see Figure 2).
All galaxies in this strip with known redshifts are plotted in redshift
space in Figure 3. Data on some of the condensations in this strip are
given in Table 1 (Vo is the redshift of the main galaxy, <Vo> the mean
redshift of the cluster, or - the velocity dispersion and nvel - the
number of galaxies with known redshifts). It is remarkable that the
redshifts of the main galaxies are very close to the mean velocities of
galaxies in the clusters with a very small rms scatter. We also note
that most of the main galaxies are supergiant ellipticals with extended
haloes, 6 of them being radio sources.
HAS THE UNIVERSE THE CELL STRUCTURE? 243

15 14 13 12 11 In 9
30'<Sc,S-

. .
.
.~

•
_-- ........
16 ~.
.
/
y/
"
/ " ,. +'4 .-;..
.. .
".~ 00 t.+
",:,.
•
"':" ..~.: ,
17
~.
~!~, .
"

'".\~'~~ $ .:,--:
~-
--
"

18 ,
10 0001<",.-'
/~ "-
...-- ~
I ,
':-.
...-- . 1". ' "- "-
19
...--
...-- -- /
/

"
: ."
Id;
\
.:~'
~
"-
"-
/
20
"" -_/
••
I
/

21 22 23 0 2
""

Figure 1. Cross-section of space in declination zone

30° < a ~ 45°. Right ascension is used as the polar angle,
redshift as the radius-vector. Abell clusters of galaxies
have been plotted as filled circles, groups of galaxies as
open circles, single galaxies as dots, Markarian galaxies
as crosses. The Milky Way zone of avoidance is shown by
broken lines.
244 MIHKEL JOEVEER AND JAAN ElNASTO

60':

& ~~'tC4 ~

5
. - .......\ .-7.5~SGB c J2.5 :. -" • ' •
-
' . ,,:, 'a. ... :.. :. \. .. -,.: t" ...... ¥

'. 0( \. .1••.• • /'

.i:·:
·~I·.~J· J,-
- ...,
". • .. \,. •
.. ... .. -. .. /

4r:! •
I. ..",.

,••
\
./
":
'/
/ 15<.)1.'40
.~
...+,
,:.
.....
~.,
\
/
/."-+.'
I·. .., "..
/ ••- oft.~.lI'
~

.
3r:! ~"~ :" •. + ":."

A lB" _ \ /';> .
.:t. -;
-:
C·":" .••
.+
.
2et :~., •\~ .. 1.' :.' ,. y
+ : #
.
•• ~ • • '
\.
i
/
/ 'r·
.\'..
. \" :.
•
'.' 'I ''/

..., "',./ ..
: • .'t .. , / +
.- '.
I'

.,+'
- ..~_\ ' :
'"
I .. '\
- ../ ..
r: ........
.~
/.
...:..... ,., - ;; ..
'

": .'
:~\
",
r:!
+ t ... ··~ t ..
.. . -.-
"
-.
.
'"
'"
/.
:i: !'
... .
\
•
22"

Figure 2. Distribution of galaxies in the Perseus region.

In declination zones from 0° to 30° galaxies in the magnitude
interval l2~ - l4~5 have been plotted, in higher declination
zones a somewhat fainter magnitude interval has been used to
compensate for the increasing galactic absorption (Joeveer,
Einasto and Tago 1977). Bright elliptical and SO galaxies
(12.0 < m < 13.9) have been plotted as small filled circles,
radio galaxies in the redshift interval 2500 < Vo < 10 000 as
crosses, clusters of galaxies as large filled circles. The
zone of width 4° associated with the Perseus chain of clusters
of galaxies and Perseus supercluster disk areas A, Band C
have been shown by solid lines and supergalactic zones by
broken lines.
HAS THE UNNERSE THE CELL STRUCTURE? 245

'. :f-':\'
.. 0'
/
":;0.,
.. ~
"/ ":...
'o~ A '2. ! •
I .. ~ .... tf":.l
lI".':'" _ .•

or "'....... • .,.:
10000 5000
Redshltt (km s")

Figure 3. Distribution of available redshifts

of galaxies in areas A, Band C (see Figure 2).

Table 1
Data on the Perseus chain of clusters of galaxies

Main galaxy Cluster

Cluster NGC Vo Vo or ~el References
(km s-l) (km s-l) (km s-l)

N 315 315 5218 - - 1

N 383 383 5095 5125 411 24 Moss & Dickens
(1977)
N 507 507 5127 5113 658 14 Tifft et al.
(1975)
A 262 708 5023 5031 438 33 Moss & Dickens
(1977)
A 347 910 5315 5520 743 4
N 1129 1129 - - - -
A 426 1275 5361 5490 1396 50 Chincarini & Rood
(1972)
Field - - 5362 786 8

Total 5l90±132 5274±211 134

246 MIHKEL JOEVEER AND JAAN EINASTO

We note an interesting detail in the structure of the clusters of

galaxies in the Perseus chain. As can be seen from the charts given by
Zwicky et al. (1961-68), practically all clusters are elongated along
the main ridge of the chain. This means that the clusters are either
elongated like cucumbers or have flattened disks. We prefer the first
explanation since in the last case all clusters would be oriented edge-
on to the observer, which occurs with small probability.

Thus we conclude that the Perseus chain is essentially a one-

dimensional formation.

Now we consider the region to the south of the Perseus chain

between declinations 0 = 0° and 0 = 28°. The distribution of radial
velocities in three sections (denoted A, Band C in Figure 2) is given
in Figure 4. We see that in the middle section B between a = 23 h Sam
and a=2 h there is a strong peak with the mean redshift <V Q> = 5220 km
s-l. Galaxies with velocities of about 5000 km s-l are dlstributed
over the whole area of the section. In the neighbouring areas A and C
the red shift distribution is more or less uniform. It is apparent that
in section B we see a thin layer of galaxies and groups of galaxies.
This layer can be called the disk of the Perseus supercluster.

~r--------.r--------,-,---------,--------.-,---------.--------- ..
N
A 8 c

lD

Figure 4. Distribution of available redshifts of galaxies

in areas A, Band C (see Figure 2).

Along the eastern boundary of the disk of the Perseus supercluster

a number of Zwicky clusters of galaxies and 5 radio galaxies are located
between clusters A 426 and A 194. All these clusters and radio
galaxies have redshifts about 5000 km s-l.

To determine the extent of the disk in the southern and western

direction, we have made cross-sections in supergalactic coordinates
(Figure Sa) and in the western part of the Perseus supercluster in a
coordinate system, perpendicular to the Perseus cluster chain
HAS THE UNIVERSE THE CELL STRUCTURE? 247

SGL
(0) ( b)
3500
-'l5"<SGB~ 12,5·
.J'()o

.2593 •
• • 2572
2657 • ~
, . / .76
68 • • 147
."9 ',
•
151 /
•

o S") o
Figure 5. Cross-sections of the Perseus supercluster in
Supergalactic coordinates (left) and in Perseus supercluster
coordinates (right, for the definition of Perseus Supercluster
coordinates see Joeveer, Einasto and Tago 1977). Designations
as in Figure 1. Abell cluster numbers have been indicated;
some clusters just outside the zone limit have also been
plotted by broken circles.

(Figure 5b). As seen from these figures approximately at SGL = 300 0

(near the cluster A 194) and in the region of the Pegasus cluster
there exist bridges and chains of galaxies and poor clusters of galaxies
which join the Perseus supercluster to the Local Supercluster. It is
natural to consider these regions as the southern and western boundaries
of the Perseus supercluster.

The distribution of galaxies and clusters of galaxies in the

Perseus region as well as in other regions of the sky resembles cells.
Cell walls can be considered as disks of superclusters. Along t~
intersections of cell walls chains of clusters of galaxies are located
which encircle supercluster disks from all sides. Cell interiors are
almost void of galaxies; they form big holes in the Universe with
diameters of 100-150 Mpc. The cell structure can be considered third
order clustering of galaxies (superclusters being second-order clusters).

4. COMPARISON WITH THE RESULTS OF OTHER AUTHORS

Chains of clusters of galaxies in the Perseus region of the sky

are visible already on the maps of NGC objects (Meyer 1908). Similar
chains of smaller size are visible on computer-processed maps based on
the Shane-Wirtanen counts of galaxies prepared by a group of Princeton
astronomers (SeIdner et al. 1977). The limiting magnitude of the Shane-
Wirtanen catalogue is 19m, giant galaxies being visible to a distance
248 MIHKEL JOEVEER AND JAAN EINASTO

of about 1000 Mpc. At this distance a chain of diameter of 100 Mpc has
an apparent diameter 5°. As can be seen from the maps published by
SeIdner et al. (1977), the smallest and most numerous chains have diam-
eters just in this range.

5. MEAN MATTER DENSITY

The total luminosity of the Perseus chain of clusters is 7 x 1013L@

and the total luminosity of the disk of the Perseus supercluster is
6 x 1013L@ (Joeveer, Einasto and Tago 1977). Adopting a mass-to-
luminosity ratio of 200 for the chain (all clusters have elliptical main
galaxies) and 100 for the disk (Einasto et al. 1976), we obtain
2 x 1016M for the total mass of the Perseus supercluster. The super-
@ 2 ·
cluster covers an area of 4000 Mpc. Adopt~ng 15 Mpc for the mean
thickness of the supercluster (in the thickest part around the cluster
A 426 the thickness is about 20 Mpc) , we obtain a volume for the super-
cluster 6 x 10 4 Mpc 3 . If the mean diameter of the cells is 100 Mpc and
the mean thickness of the cell walls is 15 Mpc, then cell walls fill
about half of the total volume of space, the other half being formed by
cell interiors - big holes. Thus the Perseus supercluster represents a
volume of 1.2 x 10 5 Mpc 3 . The mean matter density in this volume
exceeds the critical density by a factor of 3. In the Perseus chain of
clusters the mean density exceeds the critical value by 2 orders of
magnitude.

In the Perseus region the density of clusters of galaxies is higher

than on the average. However, the available data suggest that about
half of all galaxies are located in clusters of galaxies. The density
estimate of Ostriker, Peebles and Yahil (1974) and of Einasto, Kaasik
and Saar (1974), Q = 0.2, is based on Shapiro's (1971) determination of
the mean luminosity density. This density is an underestimate for two
reasons. First, as indicated by Kiang (1976) in calculations of the
effective volume of the sample, galactic absorption has not been taken
into consideration. To correct for this effect, the density has to be
multiplied by a factor of 2. Second, Shapiro omited the Virgo cluster
with the aim of removing the influence of the local density enhancement.
If half of all galaxies are located in clusters, then another factor of
2 must be applied and we obtain Q = 0.8. Similar estimates have
recently been obtained by SeIdner and Peebles (1977b) and Huchra (1977
private communication).

6. DISCUSSION

Summarising the observational evidence, we can say that the avail-

able data show that galaxies are not randomly distributed and do not
form symmetrical, more or less spherical superclusters. Superclusters
may have a form of distorted triangles, as in the case of the Perseus
supercluster and they can be surrounded by chains of clusters of
HAS THE UNIVERSE THE CELL STRUCTURE? 249

galaxies. Disks of superclusters intersect at right angles, forming

walls of cells in the Universe. In cell interiors the density of
galaxies is very small and there we see big holes in the Universe. The
mean diameter of big holes as well as superclusters is ~ 100 Mpc.

Numerical experiments show that cell structure cannot be formed by

random clustering. We believe that this structure is primordial - that
it was formed prior to the formation of galaxies and clusters of galax-
ies in the gaseous phase of the Universe after recombination. A theory
of galaxy formation leading to the formation of cell structure princi-
pally with one- and two-dimensional gas layers has been proposed by
Zeldovich and his group (Zeldovich 1970 and this volume, Sunyaev and
Zeldovich 1970, Doroshkevich et al. 1977). The predicted diameter of
cells is just of the order of ~ 100 Mpc, as observed.

The density estimate obtained is, of course, uncertain, but the

present work suggests that there exist systems with a characteristic
scale of ~ 100 Mpc where the density exceeds the critical value.

Note added after the symposium. Professor de Vaucouleurs drew our

attention to the fact that the dense part of the Perseus chain of
clusters was already discussed by Proctor in the 19th century. The
presence of holes of various diameters was demonstrated during the
symposium by B. Tully and W. G. Tifft. G. Chincarini showed that the
foreground galaxies in the direction of the Hercules supercluster are
located at distinct redshifts which are a natural consequence of the
cell structure. The fact that clusters of galaxies, in particular the
Perseus chain of clusters, form essentially a one-dimensional structure
was also emphasized by W. G. Tifft.

REFERENCES

Abell, G.O., 1958. Astrophys. J. Suppl., 3, 211.

Abell, G.O., 1961. Astr. J., 66, 607. -
Chincarini, G. and Rood, H.J.,~972. Astr. J., ll, 4.
Corwin, H.G., 1974. Astr. J., 79, 1956.
Davis, M., Groth, E.J. and Peebles, P.J.E., 1977. Astrophys. J. Lett.,
212, 1107.
de Vaucouleurs, G., 1956. In Vistas in Astronomy, ed. A. Beer,
Pergamon Press, London, 2, 1584.
de Vaucou1eurs, G., 1975a. Astrophys. J., 202, 319.
de Vaucouleurs, G., 1975b. Astrophys. J., 202,610.
de Vaucouleurs, G., 1976. Astrophys. J., 203, 33.
de Vaucouleurs, G., de Vaucouleurs, A. and Corwin, H.G., 1976. Second
Reference Catalogue of Bright Galaxies, University of Texas Press,
Austin and London.
Doroshkevich, A.G., Saar, E.M. and Shandarin, S.F., 1977. lnst. Appl.
Math., Moscow, Preprint No. 72.
Einasto, J., J6eveer, M., Kaasik, A. and Vennik, J., 1976. Proc.
Third European Astr. Meet., Tbilisi, p. 431.
250 MIHKEL JOEVEER AND JAAN EINASTO

Einasto, J., Kaasik, A. and Saar, E., 1974. Nature, 250, 309.
Faber, S.M. and Dressler, A., 1976. Astrophys. J. Lett., 210, L65.
Fall, S.M. and Jones, B.J.T., 1976. Nature, 262, 457.
Joeveer, M., Einasto, J. and Tago, E., 1977. Tartu Astr. Obs. Preprint
A-1.
Karachentsev, I.D., 1966. Astrofizika, 2, 307.
Kiang, T., 1976. Mon. Not. R. astr. Soc~, 174, 429.
Kiang, T. and Saslaw, W.C., 1969. Mon. Not~. astr. Soc., 143, 129.
Meyer, M.W., 1908. Das Weltgebaude, Leipzig and Wien.
Moss, C. and Dickens, R.J., 1977. Mon. Not. R. astr. Soc., 178, 701.
Ostriker, J.P., Peebles, P.J.E. and Yahil, A., 1974. Astrophys. J.
Lett., 193, L1.
Sandage, A. and Tammann, G.A., 1975. Astrophys. J., 196, 313.
SeIdner, M. and Peebles, P.J.E., 1977a. Astrophys. J., 215, 703.
Seldner, M. and Peebles, P.J.E., 1977b. Astrophys. J. Lett. (in press).
Seldner, M., Siebars, B., Groth, E.J. and Peebles, P.J.E., 1977.
Astr. J., 82, 249.
Shane, C.D. andlYirtanen, C.A., 1967. Lick Obs. Publ., ~, 1.
Shapiro, S.L., 1971. Astr. J., 76, 291.
Sunyaev, R.A. and Zeldovich, Ya.~, 1972. Astr. Astrophys., 20, 189.
Tifft, W.G., Hilsman, A.K. and Corrado, L.C., 1975. AstrophyS: J.,
199, 16.
Zeldovich, Ya.B., 1970. Astr. Astrophys., 5, 84.
Zwicky, F., Wild, P., Herzog, E., Karpowica-;-M. and Kmval, C., 1961-
1968. Catalogue of Galaxies and Clusters of Galaxies, l-~, Cal.
Inst. Techn., Pasadena.

DISCUSSION

Davis: ~1ost of your redshifts are derived from the second reference
catalogue of de Vaucouleurs and since the sky coverage of the catalogue
is quite patchy, one must exercise caution in jUdging the reality of
the holes between superclusters.

Einasto: Some of the distant holes may be due to the absence of data,
of course. But the big hole between us and the Perseus supercluster is
without any doubt real.

Rudnicki: Perhaps your results for the Perseus area need some corection.
As I have shown in my paper on the Perseus cluster published in 1963 in
Acta Astronomica, there,is strong patchy galactic obscuration in this
area.

Einasto: We have studied the effect of galactic obscuration (Joeveer,

Einasto and Tago, Tartu astr. obs. Preprint A-l (1977)). In calculating
luminosities differential obscuration has been taken into account. The
presence of chains of clusters of galaxies is not due to the patchy
obscuration.

Fessenko: Systematical devitations ~n the surface number density of so-

HAS THE UNIVERSE THE CELL STRUCTURE? 251

called clusters of galaxies may reflect contamination of the observation-

al material by zodiacal light. Also you do not take account of other
(local) possible effects due to variability of observing conditions.

Einasto: We have studied the distribution of galaxies along the line of

sight. It is apparent that zodiacal light will not affect the redshift
distribution of galaxies.

Tinsley: Would Drs Turner or Gott like to comment on the discrepancy

between the values of ~ obtained by them and by Dr Einasto?

Gott: Einasto has assumed a mean M/L ~ 100-200. The best way to get
the mean luminosity density is to count all galaxies in some large homo-
geneous region such as the Zwicky catalogue to the 15th magnitude.
Davis, Geller and Huchra found eL ~ 6 x 10 7 LQ Mpc- S , while Gott and
Turner found eL ~ 5 x 10 7 L0 Mpc- 3 from a similar study. These properly
average over superclusters and holes and avoid the contamination of the
local supercluster which makes Shapiro's result 3 times higher. With
the luminosity densities quoted above and Einasto's assumed M/L, one
finds ~ = 0.07 - 0.18.

Einasto: Of course the luminosity density in the vicinity of the Perseus

supercluster is higher than the mean. Our estimate of ~ was based upon
Shapiro's estimate of the luminosity density.

de Vaucouleurs: I would like to make a historical remark. The Perseus

supercluster has been known for over 100 years. It was discovered by
Proctor in 1870 and has since then been noted by many astronomers. So
we are making progress!

Kornberg: Have you tried to check the cell structure studied in your
paper with X-ray surveys and 21 em data? There are now data that show
that superclusters are X-ray sources?

Einasto: This is interesting to do but we have not done it yet.

Huchra: I would like to emphasize once again a point stressed by de

Vaucouleurs that the magnitudes and surface counts of galaxies should be
corrected for absorption and note that it is now possible to so using
the data on neutral hydrogen column densities in the Galaxy of Heiles et
al. plus a value of the gas/dust ratio. In addition, conversion to a
standard magnitude system could help remove or reduce the discordance in
the values of PL and thus ~ derived.

Peebles: The 2l-cm map of Heiles shows very little correlation with the
Lick counts at b > 40 0 • Of course, this is a useful limit only if the
dust follows the gas.
THE PHYSICAL PROPERTIES OF LARGE SCALE SYSTEMS FROM OPTICAL OBSERVATIONS

G. O. Abell
Department of Astronomy
University of California, Los Angeles

Holmberg's (1937) analysis of the distribution of double and

multiple galaxies provided what may have been the first hint of a local
inhomogeneity of greater scale than that of the Local Group. The idea
of a Local Supercluster was subsequently revived by de Vaucouleurs
(1953, 1956, 1958). The analyses of others, as well as the continuing
study of de Vaucouleurs himself (1976 and references cited therein) have
now effectively established the reality of the Local Supercluster.
Several other more remote inhomogeneities, or "clouds" of galaxies, were
described by Shane and Wirtanen (1954). The writer (Abell 1958) found
the distribution of rich clusters to be clumpy, and published a finding
list of several apparent superclusters (Abell 1961).

For the most part, however, our knowledge of large-scale

inhomogeneities in space is based on statistical analyses of the
distribution of galaxies and clusters of galaxies (e.g., Abell 1974;
Kalinkov 1972; Davis et al.1977 and references cited therein). These
statistical investigations suggest that matter is clustered on at least
two different orders, an~poss\bly on a continuum of scales up to 50 to
100 Mpc (for H = 50km s- Mpc-). What we call "superclusters" are
evidently these systems of clusters of galaxies, groups of galaxies, and
possibly individual galaxies, such systems of which have characteristic
diameters of 50 to 100 Mpc. As useful as the statistical approaches are,
however, it is of obvious interest to review what we know about the
structures and dynamics of individual superclusters.

1. THE LOCAL SUPERCLUSTER

Both de Vaucouleurs (1976) and Jones (1976) have carefully

re-examined the Local Supercluster, but largely on the basis of existing
data. Considerable new data have recently been gathered for galaxies
in a large portion of the Local Supercluster by Abell and Eastmond
(Eastmond 1977 and references c~ted therein). From extrafocal Palomar
Schmidt plates covering 158 deg centered on M87, Eastmond has determined
total magnitudes for approximately 3000 E and SO galaxies, complete to
253

M. S. Longair and J Emasto (eds.), The Large Scale Structure of the Universe, 253-262. All Rights Reserved.
Copyright © 1978 by the fA U.
254 G.O.ABELL

0
. TURNER a GOTT

EI*
(GROUPS)

i Z " A 1413 (AUSTIN a PE ACH )

0
~
o A 1930 (AUSTIN a PEACH)
-1.0

o -I -2
m-m*
Fig. 1. Integral luminosity functions for galaxies in nearby groups
and in two rich clusters, fit to that of the Coma cluster (solid line).

the limit m = 16.5, by the method of Abell and Mihalas (1966). In

addition, E~stmond made step-scale magnitude estimates for all galaxies
to the limit m = 14.0 on red priRts of the ~2 Palomar Sky Survey
fields coverin~ the region a = 11 30m to 13 50m and 6 = -2T~o to 26~o.
Eastmond calibrated his step-scale estimates against photometry by
Holmberg (1958) and also against the extrafocal photometry in the
central cluster region, and finds the statistical mean error of a
single magnitude estimate to be 0.3 mag. Finally, Eastmond made DDO
luminosity-class estimates for all Sb and Sc spirals in the 52 fields.

Eastmond then examined the Hubble diagram for the spirals of known
radial velocity and distance moduli obtained from his magnitude and
DDO-type estimates. Although the sample was heavily weighted with
Virgo cluster spirals, there was nevertheless a striking correlation
of velocity with distance. Yet, nearly all of the galaxies observed
are within the Local Supercluster; thus Eastmond's finding suggests
that the supercluster is expanding.

To examine the kinematics of the Local Supercluster more

quantitatively, Eastmond arbitrarily selected, within the large region
surveyed, several small regions where the relatively high surface
density of bright galaxies suggested the likelihood of physical
associations. Following Eastmond's notation, we identify those apparent
groups of galaxies with letter designations. The next goal is to
derive mean velocities and distances for the galaxies within each group.

We estimate the distance of each group from a plot of the integral

luminosity function of its elliptical galaxies. It is now well
established that the elliptical galaxies in rich clusters have a
PHYSICAL PROPERTIES OF LARGE SCALE SYSTEMS FROM OPTICAL OBSERV ATIONS 255

1.0 A
B 6
•
••
-
C
0
*> F D
G 0
...E
- •
H
K +
L D
Z M
C-
O p X

.2

Fig. 2. Integral luminosity functions for elliptical galaxies in 11

Eastmond groups, all fit to that of the Coma cluster (solid line).

characteristic luminosity function. Figure 1 shows the integral

luminosity function of the Coma cluster ellipticals, from Abell (1977),
super-imposed on the luminosity functions for two other clusters
observed by Austin and Peach (1974) and Austin et al (1975). The
luminosity function for all galaxies in nearby groups, as determined by
Turner and Gott (1976), is also shown. Evidently, perhaps fortuitously,
even spirals seem to satisfy the same luminosity function, but to keep
the present sample as pure as possible, we consider only ellipticals
in the groups studied here. We estimate relative distances of the
groups by the horizontal shifts necessary to match their elliptical
galaxy luminosity functions.

Figure 2 is a composite of the luminosity functions for 11 of

Eastmond's groups, prepared by the writer from the individual magnitudes
given by Eastmond. The plot for each group has been shifted vertically
(to take account of differences in richness) and horizontally (to take
account of different distances). The smooth line is the luminosity
function for the Coma cluster. The writer defines a particular point
in the Coma luminosity function to designate a magnitude, m *. The
horizontal shift of the function for each group required toVachieve a
256 G.O.ABELL

TABLE I

<V >
Group a (1950) m * l' 1 n
v (kIn s- )

A 12h20m to 12 h 35m +11°00' to +14°30' 10.0 896 33

B 12 14 12 23 + 5 00 + 7 00 12.1 2274 9
C 11 58 12 06 + 1 00 + 3 30 '\,14.1 5859 7
D 12 41 13 05 - 5 00 -11 00 12.3 3992 7
F 12 44 13 02 -11 00 -16 00 12.5 4192 2
G 12 35 12 59 + 9 00 +14 00 10.1 977 17
H 12 12 12 35 + 7 00 +11 00 10.5 1012 17
K 12 14 12 35 +14 30 +19 00 10.7 942 9
L 13 08 13 22 -15 30 -18 00 12.9 2285 5
M 12 23 12 56 - 1 00 + 5 30 9.8 1081 26
p 13 12 13 29 - 9 30 -15 30 12.6 2576 2

fit to the Coma cluster determines m * for that group. For some of the
poorer groups the data are few, and ~he fits are poorly determined;
yet it is gratifying that the procedure works as well as it does,
despite some uncertainty (up to 0.5 mag) in determining m *.
v
Radial velocity data for many of the Eastmond groups were very
sparse or absent. Consequently, new radial velocities were observed
for 142 galaxies with the Cassegrain scanner of the 3-m telescope at the
Lick Observatory in 1975-76. These new data permitted us to obtain mean
radial velocities for each of the 11 Eastmond groups surveyed. In some
cases, foreground and background galaxies were rather obviously pres~£t
in the field; Those with velocities differing by more than 2000 km s
from the mean were not used in calculating mean group velocities.

The data obtained are summarized in Table I. Successive columns

give the group designation, the range of a and 0 that define each group,
the value adopted for m *, the mean radial velocity for each group, and
the number of galaxies Rsed in computing the mean velocity. The Hubble
diagram for the groups is shown in Figure 3.

* Also shown in Figure 3 is a point representing the Coma cluster

m 14.5; <V >=6952km/s). The straight line has the cosmological slope
o¥ 0.2. WithIn the uncertainties of the observational data, all points
are consistent with a uniform expansion of the Local Supercluster
(containing Groups A, B, G, H, K, L, M, and p), the more remote groups
(C, D, and F), and the Coma cluster. In other words, despite the
apparent reality of the Local Supercluster as a spatial inhomogeneity,
there is no evidence for any local inhomogeneity in the Hubble flow--a
result in agreement with Sandage and Tammann (1974) from their observa-
tions of the Virgo cluster alone. Evidently, the Local Supercluster is
not gravitationally bound; moreover, within uncertainties (perhaps 20%)
it expands as rapidly as the universe in general.
PHYSICAL PROPERTIES Or: LARGE SCALE SYSTEMS FROM OPTICAL OBSERV ATIONS 257

4.0,..--------,-----r-----,r---,-----r----,

N 3.5
(.)

c>
o

3.0

10 12 14
m*V
Fig. 3. Hubble diagram for Eastmond groups and the Coma cluster.

2. THE COMA SUPER CLUSTER

Observations of the distribution of galaxies in the dense core of

the rich regular Coma cluster (e.g., Abell 1977) suggest that at least
some dynamical evolution has taken place and that the cluster appears
to have been stable for at least a good fraction of a Hubble time. Thus
there is no reason to doubt that this and other similar systems are
gravitationally bound. Yet, all galaxies in the field of the cluster
do not share the structure of the dense Coma core. In particular,
Abell has shown that the spirals exhibit almost no central concentration
to the cluster center, even though most of them have radial velocities
that would suggest that they are members of the Coma cluster. It is as
if there were a highly concentrated, negative-energy core of E and SO
galaxies embedded in a cloud of other galaxies, including the spirals
in the same field, which are not gravitationally bound to that core;
that is, the Coma cluster core appears to be a bound concentration
within a larger supercluster, which may not be gravitationally bound.
Tifft and Gregory (1976) arrive at a similar conclusion.

So do Chincarini and Rood (1976), who have obtained spectra of 50

of the 52 galaxies brighter than m =15.1 in Cluster 16 in Zwicky-Herzog
(1963) field 12~' They find that ialaxies in the field have velocities
near_tOOO km s --evidently members of the Local Supercl~fter--near 4000
km s --members of the NGC 4169 group--or near 7000 km s --like that of
the Coma cluster. Chincarini and Rood conclude that most of the objects
in Cluster 16 are members of the Coma supercluster (but up to 14° away
258 G.O.ABELL

1 1 _

• SA 57
• M 13
" SA 68
• A 2670

Z 2.01- -

1.0 I- 1 -
I
20 18 16
my

Fig. 4. Number-magnitude relation for field galaxies in four fields.

from the center of the Coma cluster itself), and that the entire system
includes a semi-relaxed Coma core with dynamical history, which merges
into an outer non-relaxed supercluster expanding with the Hubble flow.
They also suggest as did Abell (1961) that cluster Abell 1367, 41 Mpc
distant, belongs to the same supercluster.

3. GROUPS OF RICH CLUSTERS

Rood (1976) has attempted to delineate physical groups of clusters

in the Abell (1958) catalog by imposing the criterion that the space
density of clusters in such a group must be at least 100 times that of
the clusters in general. With this criterion he identifies 5 definite
groups of clusters belonging to Abell distance classes 0 to 2 and 39
probable groups in distance classes 3 and 4. In fact, 41% of the
clusters in distance classes 0 to 2 are in such groups. Because radial
velocities are available for all distance class 0 to 2 clusters, Rood
was able to estimate linear separations of the clusters in each group
by assuming that the groups expand with the normal Hubble flow. To
test this assumption, Rood notes that if the clusters within a group
are oriented at random, their mean radial separation, <R>, should be
related to their mean transverse separation, <T> by <R>/<T>=2/n=0.64.
For the 11 clusters of distance classes 0 to 2 in the 5 groups, the
observed value is <R>/<T> = 0.53 ~ 0.20. In short, the Rood analysis,
while hardly definitive, is consistent with the picture that clusters
tend to group in superclusters, but that the superclusters are expanding
with the universe.

4. IS THERE AN END TO THE HIERARCHY?

The statistical studies of Peebles and others indicate correlations

PHYSICAL PROPERTIFS OF LARGE SCALE SYSTEMS FROM OPTICAL OBSFRV ATIONS 259

-E >
~,

20 18 16

Fig. 5. Observed number-magnitude relation for field galaxies (dots),

compared to two different superclustering models.

in the positions of galaxies and clusters over a distance of up to

100 Mpc, but are not conclusive over larger scales. Other studies,
however, suggest that the correlations do not extend to very much
greater distances, and that, in partiCUlar, a hierarchical universe
probably does not exist.

One such study is that of Webster (1976), who concludes that faint
radio sources are distributed with remarkable isotropy. If most of
those sources are radio galaxies, Webster's analysis rules out spatial
density fluctuations of as much as 10% over a scale of 1000 Mpc; the
study is, in other words, entirely consistent with the existence of
superclusters of the sort discussed here, but not with very much
larger inhomogeneities. Similarly, the observed isotropy of the
microwave background, if the usual interpretation is assumed correct,
rules out an indefinite hierarchy of clustering.

Direct observations of the isotropy of optical sources are provided

by Rainey's (1977) counts of galaxies to various limiting magnitudes.
Rainey made counts of galaxies as a function of magnitude to the limit
m =19.5 in three widely separated fields (around Sele 2ted Areas 57 and 68
agd in the field of Ml3), each of approximately 1 deg. Rainey's counts
for his three fields are superimposed in Figure 4. His data are
supplemen~ed with counts by Mottmann and Abell (1977) of galaxies in a
0.226 deg field near cluster A2670. Counts by Brown (1976) to several
limiting magnitudes are also highly consistent with those shown in
Figure 4.
260 G.O.ABELL

The agreement of the number-magnitude relation of galaxies in

widely separated directions in the sky suggests a remarkable isotropy
in the distribution of faint galaxies, but it remains to be shown that
it rules out large inhomogeneities in the galaxy distribution. To this
end, Rainey has calculated theoretical number-magnitude relations for
several models of galaxy distribution. Figure 5 shows the observed
(composite) distribution compared with that for two models of large-scale
superclustering, both of which assume typical Friedmann cosmologies, and
a galaxian luminosity function like that of Figures 1 and 2. In each
superclustering model, the galaxies are presumed to be distributed
roughly uniformly in systems with the diameters indicated, with similar
distances separating superclusters. According to Figure 5, even
inhomogeneities of size 300 Mpc should result in easily observable
distortions of the observed number-magnitude relation. These data and
calculations suggest that inhomogeneities in the universe much larger
than 100 Mpc probably do not exist.

REFERENCES

Abell, G. 0.: 1958, Astrophys. J. Suppl. 3, pp.211-288.

Abell, G. 0.: 1961, Astron. J. 66, pp.607-613.
Abell, G. 0.: 1974, IAU Symp.63, pp.79-92.
Abell, G. 0.: 1977, Astrophys. J. 213, pp.327-344.
Abell, G. O. and Mihalas, D. M.: 1966, Astron. J. 71, pp.635-641.
Austin, T. B., Godwin, J. G., and Peach, J. V.: 1975, Monthly Notices
Roy. Astron. Soc. 171, pp.135-142.
Austin, T. B. and Peach, J. V.: 1974, Monthly Notices Roy. Astron. Soc.
168, pp.591-602.
Brown, G. S.: 1976, University of Texas, Austin (Ph.D. Thesis).
Chincarini, G. and Rood, H. J.: 1976, Astrophys. J. 206, pp.30-37.
Davis, M., Groth, E. J., and Peebles, P. J. E.: 1977, Astrophys. J.
(Letters) 212, pp.LI07-Llll.
de Vaucouleurs, G.: 1953, Astron. J. 58, pp.30-32.
de Vaucouleurs, G.: 1956, Vistas in Astronomy 2, pp.1584-1606.
de Vaucouleurs, G.: 1958, Astron. J. 63, pp.253-266.
de Vaucouleurs, G.: 1976, Astrophys. J. 205, pp.13-28.
Eastmond, T. S.: 1977, University of Calif., Los Angeles (Ph.D. Thesis).
Holmberg, E.: 1937, Ann. abs. Lund, No.6.
Holmberg, E.: 1958, Medd. Lunds, Ser. 2 13, No. 128.
Jones, B. T.: 1976, Monthly Notices Roy. Astron. Soc. 174, pp. 429-447.
Kalinkov, M.: 1972, Proc. 1st Eur. Astron. Meeting, Athens, 3, pp.142-161.
Mottmann, J. and Abell, G. 0.: 1977, Astrophys. J. (in press).
Rainey, G. W.: 1977, University of Calif., Los Angeles (Ph.D. Thesis).
Rood, H. J.: 1976, Astrophys. J. 207, pp.16-24.
Sandage, A. R. and Tammann, G. A.: 1974, Astrophys. J. 194, pp.559-568.
Shane, C. D. and Wirtanen, C. A.: 1954, Astron. J. 59, pp.285-304.
Tifft, W. G. and Gregory, S. A.: Astrophys. J. 205, pp.696-708.
Turner, E. L. and Gott, J. R.: 1976, Astrophys. J. 209, pp.6-11.
Webster, A.: 1976, Monthly Notices Roy. Astron. Soc. 175, pp.71-83.
Zwicky, F. and Herzog, E.: 1963, Catalogue of Galaxies and Clusters of
Galaxies Vol. II, California Institute of Technology, Pasadena.
PHYSICAL PROPERTIES OF LARGE SCALE SYSTEMS FROM OPTICAL OBSERVATIONS 261

DISCUSSION

Chincarini: Since George Abell referred to my preliminary work on the

various contributions to the cosmic luminosity, we have the following
results, based on Demler's total cluster luminosity determination and
estimates for groups.

Object L/L0 Total Notes

luminosity
(Units 10 12 ) L0Mpc- 3

Cluster Richness 5 17 1. 78 x 10 4
4 (11. 75) 7.24 x 10 4
3 6.85 4.79 x 10 5
2 5.96 2.4 x 10 6
1 3.28 4.2 x 10 6
0 > 2.45 > 2.5 x 10 6 uncertain
Groups outside the 3 x 10 6
~ 0.2 > very uncertain
Local Supercluster
Groups and galaxies ~
0.2 '" 3.3 x 10 7 very uncertain
in Superclusters
~

,
Adding all contributions and, depending on the relative weighting of the
contributions, we find 1.5 x 10 7 ~ Lcosmic ~ 4.6 x 10 7 L0 Mpc- 3 .

Abell: In connection with the Universal mass density, there are about
4000 great clusters (richness 1 or greater) within z = 0.2. If cluster
masses are typically 4 x 10 15 M0 , all great clusters contribute only
n = 0.004. To have n = 1, there must therefore be 250 times as much
matter outside of the great clusters as within. Hubble, Minkowski, and
others have estimated that about one tenth of the visible galaxies are
in great clusters; Abell has made a similar estimate. The numbers are,
of course, highly uncertain, but it seems unlikely that the number of
non-cluster galaxies can be high enough to make n much greater than 0.1.

Ostriker: A word on your calculation of n. The luminosity and mass in

great clusters can be made larger and larger as one defines their radius
to be larger and larger. Correspondingly, the fraction of the cosmic
luminosity (and mass) in clusters is larger. But the total cosmic light
(or mass) density calculated should be invariant if one is careful to be
consistent in the two calculations.

Abell: Of course, I fully agree.

Jaakkola: What has happened to your earlier observations? At the

Uppsala symposium you presented a diagram in which the Virgo cluster fell
262 G.O.ABELL

distinctly above the mean Hubble line and later Dr Gudhus has obtained
similar results. Now you have obtained a Hubble relation in which the
Virgo cluster contradiction has been removed.

Abell: At Uppsala I was using the published mean radial velocity for
all galaxies in the Virgo region. In the diagram presented here, I use
the mean velocity of the elliptical and SO galaxies within 3 0 of M87;
that velocity is about 1000 km s-l For that tight group of galaxies
there is no discrepancy with the mean Hubble line.

Fall: From your Hubble diagram for groups within the local supercluster,
what would you say is the maximum allowable deceleration with respect to
the centre of the Virgo cluster?

Abell: I certainly could not rule out a local perturbation of 20%, but
do not think it could be as high as, say, 50%

Silk: If you were to use distance indicators suggested by other workers

(Sandage - Tammann etc.), what is the corresponding spread in the lumin-
osity function that you have derived?

Abell: For the great and distant clusters (Coma and beyond) the differ-
ence in modulus between that found from the brightest galaxy, m1, and
from the luminosity-function fitting, m*, can be as great as a full
magnitude. The dispersion in m*-m1 is about 0.3 mag. However, for
elliptical galaxies in groups, studied here in the Virgo region, we
could not use m1 because most of the groups have small numbers of ellip-
ticals and they are nearly always contaminated by foreground and back-
ground galaxies.

Peebles: From the cross correlation of Lick counts with Abell clusters
M. Seldner and I find that the mean number of galaxies, in excess of
random, at 1.5 < hr < 30 Mpc from an Abell cluster centre, is ~20 times
the number at r < 1.5 h- I Mpc. Taking ~2 Abell clusters per super-
cluster on this scale one finds ~10 times as many galaxies in the super-
cluster as in the great clusters, in agreement with Abell's estimate.

Abell: That is encouraging.

THE STRUCTURE OF THE HERCULES SUPERCLUSTER

M. Tarenghi, LFCTR, Istituto di Fisica, Milano

w. G. Tifft, Steward Observatory, University of Arizona
G. Chincarini, Dept. of Physics and Astronomy, University
of Oklahoma and Universita di Bologna
H. J. Rood, Dept. of Astronomy, Michigan State University
L. A. Thompson, Dept. of Astronomy, University of Nebraska

Counts by Shane and Wirtanen detected a cloud of Galaxies - the

Hercules Supercluster - centered at about a = l6 h 02m; 0 = l7000~ It is
unclear whether there is a physical connection with another cloud
located to the south for which no radial velocities are yet available
and with the system A2l97/A2l99 located to the north. The latter is at
slightly smaller redshift (velocities between 8000-11,000 km sec-I) than
the Hercules cluster.

The Hercules Supercluster contains 3 Abell clusters: A2l47, F;

A2l5l, F: III; and A2l52 I : III. In A2l47 a giant elliptical g2laxy is
located at a = l6hOO~0; 0 = 16 0 :06' and the centroid of the X-ray source
at a = l6hOO~4; 0 = 16 0 25'.2 - the cluster A2l52 is of distance class 4.
Preliminary estimates of the redshifts of the galaxies and of their
morphology show that:

1. Velocities are sharply segregated into three ranges of redshift

around 2500 km s-l (local supercluster); 4700 km s-l (group) and 9000-
13,000 km s-1 (Hercules supercluster). This confirms the clumpy distri-
bution of galaxies and the low density, if any, of homogeneously distri-
buted "field" galaxies.

2. In the lresent sample various galaxies have velocities larger than

11,000 km s- (the limit of the Burbidge and Burbidge sample). The
possibility of two systems of redshifts is unlikely and cannot be suppor-
ted by the present results. The broader range of redshifts results,
therefore, in a rather high velocity dispersion (see Table 1, preliminary
analysis).

3. A2l47 and A2l52 have very similar morphology. In A2l5l there is a

smaller content of elliptical galaxies as shown ~n Table 2.
Table 2
Cluster S/E S/SO SOlE
A2l47 2 1.45 1. 38
A2l52 2 2.67 .75
A2l5l 4.6 1.77 2.6
263

M ~ 1 nnvmr find.l Finasto reds.) The [,arye S{'oie SfnJrfllrp (If thp TTnillPr(!p ?f) ~-?Ii), All Ri/!hts Reserved.
264 M. TARENGHI ET AL.

Table 1
Preliminary vi rial parameters for the Abell Clusters embedded in the
Hercules Supercluster. The separation of the clusters into velocity
groups is, at present, merely indicative and no physical significance
should be attached to it.

z. PARAMETER

R.A,),.50
A 2152

16· 31 32
vs
.. 2152

16: 3: 00
10000 V
A 2152
:>

16: 3:.tO
11000
A 2141

15: 59: 35
V'S 11000
Ito. 2147

IS: 59:42
V ~
A 2147

1600-00
11500
A 2\51
v .. 11500
A 2151

16·03; 00 1603 14
v ~ 12000
A 2151

16:03:04

0'1950
Vel
16°

,_
10526
3S'
....
16° 40'

"...
16: 26

"'.,
16°
1097'
Ie 16° 11'
10209
16·26
12221
"...,
," 57'
"368
17 57'
lorn
17: 51
12521

Vem '2.'34' 1_ 10282 l11M9

I.'" 12458
Ryjr
V,
3.49Mpc::
1965 -
'.2
81.
3.83
1937
4,01
B79
'.09
1030
'.'
1741
2.'
I039( .)
253
7',
6.51011 - 4.0110 11 1.791012 9.610'1 8.51011 2.21012 1.2310'2 , 1 1011
L("""
3.11015 - 8.1 1014 3.34 1015 7.21014 7.61014 2."'015 7.3 10'4 3.310'4

"'"
4. The P.A. distribution of the galaxies is somewhat peculiar for the
region containing the three clusters. A non-random distribution seems
to be present in A2l47, with a peak between P.A. 150 0 and 180 0 •

5. Indicative virial solutions are given in Table 1. The clusters may

be unstable and the supercluster, or its core region, may go through a
formation process similar to White's theoretical model.

6. So far we have no evidence for the likely mechanism of X-ray

emission. The X-ray source is centered on A2l47. Two wide angle tail
galaxies are in A2l5l. The differences in velocity dispersion among the
clusters are not very significant. The main difference in content is
not in the percentage of SO galaxies but in the number of ellipticals
(see Table 2).

In conclusion, the redshifts obtained in the Hercules Region confirm

the previous findings in the Coma Supercluster, indicating that galaxies
are grouped in large asymmetric structures, groups and superclusters.
Clusters are, generally, bound condensations embedded in superclusters.
The gaps in the velocity field are a very important feature and can be
used to estimate an upper limit for the density of "field" galaxies. The
existence of such galaxies is however uncertain.

REFERENCES

Burbidge, G.R. and Burbidge, E.M., 1959. Astrophys. J., 130, 629.
Chincarini, G., 1976. Frascati (Italy) Workshop on galaxies, May 1976,
Mem. S.A.I.
Chincarini, G. and Martin, D., 1974. Astrophys. J., 196, 335.
Chincarini, G. and Rood, H.J., 1975. Nature, 257, 29~
Chincarini, G. and Rood, H.J., 1976. AstrophyS:-J., 206, 30.
Cook, B.A., Maccacaro,T., Perola, G.C., Tarenghi, M., and Valentijn,E.A.,
1977. Astro. Astrophys., 58, L17.
Tarenghi, M., 1976. Read at I.~U. Colloquium No.37, Paris.
Tifft, W.G., and Gregory, S.A., 1976. Astrophys. J., 205, 696.
THE STRUCTURE OF THE HERCULES SUPERCLUSTER 265

DISCUSSION

Ozernoy: Did you try to construct kinematic models for the velocity
distribution in order to establish whether the Hercules supercluster
expands isotropically or preferentially in only one direction?

Chinaarini: I did look into the problem, of course. However, we are

still somewhat limited by the observations. We should go to a fainter
limiting magnitude to see better the Hubble flow. In fact, what we must
find is a correlation between magnitude and redshift in the supercluster.
Such a correlation is weak both in the Coma and Hercules systems, due to
the small range of magnitude observed. In Hercules, however, some
correlation may be present in spiral galaxies; the diagram for EISa
galaxies shows more scatter. Perhaps the best way to detect the effect
is to select the 5 or 10 brightest galaxies and use these to look for
evidence of the Hubble flow. This has not been done.
I did also consider the behaviour of the three clusters A2l47,
2151, 2152 in the supercluster. The three point masses have a velocity
dispersion of about 550 km s-l and a reasonable virial mass. On the
other hand, once errors are taken into account, we also have agreement
with the Hubble flow.
These tests will have higher significance as soon as the detailed
analysis of our observations is completed. We must go to fainter
magnitudes. I should add that we have a confusion problem due to the
fact that, having a magnitude-limited sample, we sample the luminosity
function to different absolute magnitudes on the front and back sides of
the supercluster.
OBSERVATIONS OF THE LARGE SCALE DISTRIBUTION OF GALAXIES

W. G. Tifft
Steward Observatory, University of Arizona, Tucson, Arizona
and
S. A. Gregory
Physics Dept., Bowling Green Univ., Bowling Green, Ohio

It has long been assumed that clusters of galaxies merge into a

smooth "field" of isolated galaxies. Early estimates put 50% of all
galaxies into the field. Using new red shift surveys over large areas
we find the real distribution of galaxies to be very different. Nearly
all galaxies are in clusters or groups covering a range of at least
1000 in mass and richness.

The first large scale survey was of a 60 radius field centered on

the Coma cluster (Tifft and Gregory 1976). This was extended by Gregory
and Thompson (1978) to cover a 250 square degree region bridging the
Coma and A1367 clusters. A1367 has properties similar to Coma (Tifft
1978). The extended survey shows that the two rich clusters lie in a
supercluster containing at least four lesser groups and some scattered
"isolated" galaxies. The foreground shows no galaxies which cannot be
associated with groups, There are regions more than 20 Mpc in radius
which are totally devoid of galaxies.

New observations of galaxies down to photographic magnitude 14.0

in a 500 square degree region surrounding the Perseus, A396, A347 , A262 ,
NGC507, and NGC383 clusters confirm the results of the Coma surveys.
The groups form an enlongated supercluster (see figures) as originally
suggested by Tifft, Hilsman, and Corrado (1975). The foreground is
again very clumpy with one major void of radius close to 40 Mpc.

Galaxies in the Coma, Perseus, and Local superclusters which are

not definite members of groups can be accounted for as 1) members of
unidentified small groups, 2) remnants of loose clouds which may have
been tidally disrupted, or 3) a small population which may have escaped
from clusters. We conclude that all galaxies were probably formed in
clusters. In a classical picture this implies that galaxies fragmented
after clusters were formed. The observations are also consistent with
galaxy formation by fragmentation and slow expansion of dense cores at
the centers of clusters.

SAG acknowledges support from NSF grant AST 74-22597.

267
M. S. Longair and J. Einasto (eds.). The Large Scale Structure of the Universe. 267-269. All Rights Reserved.
Copyright © 1978 by the IAU.
268 W. G. TIFFT AND S. A. GREGORY

~~S40'*J2
0)500"''4''4100
. . . I00< .... cl300
&,.>1000

".
RAI'-<s)

• V"ZOOOI(JnS-'
• VcIlOOK",S-'

,
RIIIo{hoonl

Figure 1. Redshift and spatial distributions (divided at Vo =

3000) of galaxies in the Perseus supercluster. Wedge angle is
taken along the line of clusters.
REFERENCES

Gregory, S. A. and Thompson, L. A. 1978, Ap.J. (in final preparation).

Tifft, W. G. 1978, Ap.J. (in press).
Tifft, W. G. and Gregory, S. A. 1976, Ap.J. 205, pp 696-708.
Tifft, w. G., Hilsman, K. A., and Corrado, L. C. 1975, Ap.J. 199, PR16-18.

DISCUSSION

Chincarini: I would like to emphasize that the gaps we observe in the

velocity field give an upper limit to the number of possible "field"
galaxies. A preliminary computation based on the velocity field in a
magnitude-limited sample gives
dN«m) < 1.0 x 10-8 10'
0 6m
drl
at about 2-30 level. Such a limit will be improved by more detailed
computations.

Ozernoy: I am worried about your conclusion that the lack of bright

galaxies in the field is evidence in favour of their origin in clusters.
In fact, in the collapse of protoc1usters, which is highly anisotropic,
only part of the collapsing matter forms a gravitationally bound system.
A significant part of the contracting protoc1usters, after the formation
of galaxies, may re-expand and become field galaxies.
OBSERV ATiONS OF THE LARGE SCALE DISTRIBUTION OF GALAXIES 269

Tifft: The observed low limits on the galaxy density outside clusters
or groups should provide important constraints on possible theories of
the origin of clusters and galaxies. I also suggest that numerical N-
body collapse studies consider formulating some indices to describe the
number of isolated galaxies left over, and the shapes of structures
formed, since these quantitites can be critically compared with the
observations. The general statistical analysis of the surface distribu-
tions of galaxies is totally insensitive to small variations in isolated
galaxy or small group populations. One of the most striking differences
between the N-body collapse models with different n appears to be the
number of "field" or isolated galaxies left over after the main collapse.

Silk: The apparently sharp boundaries and holes over large scales that
are being inferred may partly be a function of the nature of the
magnitude-limited sample. At the distance of the Coma Cluster, one is
barely at the kneee in the galaxy luminosity function. Many fainter
galaxies could be present, and it is possible that the more luminous
galaxies are only found in dense regions.

Tifft: We have a partial test of this hypothesis by looking at groups

and clusters as a function of distance. In the most nearby groups, we
can investigate the distribution of galaxies to very much fainter
absolute magnitudes than in more distant groups but there is not deteri-
oration in the definition of these nearby groups. Therefore, we do not
expect the distribution of faint galaxies to be very different from that
of bright galaxies.

Fall: Do you have a numerical estimate of the density excess in the

bridge connecting Coma and A l367?

Tifft: It is premature to give any good estimate since the sample is

incomplete. In particular the local Coma I cloud sits centrally in the
field and makes it very difficult to estimate the number of slightly
fainter galaxies belonging to the bridge population.

Peebles: If the groups and clusters tended to contract it would make

them appear narrower in redshift distribution than would be expected
under the assumption of pure Hubble expansion. Conceivably this is a
partial explanation for the apparent sharp boundaries of the clouds of
galaxies.

Tifft: Both the Coma and Perseus superclusters are quite extended
spatially, a good fraction of a radian, but show a "constant" redshift
along the entire length and very little dispersion. They therefore have
negligible "depth" compared to width. They are "lines" or "sheets"
perpendicular to the line of sight! It is true that radial collapse
could offset part of the differential Hubble flow seen through the object
but the collapse velocity could not easily match the Hubble flow at all
points so I would still expect much more spread in "velocity depth" than
is seen. I think that if we find many more cases of extended "thin"
discs or linear structures all aligned perpendicular to the line of sight
we will have a good case for an intrinsic redshift.
THE FIELD LUMINOSITY FUNCTION AND NEARBY GROUPS OF GALAXIES

John Huchra
Center for Astrophysics
Cambridge, Massachusetts

We have assembled a catalog of radial velocities and magnitudes on

a homogeneous system (the corrected Harvard, B(o) magnitudes of
de Vaucouleurs) for over 4000 galaxies. Usinq this catalog, we have
compiled a magnitude limited sample of ~ 1000 galaxies with nearly
complete radial velocity data. The magnitude limit is 13.0 and the
galaxies are primarily from the Shapley-Ames catalog plus a few low and
high surface brightness objects properly included in a magnitude limited
sample. My colleagues, M. Davis, M. Geller and I are presently analyzing
the dynamical properties of this sample. I would like to briefly
describe a new determination of the field luminosity function and
density plus our initial experiments with the use of a redshift catalog
to select groups of galaxies.

The luminosity function ~(M) dM can be derived from a magnitude

limited sample simply by counting the number of galaxies in an absolute
magnitude interval and dividing by the sampled volume (e.g., Schecter,
1976, and references therein). This method avoids the special assumption~
other than homogeneity, required in other determinations (e.g., Holmburg,
1969, Turner and Gott, 1976b). The major improvement in our estimate is
the use of a larger sampling volume, thus 4 times as many galaxies as
most previous estimates. Figure 1 shows our separ~te de!erminations o
for the North and South galactic caps Ho = 50 KIn s 1 Mpc 1, b II 1>40 , I
and galactic absorption equal to 0.2 csc (b I !). Even in this sample,
below M = -16 there are relatively few galaxies and the sampling radius
indicates that we are dominated by the Local group. In addition, the
difference between Nand S indicates that we are not yet free of the
local density enhancement. A fit of the Schecter (1976) Function to the
combined N + S sample gives a = -1.24, L* = 2.9XI0 10 or M* = -20.67,
~* = 0.0055 and a corresponding luminosity density of 1.8XI0 8 ~ Mpc- 3 •
If we assume that the shape has been derived accurately and then correct
for the density enhancement by scaling the surface counts to the
expected values from deeper samples -- e.g., the E + 2S sample of Gatt
and Turner (1976) -- ~* and the luminosity density are ~ of the above
values. A more detailed study of the luminosity function is in
preparation.
271

M. S. Longair and J. Einasto reds.), The Large Scale Structure of the Universe, 271-273. All Rights Reserved.
Copyright © 1978 by the fA U.
272 JOHN HUCHRA

(.)

(.)
-2 1
..... : .. .
.1 •• ••

-6 (.)

-14 -16 -18 -20 -22

•
Fig. 1. Log ~ in galaxies per cubic megaparsec per magnitude interval.
NGC = circles, SGC = triangles. Sample depth v megaparsecs lS shown
along the bottom.

Previous identifications of groups of galaxies have been based

either on limited and subJective data or only two dimensional criteria
(de Vaucouleurs, 1976; Turner and Gott, 1976a). M. Gel~er and I are
experimenting with algorithms for the selection of groups of galaxies
using a redshift catalog. Three problems are important. Selection
criteria should be commutative, should not impose arbitrary limits on
group size or velocity dispersion, and should attempt to separate
groups superposed on the sk~. One algorlthM is,based on the projected
redshlft separation S = (vI +v 2 2 -2V I V cos 8 1 , )2/50. Another, suggested
by p. Schechter, is based on luminosity denslty enhancements. The
simplest lS based on "boxes" In redshift space c1eflned by maXlmum velocity
and projected separation parameters - 6v 12 <DVmax ' 6D12 = 100 tan 8 12
/(v l +v 2 ) < 6Dmax ' Figure 2 shows a typlcal group catalog map.

GROUPS
Fig. 2. Groups of galaxies,
o
0" 2. Mpc
v- IOOkm a-I
n >, 4 members, selected by
24'_-~--~-'l.l"'~>"--~---O"----;O' box algorithm.

Groups are found by assoclatlon -- a galaxy is searched for companions

inside 6D,6V, then the companions are searched for additlonal companions,
etc. Our prellminary analysis shows that certain group structures (eog.,
M8l, Fornax, Centaurus, NGC 1023 groups) persist over almost all
reasonable ranges of selection parameters. Subclustering is found in
Vlrgo for small values of 6D,6v.

de Vaucouleurs, G. 1976, in Stars and Stellar Systems IX, "Galaxies

and the Universe", eds. Sandage, Sandage and Kristian, U. of Chic.
Press.
Gott, J.R. and Turner, E. 1976, ApJ 209, ~.

Holmberg, E. 1969, Arkiv for Astronomi 2' 30".

Schecter, P. 1976, ApJ 203, 297.
Turner, Eo and Gott, J.l{o 1976a, ApJ. SU12P. E, 409.
Turner, E. and Gott, o .l{. 1976b, ApJ 209, 6.
THE FIELD LUMINOSITY FUNCTION AND NEARBY GROUPS OF GALAXIES 273

DISCUSSION

Gursky: To my eye there were many de Vaucouleurs groups that were not
found in your computer searches. What is the reason for these discre-
pancies?

Huchra: There are several reasons for this. The plots do not show
groups of fewer than 4 members and we are working with a magnitude-
limited sample - thus we miss faint members in some groups. Also, some
de Vaucouleurs groups may not be real. In fact, I am impressed with the
amount of correspondence.

de VaucouZeurs: May I remind you again that in searching for physical

groups one must also take into consideration morphological types, degree
of resolution, magnitudes, diameters, evidence for interaction, etc.
The current computer methods are too crude to find all types of groups.

Huchra: The reasons for "experimenting" as we are, are that when deeper
samples are used, the only available data will be magnitudes and red-
shifts, so objective methods need to be developed.

Ostriker: A comment on selection by velocity. If the true velocity

dispersion is, say, Gaussian, the <~v2> moment will have a large contri-
bution from the small number of objects in the Maxwellian tail. If
these are thrown out simply because their contribution seems too large,
then one will systematically underestimate <~v2>. You do not make this
error, but other investigators have done it, thus partially explaining
the differences between your results and theirs.

Davis: Could you comment on the experiments you and Geller have made
concerning the effect of ~v on the measured velocity dispersions within
the observed groups?

Huchra: We find that the mean number-weighted velocity dispersion of

the groups found with reasonable values of ~v asymptotically approach
z 300 km s-1 even for very large values of ~v. We have not, however,
completely analysed all the selection effects that may be important.

Fall: How do your groups compare with Turner-Gott groups?

Huchra: I have not compared these catalogues at different magnitude

limits yet.

Abell: Hhat kind of magnitudes did you use?

Huchra: I have used B(O) magnitudes - this system is the easiest to

measure and can be easily extended to fainter galaxy catalogues because
of its close correspondence to Zwicky magnitudes. ~1y preference is for
galaxy magnitudes that can easily be measured and require little extra-
polation.
274 SHORT COMMUNICATIONS

MARKARIAN GALAXIES IN THE VICINITY OF THE COMA CLUSTER

M. A. Arakelian

The distribution of Markarian galaxies in annular zones at differ-

ent angular distances 9 (degrees) from the centre of the Coma cluster
is shown in the Table. The coordinates of the centre for the equinox
1950 are assumed to be a = l2h57~5, 0 = 28 0 15'. The third, fourth and
fifth columns of the Table contain the number of Markarian galaxies in
each zone, the solid angle in square degrees and their mean surface
densities respectively. It can be seen that an excess surface density
of Markarian galaxies exists not only in the very central zone I but in
the annular zone II as well. Indeed, the ratio n/w in zone II is more
than twice the value corresponding to the whole area surveyed (and to
zones III and IV).

Zone 9 n w n/w n' n" <9> n"/n'

I 0 1.5 7 7 1.00 6 2 0.8 0.33

II 1.5- 6.0 22 106 0.21 15 10 4.8 0.67
III 6.0 - 12.0 38 338 0.11 31 5 8.6 0.16
IV 12.0 0.10 570 71 0.12

The distribution of redshifts of Markarian galaxies indicates that

there is an excess of these galaxies round the centre of the Coma
cluster. The sixth column of the Table contains the numbers of galaxies
with known redshifts and the seventh gives the numbers of galaxies with
redshifts in the interval of 0.022 - 0.025. The eighth column contains
the mean 9 of galaxies of the seventh column. One can see from the
ninth column that the relative number of galaxies having redshifts
between 0.024 (1 - sin 6 0 ) and 0.024 (1 + sin 60 ) in zone II is five
times higher than that of the galaxies in zone IV (0.024 is the mean
redshift of galaxies of the Coma cluster).

The luminosities of eight out of the above ten galaxies with red-
shifts close to 0.024 are higher than ~ = 20~0 for H = 50 km s-1 Mpc- 1 •
Thus the density of luminous Markarian galaxies in the vicinity of the
Coma cluster is nearly an order higher than their mean space density.

DISCUSSION

Huahra: My result for the Markarian luminosity function at the bright

end agrees well with those of Dr Arakelian and I suggest that this may
be due to an increased percentage of the Seyfert galaxies.
SHORT COMMUNICATIONS 275

Arakelian: Seyfert galaxies do not show a concentration towards the

Coma cluster but the statistics are very small.

REMARKS ON THE ANGULAR DISTRIBUTION OF MARKARIAN GALAXIES

Piotr Flin and Harek Urbanik

Lists I-IX of Harkarian galaxies have been used as the basic

observational material. An attempt has been made to form a statistical-
ly complete sample. Two regions (region I 6~8 < a < 14~2, 41 0 < 0 < 77 0 ,
region II 6~8 < a < 13~9, 50 < 0 < 32 0 ) which appeared to be roughly
complete have been carefully examined. After preliminary tests the
first two hours in R.A. of the region II were removed from the analysis.
The method of statistical reduction (Flin et al. 1974, A. Zieba 1975) and
the maximum likelihood estimate of the number-magnitude counts (Crawford
et al. 1970) were applied as a combined test in their practical version
(Urbanik 1976,1977). We have examined the whole region, as well as a
number of smaller SUb-regions varying in size and pos1t10n. The angular
sizes investigated ranged from 1 0 up to 30 0 . Our analysis of 15 galaxy
sets led to the following conclusions.

A much higher tendency toward clustering in region II in comparison

with region I is accompanied by a significant decrease of surface density
and flattening of number-magnitude counts. This effect varies across
region II. This situation remains unchanged after restricting the
analysis to galaxies brighter than l6~ It has been found that multiple
re-surveying of some small fields causes a sharp increase of density
within them. These facts resemble the behaviour of radio source counts
when investigations are carried out below the completeness limit of the
radio survey. Therefore we infer that the completeness of the region II
is considerably worse than that of the region I. Moreover the incomplete-
ness is independent of the strength of the UV. The properties of the
distribution of galaxies in region I do not differ qualitatively from
those obtained for fainter normal galaxies (S. Zieba 1977). This simi-
larity disappears in region II. In the light of these conclusions it
seems that any large-scale statistical analysis of Markarian galaxies
requires improvement in the completeness of the observational data.

REFERENCES

Crawford, D.F., Jauncey, D.L. and Murdoch, M.S., 1970. Astroph. J.,
162, 405.
Flin,~, Machalski, J., Maslowski, J., Urbanik, M., Zieba, A., Zieba,
S., 1974. In "Confrontation of Cosmological Theories with
Observational Data" (ed. M. S. Longair), p.121 (D. Reidel,
Dordrecht) •
Urbanik, M., 1976. Acta Cosmologica, 2, 97.
Urbanik, M., 1977. Thesis, Jagiellonian University.
Zieba, A., 1975. Acta Cosmologica, 3, 75.
Zieba, S., 1977. Rozprawy habilitacyjne U.J., No.2, Krakow •.
276 SHORT COMMUNICATIONS

DISCUSSION

Huchra: Can you estimate a magnitude limit for the completeness of the
Markarian Surveys?

Urbanik: We estimate the limiting magnitude for completeness in region

II to be about 14m or even worse, and about 15m in region I.

Heidmann: Cannot these limiting magnitudes for completeness be simply

obtained from the log (number) versus magnitude counts?

Urbanik: The limiting magnitude for completeness can, of course, be

obtained from such counts. We did not make counts, because we were not
specially interested in a knowledge of limiting magnitude for each
region individually. We were interested in the mutual relations between
changes of the tendency toward clustering, average surface density and
the slope of the mean number-magnitude counts.

SUPERCLUSTERING OF GALAXIES
M. Kalinkov, V. Dermendjiev, B. Staikov
I. Kaneva, B. Tomov, and K. Stavrev

A new type of catalogue of extragalactic objects is nearing comple-

tion in our department. The catalogue is on magnetic tape and is a com-
pilation of data for galaxies and cluster of galaxies, together with
references, and for counts of galaxies. At present this Metacatalogue
contains about 3 x 10 5 entries - Abell and Zwicky clusters, Zwicky
galaxies, Lick counts, Jagellonian counts and some others. The first
two files of the catalogue can be obtained on request from the Centre de
Donnees Stellaire at Strasbourg.

The Metacatalogue enables -us to investigate the large scale distri-

bution of galaxies. A process for searching for second-order clusters
of galaxies has been developed. The process consists of smoothing and
filtering two-dimensional discrete fields, defined on a lattice, the
trend being attenuated beforehand. The process is described by Kalinkov,
M. (1974, Proc. First Europ. Astron. Neet.~ 3, 142; 1975, Mem. Soc.
Astron. Ital.~ 45,639; 1977, Highlights of Astronomy~ 4, part I, 279),
by Kalinkov, M.~Stavrev, K., Kaneva, r., Dermendjiev, V. (1976, in
Stars and Galaxies from Observational Point of View~ Tbilisi, 309) and
by Kalinkov, M., Kaneva, r., Stavrev, K., Tomov, B., Vlahova, K., and
Yanev, K. (1976, C. R. Acad. Bulg. Sci.~ ~, 453).

Some examples of the latest (1977) results are given in Figures 1-3.

Figure 1 shows a medium scale filtering for all Zwicky clusters

(northern galactic hemisphere). The North Galactic Pole is at the
centre and the point of view is £ = 315 0 • At least 60% of all conden-
sations are real second-order clusters with characteristic sizes of
SHORT COMMUNICATIONS 277

Figure 1

$1
I
I
I

Figure 2
278 SHORT COMMUNICATIONS

40-50h- 1 Hpc. The density contrast is between 2 and 4 for the particu-
lar filtering function which was used.

Figure 2 is the same field but for the case of large scale smooth-
ing. The density contrast is smaller (1.5-2) and the effects of
galactic extinction are present. However, there are two large condensa-
tions, which may be third-order clusters. The same condensations may be
found on the smoothed map of Abell clusters.

Figure 3 represents a large scale smoothing of the Lick reduced

counts of galaxies. Only one of the condensations corresponds roughly
to Abell and Zwicky condensations.

Our recent results support the ideas of Kiang, Saslaw, de Vaucou-

leurs, Peebles and others of continuous clustering - we find a tendency
towards clustering with certain preferred sizes of about 40-50, 90,
120-150, 200 and 300 Mpc, and with density contrast which decreases with
increasing characteristic size of the high-order clustering.

:/J.
Figure 3

DISCUSSION

de Vaucouleurs: Did you make similar plots for the old Hubble counts?
If so, how do they correlate with Lick, Abell and Zwicky clusters counts?
SHORT COMMUNICATIONS 279

Kalinkov: Yes, we have now some maps for the old Hubble counts. They
are similar to the maps for Lick counts but not to Zwicky cluster maps.

Tammann: Do you find the same superclusters from two or more catalogues?
If so are the observed identities above those expected from chance
coincidences?

Kalinkov: Yes, there are at least 20 second-order cluster coincidences

between Abell and Zwicky catalogues. The expected number of chance
coincidences is 2 or 3.

ON THE METHODS OF DISCOVERING GROUPS AND CLUSTERS OF GALAXIES

B. I. Fessenko

Groups. Even if real groups did not exist, the investigation of

large regions of the sky will reveal other dense clumpings of galaxies.
The usual estimate of the number of optical members in these clumpings
will confirm their reality. But let us remember that all these groups
are false. False groupings of galaxies should emerge in large numbers
because of the effects of interstellar absorption and of variability of
the observating conditions. Statistical considerations of the groups
detected by Turner and Gott (1976, Ap. J., Suppl. Ser., 32, 409) leads
to the conclusion that the percentage of false groups members is equal
to 40% or more.

Clusters. Here the same remarks apply. At a given apparent

limiting magnitude the distribution of galaxies (or real groups) as a
function of distance from us has a single maximum. Near that maximum
false clusters emerge with the greatest probability. With a larger
limiting magnitude for galaxies in the sample, more distant false
clusters should be obtained. Their angular diameters decrease in the
same way as if the clusters were real ones. Statistical investigations
of the Lick counts and other sources of data on the apparent distribu-
tion of galaxies lead me to the conclusion that a substantial fraction
of the clusters of Zwicky and Abell are false clusters.

DISCUSSION

de Vaucouleurs: We do not understand how you could separate out the

galactic extinction variance by restricting your study to b ) 60 0 (i.e.
cosec b .:;:; 1).

Fessenko: In the analysis of the Lick counts, absorption of light was

estimated not by means of a cosec law but by a special differential
method of processing unreduced galaxy counts.

Bolton: Is the absorption you are considering colour dependent or non-

selective?
280 SHORT COMMUNICATIONS

Fessenko: New data on the luminosities, linear dimensions and the B-V
colour indices for ScI galaxies by Rubin, Ford and others enabled us to
obtain values for the light absorption coefficient which exceeded 0.36.

THE AUTOCORRELATION ANALYSIS OF DEEP GALAXY S~MPLES

S. Phillipps
(read by G. Efstathiou)

The two point correlation function wee) has been evaluated for the
galaxies measured by the COSMOS machine at the Royal Observatory,
Edinburgh, in an area of about 2 square degrees on a 2 hour exposure J
plate and a 2 hour exposure R plate (Phillipps, S., Fong, R., Ellis,
R.S., Fall, S.H. and MacGillivray, H.T., 1977, Mon. Not. R. astr. Soc ...
in press). In each case wee) is found to be in agreement with the form
w = Ae- O• 8 found previously by Peebles and coworkers. Since the samples
are not magnitude limited the selection function, i.e. the distribution
in distance, was determined by using models of the galaxy distribution
to fit the observed angular diameter counts. However, when these selec-
tion functions are used to scale the amplitudes found for our samples,
the amplitudes are found to be lower than those expected from Peebles'
results by a factor of about 3. We consider that this is likely to be
due to a lack of clusters in the small area of sky which we have studied:
analysis of further areas should show whether this is the case.

DISCUSSION

Huchra: Have you tried to obtain an estimate of the limiting magnitude

of your plates independent of your very complicated procedure?

Efstathiou: No.

Longair: What do you estimate the typical redshifts of the galaxies you
are studying to be?

Efstathiou: Typically about 0.2 to 0.3.

 DISCUSSION OF METHODS ON DETERMINING THE MEAN
MATTER DENSITY OF THE UNIVERSE

Tammann: The question of the mean mass density in the Universe has been
mentioned many times and the organisers thought it would be useful to
look at future prospects for obtaining improved estimates of n. The
methods may be split into two types. First, the conventional methods
involve determining the luminosity density in the Universe and mUltiply-
ing by an appropriate mass-to-luminosity ratio. It must be emphasised
that the mass determinations are dynamical and that most of the mass is
not visible. There is good agreement among independent workers about
the luminosity density and this figure seems to be known within a factor
of 2. The appropriate mean value of M/L is more controversial, values
between 10 and 200 having been discussed in the preceding lectures. It
is agreed, however, that if the mean value of M/L lies in this range,
one cannot close the Universe. Perhaps one should be more cautious and
say that even granted the uncertainties in the quantities involved, it
is unlikely that the Universe is closed.
Second, there are the new methods which have been mentioned in
previous discussions. These may be called large scale dynamical methods
and we want to concentrate upon these in the discussion in the hope of
encouraging future observations to make them work in practice. Three
such methods will be described by Drs Silk, Peebles and Falls.

Silk: Some day we shall see n (or at least a solid lower limit thereon)
written on the sky! By studying the velocity distribution of galaxies
in the environs of a supercluster, where the Hubble flow is distorted by
up to 100 percent, we should be able to infer n. The models should be
simplest in regions where virialization cannot yet have occurred. The
relative mass excess must be larger within a given region in order to
account for a given Hubble flow distortion if n is small: this is
because in this case the Hubble flow is kinetic energy-dominated, and a
larger potential fluctuation is required. In the simplest model of
spherically synunetric infall, the idealized "zero velocity" shell of
matter contains an average density contrast, or mass "excess" relative
to the local background, 1 + 02/2 = w2 (Ho t o )-2(Sn)-1, where 1 ~ Hoto ~
2/3 and n ~ 1.
Some notes of cautions are in order before observers rush to obtain
redshifts. The density gradients around superclusters are extremely
insensitive to n; the perturbed Hubble flow combined with the average
density contrast within a shell surrounding a supercluster are required.
Luminosity~eighted galaxy counts actually yield a measure of the lumin-
osity rather than the density contrast, and a systematic gradient in M/L
(for example an inverse correlation with distance from the supercluster
center) could frustrate attempts to determine n: actually only a lower
limit on Q would be obtained. Finally, the observer should use isolated
clusters for this test. Gross distortions of the Hubble flow can be
induced by tidal interactions between neighbouring clusters. This is
particularly likely to have occurred at a redshift of 1 or 2, and could
result in anisotropic structure of the galaxy distribution on very large
scales. 281
M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 281-286. All Rights Reserved.
Copyright © 1978 by the fA U.
282 METHODS ON DETERMINING THE MEAN MATTER DFNSITY OF TIlE UNIVERSE

Tully: The most uncertain assumption to applying this method must be

that there is a close correlation between the observed and actual dis-
tributing of matter.

Peebles: We have heard of the problems of defining groups of galaxies

and how different workers can obtain different results from analyses of
basically the same data. We should therefore use all the possible
methods of analysing the problem until, hopefully, we get a consistent
answer. My favourite method at present is to use the relative veloci-
ties of all pairs of galaxies in a properly selected sample and to
relate this to the mass distribution. This method has the great advan-
tage that only the relative velocity of one galaxy with respect to
another is required. The mean square relative velocity of galaxies is
related dynamically to statistics we can hope to measure - in fact, the
three point correlation function for galaxies which we think we know.
I corne up with the plea that we know the right-hand side of the equation
from integrals over the three-point correlation function. What we don't
know is the left-hand side, the mean square relative velocities for
good random samples of galaxies. The determination of velocities for
complete random samples of galaxies down to 15th magnitude would give us
a credible value for <6V r 2 >. We would dearly like to have it.

Fall: I have been asked to make some remarks on dynamical methods for
estimating the local mean density of matter. (In this context "local"
refers to scales larger than individual galaxies but smaller than the
horizon.) Let me first emphasize that all of these methods apply only
to matter clustered in the same way as galaxies and that they tell us
nothing about any uniformly distributed hot components of the Universe.
Traditional methods of this sort are based on the usual virial theorem
for individual groups and clusters. Recently, Geller and Peebles and
Peebles have developed statistical versions of this method which average
over groups and clusters. Like the ordinary virial theorem methods they
are restricted to scales small enough that clustering can be considered
stationary. As Peebles has already remarked, these "statistical" or
"cosmic" virial theorems have the advantage that the required velocities
are easily measured.
The other class of dynamical methods for estimating the mean matter
density includes both the motions of galaxies in the relaxed parts of
clusters and also the decelerated motions of galaxies in the outer parts
of clusters; they rely upon deviations from perfect Hubble flow on a
variety of scales. They are based on the idea that we can measure
relative deviations from Hubble flow DHIH and relative deviations from a
uniform distribution of galaxies Dplp on different scales. Relating
DHIH and Dplp by some model for clustering then gives us the mass of
clusters and hence the (dimensionless) mean density U.
The methods involving deviations from Hubble flow also take both a
statistical form and a fOTQ applicable to individual clusters. The
simplest of the second type is a model introduced by Silk for the non-
linear evolution of a uniform density spherical cluster. Peebles'
METHODS ON DETERMINING THE MEAN MATTFR DFl\SITY OF THE UNIVERSE 283

version applies to the linear development of a spherical but non-uniform

cluster. On the theoretical side, we need more realistic models for the
development of clusters in the regime where op/p ~ 1, i.e. where expan-
sion has been halted and reversed. On the observational side, we need
more and better distance and velocity estimates of the type Abell dis-
cussed in his talk. Regions within 10 or 20 Npc of the centres of
several large clusters such as Coma and virgo are especially promising.
Finally, the statistical version of these methods is most naturally
expressed in terms of energies. With the deviations from Hubble flow
one can associate the kinetic energy T = ~<v2> where v is a non-Hubble
velocity and the average is over all galaxies. Similarly, one can
associate with clustering the potential energy W = ~pJd3r(-G/r)s(r)
where S is the galaxy pair-correlation function. The relation between T
and H then comes from a "cosmic energy equation" which simply expresses
the conservation of gravitational energy in an expanding system but does
not require the assumption of a stationary state as in all virial theorem
methods. The method is not sensitive to the shape of S nor is it sensi-
tive to the adopted cosmological and clustering models: T = -dvJ with
~ t d ~ 2/3. Thus, the major uncertainties of the method may be collec-
ted into the expression ~ IT <v 2 >/(amplitude of s).
At first sight it might seem that estimating the velocities required
for this statistical energy condition would be difficult. As Gott has
em~hasized in his talk, however, there is good reason to believe that
<v > is simply related to the relative velocities of galaxy pairs in the
models of interest. On the theoretical side it will be important to see
if this relation holds up on very large scales. The first estimate of s
came from magnitude-limited samples using no velocity data and their
major uncertainty results from uncertainties in the galaxy luminosity
function. As I hope Davis will explain in his talk, estimates of scan
be improved using redshift data from a large sample of galaxies. On the
observational side it will be important to extend existing redshift
samples to volumes of space which are large enough that they are likely
to be representative of the Universe as a whole. A fairly complete
sample of galaxies to 15th magnitude would probably be adquate for this
purpose.

Binney: If one confines oneself to scales too large to have virialized,

one should be able to obtain an estimate of Fall's peculiar velocity from
timing arguments applied to the known spatial distribution and an assump-
tion of homogeneity at recombination. The problems which one encounters
when carrying through this programme with spherically symmetric infinite
models should not arise in more realistic cosmologies.

Peebles: I think knowledge of "initial and final positions" with time is

not adequate to determine final velocities - one needs also the curvature
between, that is, ~.

Davis: One should note that there are difficulties in applying both of
the two new virial theorems described by Drs Peebles and Fall. In one,
284 METHODS ON DETERMINING THE MEAN MATTER DENSITY OF THE UNIVERSE

there are problems in determing the kinetic energy terms and, in the
other, the potential energy terms in the cosmic energy equation. In
Fall's case, it is difficult to derive the single particle velocity
dispersion from the relative velocity dispersions. In Peebles' case,
the integral over the three-point c9rrelation function is tricky and
very sensitive to the observations. It may be possible to overcome
these difficulties when we have magnitude-limited samples to 15th
magnitude.

Turner: It might be worth mentioning that the value of peculiar v rms

could, in principle, be obtained directly from observations without
intervening analYfic or N-body models relating it to relative velocity
dispersions <~vr>2. This could be accomplished by comparing the spatial
clustering deduced from the distribution of galaxies on the sky to the
clustering in "redshift space"; the characteristic distortions of the
galaxy distribution caused by non-Hubble flow motions reflect the value
of peculiar v rms ' Thus, the Fall-type cosmic virial theorems are not
intrinsically model dependent. The very great practical difficulty is,
of course, that one would need a redshift survey considerably deeper
than the largest scale on which coherent non-Hubble motions occur.

Tammann: There has been much discussion about determining redshifts for
large samples of galaxies. Let me remind you that available redshifts
are very ~oor. Few are determined with precision much better than
100 km s- mean error. Some recent compilations have greater errors.
Therefore, you should also specify that you want good redshifts.

Kornberg: It seems to me that even if we have a sufficient number of

redshifts for galaxies in groups and clusters, we shall not be able
without inclusion of additional considerations to decide which objects
are physically connected with particular complexes. Such additional
considerations might include for example (1) the relative abundance of
gas, (2) the rate of star formation and resulting from that the colours
of galaxies, (3) morphological peculiarities and peculiarities of the
distribution of various types according to the radius of the cluster or
according to the velocity dispersion, (4) the activity of nuclei and its
connexion with the fraction of gas in galactic clusters.

Longair: Can each of the proponents of the new methods estimate the
quantity and quality of data required to obtain a good result from the
methods, i.e. precision of redshifts, number of galaxies to be measured
to what magnitude, and so on?

Peebles: First, if you want to decide whether ~ is closer to 0.03 or

0.3, in the latter case the velocity dispersion of galaxies separated
by 1 Mpc should be at least 300 km s-l, perhaps even greater. Thus a
precision of 100 km s-l is adequate to test which is the better value
of ~.
Second, Zwicky's catalogue of galaxies to 15th magnitude seems to
be a fair sample of the Universe and hence several Zwicky plates would
probably be sufficient.
METHODS ON DETERMINING THE MEAN MATTER DENSITY OF THE UNNERSE 285

SiLk: I have been very impressed by Tully's data which represents a

major advance over previous knowledge of the dynamics of galaxies in
the vicinity of the local supercluster. Continued efforts in this direc-
tion may provide us with a good estimate of ~ as I outlined in my earlier
remarks.

Tammann: This raises an interesting question. Are you happy only to get
the redshifts for spiral galaxies since elliptical galaxies are weak
emitters of 2l-cm radio emission?

GeneraL repLy from audience: No!!

FaLL: In addition to measuring redshifts in the "representative" samples

for the statistical methods, I would also urge continued work in fields
near clusters such as Virgo and Coma.

PeebLes: In the method using the dynamics of galaxies in the local

supercluster, we need distances as well as redshifts unless you are
prepared to build a very detailed model of the velocity field.

Davis: I would like to note that a group at Harvard is planning to

measure the redshifts of all galaxies in the northern sky down to about
14.5 magnitude. It is estimated that 4000 galaxies will be measured and
it will take a few years.

Jones: Are magnitudes important?

Tammann: Only as far as they effect distances.

de VaucouLeurs: 1. We are mapping the velocity field in the local

supercluster within 40 Mpc from 300 spirals with good distances.
2. We are reducing (with H. Corwin and collaborators) the magnitudes in
the Zwicky catalogues to a proper magnitude scale; corrected magnitudes
will have mean errors ~ 0.15 - 0.20 mag.
3. Measurements of ~ 300 redshifts of bright galaxies are in progress.
4. Question to theorists: how important is the spherical approximation
~n the cosmic virial theorem; can you take into account sheet or string-
like structures?

PeebLes: In my version of the statistical virial theorem, these compli-

cations are automatically taken into account through using the 3 point
correlation functions.

van der Laan: He need clarification on the relative value of 21 cm

(spiral)redshifts and redshifts of ellipticals. Are they required
equally badly, is there a trade-off, are they interchangeably useful?

Huchra: Optical redshifts can be as good as 2l-cm redshifts if you use

a high enough dispersion, say using 50 or 100 ~ rom-I rather than 300 ~
286 METHODS ON DETERMINING THE MEAN MATTER DENSITY OF THE UNNERSE

mm-I. Furthermore, if the nucleus of the galaxy has an emission line

spectrum, you can determine a very good redshift. There may be asymme-
tries in the 21 cm line emission.

Ekers: Is it also useful to have rotation curves and hence mass esti-
mates for spiral galaxies since this can be done with synthesis tele-
scopes at 2l-cm? It would be very time consuming but it is possible.

Fall: In their present form, the methods we have been discussing make
no explicit use of information on the masses of individual galaxies. If
this information were readily available, however, it could certainly be
incorporated into these methods.

Ginzburg: It is not clear why at a discussion of the value of ~ nobody

touched upon the density of the intergalactic environment (primarily
gas). It is natural that in a mathematical simulation all the mass
between the clusters is considered. However, when the dynamics of
clusters is discussed and large superclusters are not considered, the
case will be different. According to the estimates that we made, the
density of the intergalactic gas may be great enough to increase the
value of ~ ~ 0.1 many times. What is the new evidence here and what are
the restrictions to the amount of the intergalactic gas?

Silk: In fact, the intensity and spectrum of the X-ray background would
be consistent with the breumsstrahlung emission of hot intergalactic gas
of density ~z 1 if its temperature were = 3 x 10 8 K.

Tinsley: Intergalactic matter is counted 1n the cosmic virial theorems

provided it is distributed like galaxies.

Gatt: Statistical virial theorem estimates made by measuring velocity

differences between pairs of galaxies separated by 1 Mpc give informa-
tion on the mass distribution on scales of this order. Group catalogue
techniques give similar information. The tests discussed by Silk on
uniformity of the Hubble flow in the Local Supercluster measure mass
distributions on scales of 20 Mpc. With the large space telescope,
accurate distances will become available throughout the supercluster and
the uniformity test can be carried out with confidence. It will be
quite interesting to see if this test gives a value of Q in agreement
with that deduced from the group catalogue and statistical virial theorem
methods.

Tinsley: In response to Dr Ekers' question, one should note that the

masses of galaxies are of the greatest interest for many other reasons,
for example for understanding their stellar populations and their values
of MIL.
IV

OBSERVATIONAL EVIDENCE FOR COSMOLOGICAL EVOLUTION

 QUASARS AND COSMOLOGICAL EVOLUTION

Maarten Schmidt
Hale Observatories
Pasadena, California, U.S.A.

The first evidence that quasars exhibit a non-uniform

distribution in the universe was obtained ten years ago from a sample
of 33 quasars in the 3CR catalogue. Much larger samples of radio
quasars are now available and we will discuss the results in terms of
an exponential decrease of quasar numbers with cosmic time. Only a
few samples of optically selected quasars exist. We will show that
their counts indicate a non-uniform space distribution, even if the
redshifts are non-cosmological.

1. RADIO QUASARS

For samples that are complete to given limits of optical flux and
radio flux one can test the hypothesis that the objects have a uniform
distribution in space. For each object one can derive V/V , where
V is the volume of the Universe out to the distance of the max object,
and V the volume corresponding to the maximum distance at which the
objecra~ou1d still be observed within the sample limits.

Table I lists values of <V/Vmax > obtained for five complete

samples. The 3CR sample (Schmidt 1968) is now complete to about 20-th

Table I

Radio Quasar Samples

Catalog Sample v n <V/V >

max
3CR 178 MHz 44 0.64
4C Olsen, Schmidt 178 51 0.64
4C Wills, Lynds 178 117 0.69
PKS+4° {MaSSOn, Wall 2700 57
Wills, Lynds 2700 60 ~:~~} 0.62
Sl,S2,I Schmidt 5000 51 0.61

289

M S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 289-293. All Rights Reserved.
Copyright © 1978 by the IA U.
290 MAARTENSCHMIDT

magnitude. The Olsen (1970) 4C sample between declinations 20 0 -40 0

has been investigated to about the same optical limit (Schmidt 1974).
Wills and Lynds (1978) have studied several 4C samples represented by
one entry in Table I. The PKS±4° quasar sample has been discussed by
Masson and Wall (1977) and Wills and Lynds (1978) with rather different
results. Dr. D. Wills has attempted a reconciliation between the two
studies (private communication) from which I estimate <V/V > = 0.62
for qo = O. The last sample (Schmidt 1976, 1977) is basedmaxon the
Sl, S2, and I parts of the NRAO 6-cm survey.

The <V/V > values shown in Table I have mean errors of around
0.03 - 0.05. max Each of the entries is significantly larger than the
value 0.50 expected for a uniform space distribution. This constitutes
the evidence for cosmological evolution of radio quasars.

Flat-spectrum radio quasars appear to have smaller values of

V/V . Table II gives relevant results for four complete samples.
Themaxfinal value given for the PKS+4° sample is again an estimate,

Table II

Quasars with Flat Radio Spectrum

Sample Spectral Index n <V/v >

max
3CR, 4C, Sl, S2, I (Schmidt) a > -0.2 28 0.52+0.05
a > -0.5 42
PKS+4° (Masson, Wall)
PKS+4° (Wills, Lynds) a > -0.23 30
~.~~ } 0.58±0.04
. (a > -0.5)
S4 (Kuhr, Schmidt) a > +0.2 29 0.58+0.05

for qo = 0, based on a discussion by D. Wills. The last entry is the

result of an unpublished study of quasars with inverted spectra between
6 and 11 cm in the NRAO-Bonn S4 catalogue.

Flat-spectrum quasars exhibit <V/V > ~ 0.57 while steep-spectrum

3
quasa s have <V/V
p = e T and
> arouns 0.67. Them~3rresponding density laws are
maw p = e l 1respectively, where T = z/(l+z) is the
light-travel time in a q = 0 cosmology. We have assumed here that
the density depends expoRentially on cosmic time. Larger radio quasar
samples will be required if we want to determine, rather than assume,
the shape of the density law eventually.

2. OPTICALLY SELECTED QUASARS

Radio quasars constitute only a small fraction (around one percent)

of quasars selected by optical means. The initial optical selection is
based on the ultraviolet excess, i.e., the criterion U-B < -0.4 which
is fulfilled by almost all radio quasars at high galactic latitude
QUASARS AND COSMOLOGICAL EVOLUTION 291

with redshifts less than 2.5, or on evidence for emission lines on

objective prism exposures. The two techniques tend to complement each
other since the objective prism should be effective for redshifts of
2 and more since La emission is usually strong. All candidates selected
by color or objective prism spectra require slit spectroscopy to confirm
the quasar nature and to measure the redshift.

Table III

OPTICALLY SELECTED QUASARS

Area Mag. Limit Surface density
-1
Braccesi 36 sq. deg. B < 18.0 0.47 (sq deg) -1
Green, Schmidt 1434 sq. deg. B < 15.7 0.0035 (sq deg) 1
Sandage, Usher 40 sq. deg. B < 18.5 1.3-1.6 (sq deg)-

We list in Table III results of surface densities for three studies.

For the Braccesi survey (Braccesi, Formiggini, and Gandolfi 1970) we
have used a somewhat conservative magnitude of completeness B = 18 to
derive the surface density. The result by Green and Schmidt is based
on a discussion of part of the Palomar Bright Quasar Survey (Green 1976)
designed to detect all quasars brighter than B = 15 - 16 in an area of
10,000 square degrees on the basis of their ultraviolet excess. The
last entry concerns an as yet unpublished result of a study by Sandage
and Usher.

The values for the surface density shown in Table III vary very
steeply with magnitude. In fact, the numbers increase by a factor of
about 8.5 per magnitude. The expected increase for a uniform distribution
of quasars is only a factor of around 2.2 per magnitude (Green and
Schmidt 1977). This discrepancy constitutes solid evidence for
cosmological evolution of optically selected quasars. ThrOnumbers given
in Table III lead to a density law even steeper than p= e T, found for
steep-spectrum radio quasars. The results from the entire Green survey
will eventually allow a determination of the cosmological evolution as
a function of optical absolute luminosity.

The very steep slope of the optical counts has a bearing on the
local hypothesis of quasars, which assumes that their redshifts are
non-cosmological and that their distances are relatively small. In
this case the distribution of quasars would presumably be similar to
that of galaxies, i.e., approximately uniform. A uniform distribution
in space yields counts that rise by a factor of 4 per magnitude,
regardless of the shape of the luminosity function. Green and Schmidt
(1977) show that such a slope is definitely incompatible with the
surface densities given in Table III. Hence, quasar counts do not allow
a uniform space distribution in the local hypothesislo~ quasars. Rather,
the space density has to increase approximately as r · (where r is the
distance). Any version of the local hypothesis that does not incorporate
such an increase of space density with distance will be. unable to reproduce
the counts as given in Table III.
292 MAARTENSCHMIDT

REFERENCES

Braccesi, A., Formiggini, L., and Gandolfi, E. 1970, Astronom. and

Astrophys., l, 264.
Green, R. F. 1976, Pub. Astronom. Soc. Pacific, 88, 665.
Green, R. F., and Schmidt, M. 1977, in preparation.
Masson, C. R., and Wall, J. V. 1977, Mon. Not. Roy. Astronom. Soc.,
180, 193.
Olsen, E. T. 1970, Astronom. J., ll, 764.
Schmidt, M. 1968, Astrophys. J., 151, 393.
Schmidt, M. 1974, Astrophys. J., 193, 505 and 1975, Astrophys. J., 195,
253.
Schmidt, M. 1976, Astrophys. J. (Letters), 209, L55.
Schmidt, M. 1977, Astrophys. J., 217, 358.
Wills, D., and Lynds, R. 1978, submitted for publication.

DISCUSSION

Peterson: It has been known for some time from the work of Bolton and
his co-workers at Parkes that the source counts of quasi-stellar radio
sources with flat radio spectra differ from the source counts of the
steep spectrum quasi-stellar sources. In order to see whether or not
the difference in counts implied a difference in their space distribu-
tion, Jauncey, Wright, Condon and myself, during the past two years,
have obtained redshifts with the 4-metre Anglo-Australian telescope of
Parkes quasi-stellar sources with flat radio spectra. We have more than
100 redshifts, of which about 1/3 form a complete sample. These give a
<V/Vm> for quasi-stellar sources with a > -0.5 of 0.56 ± 0.05, which is
similar to Prof. Schmidt's result.

Ostriker: How does the slope of your log N versus B relation for
optically selected quasars compare with that found by Sandage and
Luyten?

Schmidt: The early Sandage-Luyten slope of 0.75 was based on incomplete

observations of their l-hour field and on very incomplete observations
of a preliminary version of the Braccesi list of objects with ultra-
violet excess. This slope may well be approximately correct over the
magnitude range (B = 19 - 19.5) for which it was originally derived.

ZeLdovich: What is the effective z to which the e 10T law is valid? How
does it change if qo = 1/2 instead of O?

Schmidt: The e 10T law is probably valid at least to z = 1 or 2. The

exponent would probably differ by only 1 or 2 units for qo = 1/2.

SiLk: Are there any other morphological differences for sources with
different values of V/Vm• for example, radio structure and frequency of
optical absorption systems?
QUASARS AND COSMOLOGICAL EVOLUTION 293

Schmidt: There is the well known correlation between small radio size
and flat radio spectrum. I am not aware of anyone having investigated
correlations with absorption line systems.

Petposian: Is there any difference between the redshift distribution of

steep and flat spectrum radio quasars?

Schmidt: At an optical magnitude of around 18 or 19 the redshift distri-

bution of steep and flat spectrum radio quasars are indistinguishable at
the present time.
THE SURFACE DENSITY OF QUASARS

J.G. Bolton
Division of Radiophysics, CSIRO, Sydney, Australia

Ann Savage
Royal Greenwich Observatory, Hailsham, Sussex, U.K.

A knowledge of the surface density of quasars as a function of

magnitude is important for two reasons. Firstly it is necessary in
order to assess the possible statistical significance of close pairs of
quasars or the association between quasars and bright galaxies. Secondly
it is a necessary step in the determination of the space density or
luminosity function of QSOs. We have carried out what we believe to be
currently the most comprehensive investigation into the surface density
of quasars. Three techniques have been used in this investigation.
These are -

(1) A search for ultra-violet excess stellar objects and their

subsequent classification.

(2) A search for emission-line stellar objects on objective prism

plates.

(3) Identification of radio sources from deep radio surveys.

The investigation has made use of five optical and one radio tele-
scope: the SRC and Palomar 1.2-m Schmidt telescopes, the Anglo-
Australian 3.9-m telescope, the 1-m and 0.5-m telescopes at the
Sutherland Observing Station in South Africa, and the Parkes 64-m radio
telescope.

The regions selected for study are two areas of 25 square degrees
each centred at 22 h 04 m, -18°55', and 02 h OOm, -50°00'. Pairs of blue and
UV plates were taken for us by the UK Schmidt Unit in five areas near the
south galactic pole and the final selection of two areas was based on
plate quality. In the 22h area a two-colour (blue and UV) plate which
had been taken with the Palomar Schmidt in 1966 was available for com-
parison. The two-colour plates were blinked using a closed-circuit TV
combined blink and coordinate measuring machine. Each plate is viewed
through a TV camera and the outputs of the two cameras are combined
electronically to present the two images of the same star side-by-side.
The desired colour balance can be set in the electronic mixer. A target
295
M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 295-303. All Rights Reserved.
Copyright © 1978 by the [AU.
296 J. G. BOLTON AND ANN SAV AGE

viewed through a third camera can be switched in for coordinate measure-

ment using precision screws on the X-Y carriage which bears the plate-
viewing cameras. Areas of 6' x 6' arc were inspected at a time and
inspection of one set of plates took approximately nine days. Each
plate gair was scanned by A.S. and again by A.S. and J.G.B.; thus for
the 22 field, where the additional Palomar plate was available, four
independent examinations were made. About one-third of the UVX objects
brighter than B = 19m detected in the plate scanning were galaxies but
about 250 apparently stellar objects with UVX were found in each area.
(UVX was defined as U-B ~ -0.4.)

Blue and ultra-violet magnitudes for these objects were established

in three stages. Photoelectric magnitudes were measured for about 20
objects between 10m and 17m in the centre of each field using the South
African 0.5-m and 1.0-m telescopes. A photographic sequence extending
to 20 m was then obtained through the use of a 3m obj ecti ve grating built
for use with the SRC Schmidt telescope. The grating is formed from
0.15 cm nylon lines 0.7 cm apart stretched across a circular aluminium
angle frame. Several plates in both colours were taken of the two
fields and iris photometry of these plates was used to form the extended
photographic sequence. Finally, iris photometry of the two-colour plates
provided the magnitudes for the UVX objects. R.M.S. errors in these mag-
nitudes are ~Om.15 in Band U and the major contribution to them arose
from uncertainties in the final stage.

Low-dispersion spectra of about 30 objects in each field were

obtained with the Robinson-Wampler image dissector scanner at the
Cassegrain focus of the 3.9-m Anglo-Australian telescope. Half of the
objects were brighter than B = 18m and most of these proved to be stars,
generally dwarfs or sub-dwarfs; most of the objects fainter than
B = 18 m proved to be QSOs. Classification of the remaining UVX objects
in the two fields was made from deep 111a-J plates and objective prism
plates taken with the SRC Schmidt telescope. Four classes of objects
were distinguished:

(1) Compact galaxies. Although these appear to be stellar on the

11a-0 or 111a-0 two-colour plates both the high resolution of the 111a-J
direct and objective prism plates clearly indicate that they were
galaxies. These were almost all fainter than B = 19m.

(2) Galactic stars. These could be recognized as such from the

form of their continuous spectra on the objective prism plates by com-
parison with those confirmed as stars with the Anglo-Australian tele-
scope.

(3) Emission-line QSOs. Examples of objective-prism spectra of

some of these objects are shown in Figure 1. Comparison of objective
prism spectra and the Anglo-Australian telescope scanner spectra of the
sam~ objects showed that lines with an equivalent width of as little as
15 A could easily be seen on the objective prism spectra.
THE SURFACE DENSITY OF QUASARS 297

(4) Unknown. These objects have continuous spectra similar to

those of QSOs but no emission lines. Most of them are fainter than
B = 19m and although some may be QSOs we feel that the majority are
very hot stars.

Fig. 1 - Reproductions of six QSO

spectra from the SRC Schmidt objective
prism plate of the 22h field. Magni-
tudes and redshifts for the objects
from top left to bottom right are:

B 17m.S, z = 2.092
B lSm. 3 , z 2.2S or 0.43
B lS m.S, z 2.04
B lSm. 9 , z not known
B lS m.9, z 2.31
B 19m.S, z = 2.31

The distribution of UVX objects in the four classes as a function

of magnitude for the 22h field is shown in Figure 2. Galactic stars
dominate the UVX population for objects brighter than Ism; then the
QSOs take over. However, compact galaxies make some impact near plate
limit, and presumably this effect has also been present in earlier
studies using similar plate material.

Fig. 2 - Number-magnitude
distribution of UVX objects
in the 22h field. Stars are
shown with diagonal hatching,
compact galaxies with hori-
zontal hatching, QSOs in
black and unclassified ob-
jects clear.

MAGN ITUDE
 298 J. G. BOLTON AND ANN SAV AGE

The objective prism plates were also scanned for emission line
QSOs in an area of 30 square degrees. Some 70% of the UVX QSOs in the
central 25 square degrees - generally the stronger-line objects - were
rediscovered in this investigation. In the 22h field only one UVX QSO
was found which had been missed but in the 02 h field 10 such objects
were found. Clearly the two plate pairs in the 22h field resulted in
greater completeness. In the 22h field about 20% as many non-UVX QSOs
were found as UVX QSOs but in the 02 h field almost 60% as Qany non-
UVX QSOs as UVX QSOs. This result was not due to a difference in colour
balance between the two-colour plates of the two areas, since the
individual spectra clearly show whether UVX would be expected. The
difference in the fraction of non-UVX objects in the two fields is re-
f~cted in their number-magnitude distributions shown in Figure 3. The
02 field contains a large number of generally high redshift QSOs with
very strong emission lines - fortuitously in the B-band. Sgme of the
lines have equivalent widths which are of the order of 500 A and the
contribution from the emission line considerably brightens the
B magnitude. If the number-magnitude diagram for the 02 h field is
plotted using U magnitudes rather than B magnitudes its form is changed
to resemble the B magnitude diagram for 22h.

ZZh field OZh field

20-

17 18 19 17 18 19 20

MAGN ITUDE

Fig. 3 - Number-magnitude distributions for QSOs in the 22h and 02 h

fields. UVX objects are shown cross-hatched and non-UVX objects clear.
THE SURFACE DENSITY OF QUASARS 299

The surface densities of UVX QSOs and UVX plus non-UVX QSOs (in
parentheses) to magnitudes 19.5 and 19.75 for the two fields are as
follows.

22h field 1.8(2.2) 2.4(2.8)

No. per square degree
02 h field 2.2(3.6) 2.5(4.0)

The surface densities of the UVX objects are somewhat lower than those
used by Bolton et al. (1976) in their claim of a significant pairing
between radio and radio-quiet QSOs. A nearest-neighbour analysis of the
distribution of the QSOs in the 22h field (shown in Fig. 4) shows
significant palrlng for separations of the order of 2' arc; however,
there is no similar effect in the 02 h field.

The combination of the UVX and objective prism searches of the

central 25 square degrees and the objective prism searches in the
peripheral 5 square degree areas yielded a total of about 120 emission-
line QSOs in each field. For measurement of line wavelengths the use-
ful spec~ral range of the IIIa-J objective prism plates is abou~ 3300
to 5100 A (the cut-off of the IIIa-J emulsion is close to 5200 A but the

•
.
AN • 01
..J
,. - -A._ __
.J
• • • /
• 0

•
, •• • •
, .. .- .. ,
• • • • II ·0 • • • ...1
0

..
••
o • A • I
•
0
••
• • •
•
0
, •
• • • 0 0

[I •
• • •• •• • IW
• o• A I •
Q)

I 0

• • •
• • •
I
·0
• • • •
• • fP • • • •~
•
0
I
•• • •• •
• • I
-I -. - 5 I
Fig. 4 - The distribution of QSOs in the 22h field.
The 50 square area of the search for UVX objects
lies inside the dashed lines. UVX objects are
shown as filled circles and non-UVX objects as open
circles. Radio sources are indicated by triangles.
 300 J. G. BOLTON AND ANN SAVAGE

resolution is so low at the red end that lines are very difficult to
distinguish or measure). Wavelength calibration of the prism was
determined by using the measured or deduced line wavelengths of QSOs
which had been observed with the Anglo-Australian telescope. Unambigu-
ous redshifts could be determined for most QSOs with two or more lines
and for one-line objects where that line occurred at a wavelength where
all but one line identification could be excluded by the absence of
other lines. The combination of Anglo-Australian telescope and
objective-prism spectra yielded redshifts for 90 QSOs in the 02 h field
and for 60 QSOs in the 22h field, and magnitude-redshift diagrams for
the two fields are shown in Figure 5. The obvious gaps in certain
redshift ranges occur where redshifts for single-line objects on the
objective prism spectra are indeterminate. The diagrams reflect the
familiar scatter of other QSO samples. Perhaps significant is the
fact that the few radio emitters lie in the lower halves of these
diagrams. The decrease in the number of QSOs with redshifts greater
than ~2.5 is probably real, since the strong Ly-a line can be dis-
tinguished on the objective prism spectra up to a redshift slightly in
excess of 3.

Two radio investigations were made in two fields. Deep radio

surveys with a lower limit of detection of 100 mJy were carried out in
both fields with the Parkes 64-m telescope at a wavelength of 11 cm.

22h field 02 h field

ll_
20 r ••
..
0
·
•• :0
0
20

. :. •
.. .
o

'" •....·
w .0 o· 00 00 ~~ o•
Cl
:::J 19 19
0

A
0
...:!t
... 0

..
I- A
o
z:
c..!:l
~
::;::
18
0
•
..
.0
18
0 ... eo

, .
~4
A

o 0
• o·
A
A
A

0.5 1.0 2.0 10 0.2 0.5 1.0 2.0 10

REDSHIFT

Fig. 5 - Redshift-magnitude diagrams for QSOs in the two fields. UVX

objects are shown as filled circles, non-UVX objects as open circles and
radio sources as triangles.
THE SURFACE DENSITY OF QUASARS 301

Forty-three sources were found in the 02 h field and 37 in the 22h field.
Their positions were examined on the two-colour, the deep IIIa-J and the
objective prism plates. Ten radio galaxies were identified in the 02 h
field and six in the 22h field. Six QSOs were identified in the 02 h
field and four in the 22h field. The four in the 22h field had all been
recognized as QSOs from the UVX or objective prism searches as had four
in the 02 h field; the other two were objects with low UVX and very weak
emission lines. One of the objects in the 22h field - PKS 2203-18, long
known as a QSO - has UVX but all attempts to detect emission or absorp-
tion lines have failed.

In the second radio investigation an attempt was made to detect

radio emission at a very low level, at 6 cm, from all the UVX objects
in the two areas. This was made in the hope that it would serve to
distinguish between stars and QSOs. The 64-m telescope with the
receiver switched between two feeds was used in a manner analogous to
an optical telescope with a two-beam photometer. The system permits a
very low signal to be detected - 20 mJy in only 100-s integration with
an error of ±7 mJy; however, no position information is available.
Very few possible detections were made and most of these were certainly
chance detections of a radio source not coincident with the UVX object -
the apparent detection rate for UVX objects subsequently found to be
stars was as high as that for objects subsequently found to be QSOs!
Additional observations at 6 cm were made on the non-UVX emission line
QSOs with similar results. Finally all the QSOs in the 22h field and
half of those in the 02 h field were investigated using a new 2 cm
receiver. Here the telescope beamwidth is much smaller, the confusion
very much lower, and the detection limit 8±4 mJy for 300-s integration -
but the results were again almost zero. To summarize: the fraction
of optically selected QSOs which are radio sources shows almost no change
as the radio detection limit is decreased by a factor of five or more
below 100 mJy. This result is consistent with the explanation put forward
by Bolton (1977) for the peak in the optical magnitude distribution for
radio QSOs - QSOs either have radio emission which is fairly closely
linked to their optical emission or none at all (though presumably free-
free emission from the line-emitting region would exist, but at a level
too low for detection).

A final interesting result from this investigation concerns the

optical variability of QSOs. We had hoped that for the 22h field where
we had plates taken 10 years apart variability could be used to dis-
criminate between stars and QSOs. In fact, only two variable QSOs were
found: one is one of the four radio emitters in the field and the
other one of the hundred radio-quiet QSOs. This result on radio-quiet
QSOs is in strong contrast with extensive data on the optical variability
of the radio QSOs. In a recent summary Heckman (1977) states '(radio)
quasars are found to vary typically 1-3 magnitudes on time scales of
tens of years'. Radio and radio-quiet QSOs, it would appear, can be
distinguished statistically in two ways. Firstly there is a difference
in the number-magnitude distribution and secondly there is a difference
302 J. G. BOLTON AND ANN SAVAGE

in the incidence of optical variability. Subjectively, we believe that

there is a third distinction - the equivalent widths of the emission
lines in the radio-quiet objects are greater than those in the radio
sources. It would not be unreasonable to relate this effect to the
difference in optical variability.

We acknowledge with great pleasure the assistance given to us in

this investigation by Keith Tritton and all members of the UK Schmidt
Unit.

References

Bolton, J.G., Peterson, B.A., Wills, Beverley J. and Wills, D.: 1976,
Astrophys. J. (Lett.) 210, p. L1.
Bolton, J.G.: 1977, in Proc. IAU Symp. No. 74 (Ed. D.L. Jauncey),
Reidel, Dordrecht (in press).
Heckman, T.M.: 1977, Publ. Astron. Soc. Pac. 88, p. 844.

DISCUSSION

Suehkov: vfuy is there a significant deficiency of objects on the red-

shift magnitude diagram in the range of redshifts from 0.5 to 1.0?

BoLton: This is one of the redshift ranges in which only one line
occurs in the spectral range (3300 to 5200 R) of the objective prisn
spectra. It is not a real gap, just an artifact of our instrumentation.
There are other small gaps but none between z = 1.7 and 3.1. At
z = 3.1, we have an instrumental cut-off as Ly-a disappears beyond the
red end of our spectra.

TinsLey: Have you a number for the density of galaxies at 19~5 on your
IIIaJ plates?

BoLton: For UV excess galaxies it is approximately the same as that

for the quasars.

Jones: I am particularly struck and surprised by the great differences

between your two fields. According to usual cosmological ideas there
should be no difference between such fields. Is the difference, in
your opinion, significant and what may it be attributed to?

BoLton: The differences are undoubtedly real. We would have to look at

a number of other fields to see whether these two are anomolous or not.

TuLLy: Allan Stockton at Hawaii has to my mind definitive new evidence

regarding the cosmological nature of QSO's. He has now about eight red-
shifts for galaxies near QSO's, which agree with the QSO redshift to
within roughly 200 km s-l, or one part in three hundred. Most of the
redshifts are determined from the continuum break at 4000 Rand H+K
absorption lines.
THE SURFACE DENSITY OF QUASARS 303

van der Laan: Would Dr Schmidt care to comment on the results of Green
and himself in the light of Dr Bolton's results?

Schmidt: The steep gradient of optical counts of quasars with

magnitude, between the Green-Schmidt (B < l5~7) and the Braccesi
(B < 18~0) surveys, applies only to objects with ultraviolet excess,
since quasars in both surveys fulfil (U-B) < -0.5. We can say nothing
as yet about the quasars with redder (U-B) colours (with redshifts less
than 2.5), if they exist.

Bolton: In our 2-hour field non UV excess objects with strong Mg II in

the B band make an appreciable contribution to the low redshift numbers.

Tammann: I would like to add that Dr Stepe at Basle has been using UG as
well as UV plates to search for ultraviolet excess objects. Using the
UG plates which are taken with narrower filters than the UV plates, he
finds 20% more ultraviolet excess objects than with the normal UV plates,
in agreement with what you said.

Khachikian: I should like to point out that there are also strong
differences between Markarian galaxies which have strong uv continuum
and those which have not. The galaxies with strong UV continuum usually
have active nuclei and most of them are the Seyfert galaxies.

A SURVEY PROGRAl~ FOR QSO AND RELATED OBJECTS

H. Lorenz

I would like to give the first results of a survey programme for

high redshift quasars and related objects on the Tautenburg-Schmidt
plates. It is known that objective prism plates are very useful for
detecting these objects. The smaller the dispersion of the spectra,
the fainter are the objects that can be found. Our Schmidt correcting
lens gives spectra with the dispersion of 2500 R
mm- I at Hy. The limit-
ing magnitude is about 19m in the B range. At this very low dispersion
reliable classification of the objects requires the extraction of all
the information stored on the plate. That means that a quantitative
evaluation should supplement visual insepction.

First we detrmine the wavelength of spectral features in the

suspected object using a wavelength scale, which has been determined by
the position of reference stars on a direct Schmidt plate and the posi-
tion of spectral features of these stars on the prism plate. The accur-
acy is sufficient to determine redshifts larger than 0.04. The main
purpose is the determination of the intensity distribution in the
spectra. The characteristic curve (as a function of A) is derived from
the known average intensity distributions in stars of the same spectral
types as the references. The relatively crude estimate of the spectral
304 SHORT COMMUNICATION

types is compensated by the large number (~ 30) of stars. The measure-

ments of the plates are performed on an automatic microdensitometer.

Up to now we have surveyed visually a field of 90 square degrees

and found 15 emission line objects. If these are really quasars the
surface density is 0.16 per square degree. The first application of the
method to these quasars described above proves that the quantitative
analysis of very low dispersion spectra increases considerably the power
of visual inspection.
THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES

M. S. Longair
Mullard Radio Astronomy Observatory
Cavendish Laboratory, Cambridge

1. INTRODUCTION

In this brief survey, three topics will be described:

(i) the large scale distribution of extragalactic radio
sources on the celestial sphere;
(ii) the problems of identifying optically quasars and radio
galaxies in that region of the source counts where they
diverge most significantly from the predictions of all
uniform world models;
(iii) the problems of interpretation of the source counts, some
models for the spatial distribution of sources and the
most important observations for defining more precisely
the evolution of the radio source population.

Nowadays, all source counts are presented in normalised, differen-

tial form, meaning that the number of sources ~N in the flux density
interval S to ~S is normalised to the prediction of a locally Euclidean
world model ~No cr S-5/2~S. A number of recent source counts at frequen-
cies 408, 1400, 2700 and 5000 MHz are presented in the form ~N/~No in
Figure 1; in this diagram, the normalisation at different frequencies
is arbitrary. The most recent data were presented at IAU Symposium
No.74 "Radio Astronomy and Cosmology" which has just been published
(Jauncey 1977).

It is well known that all the counts shown in Figure 1 contradict

the predictions of uniform world models. As an example, in Figure 2,
the most recent counts at 408 MHz are compared with the predictions of
a uniform Friedmann world model having ~ = 0; similar results are
obtained for all values of~. The prediction is based upon a knowledge
of the radio luminosity function derived from samples of sources at high
flux densities for which the optical identification percentages are
greater than 90%. A similar prediction is expected for counts at the
other frequencies shown in Figure 1.
305

M. s. LongGlr and J. Einasto (eds j, The Large Scale Structure of the Universe, 305-314. All Rights Reserved.
Copyright © 1978 by the [AU
306 M. S. LONG AIR

10 ~ __ 408 MHz

~ ~
:rrf' . ,_ ~ 1410 "-

~-'I"~
, ,~'...I.
____ T ' TT~
2700

~-~~
1·0
FLUX DENS lTV (Jy)

Figure 1. Differential counts of radio sources at 408, 1400, 2700 and

5000 MHz. For references, see Wall (1977). For more recent data, see
Jauncey (1977).

h-N
~o

10
""""
100 1000 "lUI

Identification
Percentoge

I-
-25% 65-70% 100%
Ion deep 4m plales ~40·1.,
de RUTler elol.19771

Figure 2. Comparison between the counts of radio sources at 408 MHz

and the predictions of world models (Pearson 1977). The solid line is
for a uniform model having ~ = O. The dashed line is the prediction
of the evolutionary model described in Section 4. At the bottom of the
diagram, the integral percentage identifications at different flux
density levels are indicated.
THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES 307

It will be noted that the maximum in the differential counts is

most pronounced at low frequencies. At 2700 and 5000 MHz, the counts
are much "fl1tter" and indeed at the latter frequency, they follow the
law ~N oc S-5 2~S over a wide range of flux densities. However, because
the sources counted extend to large redshifts, this result contradicts
the predictions of uniform models in which the exponent of the differ-
ential distribution is significantly smaller than 5/2 as shown in
Figure 2.

2. THE ISOTROPY OF THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES

The source counts shown in Figure 1 are nowadays based upon very
large samples of radio sources, mostly lying in directions away from
the Galactic plane, and hence they provide important information about
the isotropy of the Universe on a large scale. In Table 1, the numbers
of sources at various flux densities and frequencies suitable for such
studies are listed.

Table 1
Large surveys of radio sources

Frequency Survey Flux density range Numbers of

(MHz) (Jy) sources

178 2 - 10 5000 PS
4C
'"
408 Bologna B2 ~ 0.2 '" 9000 PS
Molonglo ~ 0.1 '" 4000 PS
5C 0.01 1000 PS
~
'"
1400 GB 0.1 1000 PS
~
'"
Westerbork ~ 0.01 '" 1800 B

2700 Parkes 0.1 7500 PS

~
'"
Greenbank
5000 ili 0.1 '" 800
& Parkes

The surveys indicated by the letters PS have been analysed by

Webster (1977) using the technique of power spectrum analysis, similar
to that developed by Peebles and his colleagues for studying the distri-
bution of galaxies on the celestial sphere. This technique is very
sensitive to any anisotropies in the source distribution and Webster
shows that, except on the largest scales, it is the optimum test of
isotropy. The Westerbork deep surveys, marked B, have been analysed by
308 M. S. LONGAIR

the technique of mUltiple binning analysis (Willis et al 1977). (These

surveys have now also been subjected to power spectrum analysis with
similar results - Oosterbaan, private communication.)

According to these analyses, there is no evidence that sources are

not distributed independently at random on the sky. The conversion of
this result into a limit to the statistical fluctuations in the source
distribution in space depends upon a knowledge of the typical distances
of the sources. According to the models described below (and on
general grounds), the majority of sources studied probably lie in the
redshift range 1 ~ Z ~ 3. On this basis, Webster (1977) quotes a limit
to the amplitude of the fluctuations ~N/N ~ 3% on a scale of 1 Gpc,
i.e. if one moves a cube of side 1 Gpc about the Universe, the varia-
tions in the number of sources counted is less than 3%. This limit is
derived from the largest sample of sources studied, the Bologna B2
catalogue (see Fanti, Lari and Olori 1977), and is limited only by the
finite size of the sample (N ~ 10 4 ). To obtain stronger limits, much
larger surveys of ?ources are required, the limit to the isotropy being
proportional to N-2. In addition to the above limit, limits to the
isotropy on a wide range of physical scales can be obtained down to the
typical angular distance between sources (Webster 1978).

The significance of these results is twofold. First, they provide

upper limits to the covariance function for the distribution of matter
in the Universe on much greater physical scales than has been possible
for galaxies and clusters of galaxies. Second, the obvious comparison
is with the upper limits to fluctuations in the microwave background
radiation. It must be emphasised that the present result refers to
limits to the fluctuations in the matter distribution whereas the
microwave background radiation tells us only about the radiation content
observed by us now. It is well known that if there is early reheating
of the intergalactic gas, the amplitude of temperature fluctuations of
the background radiation can be strongly damped and hence the present
limits to these fluctuations could be consistent with large fluctuations
at the epoch of recombination, Z ~ 1500. In some ways, the present
limit ~N/N ~ 3% is therefore stronger than the limit from the background
radiation because fluctuations in the matter distribution grow with time
as ~N/N ~ (l+Z)-l. Thus at Z = 1500, (~N/N)matter~ 0.03 x (3/1500) x
10- 4 . Notice that at Z = 1500, this limit refers to scales much larger
than the ·horizon.

3. PROBLEMS OF INTERPRETATION OF THE SOURCE COUNTS

If the redshifts of all the sources now observed were known, it
would be a relatively straightforward matter to derive directly from
observation the evolutionary history of the radio source population.
The best method would involve a variant of the V/Vmax technique
described by Schmidt (this volume). Unfortunately, at present,
distances can only be measured by first identifying the sources optical-
ly and then measuring the redshift of the optical object. The problem
THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES 309

can be understood from Figure 2 in which the optical identification

percentages at different flux density levels are given. At the very
highest flux densities, S408 ~ 15 Jy, optical identifications and
redshifts are more or less complete. However, at lower flux densities
fewer and fewer sources can be identified optically and very few
redshifts are available. At the lowest flux densities, only about 25%
of the sources can be identified using deep 48 inch Schmidt plates;
higher percentages, ~ 40%, have been reported by de Ruiter et al (1977)
who obtained very deep plates with the Mayall 4-m telescope. Quasars
can now be identified relatively easily because the radio positions are
of high precision and,because of their strong emission-line spectra,
redshifts can be measured without much difficulty. Schmidt (this volume)
has described how successfully this work is proceeding. For radio
galaxies, however, the situation is much less satisfactory. Even at the
limits of the largest telescopes under conditions of excellent astronom-
ical seeing, radio galaxies can only be identified to redshifts Z ~ 1
and it is very difficult to measure their redshifts.

A further problem in interpreting the counts even at high flux

densities can be seen in Figure 2. Just at the point where the diver-
gence between the predictions of world models and the observations
becomes large (S408 ~ 5 Jy), the optical identification percentage
decreases. It is certain that part of this divergence is due to the
steep source count of the quasars (see Schmidt, this volume). However,
the remaining unidentified sources also have a very steep source count
and it is important to discover the nature of these sources.

We have recently completed a deep optical survey of the fields of

unidentified 3CR radio sources which form part of a complete statistical
sample of sources with S178 ~ 10 Jy, corresponding roughly to S408 > 5
Jy (Laing, Longair, Riley, Kibblewhite and Gunn 1978). We were lucky to
have one night of excellent astronomical seeing at the Hale 200-inch
telescope which enabled us to make 10 new identifications of very faint
radio galaxies having 20 < m < 23; no new quasars were found. The
resulting optical identification statistics for a complete sample of
60 3CR radio sources is given in Table 2. The source counts (N(~ S) ~
S-B) and V/Vmax test for this sample are given in Table 3. For radio
galaxies without redshifts of which there are 14 examples in the sample
a conservative lower limit to their redshift of 0.3 has been adopted;
the values of V/Vmax for radio galaxies are therefore lower limits. It
can be seen that the slope of the integral source count B and the value
of <V/Vmax ) for all sources and quasars are typical of much larger
samples of sources. The radio galaxies have a source count and
(V/Vmax) greater than those expected for a uniform distribution although
as yet not with great statistical significance. The importance of the
present work is, however, that only 4 out of 60 sources in the complete
sample are either doubtful or unidentified and hence there is unlikely
to be any other unknown class of source contributing to the overall
source counts in addition to radio galaxies and quasars. If this result
is typical for all bright radio sources, it can be seen that strong
310 M. S. LONGAIR

Table 2
Optical identifications in a complete sample of 60 3CR radio sources

Type of object Number Percentage of total

Quasars 23 3st )
) 93!
Galaxies
(Certain + Confirmed + New) 33 55 )

Possible identifications
with galaxies 2 3!

Empty fields 2 3-~

60 100

Table 3
Source counts and the V/V a test for the statistical sample of
60 3~Rxradio sources
i3 (V/Vmax ) 0*

All sources 1.88 ± 0.25

Quasars 2.21 ± 0.5 0.709 3.5
Galaxies 1.59 ± 0.29 0.581 1.6
Weak radio galaxies
1.06 ± 0.42 0.52 0.2
P178 < 10 26 WHz-1sr- 1
Strong radio galaxies
1. 92 ± 0.42 0.610 1.9
P 178 > 10 26 WHz-1sr- 1

*0 = significance in standard deviations of difference of

(v/vmax) from 0.5.

radio galaxies exhibit strong cosmological evolutionary changes with

cosmological epoch, similar to those of quasars.

4. INTERPRETATION OF COUNTS OF RADIO SOURCES

The general features of satisfactory models of the radio source
population are well known:

(i) The evolution must be strong and for the most powerful sources an
evolution law similar to that of quasars is satisfactory f(Z) ~ e-mt
where t is cosmic time and m = 10 Ho for n = 0;
THF LARGE SCALE DISTRIBUTION OF RADIO SOURCES 311

(ii) Only the most powerful sources evolve in this manner or else the
convergence of the counts at low flux densities cannot be reproduced;
(iii) A cut-off to the source distribution at large redshifts may be
required in some models to achieve satisfactory convergence of the
source counts.

Wall, Pearson and I (1977) have developed a fast numerical proce-

dure for testing models of the source population and comparing the
predictions of the models statistically with all the avilable observa-
tional data. This work was motivated (i) by the recent vast improvement
in the statistics of the source counts, (ii) the wealth of new identifi-
cation data at high and low flux densities and (iii) the need to know
which are the most sensitive tests of the models.

We found that none of the many models in the literature was

acceptable when subjected to a proper statistical comparison with the
present source counts. We found a number of models, some of which
incorporated a cut-off in the source distribution at large redshifts.
I find most intriguing a model in which one uses the V/Vmax of all
sources in complete samples at high flux densities to define the varia-
tion of the exponent of an exponential evolution function M(P) as a
function of radio luminosity. A best fitting model has M(P) = 11 for
P408 > 10 27 WHz-1sr- 1 , M(P) = 0 for P408 ~ 10 26 WHz-1sr- 1 and varies
linearly with log P 408 between these values. This gives a remarkably
good fit to the overall counts (see Figure 2).

The most sensitive test of the

models is the identification content
of samples of faint radio sources.
For example, the predicted redshift
distributions for three of the models
at 8 408 = 0.01 Jy are shown in
Figure 3. What is at present avail-
.01 0.1 to 10
able from observation are identifica-
10- tion percentages for some of the 5C
and Westerbork surveys which have
been studied with deep 53-inch
5- Schmidt and 4-m plates respectively.
Richter (1975) finds that more than
35% of the sources are associated
0
.01 OJ to 10 with red objects; de Ruiter et al
(1977) find that 28% are galaxies,
10-

Figure 3. Predicted distribution of

5-
redshifts for sources with 10 < 8 408
< 50 rnJy. The models are described
0 in Wall, Pearson and Longair (1977).
.01 Q1 to 10 In model (b) there is a cut-off in
the source distribution at Z = 3.5.
312 M. S. LONGAIR

11% quasars and 7% unclassifiable. These results suggest that in

satisfactory models, 30-40% of the sources should have redshifts Z ~ 1.
Interpreted literally, this would exclude models such as (a) and is
barely consistent with model (c). These results are not yet strong
enough to prove that there must be a cut-off at large redshifts as in
model (b) but they are suggestive and indicate the areas in which
further identification work is urgently needed.

5. FUTURE OBSERVATIONS

The importance of performing deep optical identification and

redshift observations at all flux densities has been emphasised above.
Particular attention should be paid to the completeness of the samples
and as a first step intensive studies of complete samples of, say, 100
sources at flux densities S ~ 1 Jy are most valuable. In addition to
optical identifications, the radio properties of these sources should
be studied to provide further physical information about how the sourcef
themselves change with luminosity and redshift. It is to be hoped that
eventually we will obtain an overall picture of how radio-source
activity has changed with cosmological epoch. This evidence will pro-
vide complementary information about the evolution of the Universe as a
whole to that obtained from optical and X-ray studies of galaxies and
quasars.

The above analyses have been restricted to the interpretation of

low frequency surveys (v ~ 1000 MHz). There is no simple ,way of relat-
ing these results to high frequency source counts (see Wall, Pearson
and Longair 1977 for details). However, we can assert with confidence
that the same types of programme outlined above will be of the greatest
importance at these frequencies in order to obtain a complete picture
of the evolution of the radio source population at radio wavelengths.

REFERENCES

Fanti, C., Lari, C. and Olori, M.C., 1977. Astron. Astrophys., in

press.
Jauncey, D.L. (ed.), 1977. "Radio Astronomy and Cosmology", IAU
Symposium No.74, Reidel Publishing Company.
Pearson, T.J., 1977, Ph.D. dissertation, University of Cambridge.
Richter, G.A., 1975. Astron. Nachr., 296, 65.
de Ruiter, H.R., Willis, A.G. and Arp,~C., 1977. Astron. Astrophys.
Suppl., 28, 211.
Schmidt, M., 1977. This volume.
Wall, J.V., 1977. "Radio Astronomy and Cosmology", op. cit., 55.
Wall, J.V., Pearson, T.J. and Longair, M.S., 1977. "Radio Astronomy
and Cosmology", op. cit., 269.
Webster, A.S., 1977. "Radio Astronomy and Cosmology", op. cit., 75.
Webster, A.S., 1978. Mon. Not. R. astr. Soc., in preparation.
Willis, A.G., et al., 1977. "Radio Astronomy and Cosmology", op.cit. ,39.
THE LARGE SCALE DISTRIBUTION OF RADIO SOURCES 313

DISCUSSION

ZeLdovich: Is the cut-off at large redshifts certain? I remind every-

one that power-law evolution f(z) : (l+z)n needs a cut-off, because as
z + 00, f(z) + 00. An exponential law is a remedy, but does it need a
cut-off? Is the cut-off in optical QSO identifications due to observa-
tional difficulties because Ly-a is shifted into the near infrared part
of the spectrum?

Longair: I emphasised that there is no unambiguous answer to this

question. All we have done is to indicate how we may be able to detect
a cut-off using observations of complete samples of radio sources. The
present data are more easily accommodated in exponential models, which
include a cut-off but I would not exclude the absence of a cut-off yet.

Schmidt: While there may be a need to introduce a cut-off in the density

law to explain radio source counts, as mentioned by Dr Longair, there is
at present no evidence of a cut-off for quasars from statistical samples.
From the redshift distribution at B : 17.5, we can predict that quasars
with z > 3.5 should appear in significant numbers beyond B : 19.5. Since
the ultraviolet excess will not show at these redshifts, we have to rely
on objective prism or grating surveys. Osmer has suggested that the
Hoag-Smith 4-metre survey may be incomplete beyond B : 19.2. If so, the
absence of z > 3.5 quasars in this survey is not (yet) a serious argument
for a cut-off at this redshift.

Ostriker: It is easy to show that, in general, galaxy formation was

relatively recent; if galaxies formed at z ~ 10 the cosmic density was
so high that the resulting system would have a velocity dispersion much
larger than the values of ~ 200 km s-1 seen in typical galaxies.

Ozernoy: Could you give an upper limit to the exponent of the evolution
law for weak radio sources using the condition of convergence at small
radio flux densities?

Longair: Not without quite a lot of model computations. Weak evolution

is allowed, but I am most reluctant to quote any specific figure.

SiLk: What is the angular scale associated with the upper limit of
microwave background anisotropy that was inferred from the study of the
source distribution?

Longair: Webster quotes a figure of ~N/N < 3% on a scale of 1 Gpc at a

typical distance of 6000 Mpc, i.e. an angular scale of 10 0 -20 0 • "Notice
that fluctuations on these scales only came within the horizon at late
epochs, z '\, 6.

Peebles: I might remark that the observed pos1t1ve "cross-correlation"

between Lick galaxy counts and 4C radio-position does not require that
the 4C objects have optical brightness greater than B : 19, only that
314 M S.LONGAIR

some fraction of these objects belong to groups or clouds of galaxies

having some members brighter than 19.

Longair: This may certainly help resolve some of the discrepancy. It

is generally the rule that radio sources are associated with galaxies in
small groups as mentioned by Dr Bolton.
SEARCH FOR PRIMORDIAL PERTURBATIONS OF THE UNIVERSE:
OBSERVATIONS WITH RATAN-600 RADIO TELESCOPE

Y. N. Parijskij
Special Astrophysical Observatory,
Stavropolskij Kraj, USSR

All kinds of primeval perturbations of the Universe should result

in fluctuations of the microwave background radio emission. Here we
report our latest upper limits to these fluctuations on scales 5' to 30 •
Using the new 600-m Soviet Radio Telescope we obtained a mean tempera-
ture profile of the region from OSh to ISh in R.A., centred at the
declination of the Coma Cluster. 20 good records of this region were
used in the final reduction of the data. After "normalization" of these
data by filtering out low-frequency atmospheric noise and "bursts" which
exceed the 4a level we calculated an upper limit to the fluctuations of
the microwave background radiation.

The following

Here a(T A) is the observed dispersion of the antenna temperature

(on a given angular scale);
a(T B) is the dispersion of the intensity of the microwave
background radaition;
nl,n2 are the instrumental correction factors (spillover effect
and finite beam size effect);
n is the number of independent points Ln the scan;
r is the mean correlation coefficient of two independent
sets of measurements;
up is the quantile of the distribution of deflections of the
measured antennna temperature from the mean.

The results of these calculations are shown in Table 1.

All values of a(TB)/T are lower than the predictions. The smooth-
ness of the microwave background radiation is amazing and needs to be
explained.

315
M S Longalrand J. Emasto (eds.), The Large Scale Structure or the Unzverse, 315-316. All Rights Reserved.
COPYrIght © 1978 by the [AU
316 Y. N. PARIJSKIJ

Table 1

Scale cr(TA) cr(TB)/T

(arcmin) (xlO- 4 K) (20'-level)

5 5 < 0.8 x 10- 4

10 3.5 < 4 x 10- 5
20 2.4 < 3 x 10- 5
50 1 < 2.4 x 10- 5

75 0.8 < 1.9 x 10- 5

125 0.6 < 1.8 x 10- 5
150 0.4 < 1.3 x 10- 5

DISCUSSION

Boynton: What do you do about discrete sources below the 40' level
(i.e. the level above which discrete sources are removed)? You might
expect to be limited by such discrete sources at this wavelength?

Parijskij: We now have a new programme for finding discrete sources.

The source density which we find at about 1 mJy or less is about
10 4-10 5 sr- 1 which is consistent with the extrapolation from source
counts at higher flux densities. There has been some misunderstanding
about how discrete sources affect our previous results. I may comment
on this later in the general discussion.

Davis: Have you performed an F test on your data to determine whether

the null hypothesis (i.e. no excess noise above receiver noise) is
acceptable?

Parijskij: F test is not the best method in our case. We have checked
the statistics of the deflections by a X2 test and shown also that the
correlated part of the observed fluctuations in different sets of obser-
vations is uncorrelated, the correlated part of the noise being ~ N-l/2
on a relative scale, N being the number of independent samples.

Zeldoviah: What is the beam-shape of the telescope with which you made
the observations?

Parijskij: About 2' x 20" arc at 4 em wavelength. In both directions

the beam is roughly Guassian.
THE QUEST FOR FINE-SCALE ANISOTROPY IN THE RELICT RADIATION

Paul E. Boynton*
Department of Astronomy, Harvard University, and
Harvard College Observatory
Cambridge, Massachusetts

Up to Symposium Session IV, the observed properties of the large

scale structure of the Universe have been discussed in terms of studies
based on the spatial distribution of galaxies with redshift less than
one; studies which indicate clustering on characteristic scales up to
~20 Mpc or equivalently ~1015MQ. Such a survey is rather local in scope
compared to what in principle could be inferred from the measurement of
temperature fluctuations in the relict radiation surviving from Z ~ 1000
as the fossil imprint of "primordial" density fluctuations on scales up
to the horizon at the epoch of decoupling, M ~ 1019Me. However, the
investigation of structure over a still-larger sample volume and scales
greater than 1015Mg is not a possibility which I wish to stress in this
paper. Rather, given our present state of ignorance about the forma-
tion of structure, perhaps a more fruitful first approach would be to
examine small angular scales in the relict radiation for insight into
the evolution of density perturbations by a careful comparison between
the observationally inferred "initial" spectrum of mass inhomogeneities
present back at the epoch of decoupling, and the mass spectrum of clus-
tering which characterizes the present Universe.

Unfortunately, reference must continually be made to the possibil-

ities implicit in a study of fine-scale temperature anisotropies because
none have yet been observed. However, the upper limits on amplitudes
~ various angular scales are being forced lower and lower each year
by determined observers; and if this process continues, we may be forced
into a situation reminiscent of the paradox posed by the extreme iso-
tropy of the relic radiation-on large scales (that is,'on scales large
enough to comprise regions which we naively expect to be acausally re-
lated). The corresponding fine-scale predicament might be stated: "How
can the early Universe appear so isotropic on angular scales correspond-
ing to currently existing mass associations?" Or,"How can we explain
the evolution of the highly structured Universe observed today from a
virtually featureless past?" In fact, we are not yet compelled to take
such an extreme position for, as shown below, the current upper limits

*On leave from the University of Washington, Seattle, Washington

317

M. s. Longair and J. Emasto (eds.), The Large Scale Structure of the Universe, 317-326. All Rights Reserved.
Copyright © 1978 by the 1AU.
318 PAUL E BOYNTON

are not so low as to pose a paradox. On the contrary, most of us prob-

ably expect that an order of magnitude improvement in sensitivity will
inevitably reveal the texture of the early Universe. Despite that mys-
tical confidence, the detection and measurement of temperature/density
fluctuations present during the recombination era remains a primary goal
of observational cosmology (Rees and Sciama 1969, Weinberg 1972).

The following sections of this brief review are devoted first to a

summary of those elements of gravitational instability theory which are
necessary to establish a meaningful comparison between theory and fine-
scale anisotropy data, and also to motivate the direction of future ob-
servational effort. Currently available upper limits on temperature
fluctuations are then tabulated; many announced within the past few
months are still unpublished at the time of this Symposium. Finally, a
possible "next-generation" fine-scale observing technique is suggested;
one with adequate sensitivity to force a confrontation between theory
and fact.

I. GRAVITATIONAL INSTABILITY - OBSERVATIONAL IMPLICATIONS

Many currently envision the evolutionary transition from the rela-

tively featureless, homogeneous, early Universe to the manifestly inhomo-
geneous cosmos of the present epoch, to proceed through the growth of
density perturbations due to gravitational instability. Within the con-
text of the hot Big Bang, this process yields a rather rich variety of
behavior with details which depend on the mass scale of the particular
perturbation (Field, 1967; Silk, 1968; Rees and Sciama, 1969; Sunyaev
and Ze1'dovich, 1970; Peebles and Yu, 1970; Silk, 1974). However, for
purposes of the following discussion, we consider only the final phase
of the condensation scenario beginning with the recombination of hydro-
gen and the consequent decoup1ing of matter and radiation. Having de-
fined the problem in this way, we must postulate the presence of "seed"
perturbations at decoup1ing (a self-serving gesture, since their origin
is considered by many to be inscrutable). Even so, the existence of
such perturbations at some minimum amplitude is clearly a logical neces-
sity, and observational verification (at Z ~ 1000) of a mass spectrum of
density inhomogeneities appropriate to the precipitation of the current
structural hierarchy is a fundamental test of not only the instability
picture, but indirectly of many of the details of the Big Bang hypothesis.

An essential question to ask of the theory is to require an estimate

of that minimum necessary perturbation amplitude. The value turns out to
be rather large because of a curious "difficulty" with the emergence of
structure through gravitational instability; namely, that the growth of
density perturbations in a Friedmann Universe is merely a power law
(Lifschitz, 1946). In the post recombination era perturbations behave as:

(1 + Z)-l (1)
THE QUEST FOR FlNF-SCALE ANISOTROPY IN THE RELICT RADIATION 319

from Z = Zrec up to Z = Zf where 1 + Zf = l/~o' That is, in the linear

theory (low density contrast) perturbations cease to grow for Z < Zf'
Therefore, to assure the formation of high-contrast structure, the non-
linear regime, 00/0 ~ 1, must be approached for Z ~ Zf. Through this
simple condition, the magnitude of density perturbations at recombina-
tion (Zrec ~ 1000) is approximately specified:

[o~jz >
1 + Zr:
1 + Zf [~1
(2)

r f
Thus, for
[o~z f
~
1

.§£ > 10- 2 for ~ 0.1 (3)

0
0
or, .§£ > 10- 3 for ~ 1.0 • (4)
0
0
These are hardly small perturbations in the customary sense; and corres-
pondingly, such density perturbations, whether adiabatic or isothermal,
are expected to produce sizeable temperature anisotropies in the (decoup-
ling) radiation field.

This coupling between density and temperature fluctuations, either

through the establishment of LTE conditions and/or through scattering off
matter participating in large-scale motions, has been discussed by several
authors (Silk, 1968; Sunyaev and Zel'dovich, 1970; Silk, 1974), who also
point out a variety of damping and averaging processes which selectively
diminish density and temperature fluctuations on mass scales smaller than
1015Me. A representation of the relationship between observable frac-
tional rms temperature fluctua~ions, ~T/T, and perturbation mass (or
equivalent angular scale, eaMl/3) due to Sunyaev and Zel'dovich (1970)
is shown for several values of ~o in Figure 1. Each curve is defined
so that 00/0 = 1 at 1 + Z = l/~o'*

II. OBSERVATIONAL TEST - THE STRONG FORM

In Table 1 are listed the currently available upper limits on fine-

scale fluctuations in the relict radiation. The more stringent of these
limits are also plotted in Figure 1 and labeled by the Table entry number.

*The possible consequences of reheating of the intergalactic medium are

not discussed here. Although it seems difficult energetically to produce
an optical depth greater than unity back to Z ~ 1000, even in that case
the imprint of early density perturbations should be observable in the
relic radiation (see paper by Sunyaev in this volume).
320 PAUL E. BOYNTON

1----------,,~----_r'-------(9)

1----~--~--~---(5)

I---_r--------:;r----(IO)

p'Sp = lot no
I+Z=..!..

2 5 20 50 100
ANGULAR SCALE, 8 (orcminutes)

Figure 1. Fine-scale temperature fluctuations, ~T/T, as a function of

angular scale. Bold curves represent the calculations of Sunyaev and
Zel'dovich (1970). Horizontal bars indicate current observational lim-
its identified in Table 1. The critical region is defined in the text.

Within the context of the theoretical expectations presented in Figure 1,

these 6 observational limits argue against the evolution of high contrast
mass associations on scales greater than ~1016MQ for 0.03 < Qo < 1.0.
Although this is an interesting result, there is no compelling evidence
for or against such associations at the current epoch.

As already suggested, perhaps the crux of the confrontation between

observation and theory lies in searching for primordial fluctuations
(those present at decoupling) on mass scales for which we know high con-
trast associations exist at the current epoch, that is, fo~~ 101~.
The theory then demands that temperature fluctuations be observed at a
specified amplitude for a given Qo ' This "strong form" of the test for
primordial fluctuations is characterized (for Qo = 0.1) by the double
hatched region in Figure 1. That is, in an no = 0.1 universe, any obser-
vational effort which penetrates this region, or its lower extension,
should detect fluctuations. The right-hand boundary is somewhat con-
servatively defined in the sense that the fluctuation spectrum is
assumed to plunge to zero for M > 1015 MQ. The union of all such regions
THE QUEST FOR FINE-SCALE ANISOTROPY IN THE RELICT RADIATION 321

Table I

Observers Angular Scale A (cm) nT/T t

1. Conklin and Bracewell (1967) 10' 2.8 <1.8 x 10- 3

2. Penzias et al (1969) 2' 0.35 <6.0 x 10- 3
3. Boynton and Partridge (1973)* 1!5 0.35 <1.9 x 10- 3
4. Carpenter et a1 (1973) >2' 3.6 <7.0 x 10- 4
5. Parijskij (1973)** >5' 2.8 <4.0 x 10- 4
6. Stankevich (1974) 10'-20' 11.1 <1.5 x 10- 4
7. Caderni et al (1977) 30' 0.13 <1. 2 x 10- 4
8. Partridge (1977)*** 4' 0.9 <5.0 x 10- 4
9. Pigg (1977) >1!25 2.0 <7.0 x 10- 4
10. Parijskij (1977)**** >5' 2.8 <8.0 x 10-5

tGenerally quoted as 2a or 95% limit.

*Revised downward, see Partridge (1977).
**Revised upward, private communication (1977).
***Tentative upper limit pending completion of analysis.
****See paper by Parijskij in this volume.

for 0.05 < n < 1. 0 then generally defines the "critical region" for
such observa~ions, as labeled in Figure 1. The shape of the upper bound-
ary of the critical region is consistent with the previous simple argu-
ment about the requisite amplitude of density perturbations. That 1s,
to achieve high contrast mass condensations,

1
> (5)
'V
(1 + Z )
r
n
0

and since nT/T is a monotonic function of np/p, as the upper limit on

nT/T is forced lower and lower, no is restricted to larger and larger
values. As seen in Figure 1, for an observed nT/T < 3 x 10- 5 on all ang-
ular scales, no must be greater than one. In this way one sees that the
measurement of the amplitude spectrum of temperature fluctuations nT/T(G)
not only promises insight into the nature of the growth of mass associa-
tions, and possibly yields some clues to the origin of the seedperturba-
tions, but also may provide interesting constraints on that ever-elusive
label of universes, no.

III. CONCLUSIONS

1. Comparison of data and theory possible at this time suggests

that the density contrast of mass associations may be significantly de-
pressed for M > 1016Mo

2. The three-way confrontation between evidence for clustering

at Z ~ 1, the properties of density perturbations at Z ~ 1000, and the
322 PAUL E. BOYNTON

theoretical evolutionary connection between these observations--what I

have called the "strong" form of the test of these ideas--has yet to
take place. We need fine-scale anisotropy measurements on angular scales
down to one arc minute with an rms precision of ~T/T ~ 10- 5 •

3. The customary radio astronomy techniques previously employed

may not be adequate to carry these studies to the prec1s10n required
above. This last, as yet unjustified, conclusion motivates the follow-
ing discussion.

IV. NEW DIRECTIONS

How do we make observational incursions into the critical region

defined in Figure I? Temporarily putting aside limitations posed by dis-
crete emission sources, the major difficulty in achieving sensitivities
on the order of ~T/T ~ 10- 5 is the impracticality of attempting to sched-
ule adequate observing time on one of the few filled-aperture radio tele-
scopes capable of l' to 2' resolution. The integration time requirement,
T, to achieve ~T/T 10- 5 for N independent pairs of sky elements is given
by:
10 -1 2
T ~ 10 NB T seconds (6)
rec
where B is the receiver pre-detection bandwidth in Hz, and T the re-
ceiver noise temperature. Currently the best maser receiver~e~ield TIN
values of around 24 hours. Improvements in the foreseeable future might
reduce that time to ~ 12 hours. Thus a statistically meaningful sample
(N ~ 100) of sky elements would require concurrent use of a state-of-the-
art receiver and one of the largest radio telescopes for a large fraction
of a year even under optimal observing conditions.

Far infrared techniques are an alternative approach to the observa-

tional challenge posed by the critical region in Figure 1. Existing
high-throughput bolometer and filter combinations allow TIN values as
low as 10 minutes with a bandwidth narrow enough to utilize the atmos-
pheric transmission window near A = 1mm. Ground-based observing seems
an inescapable constraint because the requisite angular resolution (at
approximately unit throughput) implies apertures too large (i.e. ~ 10m)
to consider airborne or satellite platforms for at least the next decade.

Although detector noise is quite favorably small, customarily there

have been two principal objections to far infrared observations of fine-
scale anisotropy: the first is the spectre of atmospheric emission fluc-
tuations. However, work by Corsi et al. (1974) at A ~ lmm suggested
that low frequency fluctuations are not substantially larger than detec-
tor noise at a good site. Moreover, the use of beam switching comparison
of sky elements separated by only a few minutes of arc reduces atmos-
pheric fluctuation noise to the extent that such fluctuations are highly
correlated on small angular scales. Recent beam switching measurements
THE QUEST FOR FINE-SCALE ANISOTROPY IN THE RELICT RADIATION 323

over angles of 30 arc minutes (Cademi et a1., 1977) failed to record

any excess noise due to atmospheric effects at A ~ 1mm; and their aniso-
tropy limit of 6T/T < 10- 4 was consistent with detector noise limited
performance. These rather surprising results suggest that 6T/T ~ 10- 5
might well be attainable for beamswitching angles less than 3 arc minutes
under suitable observing conditions.

The second of the two objections concerns the contribution of dis-

crete emission sources to fine-scale anisotropy. The early investiga-
tion of observational constraints on fine-scale anisotropy measurements
by Longair and Sunyaev (1969) indicated that any reasonable extrapola-
tion of radio source counts to A = 1mm produce a confusion limit (one
source per beam) at a negligible level, ~T/T ~ 10- 9 • However, they cor-
rectly pointed to possible severe difficulties with "inverted spectrum"
sources and the then newly-discovered "infrared galaxies". Their con-
servatively estimated one-source-per-beam contribution at a level of
~T/T ~ 10- 5 reflected understandable pessimism. However, we now have
the benefit of recent A = 1mm survey by Ade, Rowan-Robinson and Clegg
(1976) augmented by further observations of Hi1deband et a1. (1977)
which allow significant constraints on the bright end of the source-
count function. Utilizing the Poo1ey-Ry1e N(S) function in the same
fashion as Longair and Sunyaev, we find that the source confusion limit
at A = 1mm is much smaller (6T/T < 8 x 10- 7 ) than originally estimated,
a result which is clearly crucial to this study.

In summary, there may be no severe practical or fundamental limita-

tions to achieving considerably improved measurements of fine-scale
anisotropy through a ground-based, high throughput, far-IR bo10metric
system. However, no survey of this type is meaningful unless the spec-
tral properties of the anisotropy are measured, thereby allowing the
identification of non-cosmological components. This crucial extension
of the simple technique proposed here implies exploitation of the 2mm
and 3mm atmospheric windows and possibly supplemental studies on larger
angular scales from the NASA C141 IR Observatory and even the proposed
NASA SIRTF Orbiting Infrared Telescope.

As determined observers continue to force the anisotropy limits

lower and lower during the next decade, we wait with that aforementioned
mystical assurance for the incipient texture of the early Universe to
come into view ••• but will it?
This work was supported largely by the Alfred P. Sloan Foundation.
1. Ade, P.A.R., Rowan-Robinson, M., and Clegg, P.E; 1976, Astr. Ap.
53, 403.
2. Boynton, P.E. and Partridge, R.B.; 1973, Astrophys. J. 181, 243.
3. Boynton, P.E.; 1974, in Confrontation of Cosmological Tb;Ories
with Observational Data, M.S. Longair (ed.), IAU Symposium 63.
4. Cademi, N., De Cosmo, V., Fabbri, R., Me1chiorri, B., Me1chiorri,
F., and Natale, V.; 1977, Phys. Rev. D. (to be published).
5. Carpenter, R.L., Gu1kis, S. and Sato, T.; 1973, Astrophys. J.
Letters 182, L61.
324 PAUL E. BOYNTON

6. Conklin, E.K. and Bracewell, R.N.; 1967, Nature 216, 777.

7. Corsi, R., Dall'Oglio, G., Fonti, S., Guidi, I., Melchiorri, B,
Melchiorri, F., Natale, V., Mencarag1ia, P., Lombardini, P.,
Trivero, P., Sivertsen, S.; 1974, IEEE Trans. on Microwave
Theory and Techniques, MTT-22 , 1036.
8. Field, G.B.; 1967, in Stars and Stellar Systems, ed, A. R. Sandage
and M. Sandage (University of Chicago Press), Vol. IX.
9. Hildebrand, R.H., Harper, D.A., Whitcomb, S.E., Winston, R.,
Stiening, R.F. and Moseley, S.H.; 1977, Ap. J. 216, 698.
10. Lifschitz, E.M.; 1946, J. Phys. USSR 10, 116. ---
II. Longair, M.S. and Sunyaev, R.A.; 1969:-Nature 223, 719.
12. Parijskij, Yu. N.; 1973, Astrophys. J. Letter 180, L47.
13. Parijskij, Yu. N.; 1977, to be published. ---
14. Partridge, R.B.; 1977, in preparation.
15. Peebles, P.J.E., and Yu, J.T.; 1970, Ap. J. 162, 815.
16. Penzias, A.A., Schram!, J., and Wi1sm,R.W.;"1969, Astrophys. J.
Letters 157, L49.
17. Pigg, J.C.; 1977, in preparation.
18. Rees, M.J., Sciama, D.W.; 1969, Comments Astrophys. Space Phys.
1, 140.
19. Silk, J.; 1968, Ap.J. 151, 459.
20. Silk, J.; 1974, Ap. J."194, 215.
21. Stankevich, K.S.; 1974, Sov. Astron. 18, 126.
22. Sunyaev, R.A., Zel'dovich, Ya. B.; 1970, Astrophys. Space Sci 2, 1.
23. Weinberg, S.; 1972, Gravitation and Cosmology, J. Wiley, N.Y.

DISCUSSION

Longair: Will the submillimetre experiment suffer from problems asso-

ciated with water-vapour fluctuations in the atmosphere? I recall that
Webster's experiments at 9 mm from White Mountain were ruined by them.

Boynton: Yes, but recall that Webster's two beams were separated by
many degrees, 90 0 I think, but one expects a high degree of correlation
between atmospheric fluctuations in beams that are separated by only
three or four arcmin. In fact, we have already found that the atmo-
spheric contribution to fluctuations in ~T/T is below 10- 4 on scales of
30 arcmin near A = 1 mm.

Fall: How much do your conclusions about the inevitability of seeing

small scale fluctuations depend on the assumption that no reheating
occurred?

Boynton: Even in the case of reheating one expects fluctuations stem-

ming from the doppler shift in scattering radiation off mass associa-
tions which have inevitably acquired motions through the condensation
process.

Fall: Yes, but if rehating lasted for a considerable period of time

then these mass motions would be nearly uncorrelated through the depth
THE QUEST FOR FINE-SCALF ANISOTROPY IN THE RELICT RADIATION 325

of the reheated gas and the observed fluctuations would be much reduced.

Boynton: It is certainly true that the "averaging" process you mention

will reduce the amplitude of fluctuations, and even more so if the
motions are anti-correlated as suggested by Sunyaev and Zeldovich. But
at some level, the fluctuations still seem inevitable even in the case
that reheating produces substantial opacity.

Gott: Have you computed a similar curve for an isothermal density

fluctuation spectrum?

Boynton: According to the treatment of Sunyaev and Zeldovich there is

only a factor of two or three reduction in 6T/T, on the angular scales
considered here, resulting from isothermal perturbations. Otherwise,
the curve is quite similar.

Parijskij: What is the best limit to 6T/T in the infrared region of the
spectrum?

Boynton: 6T/T < 10- 4 on a scale of 20. The limit is 6T/T < 2 x 10- 4
on a scale of 2 arcmin (see the written version of the lecture).

GENERAL DISCUSSION OF PAPERS BY PARIJSKIJ AND BOYNTON

Ozernoy: Recently Kurskov and myself have estimated the angular scale
on which measurements of temperature fluctuations are optimal in order
to discover primeval density or velocity perturbations. This angular
scale corresponds to the maximum value of 6T/T. For adiabatic density
perturbations the angular scale on which the expected value of 6T/T
attains a maximum corresponds to a mass approximately equal to the
Jeans' mass at the moment of recombination and is about e ~ (15'-30')h*,
independent of Qh2 and of the initial spectrum of perturbations which,
generally speaking, is uncertain in the adiabatic model. For turbulent
velocity perturbations, the value of 6T/T is expected to have a maximum
value for whirls of the maximum scale which corresponds approximately to
the same angle e ~ (15'-30')h.
It should be emphasized that when one uses beam switching, it is
necessary to have the antenna beam width not larger than eant ~ e/n ~
(5'-10')h where e is the angular size mentioned above because otherwise
appreciable averaging of fluctuations inside the antenna beam will be
important. This averaging is significant if eant ? (l+zemission) k- 1
(QHo/2c) where k is a wave number of the perturbations at the redshift
of emission.
*We use h = Ho /75 km sec-1Mpc- 1 .

Silk: I would like to mention that Richards and his colleagues at

Berkeley have flown a further balloon-borne telescope to measure the
spectrum of the microwave background radiation beyond the maximum of
326 PAUL E. BOYNTON

the Planck curve. Even the raw unreduced data show definite evidence
for the abrupt decrease in intensity at wavelengths ~ 1 mm expected of
a true Planck distribution.

Boynton: I would like to comment that although Prof. Parijskij's latest

results do not violate the "critical region" which I have defined in my
presentation, I do not mean to minimize their importance in possibly
limiting the perturbation amplitude spectrum for masses greater than
10 16 M0 • His is a significant contribution and I earnestly hope that
Prof. Parijskij will very soon publish a full account of his data, his
analysis procedure and his conclusions so that we may all examine the
methods by which he has reduced the fine-scale fluctuations limits by
almost an order of magnitude over existing observations.

Parijskij: Full details of the experiments and reduction procedures

will be published in Astron. Zh. Letters in December of this year.
COSMOLOGICAL INFORMATION FROM X-RAY OBSERVATIONS

Herbert Gursky
Center for Astrophysics
Smithsonian Astrophysical Observatory/
Harvard College Observatory
Cambridge, Massachusetts 01776,
USA

INTRODUCTION

The fact that x-ray astronomy could yield cosmological information

was apparent in the earliest rocket flights carried out in 1962 and 1963
(Giacconi et aI, 1962, Bowyer et al,1963). These flights established
that diffuse radiation was present around the sky; indeed one of the
first theoretical papers commenting on the x-ray results was by
Fred Hoyle (1963) showing that the observed background flux was below
that predicted for the hot, steady-state universe. A few years later,
the first discrete source was credibly identified with an external
galaxy (Byram, et a1. 1966).

Figure 1 illustrates the primary data that provides the basis for
an optimistic view that cosmological information is present. The
figure contains a single day of data from the Uhuru satellite taken
along a 50 great circle band, passing near the galactic pole. The
data show two prominent extragalactic objects, the Virgo cluster and
Centaurus Ai equally distinct is the high radiation level present at
all galactic latitudes with no perceptible variation in intensity.
This radiation is the diffuse x-ray background and in the Uhuru
detectors with ~ 25 deg 2 solid angle dominates the appearance of the
data.

It is now known that the diffuse radiation is isotropic and free

of small scale fluctuations with a high degree of precision. It is
also known that the superposition of known extragalactic x-ray sources
cannot account for the radiation. Thus, cosmological information
should be available on several counts.

1. The diffuse radiation must be originating at g >1 and may be

as remote as B ~ 10. It may be revealing conditions in the inter-
galactic medium that are not otherwise known.

2. The x-ray emission from individual extragalactic sources may

yield information about galaxies or other discrete sources bearing on
327

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 327-337. All Rights Reserved.
Copyright © 19 78 by the fA U.
328 HERBERT GURSKY

the large scale structure of the universe, the evolution of cosmic

objects or the chemistry of the universe.

In this paper, I will briefly review the present state of

understanding of extragalactic x-ray astronomy. I will present what
little cosmological information we have obtained to date and I will
discuss how and possibly where new information of this kind may emerge.
I am going on the presumption that the key areas are the nature of the
diffuse x-ray background and of the discrete x-ray sources. There have
been recent reviews of extragalactic astronomy by Horstman, Cavallo and
Moretti-Horstman (1975), Gursky and Schwartz (1977), and Culhane (1977).

THE DIFFUSE X-RAY BACKGROUND

The data shown in Figure 1 are in the energy range 2-10 kev.
There is also a strong component present below one kev that for a time
was believed to be extragalactic, however that component is strongly
correlated with respect to the Milky Way and it is not evident what
fraction originates outside the Galaxy. There is background radiation
extending to high energies as seen in Figure 2, which is from a
compilation by Schwartz (1974). Apparently the radiation extends to ~
100 Mev, however information on its isotropy is meager above about 100
Kev. cowsik and Kobetich (1972) demonstrated that the spectrum could
not be matched to an inverse compton process, but that it could be
tolerably well fit by a thermal spectrum with a temperature of ~ 3XI0 8 K.
This supports the idea that the radiation originates from a hot, inter-
galactic medium. The high energy data (above ~ 100 kev) probably
means that a second process is operating as well.

Fig. 1. Great circle distribution of x-ray counts seen by Uhuru. The

upper plot (x-2) is obtained by the SX5° collimator and ~ 80% of the
observed flux (exclusive of point sources) originates in the diffuse
x-ray sources. The lower plo.t (x-l) is from the ;;X5° collimator and
show a residual count rate of ~ 7 ct/sec almost all of which can be
attributed to non-x-ray events.
GREAT CIRCLE DISTRIBUTION
O,bits 201-215
Xl
SUMMED OVER 5°

,.
~ I

x,
lOr SUMMED OVER 5°

II
I
~~r
. ._ ..,~
....... ""-'- ..........h ............
,,,n.,+,,............ ~,.-, +~H"~'''''. '"
t I
'''''1
rI\ I
COSMOLOGICAL INFORMATION FROM X-RAY OBSERV ATIONS 329

Fig. 2. The Spectrum of the Diffuse X-ray Background. The dashed

line is an approximate fit of a thermal spectrum.

1- oso-m
O-LLL
~-ASE
~ - LEIDEN-NAGOYA

In the range of 2-40 kev the large scale isotropy (~ 20 0 ) is

established to about 5% (Schwartz, 1970). On a smaller scale, there
are present fluctuations of 3.6% on a scale of 25 deg 2 as shown in
Figure 3 (Schwartz, et aI, 1976, Schwartz, 1977)0 The upper limit to
the fluctuations is 4.6%. This result is consistent with the 10gN-
logS relation as derived from the 3U catalog to about a confidence 10%
level. The discrepancy, if any, is that the fluctuations are somewhat
smaller than is predicted from the source counts. However, considering
the possibility of systematic errors, the discrepancy is not to be
taken as an indication of a deviation from a uniform distribution of
discrete sources obeying a 3/2 power law. The more recent 10gN-loqS
curve derived from the 4U catalog (Forman, et aI, 1977) is consistent
with the earlier results. The observation of fluctuations cannot be
used to derive any information regarding the diffuse x-ray background.
In particular, we cannot derive the fraction of background originating
from a truly diffuse component. The reason for this is simply that
fluctuations must result from the stronger of the unresolved sources;
e.g., those only a factor of ten fainter than the ones presently
observed. On the other hand, to make up the entire diffuse background,
the observed 10gN logS curve must be extrapolated by ~ 10 3 •

Some fraction of the diffuse x-ray background must originate in

the fainter, unresolved sources. In fact, the observed high latitude
Uhuru sources contribute about 2% of the background; i.e., a coarse,
lower sensitivity experiment would measure the background flux about
2% higher than did Uhuru.
330 HERBERT GURSKY

Fig. 3. Observations of the diffuse x-ray background with Uhuru

(5 0 X5 0 collimator) around a number of great circles in the sky. Points
containing known sources are excluded. The point-to-point fluctuations
are greater than can be accounted for by poisson statistics.

UHURU SCANS

"f
200
190

"t
200
190
++t
+ ++++++
++

+++t+/t tt
ORBIT 98

t t+ ORBIT 145
++ + +++++t
+
1
-f'
' ... 210 + + ttt +
:: 200 ++ +t ORBIT 157
+
~ 190 +
(!) 210 +/t ORBIT 171 +
~ tttt ++++ tt tttt+t+
+
200
§ 190

+
++t / ++++ ORBIT 265+
200
t+t +
190 +
l-
+ \++ t+
t
+
tt
210 ORBIT 276
200 ++ + t H++++ ++++ ++
190
180
+ tt +
60 120 180 240 '00
RELATIVE AZIMUTH

The diffuse radiation may originate in the intergalactic medium.

As noted, the spectral distribution of the radiation is consistent with
a thermal origin; if so, it is difficult to avoid the existence in the
intergalactic gas of considerable mass, perhaps equivalent to ~ ~ 0.5.
However, the problem of heating the gas is not trivial and may comprise
evidence of an energy producing stage in the evolution of matter that
we have no evidence of now. (cf. Field and Perrenod, 1977)

DISCRETE EXTRAGALACTIC X-RAY SOURCES

The Third Uhuru catalog revealed the existence of a significant

population of x-ray sources at high galactic latitudes comprising
examples of normal galaxies, active galaxies and clusters of galaxies
(Giacconi, et al, 1974). The most extensive kind of source was that
associated with rich clusters of galaxies, as first discussed by
Gursky, et al (1972), and extended by Kellogg, et al (1973). There was
also revealed a class of sources that could not be identified with
either prominent galaxies or quasars. These results have been
extended in the Fourth uhuru catalog (Forman, et al 1977) and the
Ariel catalog (Cooke, et al 1977). These latter catalogs have revealed
the Seyferts (Class I) to be prominent X-ray sources as well. The
present results reveal that in a region of the sky surveyed to a
reasonable degree of completeness for both x-ray sources and at least
certain kind of Seyfert galaxies, 18 of 89 Seyferts can be identified as
x-ray sources (Tananbaum, et al 1977). For clusters of galaxies, the
COSMOLOGICAL INFORMATION FROM X-RAY OBSERVATIONS 331

results are even more dramatic. Jones and Forman (1977) using
combined Ariel 5 and Uhuru results, conclude that 33 Abell clusters
within distance class 4 are x-ray sources. An additional 15 clusters,
mostly in the south, are also found to be x-ray sources. As discussed
by Tananbaum, et al (1977), the unseen Seyferts will make up 10-20%
of the diffuse background, depending on the completeness of the
present optical surveys. The clusters will contribute an additional
~ 15% as discussed in this volume by Culhane. Thus, the identified
sources fall short by a factor of 3 or more in accounting for the
diffuse x-ray background.

More than half of the high latitude x-ray sources, presumably

extragalactic, are still "unidentified". For example, in the Fourth
Uhuru catalog, of ~ 200 sources for which ibi ~ 20 0 , only 92 carry
identifications. However, many of these identifications must be
chance coincidences. Some will be Seyfert galaxies since the optical
surveys for these objects are far from complete. Some of these may be
galactic as indicated by the recent discovery of soft x-ray sources
such as AM Hercules, but the distribution of the sources argues against
a galactic origin. What are the unidentified sources and will they
yield any cosmological information? The study by Bahcall et al (1975),
of ten such sources, found at best two quasar candidates (m ~ 17.5)
and one peculiar galaxy. Their conclusion was that if the objects are
associated with galaxies or blue stellar-like objects, the ratio of
x-ray to optical luminosity is at least 10 and possibly 100. Another
clue, although a weak one, is found in the logN-logS curve which is
consistent with a 3/2 powers law, and the small scale fluctuations in
the x-ray background which is consistent with a "naive" interpretation
of the logN-logS cun'e. These facts imply that the x-ray sources are
relatively nearby and not subject to strong evolution.

The unidentified sources may simply be telling us that active

galaxies are far more extensive than we now believe and there is a very
broad spread in the ratio of x-ray to optical luminosity of these
objects just as is the radio to optical luminosity. If this is the
case, at a minimum, a more complete study of extragalactic x-ray
sources should significantly improve our understanding of the nature
and evolution of active galaxies and since we know this extends to e
% 3 at least, provide some information on the early evolution of
galaxies generally.

CLUSTERS AND INTERGALACTIC GAS

There are numerous indications of gas in between the galaxies,

such as the morphology of double radio sources and the radio tails on
certain galaxies, however the most extensive evidence for such gas has
been the discovery of x-ray emission from clusters of galaxies. The
emission correlates well with cluster richness and velocity dispersion
and can be sensibly accounted for as resulting from a hot gas. The
discovery of iron line emission from several of these cluster sources
332 HERBERT GURSKY

is the culminating evidence for a thermal origin (see Culhane's

contribution to this volume for detailed discussion of x-ray clusters).

Since the angular size of the emitting region has been found
for several of the clusters, the mass of radiating gas is determined.
It is typically l013_l0 14m ~, namely comparable to the mass of the
visible galaxies, but still a factor of ten below the virial mass
for the clusters. Thus the gas does not resolve the problem of the
discrepancy between the vi rial mass and the observed mass, but it does
dramatically alter ones concept of a cluster of galaxies. Since there
is so much mass in hot gas, one can no longer look at the galaxy
distribution and make conclusions about the net gravitational field or
the degree of relaxation in the cluster.

The fact that iron is present in that gas at near normal abundance
is obviously of great significance. It is easy to say that the iron
means that the gas is derived from the galaxies themselves. However
with still 90% of the matter yet to be accounted for, such an easy
way out may be naive in the extreme.

Thus, the existence of hot gas in the clusters of galaxies is of

obvious significance to the large scale structure of the Universe. One
can for example imagine a society with only x-ray eyes to study the
Cosmos. Their view of the universe, at least as measured by "catalogs
of clusters" (in their parlance "hot gas clouds") would not be terribly
different from ours as derived visually. However, for these individuals,
galaxies would be virtually absent.

Now we come to a remarkable discovery by Murra~et al (1977),

albeit a somewhat tentative one. The Fourth Uhuru catalog revealed
six instances in which several distant Abell clusters were within or
near, the error box of a single x-ray source. Further examination
revealed that these clusters comprised associations listed by Abell
as superclusters. Murray,et aI, using a rigorous definition of "super-
cluster", found three whose centers were coincident with x-ray sources.
The result is statistically significant but still must be regarded as
tentative since other possibilities arise immediately. For one, it is
possible that a single cluster is the x-ray source. This cluster
would have a luminosity of order 10 46 erg/sec, far in excess of what
is seen for other clusters. The second possibility is that these are
very luminous x-ray sources associated with a galaxy within one of the
clusters. Also, the presence of the sources has yet be confirmed by
independent measurements. The crucial data will be the angular size of
the x-ray emitting regions. If the emission does originate in the super-
cluster, then the sources must show angular sizes of several degrees.

This could be a result of far reaching significance. For one the

mass of gas is estimated by Murray, et al to be ~ 10 16 M e, far in
excess of what is present in the visible galaxies in the region.
Secondly, it makes more plausible the idea that hot gas fills the
entire Universe, yielding up the diffuse x-ray background. Third, since
COSMOLOGICAL INFORMATION FROM X-RAY OBSERVATIONS 333

there is so much gas, it can hardly be argued that it originates from

galaxies. The gas could be primeval or it could result from a period
of collapse prior to the one that led to the observed galaxies. Either
way, a study of its chemical makeup, as is possible through x-ray spec-
troscopy, could reveal the conditions in a very massive component of
the early universe. Finally, it provides more reality to the concept
of supercluster, in fact the supercluster is the gas cloud since that
is where the mass is o

FUTURE PROSPECTS

So far, only a few hard pieces of cosmological information have

emerged from x-ray astronomy. Specifically, if there is a large amount
of intergalactic gas, its temperature must be a few times 108 K•
Furthermore, there is a substantial mass of hot gas associated with
rich clusters of galaxies and possibly with superclusters o

Also, there are aspects of x-ray astronomy that may eventually

yield cosmological information. Active galaxies (at least S~yferts)
and rich clusters of galaxies are x-ray sources as a class and also, a
large number of x-ray sources exist with Lx/Lopt > 10 and are not yet
identified. Iron line emission is seen in the x-radiation from the
cluster sources. Furthermore, the diffuse x-ray backgrounds is not
yet accounted for.

The prospects for extending the present results are very good.
The results so far have been obtained with very modest instruments
compared to what we know is possible. HEAO-l recently launched by
NASA is the first spacecraft that carries substantially more sensitive
instruments. The experiments on HEAO-l will extend the source lists
by at least a factor of 5, will extend the spectral coverage both at
low and high energy range, and will allow for more precise locations
and spectral data than has been possible to date. However, it is likely
to be HEAO B, scheduled for launch during Fall 1978, which may provide
major advances in cosmology. The HEAO-B instrument comprises a large
focussing x-ray telescope and four focal plane instruments o The tele-
scope is of quality comparable to that of ordinary optical telescopes --
the angular resolution over the central few arc minutes will be 2-3"
and the net field of view will be about one degree. The power of this
instrument comes from its great sensitivity, which derives from the
simple fact that the mirror concentrates the flux incident on its 1000
cm 2 aperture onto a resolution element of only 30~m diameter. The
result is that background is effectively eliminated entirely and sources
whose strength is only 10- 5 to 10- 6 photons/cm- 2 should be detectable.
This is between 10- 3 and 10- 4 of the current faintest detectable
sources. This sensitivity is so great that the question of the
diffuseness of the x-ray background can be resolved unambiguously.
Seyfert galaxies and clusters of galaxies can be seen to such great
distances that evolutionary effects must become evident. With positional
accuracy of 1", optical and radio identifications can be made
334 HERBERT GURSKY

independent of assumptions regarding the nature of the object. In fact,

it can be demonstrated that the ultimate limit to extracting cosmology
from certain of the x-ray data may be the inability to detect the
visible emission from the objects. However, our experience in situa-
tions where the quality of observational data is vastly improved is
that entirely novel results emerge. By definition, it is impossible to
predict where this might occur.

I have been aided significantly in the preparatlon of this paper

by discussions with Drs. William Forman, Riccardo Giacconi, Stephen
Murray, Daniel Schwartz, and Harvey Tananbaum of the Center for Astro-
physics.

Bahcall, J.N., Bahcall, N.A., Murray, S.S., Schmidt, Mo: 1975, "Ap.J.
{Lett.)", 199, L9.
Bowyer, C.S., Byram, E.T., Chubb, T.A., and Friedman, H.: 1964, "Nature",
201, 1307.
Byram, E.T., Chubb, ToA., Friedman, H.: 1966, "Science", 152:66.
Cooke, B.A., Ricketts, M.J., Maccacaro, T., Pye, J.P., Elvis, M.,
Watson, MoG., Griffiths, R.E., Pounds, K.A., McHardy, I., Maccagni,
Do, Seward, F., Page, C.G., and Turner, M.J.L.: 1977, submitted to
"Mon. Not. k. Astro. Soc.".
cowsik, R., and Kobetich, E.J.: 1972, "Astrophys. J o", 177, 585.
Culhane, L.: 1977, to be published in "Quart. Journ c , Astron. Soc.".
Field, G.B., Perrenod, S.C.: 1977, "Ap.J." In Press.
Forman, W., Jones, C., Cominsky, L., Julien, P., Murray,S., Peters,
Go, Tananbaum, H., and Giacco:li, R.: 1977, submittEd to
"Astrophys. J. {Suppl)".
Giacconi, R., Gursky, H., Paolini, F c , Rossi, B.: 1962, "Phys. Rev.
Lett.", 9:439.
Giacconi, R., Murray, So, Gursky, H., Kellogg, E., Schreier, E., Matilsky,
T., Koch, D., 'l'ananbaurn, :1.: 1974, "Ap.J. Suppl.", 27, No. 237,37
Gursky, H., and Schwartz, D.: 1977, "Ann. Rev. Astron o Astrophys.",
15, 541-
Gursky, Ho , Solinger, A. , Kellogg, E., Murray, S., Tananbaurn, H.,
Giacconi, R., Cavaliere, A.: 1972, "AJ).J.(Lett.)", 173, L99.
Horstman, H.M., Cavallo, Go, Moretti-Horstman, E.: 1975, "Riv. Nuovo
Cimento", 5:255.
Hoyle, F.: 1963, "Astrophys.J.", 137, 993 0

Jones, C., iilld Forman, W.: 1977, subL1itted to "Ast.rop'lys.Jo"

Kellogg, E., Hurray, So, Giacconi, R., Tananbaum, H., Gursky, H.: 1973,
"Ap.J. (Lett o )", 185, L13.
Murray, S., Forman, W., Jones i C., and Giacconi, R.: 1977, subr,\i tLed to
"AstrOp!ly.,T. (Lett)".
Schvartz, D.: 1974, "Int. Confo on x-rays in Space", Univ. of Ca1sary,
Calgary, Alberta.
Schwartz, D.: 1977, "Private Communication"o
Schwartz, D.A., Murray, S.S., Gursky, Ho, 1976 0 : "Ap.J.", 204, 315 0
Tananbaum, H., Peters, G., Forman, w., Giacconi, R., Jones, Co, and
Avni, Y.: submitted to "AstrophysoJ o (Lett) "0
COSMOLOGICAL INFORMATION FROM X-RAY OBSERV ATIONS 335

DISCUSSION

Ostriker: Concerning your remark on the contribution of gas to the

total mass of clusters, the gas contributes only about 10% of the total
mass which is determined by dynamical arguments.

Gursky: Where is all the hidden mass?

Ostriker: We do not care. The mass is determined from dynamical argu-

ments.

Silk: Two comments and a question. The supercluster X-ray sources, if

real and thermal, should be very extended and may therefore show up when
comparing observations ~ith different fields of view. The X-ray lumin-
osity-velocity dispersion correlation may largely be due to the richness
correlation, if velocity dispersions are even a very weak function of
richness. Would you comment on the present observational status and
relevance to cosmology of the hard X-ray background above 100 keV?

Gursky: We know very little about the sky above 100 keV. The situation
is similar to what was known in the 1-10 keV range in 1964; namely we
see a few galactic sources in this energy range and see a background of
radiation. In fact we have more information at 50-100 MeV than we do at
around 100 keV.

Morton: I could imagine several chains and at least one large area
clear of sources on the 4th Uhuru map. Is there any reality to these
features?

Gursky: The sky coverage by Uhuru was highly nonuniform because of the
necessity of studying specific galactic X-ray sources (e.g. Hercules
X-I) for very long intervals.

Zasov: What part of X-ray background could originate in the vicinity of

the Galaxy?

Gursky: A very small amount can be attributed to the Galaxy. Away from
the Galactic centre, the contribution is only a few per cent of the
diffuse background and does not extend more than a few degrees from the
galactic plane.

Culhane: Regarding Dr Silk's comment on possible differences in detec-

tor fields of view as a possible explanation of the inability of the
Ariel V and OSO-8 detectors to see the Uhuru superclusters, I would
point out that the Ariel and OSO detectors have 3.5 0 and 50 circular
fields of view, which are very comparable to the supercluster sizes and
it is therefore most unlikely that these instruments would have failed
to detect these sources at the Uhuru level. I would also comment on
Dr Gursky's suggestion on possible systematic differences between data
from different experiments; I know of no case in X-ray astronomy where
336 HERBERT GURSKY

differences of between 3 and 6 times have later been proved due to

systematic error.

Gursky: I understand Dr Culhane's point regarding the failure of Ariel

5 to confirm the superclusters. However, it is very unlikely that six
Uhuru sources could be spurious. This would imply that most of the
unidentified sources in the 4U catalogue (many of which also appear in
the 2A catalogue) are also spurious. Fortunately, new satellite data
will resolve the question within the year.

Kellogg: Regarding the existence of X-ray sources In superclusters, if

we accept the existence of the large scale structures reported at this
conference by Einasto and Tifft as superclusters, then there are lumin-
ous X-ray sources in these superclusters. In the Coma supercluster we
see the X-ray sources associated with Abell 1656 and Abell 1367. In the
Perseus supercluster we see Abell 426 and Abell 262 and perhaps others.

Gott: Regarding the effect of gas on the validity of N-body simulations,

the models should give reasonable results if the total mass is distribu-
ted in a similar way to the galaxies. In Coma and other great clusters
the gas and galaxies are similarly distributed. The cooling time of the
gas is longer than the Hubble time, so this gas has suffered no signifi-
cant dissipation relative to the galaxies. If most of the mass in the
great clusters is in dark stellar halos of galaxies then the two-body
relaxation effects in rich clusters will be less than in the N-body
simulations.

de Vaucouleurs: Has a search been made for an enhancement of the diffuse

X-ray backgroun in the supergalactic plane?

Gursky: We attempted with the Uhuru data to construct a map In order to

search for such large scale features; however, we ran into systematic
differences of ~ 5% from day to day which made it impossible to do so.

Zeldovich: What gas temperature is needed to make a diffuse intergalac-

tic gas observable at X-ray wavelengths?

Gursky: It is very difficult to be certain that one has observed the

diffuse emission from hot intergalactic gas. One could only attribute
the background to this mechanism if all other possibilities have been
eliminated. It is probably best to proceed from observations of regions
where the gas is known to be clumped as in clusters or superclusters.

Sunyaev: I want to mention that observations of fluctuations of the

intensity of the X-ray background are approximately 4 times more inform-
ative (at the same level of ~I/I) than the observations in the microwave
waveband. This is due to the well-known Compton-Getting effect, and is
connected with different spectral indices J v cr va; a ~ 2 for the
Rayleigh Jeans spectral region in the microwave waveband and a ~ -1 In
the X-ray band. For example, in the case of motion of the Earth in the
COSMOLOGICAL INFORMATION FROM X-RAY OBSERVATIONS 337

background field with velocity v « C

(3 - a) vic cose.
It is obvious that observations in the microwave band are now more
sensitive than in the X-ray band. However, it is very pleasant to know
about this factor for future observations.

van der Laan: I understood you to say that between 20 to 40% of the
X-ray background in the 4-10 keV window may be contributed by the two
known categories of extragalactic sources, viz. clusters and Seyfert
galaxies. For the unidentified sources (probably unidentified only
because of inadequate positional accuracy) we have no information at all
on their X-ray luminosity function. If they have a z-dependent popula-
tion evolution comparable to strong radio sources, they may contribute
the remainder of the background. Do you have any reason to rule this
out?

Gursky: No reason at all. In fact, Silk in ~ 1965 developed the

formulae for the contributions to the diffuse X-ray background from a
population of sources which evolved as did the quasars.
COUNTS OF FAINT GALAXIES

I. D. Karachentsev and A. I. Kopylov

Special Astrophysical Observatory
USSR Academy of Sciences
st. Zelenchukskaya
357140 Stavropolskij Kraj
USSR

The number-magnitude relation for galaxies is an insensitive test

for discriminating between cosmological models. Nevertheless, counts
of faint galaxies enable us to investigate possible evolutionary effects,
which may be caused by time dependence of the luminosity and/or number
density of galaxies.

The detection of evolutionary effects is masked by the following

factors: photometric scale errors in measuring extended objects,
inaccuracy of the K-correction for large redshifts, and uncertainties
in the luminosity function of galaxies, and in the mean relative abun-
dance of different morphological types of galaxies.

Numerical results on the counts of galaxies to mB = 24 m have been

presented by Karachentsev and Kopylov (1977). We used data published
by different authors and also counts of galaxies on plates obtained at
the prime focus of the 6-metre telescope. The photometric accuracy of
our counts is about a quarter of a magnitude.

To compare the observational data with the theoretical relations,

the following assumptions have been adopted:

(a) A Friedmann cosmological model with decleration parameter

qo 0
(b) K-corrections for elliptical E and spiral S galaxies according
to Pence (1976); it was also assumed that among field galaxies the
morphological types, E and S, have a relative abundance of 1:3, the same
as in the neighbourhood of the Galaxy.
(c) The luminosity function of galaxies according to Abell (1974).
(d) Intergalactic absorption has been assumed to be zero.

Figure 1 shows the counts of galaxies to different limiting magni-

tudes in the form log(N/N o ) vs. blue apparent magnitude, where
339
M. S. LongUlr and 1. Einasto reds.}, The Large Scale Structure of the Universe, 339-342. All Rights Reserved.
Copyright © 1978 by the [AU.
340 I. D. KARACHENTSEV AND A. I. KOPYLOV

Z '"'3.5

-'L-~45~--~~~--,~,----,~,----,L9--~27D--~a~'--~ZZ~--~2a~~2~.~~'~
m.
Figure 1

No = 100.6m is the number in a static, Euclidean model. The observa-

tional data are marked by full circles and by crosses (indirect esti-
mates). Some theoretical results are shown in Figure 1. The lower
solid line corresponds to no evolutionary effects. The lower dashed
curve represents number density evolution according to the steady-state
cosmology, n{z) = n o (1+z)3. The luminosity evolution, L = Lo(l+z),
which was proposed by Brown and Tinsley (1974), is indicated by the
upper dashed line. The upper two curves are for strong evolution with
different epochs of galaxy formation, ZF = 3.5 and 6.2, according to
Tinsley (1977).

The present data indicate no strong evolutionary effects for the

number density and luminosity of galaxies.

According to Dodd et al. (1975) the effect of clustering is more

pronounced for remote than for nearby galaxies, which can be explained
by the dynamical evolution of systems of galaxies. Since, the scatter
of log N(m) is approximately the same for faint and medium bright
galaxies, we conclude that there is not any increase in clustering for
faint galaxies up to z = 0.3 - 0.4.

REFERENCES

Abell, G.O., 1974. lAU Symposium "The Formation and Dynamics of

Galaxies", Dordrecht, p.225.
Brown, G.S. and Tinsley, B., 1974. Astrophys. J., 194, 555.
COUNTS OF FAINT GALAXIES 341

Dodd, R.J., Morgan, D.H., Nandy, K., Reddish, V.C. and Seddon, H., 1975.
Monthly No. Roy. Astron. Soc., 171, 329.
Karachentsev, I.D. and Kopylov, A.I.:-!977. Letters to Astron. Zh.,
3. 246.
Pence~ W., 1976. Astrophys. J., 203, 39.
Tinsley, B., 1977. IAU Colloquium No.37, "Redshifts and Expansion of
the Universe", Paris, p.223.

DISCUSSION

Abell: How large a total field was counted?

Karachentsev: The diameter of the field observed at the prime focus of

the 6-metre telescope is 12 arcmin.

Abell: How did you measure your faint magnitudes? Were they photoelec-
tric, extrafocal, iris diaphragm, or what?

Karachentsev: The counts were made visually in the regions of well-known

photometric standards: M3, M3l, MIDI, Abell 4, NGC24l9, and also in
other regions at high galactic latitude. The plates were obtained on the
same nights as those taken for the photometric standards. In counting
faint galaxies we stopped counting at ~m = 0.7 magnitudes above the plate
limit. We believe that the observation of 50% of light of a galaxy
beyond a stellar image allows us to distinguish between stellar and
galactic images. Of course, this criterion is not justified in the case
of compact galaxies.

Tinsley: What types of plate were used in these observations and what is
the magnitude system?

Karachentsev: Generally, we have used IIaO plates. For some plates we

have used a B magnitude system; for others, we reduced the magnitudes to
Zwicky photographic magnitudes. In comparing the counts with the pre-
dictions of your models, we took into account differences in the magni-
tude systems we used and those of the models.

Chincarini: My experience of 4-m plates taken at Cerro Tololo with good

seeing is that it is rather difficult to distinguish between stars and
galaxies at the plate limit. My questions are:
(1) The determination of the magnitude scale between B
B = 24m.
(2) How were the counts corrected for the effects of contamination
and what is the size of error?
(3) What kind of correction for foreground stars was used, how was
it determined and what is the error?
(4) What is the sensitivity correction across the field of the 6-m
telescope?
342 I. D. KARACHENTSEV AND A I. KOPYLOV

Karachentsev: Please see my answers to Dr Abell. Details about the

galaxy counts have been described by us in a paper published this year.

van der Laan: Can you give more details of how the counts were made,
in particular, what measuring machines were used?

Karachentsev: The counts were made by eye.

Tinsley: The theoretical curves that Dr Karachentsev derived from my

1977 Astrophys. J. paper assume that all galaxies above a given magnitude
limit are counted, regardless of their angular size. It should be point-
ed out that a large fraction of the galaxies comprising the excess with
"strong evolution" will be of very small angular diameter and could have
been rejected as stars.
Another point is that much smaller evolutionary excesses are pre-
dicted in the galaxy models discussed in my paper at this conference.
For example, an excess of a factor 'V 2.5 over the "local Euclidean"
value N(m) ~ 10o.6m is predicted at B = 24m.
R. G. Kron, at Berkeley, has counted galaxies to 24m and reports
approximately N ~ 100.6m even at 24 m in blue light. It would be inter-
esting to understand this discrepancy with Dr Karachentsev's results.
It should be noted that Kron's result is inconsistent with the models I
discuss at this conference.

Karachentsev: We find that for mB ? 22~5 the number of galaxies is

larger than the number of stars and for mB ~ 24~5 this ratio amounts to
a factor of 3. That is why in the counts of very faint galaxies the
problem of distinguishing between stellar and galactic images loses its
significance as a source of systematic errors.

de Vaucouleurs: The critical point in all log N(m) plots is the deter-
mination of m. How to measure it correctly is discussed at length in
the dissertation of G. S. Brown to be published shortly in the University
of Texas Astronomy Publications. Counts of galaxies down to B = 18m,
19m, 20 m and 21m in 13 fields at b = +70 0 , +80 0 and 90 0 are reported.

Karachentsev: I have had an opportunity to compare our galaxy counts

with those by King and Kron in the blue region. The difference in 19 N
(m = 24m) is only about 20%. The problem is the difference for brighter
(m'~ 21m) galaxies. I would like to emphasize that the data we presented
today included galaxy counts by different authors, namely, by Harris and
Smith, by Partridge, etc.
THE EVOLUTION OF GALAXIES: EVIDENCE FROM OPTICAL OBSERVATIONS

Beatrice M. Tinsley
Yale University Observatory

The clearest evidence that galaxies evolve comes from the Milky
Way: here we see stars with ages and lifetimes differing by 4 orders of
magnitude, and with chemical compositions indicating a hundred-fold en-
richment in metals between the times of formation of the halo and most
disc stars. The general resolution of Olbers' Paradox also gives evi-
dence for cosmological change: the sky is dark because galaxies turned
on a finite time ago (Harrison, 1974). Here I discuss some of the
further evidence for galactic evolution that is emerging from observat-
ions probing billions of years in lookback time.

1. COLOR CHANGES

Color evolution gives the least ambiguous evidence for changing

stellar populations in galaxies, but the changes are expected to be slow
since the types of stars contributing most light vary slowly if star
formation is proceeding at a gradually decreasing rate or if it has
stopped. The latter case, corresponding to normal elliptical galaxies,
is predicted to give the faster rate of evolution - with bluer colors in
the past because of the earlier main-sequence turnoff - but still only
an overall decrease to redshift z'VO.7 of about 0.15 mag in U-B (intrin-
sic) and about 0.07 in B-V. Spinrad (1977) has reviewed the current
status of searches for color evolution in ellipticals. In summary, the
only clear positive result comes from Spinrad and his colleagues, who
find about the predicted change in a U-B index derived from spectrophoto-
metry; null results in other cases are consistent with the predictions.

There is, however, statistical evidence for some much more dramatic
cases of color evolution. Butcher and Oemler (1977) have measured color
distributions for galaxies in rich, centrally condensed clusters at z =
0.39, 0.46, and 0.95 (see also Spinrad, 1977, Fig. 10). The range of
color in each case is consistent with that expected for normal galaxies
from irregulars to ellipticals. But, surprisingly, the colors are spread
broadly across the range, in marked contrast to the morphologically sim-
343

M. S Longair and J. Eznasto (eds.). The Large Scale Structure of the Umverse. 343-355. All Rights Reserved.
COPYright © J 978 by the fA U.
344 BEATRICE M. TINSLEY

ilar Coma cluster where the color distribution is sharply peaked at the
red end. Presumably the distant cluster members are blue because of
active star formation, so if the clusters are to become like Coma most
members are destined to stop forming stars. This idea agrees with the
standard picture in which SOs (the dominant type in Coma) form from spi-
rals by loss of interstellar gas in a dense cluster environment, but the
remarkable discovery is how late the cutoff occurs: the nearer clusters
are seen with only ~ 6 Gyr (6 x 10 9 yr) lookback time, and all three have
already collapsed to a condensed form. Further evidence for some gal-
axies being much bluer in the past appears in Kron's (1977) deep counts:
he finds that between mB = 22 and mB = 24, the mean color of counted
galaxies gets bluer by 0.35 mag. Possibly some early period of rapid
star formation is coming into view, as will be discussed below.

Sandage (1973) based an interesting argument for an evolutionary

universe on colors of elliptical galaxies. He estimated the spread of
colors that would result from the age distribution of galaxies in a
Steady State universe. This was much wider than the observed narrow
color range, which suggests that E galaxies are approximately coeval.

2. LUMINOSITY CHANGES

Cosmological tests using 100kback observations to derive the past

expansion rate (i.e. the deceleration) of the universe provide in fact
functions of both qo and galactic evolution. The standard approach has
been to determine an apparent value of qo and then to apply an evolution-
ary correction. Results have so far been inconclusive, partly because
the so-called "correction" is theoretically of the same order of magni-
tude as qo and uncertain by a similar amount! An alternative approach
is to use local cosmological data (density, stellar ages, etc.) to est-
imate q , so that the 100kback tests give information on luminosity
evo1utign (Tinsley, 1977b, c; Spinrad, 1977).

It is interesting to see what parameters dominate the results of

such tests. Here I extend some earlier results to include a particular
form of dynamical evolution: the growth of central cluster galaxies by
accretion of smaller members (Ostriker, this conference), under the sim-
plifying assumption that the shape of the cannibal galaxy remains in-
variant while its characteristic scale (specifically, the Hubble core
radius a) grows. Only first-order results will be given; they indicate
all the information that can reasonably be sought at z < 1.

The rate of dynamical evolution can be expressed in terms of the

derivative (at z = 0) fa = d In a/dz. Note that fa < 0, and may be ~
roughly estimated as ~ -1 (from the crude model in Tinsley [1977b]).
Ostriker and Hausman (1977) give detailed models of this type of evolu-
tion and note two empirical tests: (1) changes in a are related to
changes in a , the logarithmic slope of the luminosity-radius relation
at a given radius r, and (2) qualitatively, more evolved clusters have a
THE EVOLUTION OF GALAXIES· EVIDENCE FROM OPTICAL OBSIRV ATIONS 345

greater magnitude contrast between the first and second brightest gal-
axies, i.e. a lower Bautz-Morgan class. Luminosity changes due to evo-
lution of the stellar population can be expressed in terms of fL =
dln Lldz (stellar evolution). Models predict fL ~ +1; in particular,
fL > 1.2 is predicted by models that are consistent with the observed
slow color evolution (Tinsley 1977b, c). It can be shown that, to first
order, departures of the Hubble diagram from a straight line (m = Slog
z + const.) measure the quantity
2 2-2a
QH = qo + 2-0: fL + 2-0: fa'
which is evidently the "apparent qo" that one would get by ignoring evo-
lution. For example, with a = 0.6, fL = 1.2, fa = -1, one would find
QH = qo + 1.1. A similar analysis of the isophotal diameter-redshift
relation shows that the "empirical" quantity determined is

e.g. QS qo + 0.4 at ria = 10 with the above parameters. Interestingly,

the same quantity Qs is determined by Spinrad's (1977) "metric surface
brightness" (SB) test; the SB of a galaxy is measured at an angular
radius that would correspond to a fixed metric radius if qo had some
chosen fiducial value (cf. Tinsley, 1976). Note that the three tests
have similar relative sensitivities to qo and the two types of evolution.
In particular, the Hubble diagram is not especially favored as a test
for qo and metric SBs are not especially good cosmology-independent
measures of evolution.

A hope for the future is to check theoretical predictions of the

evolution of elliptical galaxies using lookback data. We need, first,
accurate enough estimates of qo from local data and well-determined re-
lations between magnitude, SB, etc. and redshift. Scatter and selection
biases in currently available data preclude any detailed comparisons
yet, but the situation is promising.

Counts of galaxies to limiting magnitudes > 18m are sensitive to

luminosity evolution of field galaxies and they are very insensitive to
qo; thus counts are a promising probe of evolution - as has long been
clear in radio astronomy. K-dimming is predicted to reduce the counts
at 24 m by a factor'\, 20 below the "static Euclidean" value No a: dex(. 6m).
It is therefore remarkable that Kron (1977) finds almost the Euclidean
slope for the N-m relation at a blue limit mB = 24. The exact interpre-
tation is complicated by uncertainties in the ultraviolet spectra of
galaxies (i.e. in the K terms), but these preliminary data almost cer-
tainly reflect luminosity evolution. An extensive study of the color-
magnitude distribution of faint galaxies is in progress, by Kron and
others, from which important information on evolution can be expected.
346 BEATRICE M. TINSLEY

3. PRIMEVAL GALAXIES: ARE THEY VISIBLE YET?

The possibility of detecting very young galaxies undergoing a bril-

liant first flash of star formation was explored theoretically by Part-
ridge and Peebles(1967a). Their models predicted objects with very high
luminosities, because a large fraction of each galaxy's mass was suppos-
edly made into stars in ~ 10 8 yr; they would be of uniform surface
brightness, with angular diameters ~ 10", and there would be thousands
per square degree on the sky. Subsequently, Partridge (1974) and Davis
and Wilkinson (1974) searched for primeval galaxies (PGs) with such
properties. Their negative results set limits on possible redshifts and
luminosities, and have been useful in testing later PG models.

Kaufman and Thuan (1977) constructed models for the formation of

stars in massive galactic halos. The primeval halos would not have been
detected unless qo is ~ 0.5. As yet, massive halos are no more visible
in their formative stages than around nearby galaxies!

Models for giant elliptical PGs with a very different appearance

have been developed by Meier (1976a, b). Based on Larson's (1974) col-
lapse models for the formation of ellipticals, these PGs reach their
brightest stage after they are already nearly as centrally condensed as
an E galaxy, and their star formation takes longer than in Partridge
and Peebles' models, so they would look smaller and fainter. If the
bright stage occurred two free-fall times after the big bang, they would
have been too faint for the foregoing searches to detect. But if the
collapse occurred later, these PGs would be bright enough but with al-
most stellar images, so the searches would have rejected them. Meier
suggests that some so-called quasars - with suitable thermal spectra
and no variability - might in fact be PGs.

The models suggest that only massive and late-forming galaxies may
be visible as PGs. If many galaxies form at redshifts < 5 and make a
large fraction of their stars in an early burst, they would show as pro-
nounced excesses in galaxy counts at certain magnitude levels (Tinsley
1977a) - unless of course they were discounted as stars. But now I
argue that maybe few young galaxies had very high star formation rates.

The UBV colors of typical Sc and later galaxies are consistent with
a constant star formation rate (SFR), and any substantial past enhance-
ment in the rate would lead to too red colors (Larson and Tinsley 1978)
- unless of course only massive stars formed at early times. Similarly,
earlier spirals can have had somewhat higher SFRs in the past than at
present, and E/sOs have UBV colors consistent with negligible star form-
ation for the last several Gyr. Thus if the first generations of stars
had a normal mass spectrum, present colors indicate that only Es and
SOs can have been very bright as PGs, and only for ellipticals do dyn-
amical arguments strongly favor rapid formation. The formation of disc
galaxies by gradual collapse and star formation is at least consistent
with their general morphology (Larson 1976) and with various chemical
THE EVOLUTION OF GALAXIES: EVIDENCE FROM OPTICAL OBSERVATIONS 347

and kinematic properties (Tinsley and Larson, in preparation); it is

also relevant that the average age of disc stars in the solar vicinity
appears to be less than 5 Gyr (Demarque and McClure, 1977).

The following picture is therefore suggested. Later types of spi-

rals have had almost constant SFRs, so their luminosities have increased
continuously (Fig. 1). Earlier spirals have had SFRs decreasing on time
scales of several Gyr, leading to an increase in luminosity for much of
the lifetime followed by a moderate recent decline (an example is shown
in Fig. 1). SO galaxies are former spirals in which star formation has
been cut off; their luminosities must drop rapidly thereafter, so they
were brightest near the cutoff time. Only ellipticals made nearly all
their stars during the initial collapse; their early luminosities were
high, in inverse proportion to the time available for star formation.
The cases illustrated in Figure 1 show that only SOs with early cutoffs
and ellipticals would be bright enough when young to qualify as PGs.
SOs with later cutoffs would appear as luminous blue objects at inter-
mediate redshifts (see upper scale, for example); these are candidates
for the excess blue galaxies found in clusters and in the field (§ 1).

Redshift
4 3 2 1 0.5 0

-4
./· . . .i,
j

-2
....
,
..........
I
>
::t
i
0

""
0 """
",/,

-I
.,.- ..,.- ......... "", .. !
---- ......

.....
,. ...- ...-
,. ,. .
'"
2L-~O~.~I--------~--------~IO~~

Age of Galaxy (Gyr)

FIG. 1. Evolution of absolute magnitude, relative to its present value,

in typical galaxy models consistent with observed colors. Upper scale:
redshift for qo=0.02, Ho=50, and formation at zF=5. Short dashes: con-
stant SFR (late spiral). Long dashes: SFR oc exp(-age/15 Gyr) (early
spiral). Other lines: constant SFR to cutoff age ~c = 0.2,1, or 10
Gyr, dropping to a model with SFR = 0 thereafter (E/SO galaxies).
348 BEATRICE M. TINSLEY

... 1.5
E-
o
U 1.01------
1/1
, ~ -) ......
.0 8-V ~~" " .... • ···.V-r
o ====- __ c::: ____ -,
~ '0

0.5
"0

' . ....

V=20
8

N ,,
::! 6
z
ICI
,
CI
\
E- 6 KEY
V= 16 Sdm
Sbc
E. after cut - -
4 E. before cut.
0.2-
Tc (Gyr) = 1 - ' -
10-"-
2
0.01 0.1
Redshift

FIG. 2. Center and bottom panels: computed redshift distributions for

galaxies at mv = 16 and 20 respectively. Ordinate is no. per unit mag.
and unit log z (with a fudge factor g ~ 1, depending on the normalizat-
ion to bright galaxy counts). Representative spirals and E/SOs are
shown; see key. Each E/SO alternative is scaled as though all were
cut simultaneously. Upper panel: colors, including K-corrections and
evolution, for the reddest and bluest galaxy models. Solid lines: E/
SOs after cutoff (no star formation); broken lines: constant SFR.

Figure 2 shows some consequences of this scenario for the redshift

and color distributions of galaxies (Tinsley, in preparation). At mV =
16, the z distribution of each type of galaxy is simply a smooth re-
flection of the width of its luminosity function, and the observed range
of colors depends mostly on K-corrections, with evolutionary changes
< O.OS.mag. At mV = 20, the spirals (dashed lines) and E/SOs after
cutoff (thin solid lines) are still well-behaved. But the latter types
before cutoff contribute various features depending on the cutoff age 'c
THE EVOLUTION OF GALAXIES. EVIDFNCE FROM OPTICAL OBSFRV ATIONS 349

(see key on figure). Note that the colors before cutoff could be as
blue as those shown by the broken lines in the top panel, but after cut-
off they would rapidly become as red as the solid lines. Rather than
the sharp features illustrated, one might expect a range of cutoff ages;
perhaps there would be some very bright primeval ellipticals near the
formation redshift, and active proto-SOs at a great range of redshifts.

Redshift data for a sample of the faint blue galaxies seen by Kron
(§l) would be very valuable for testing this type of picture. Another
handle on the early luminosities of galaxies comes from the extragalac-
tic background light (EBL) , which is the last topic to be discussed.

4. THE EXTRAGALACTIC BACKGROUND LIGHT

Dube et al. (1977) have measured an intensity of 1.0 + 1.2 S (10m

stars per square degree) for the EBL at 5100 A. On comparison wieR
published models, they found this low value to be consistent either with
galaxy formation so early that almost all emission from PGs is redshift-
ed beyond visible wavelengths (Partridge and Peebles, 1967b), or with a
closed cosmological model in which the lives of galaxies are too short
for later emission to contribute too much light (Tinsley 1977a). A
third alternative, suggested by the discussion in § 3 above, is that
only a small fraction of galaxies were bright when young. (Note that,
in contrast to Figure 1, both sets of models considered by Dube et al.
assumed that all galaxies were maximally bright near formation time.)

Using galaxy models like those illustrated in Figure 1, with SFRs

consistent with present colors, I have predicted the EBL for a variety
of formation redshifts, cutoff ages for E/SOs, etc. In all cases, the
visual EBL is at least 3 times fainter than predicted by the correspond-
ing models in Tinsley (1977a). Normalizing the local luminosity density
to reproduce counts of bright galaxies, I predict intensities of 0.6 -
1.5 SlO' even with an open cosmological model (qo=0.02, galaxy ages ~
13 Gyr) and with formation redshifts as low as 2 to 5.

The conclusion is that if E/SO galaxies alone were much brighter in

the past, a visual EBL ~ 1 SIO does not demand very short-lived galaxies
and/or very high formation redshifts. This possibility gives a footnote
to the solution of Olbers' Paradox mentioned in § 1: perhaps the sky
is very dark because most stars are much younger than the oldest ones.

5. CONCLUSION

At last, lookback observations confirm that the stellar populations

of galaxies change with time. Color changes in ellipticals correspond
to the predicted evolution of the main-sequence turnoff point, and ex-
cess blue galaxies in distant clusters and faint field samples indicate
that many SOs were actively forming stars in the visible past.
350 BEATRICE M. TINSLEY

Information on the primeval stages of galaxy evolution is still in

the form of upper limits to the numbers, formation times, and luminosi-
ties (or surface brightnesses) of galaxies that were extremely bright.
Theoretical "predictions" for such stages will surely prove to be simply
idealizations with which the real primeval chaos can be contrasted. It
will be exciting to see what is discovered as observations probe
further into the past.

It is a pleasure to thank R.G. Kron, R.B. Larson, A. Oemler, J.P.

Ostriker, H. Spinrad, and D.T. Wilkinson for valuable results and dis-
cussions. This work was supported in part by the National Science
Foundation (Grant AST76-l6329) and the Alfred P. Sloan Foundation.

REFERENCES

Butcher, H. and Oemler, A.: 1977, preprint.

Davis, M. and Wilkinson, D.T.: 1974, Astrophys. J. 192, 251.
Demarque, P. and McClure, R.D.: 1977, in B.M. Tinsley and R.B. Larson
(eds.), 'The Evolution of Galaxies and Stellar Populations,' Yale
University Observatory, New Haven, p. 199.
Dube, R.R., Wickes, W.C., and Wilkinson, D.T.: 1977, Astrophys. J.
(Letters) 215, LSI.
Harrison, E.R.: 1974, Physics Today 27, No.2, p. 30.
Kaufman, M, and Thuan, T.X.: 1977, Astrophys. J. 215, 11.
Kron, R.G.: 1977, in B.M. Tinsley and R.B. Larson (eds.), 'The Evolution
of Galaxies and Stellar Populations,' Yale University Observatory,
New Haven, p. 334.
Larson, R.B.: 1974, Monthly Notices Roy. Astron. Soc. 166, 585.
Larson, R.B.: 1976, Monthly Notices Roy. Astron. Soc. 176, 31.
Larson, R.B. and Tinsley, B.M.: 1978, Astrophys. J., in press.
Meier, D.L.: 1976a, Astrophys. J. (Letters) 203, Ll03.
Meier, D.L.: 1976b, Astrophys. J. 207, 343.
Ostriker, J.P. and Hausman, M.A.: 1977, preprint.
Partridge, R.B.: 1974, Astrophys. J. 192, 241.
Partridge, R.B. and Peebles, P.J.E.: 1967a, Astrophys. J. 147, 868.
Partridge, R.B. and Peebles, P.J.E.: 1967b, Astrophys. J. 148, 377.
Sandage, A.: 1973, Astrophys. J. 183, 711.
Spinrad, H.: 1977, in B.M. Tinsley and R.B. Larson (eds.), 'The Evolu-
tion of Galaxies and Stellar Populations,' Yale University Observatory,
New Haven, p. 301.
Tinsley, B.M.: 1976, Astrophys. J. (Letters) 210, L49.
Tinsley, B.M.: 1977a, Astrophys. J. 211, 621 (erratum, ibid. 216, 349).
Tinsley, B.M.: 1977b, in C. Balkowski and B.E. Westerlund (eds.), 'De-
calages vers Ie Rouge et Expansion de l'Universe. L'Evolution des
Galaxies et ses Implications Cosmologiques,' CNRS, Paris, p. 223.
Tinsley, B.M.: 1977c, Physics Today 30, No.6, p. 32.
THE EVOLUTION OF GALAXIES; EVIDENCE FROM OPTICAL OBSERV ATIONS 351

Note added by the author: In the printed version of my paper I inadver-

tently forgot to mention that Longair and Sunyaev (IAU Symposium No.74
"Radio Astronomy and Cosmology", 353) developed models for primeval
galaxies similar to Meier's models. Longair and Sunyaev also suggested
that some "quasars" could be primeval galaxies.

DISCUSSION

Abell: In my paper I showed integral number-magnitude counts in differ-

ent directions by George Rainey; my purpose was to show the isotropy of
the counts. Rainey's counts are in excellent agreement with those of
G. S. Brown, and in qualitative agreement with the data we have just
seen from 6-m telescope counts. They are not compatible with
Dr Tinsley's older strong evolution models:-and certainly not with the
Euclidean slope that Kron's counts are reported to suggest.
However, contrary to what I said in Paris in 1976, a careful
inspection of Rainey's data does show evidence for some evolution.
Rainey's counts are roughly halfway intermediate between the predictions
of the no-evolution Friedmann models and the Euclidean model. In other
words, Rainey's data suggest a conservative amount of galaxian evolution.

Tinsley: It sounds as though Rainey's data may give some support to the
models I have discussed here, because they predict a rather small evolu-
tionary enhancement in the counts at magnitudes fainter than 22 m in blue
light or 20m in red light. (Details will be published elsewhere.) The
main uncertain parameters in the calculations are the UV spectra
(K-corrections) of Elsa galaxies, and the range of redshifts at which
star formation takes place in Elsa galaxies.

Bolton: I now understand the point of your question to me this morning!

From the few redshifts we obtained of UVX galaxies between 19m and 20m
the absolute luminosities of these objects are of the same order as
strong radio galaxies.

Ostriker: Didn't Sargent and Searle draw the same conclusion that the
rate of star formation was relatively constant from the colours of
spiral galaxies?

Tinsley: Yes. I did a similar calculation in 1968 but I did not

realise that it was a theorem!

Zeldovich: In your models do you take account of the evolution of the

chemical composition of galaxies? What would be the result if galaxies
consisted only of hydrogen and helium and no heavy elements?

Tinsley: If galaxies consisted only of hydrogen and helium they would

be much bluer. However, since the stars in most of the galaxies we see
have normal chemical compositions, they probably formed from material
with the same element abundance. For any reasonable change in the
352 BEATRICE M. TINSLEY

chemical composition, the changes to the models would be small.

Gott: Spiral galaxies consist of both spheroidal and disk components.

The spheroidal bulge components look just like elliptical galaxies (i.e.
colours,velocity dispersions, light distributions as a function of
luminosity are the same). If primeval elliptical galaxies are 100 times
brighter than at present, wouldn't one expect that the bulge components
of spiral galaxies would be similarly brighter in the past and dominate
the total light of the spirals? Wouldn't this make all galaxies bright-
er in the past?

Tinsley: This is an interesting question and the answer is a bit sur-

pr1s1ng. The light of the bulge is included in the integrated colours
of a spiral galaxy, so the limit that we get from colours of the past
star formation rate includes early, rapid formation of the bulge. For
example, I showed the luminosity evolution of an early (Sb) spiral with
an exponential star formation rate; a very similar result would be
obtained if star formation was an initial burst followed by a constant
rate, scaled to give the same present colours. The system would get
brighter during most of its life and never be much brighter than at
present. Pushing as much star formation as possible into rapid bulge
formation, I still get only 50% more EBL than in the cases illustrated
here.

Chernomordik: How can you explain in your model the deficiency of G-

dwarfs with low metal abundance in the disk of the Galaxy?

Tinsley: There are several ways of explaining the paucity of metal-poor

stars in the solar neighbourhood. The idea suggested by Schmidt (1963)
was an initial burst of massive stars; Dr Chernomordik's question refers
to the fact that my current models for disk galaxies do not have such a
burst. An alternative explanation of the G-dwarf metallicities involves
infall of metal-poor gas (Larson, Nature Phys. Sai. 1972). The models I
discussed here for slow disk formation, although not explicitly depend-
ent on any gas flows, could be associated with dynamical models of disk
formation by infall, which would be consistent with the metallicity
distribution. Larson and I have calculated some detailed models
(preprint) that illustrate this statement. To summarize, there is no
inconsistency between the models I use here of disk galaxies and the
empirical metallicities of stars in the solar neighbourhood.

Silk: One may have difficulties in appealing to ram pressure stripping

to account for the colour evolution at redshifts as low as 0.4, because
a considerable number of characteristic galaxy crossing times (the
relevant time-scale for stripping) have already occurred in a cluster
at this redshift. Another possibility may be due to the effects of
enhanced star formation that follows the accretion of intracluster gas.
Would you comment on the effect of assuming different forms for the
~nitial mass function of your models?
THE EVOLUTION OF GALAXIES: EVIDENCE FROM OPTICAL OBSERV ATIONS 353

TinsZey: This is another way of saying the same thing. The blue
galaxies are seen at a redshift z = 0.39 and these must turn into syst-
ems like Coma by the present day. We have considered models similar to
the one you describe. In answer to the question, the only drastic
effect of changing the initial mass function is that you can make young
galaxies as bright as you like and not affect the present colours.

Jaakkola: A difference in the colour distribution between the Coma

cluster and some distant clusters was your empirical argument in favour
of evolution of galaxies over cosmological distances. Now there must be
serious selection effects present, making a cluster at z = 0.95 - which
is the only one observed out of the thousands at that distance - uncom-
parable with Coma which is our nearest cluster. I remind you in passing
that Zwicky mentions in Morphological Astronomy ten parameters of
galaxies and of clusters which are similar at large and at small
distances.

Tinsley: I believe that the clusters studied by Butcher and Oemler were
selected for study because of the radio or optical luminosity of their
first-ranked galaxies, not because of the blueness of other cluster
members. Possibly some such selection effect is present, however, and
equivalent data for many more clusters would be welcome.

Suchkov: I would like to make a comment on the IMF which is important

for some points discussed by Dr Tinsley. We now have some evidence from
the distributions of old population stars in metallicity and in eccen-
tricity, that the IMF in the early Galaxy favours large-mass stars, as
was first suggested by Dr M. Schmidt in 1963. This means that a major
part of the metal content of the Galaxy was produced during the period
of formation and evolution of halo stars, no matter what fraction of the
mass of the Galaxy they constitute. This is important for the luminos-
ity evolution of the Galaxy.

Tinsley: Unfortunately I have not had a chance to study Dr Suchkov's

data. The models that I have discussed, with a constant IMF, do give
stellar metallicities in accord with earlier data on galactic stars
(Tinsley & Larson, 1978 preprint).

Sunyaev: There are many observers of quasars here. Therefore I would

like to amplify the problem mentioned by Dr Tinsley in her report. We
heard how bright elliptical galaxies might be in the early stages of
their evolution when they were young and had a lot of bright hot stars.
Their luminosity during this stage might exceed their present luminosity
by 100 or 1000 times. If our Universe is open (~ « 1), the angular
dimensions of the central part of giant galaxies (~10 kpc) may be very
small -6 1". Therefore they may appear as quasistellar sources. The
next step is: they may be hidden among optically selected quasars.
Hence the problem is how we can distinguish "real" quasars from young
galaxies. In the case of young galaxies (1) the range of variability
must be very small, (2) light must be unpolarized, (3) there may be
354 BEATRICE M. TINSLEY

spectral features common to very hot stars. Young galaxies might be

rather bright in the CO line A 2.6 mm. In the case of young galaxies
with z ~ 3-5 the line must be redshifted to the waveband 2.6(1+z) mm ~ 1
cm. Observations of this line in the most distant quasars might help to
prove if there is a galaxy with stars and gas around them. The line
might be rather bright because the galaxy has finite angular dimensions
and they must be saturated even in the case of a low abundance of heavy
elements.

Huchra: A comment on the Butcher-Oemler result - the clusters were

chosen to have similar morphological properties (density, etc.) So the
comparison made is not between a Virgo type cluster at z = 0.4 and Coma
now - but rather a Coma then.
A complaint I have with your models, however, is that good evidence
exists for star formation with an IMF (initial mass function) other than
the Salpeter function. People should not blithely use a single value.
Your results for the luminosity evolution change drastically with other
IMF's, in particular, with a flatter IMF there can be almost no luminos-
ity evolution!

Tinsley: I am fully aware of the sensitivity of my models to the IMF,

and of evidence for variations of the IMF from place to place (e.g. see
my review in IAU Symposium No. 77). In a 30-minute talk, one has to
sound blithe about complicated issues! There are several reasons for
showing here models with an invariant IMF. (1) The colours of "normal"
galaxies are consistent with the local (solar-neighbourhood) function
holding in typical cases (as I showed in 1968 and has been emphasized by
others since). Therefore this seems a useful first approximation.
(2) I cannot predict or derive from observations the incidence or nature
of departures from the local IMF in the population of galaxies.
(3) Models do not pretend to be exact representations - they are examples
of possibilities, and here I have chosen to see what happens with the
most economical set of hypotheses. I wanted to discuss the significance
of star formation rates, and it would confuse the issue to vary the IMF
too. Of course, you are right that one would get different results with
different IMFs. (Incidentally, the models discussed here use not the
Salpeter power-law IMF, but a more detailed function based on recent
stellar statistics.)

Gursky: It is my understanding that a limit on the background light at

a particular frequency simply puts the burst of star formation back to
an earlier epoch. What is the limiting epoch in this instance?

Tinsley: As described in my paper (and more fully in a current pre-

print), the EBL in this picture is insensitive to the turn-on redshift.
This is because only a small fraction of galaxies have a burst of star
formation. A redshift of 2 for the time of first star formation (in all
galaxies) does not give too high a background intensity.
THE EVOLUTION or GALAXIES. EVIDENCE FROM OPTICAL OBSERV ATIONS 355

Kornberg: Do you believe that quasars are the whole galaxies or the
nuclei of E galaxies?

Tinsley: The idea that I mentioned, due to Meier and to Sunyaev and
Longair, is that the integrated starlight of a young massive E galaxy
would look like a quasar. The angular radius of these models of young
galaxies would be { 1 arc sec. The authors do not propose that all
quasars are young ellipticals, but that some so-called quasars are in
fact primeval galaxies. Dr Sunyaev's comment during this discussion
amplified these points.

Jones: Given your explanation of the Butcher and Oemaler effect, do we

have to come to terms with the idea that galaxy clusters formed as
recently as z ~ O.S? The point is that those galaxies have not had
their first trip through the central regions of the cluster or are
accreting a lot of material.

Tinsley: I think I described approximately Butcher and Oemler's inter-

pretation of their data. None of us has thought of a less surprising
explanation.

Jones: Astonishing.

Tinsley: I agree.
 ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES

Jeremiah P. Ostriker
Princeton University Observatory
Princeton, New Jersey

1. INTRODUCTION: STELLAR SYSTEMS OF POINT MASSES

The theory of the dynamics of star clusters (cf. Spitzer 1975 for
a review) is by now so well developed that we have, or think we have,
a moderately accurate picture of the physical processes acting in and
the overall evolution of spherical systems. In contrast, flattened
and/or rotating systems are apparently subject to a variety of ill-
understood instabilities which ultimately are a manifestation of the
second law of thermodynamics; at given total energy, a system will tend
to increase the fraction of its kinetic energy in disordered rather
than ordered form. But spherical systems (globular clusters, ellipti-
cal galaxies, Morgan cD clusters of galaxies) are relatively smooth and
featureless; they show little substructure indicating, presumably, that
they are quite stable to perturbations of their fundamental normal
modes, and they are normally modeled as rather "hot", pressure supported
systems.

In fact, the confidence we have in our understanding of spherical

systems has not been very well tested empirically. Not only do we have
no direct information concerning dynamical evolution (less even than
the indirect information available about evolution of stars from clus-
ter H-R diagrams), but our information concerning equilibrium is quite
poor. In the best observed systems we know a central velocity disper-
sion and the projection on the plane of the sky of the density dis-
tribution of a subsystem of tracer objects. In addition to the obvi-
ous loss of information due to projection and the unavailability of
Eev ll (r)2>, <vJ..(r)2>] in any system* we have growing evidence that the
observable tracer population is unrepresentative (Spinrad et al. 1978),
that its number distribution does not reflect the underlying mass dis-
tribution. The problem, however, of the evolution of a more or less
spherical N-body system is essentially simple, it might, in principle,
have been solved by Newton. And the well developed (if untested) the-
ory for the dynamics of N point masses should provide a good starting
*In our galaxy, we do know that, locally <VII 2> / <v.1. 2> 'V 2 for the
Halo (spheroidal) population (Oort 1965).
357
M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 357-375. All Rights Reserved.
Copyright © 1978 by the [AU.
358 JEREMIAH P. OSTRIKER

point for the more complicated dynamics of N galaxies, each having many
excitable internal degrees of freedom.

For the average star it is customary to distinguish between pro-

cesses occurring o~ twq widely separated time scales. On the dynamical
time scale TD ~ (Gp)-1/2, where p is the mass density, a typical star
will cross the system, purely statistical fluctuations will change,
and the system will respond to any gross departures from (virial)
equilibrium. On the much longer relaxation time scale, TR ~ (N/~nN)TD'
a star will interact with other individual stars (rather than the clus-
ter as a whole). If external force fields (e.g., "tidal forces") sig-
nificantly affect the orbits of test particles then there will be time
scales associated with these as well.

Various stages in the evolution of an N-body system are custom-

arily enumerated which occur on successively longer time scales. The
first two listed are highly conjectural and other scenarios (cf.
Doroshkevich et al. 1974) are just as plausible.

a) Uniform isotropic expansion.

Globular clusters, galaxies, cluster stars of galaxies are all
thought to arise from perturbations growing in a standard Fried-
mann-Robertson-Walker cosmology.
b) Maximum size reached, star formation, collapse.
If a given perturbation has mean density greater than the cos-
mological critical density at a given epoch, Pcrit = (3H2/8~G),
then a maximum size will be reached, the perturbation will sepa-
rate out from the general Hubble flow, and it will recollapse in
a probably anisotropic fashion with further material falling into
the deepening potential well (Gott and Gunn 1972). Much of the
material may be processed into "point" objects (those of much
higher than mean density) during these phases. That is, much
star formation must occur during the collapse phase of galaxy or
globular cluster formation (Eggen et al. 1962, Binney 1977, Peebles
and Dicke 1968) and much galaxy formation must occur during or be-
fore the collapse phase of cluster formation (Rees and Ostriker
1977) .

c) Violent relaxation.
The system of point masses so formed is generally out of equilib-
rium at first, and the violent large scale and probably asym-
metrical motions which ensue produce strong gravitational fields,
fluctuating in space and time. Individual particles interacting
with these fields will clearly not pres erve "integrals" of motion
such as energy and angular momentum, and arguments can be given
(Henon 1964, Lynden-Bell 1967) for the establishment of quasi-
Maxwellian distribution functions.

These first three phases occur on the dynamical time scale es-
tablishing a system of point masses more or less resembling the
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 359

classical Emden isothermal sphere (cf. Zwicky 1957) but, of course,

truncated in some way.

d) Eguipartition, dynamical friction.

The (presumably few) relatively massive objects (mass Mh ) in the
central regions of the cluster interacting with the swarm of
average particles of mass M (number N) suffer numerous gravita-
tional collisions which lead towards equipartition. On the time
scale of (M/Mh)T R they fall towards the center of the system (cf.
Spitzer 1969, Tremaine et al. 1975).

e) Core contraction.
Even in a system of particles all of the same mass there is a
tendency for the central density to increase. The exact physical
characterization of the process is still under debate (cf.
Antonov 1962, Lynden-Bell and Wood 1968, Spitzer and Thuan 1972,
Lightman and Shapiro 1977) but the results are agreed upon. On a
time scale of ~ 100 TR evaluated at the center, the central den-
sity approaches a singularity with the mass in the high density
core approaching zero and the energy (kinetic or gravitational) of
the core changing relatively slowly.

f) Then,on the considerably longer time scale of ~ 100 TR evaluated

at the half mass point (rather than the center),the cluster as a
whole will change significantly, although it is not at all clear
yet whether that change is better characterized as collapse or
evaporation, simply because no treatments, analytical or numeri-
cal, have been able to reach past stage e).

2. PHYSICAL INTERACTIONS AMONG "SOFT" OBJECTS

All of the stages of evolution enumerated above would presumably

occur in the great clusters of galaxies were there enough time. For
Coma the dynamical time scale phases a) - c) is TD ~ 10 9 • 3 yrs, the
equipartition time scale d) for the most massive central galaxies
(M/Mh)T R ~ 10 10 • 1 yrs and all the other processes are too slow to be
of interest on the Hubble time scale. The dynamics of galaxies in
galaxy clusters is, however, far more interesting than that of stars
in star clusters. The reason for this is simply that the ratio of the
size of the typical member to the distance of its nearest neighbor is
~ 1/5 for the galaxy in the center of a rich galaxy cluster but
< 1/10 5 in even a very condensed globular cluster. Thus, gravitational
collisions between galaxies can excite their internal degrees of free-
dom producing many effects not present in the point Newtonian dynamics
of star clusters. In some respects the problems encountered are remi-
niscent of those in atomic physics in that analogs of "exchange col-
lisions", "excited states", "pressure ionization", etc. must be con-
sidered. The "standard" processes enumerated in the previous section
have, of course, been considered with reference to clusters of galax-
ies. For example, Peebles (1970) deals with (a)-(c) and White (1976)
360 JEREMIAH P. OSTRIKER

with (d). However, there are, in addition, three kinds of processes

which are peculiar to, or far more important in galaxy clusters than
in star clusters; and I shall concentrate here on those. It is possi-
ble to maintain the hypothesis that the apparent ordering of cluster
properties into a sequence of types by Bautz and Morgan (1970) or
Oemler (1974) is only a manifestation of the greater or lesser degree
of dynamical evolution produced by these three processes: dynamical
evolution gradually transforms Bautz-Morgan III clusters (like Virgo)
to type II systems (like Coma) or type I systems (like Abell 2199).

a) Galaxy interactions with gas

It is well known that e11iptica1s and SO galaxies are more common

in clusters than they are in the field, relatively more common in con-
densed clusters than in low density systems, and relatively more com-
mon in the central than the outer regions of rich clusters (cf. Melnick
and Sargent 1977 for a recent discussion). Spitzer and Baade (1951)
suggested, some time ago, that collisions between spiral galaxies could
sweep out the gas, prevent further star formation and allow the spirals
to gradually be transformed to SO systems. The discovery of thermal
X-rays from galaxy clusters (Kellogg 1973) allowed variants of this
process to be considered. Gunn and Gott (1972) noted that the ram
pressure of the ambient gas would tend to sweep spirals clean of gas,
an idea developed in more detailed calculations of Gisler (1976). Re-
cently, Cowie (1977) has considered the thermal effects and showed that
interactions with the very hot ambient gas will tend to heat and evapo-
rate gas from galaxies in X-ray clusters.

Any and all of these processes appear capable of stripping gas

from spirals and producing SO systems. If, further, Ostriker and
Thuan (1975) are correct in their contention that the discs of spiral
galaxies are to a significant extent secondary, produced by infa11 of
gas processed through halo stars, then early and efficient gas removal
by any of the above mechanisms would prevent formation of the discs at
all, leaving only the spheroidal bulge components of the would-be
spiral galaxies which, as many investigations have shown, are indis-
tinguishable from e11ipticals. Thus, the same process which increases
the proportion of SO systems in rich clusters, will also increase the
proportion of e11iptica1s. This chain of argument has been strongly
reinforced by the finding (Serlemitsos et al. 1977) of X-ray iron emis-
sion lines from three rich clusters. The quantity of metal rich ma-
terial seen in the clusters is, within the observational and theoreti-
cal uncertainties, just that which was ejected from the elliptical gal-
axies according to the models of Ostriker and Thuan (1975) and just
that which, were it allowed to fall back into the elliptical galaxies
as secondary discs, would give the latter the appearance of normal
spiral or SO systems.

Stellar dynamical processes to transform a large fraction of

spirals to SO or elliptical systems in clusters are not an attractive
alternative. The principal reason is that the galaxies ultimately
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 361

found in the central regions of great clusters cannot be distinguished

from field galaxies in numerical simulations of galaxy clustering (cf.
Aarseth, Gott and Turner 1977) until the galaxy is actually in a region
of quite high velocity dispersion. In that environment, as many stud-
ies have shown, interactions tend to be weak and dispersive. High
velocity collisions do not produce mergers and have little effect on
the inner parts of galaxies; the main result of such encounters, as
we shall describe in the next section, is to produce tidal stripping.

b) High velocity tidal interactions

Dynamically, it is useful to define "clusters" vs. "groups" by the

velocity dispersions of the members, rather than by the (somewhat arnr
biguous) number counts, and we shall call a cluster of galaxies an as-
semblage within which the velocity dispersion among the members is sub-
stantially larger than the velocity dispersions of the stars within
the constituent galaxies; for the Coma cluster, the two numbers are
Vrms,cluster ~ 1,000 km/s and Vrms galaxy Z 200 km/s along the line of
sight. Thus, since galaxies are ";oft" and have maximum central poten-
tials comparable to Vrms,galaxy' collisions between galaxies will gen-
erally by hyperbolic regardless of impact parameter, and deviations from
rectilinear motion of the galaxy centers will be small. Gallagher and
Ostriker (1972) and others have treated the problem in the impulsive
approximation. Richstone (1975, 1976) analyzed the fast collisions
problem with essentially the same scheme of approximations as Toomre
and Toomre (1972) used for the slow, nearly parabolic, collisions
appropriate for field galaxies. First the relative orbital motion of
two rigid diffuse galaxies is computed. Then, within each galaxy, stars
are sampled by a Monte Carlo technique and their orbits calculated dur-
ing the galaxy-galaxy encounter as a restricted three-body problem.
Then, by integrating over the orbital parameters in the final state and
comparing ~Jith the comparable integrals in the pre-collision state, one
can estimate the change induced in the galaxy by the collision. Speci-
fically, since galaxies can be minimally characterized by a mass M,
central density Pc and two radii, an inner core radius Rc and an outer
tidal cutoff radius Rt , one can compare the changes (oM, oP c ' oRe, OR t )
induced by a specific collision. Then one can integrate over the dis-
tribution of impact parameters, relative velocities and masses of the
perturbing galaxies. Richstone found that the changes in the core
properties were small and uncertain. The envelopes could be changed
significantly. In general, the rate of change of any quantity Q(= M,
Pc' etc.) can be represented approximately, but simply, by the follow-
ing formula:

!E.9. [en M )/M]R

2
V1 C , (1)
Q dt p p t t,t c ,rms Q

where (np,M ) are the number density and mass of the perturbing galax-
ies (enteriRg only as their total mass density) and (Mt,Rc t) are the
mass and tidal radius of the test galaxy being stripped. Note that
this result is independent of Vrms,gal here in the limit that
362 JEREMIAH P. OSTRIKER

(V ga1 ,rms/ Vc1 rms) «1. The parameters CQ represent the efficiency
of the collisions and are,according to Ricnstone's Monte Carlo calcu-
lations,

CM = -0.015 ± 0.003
(2)
CRt = -0.014 ± 0.006

Richstone also finds surprisingly that Cp > 0, collisions tend to in-

crease the central densities; the result is however quite uncertain.
The rate at which the tidal radius is decreased can be rewritten simply
as

dRt t P
~= -3.3 x 10- 3 ~ V ( 3)
dt < > c1, rms
Pt
where <Pt> is the mean density of the galaxy within its tidal radius
and P c 1 is the local cluster density. Clearly the tidal radius will
be reduced until (THubb1e/Tdyn) = 300 «Pt>/Pc1) (Rt/Rct )· Such an ef-
fect may already have been detected in Strom and Strom s (1977) obser-
vations of the Coma cluster.

c) Accretion and Cannibalism

As noted earlier dynamical friction (Chandrasekhar 1943) will

cause massive satellites to spiral into their parent galaxies
(Ostriker and Tremaine 1975; Tremaine 1976) and in clusters the mas-
sive galaxies will tend to accumulate in the center (Ostriker and
Tremaine 1975; White 1976, 1977; Ostriker and Hausman 1977). Both ef-
fects tend to increase the observed luminosity of galaxies. The latter
process (the merging of galaxies in clusters) will tend to produce
supergiant systems of low central surface brightness like the known cD
systems. One can also show that the apparently non-statistical fea-
tures seen at the bright end of the cluster luminosity function can be
caused by cannibalism. What is the "normal" luminosity function?

Although galaxies of apparently normal character exist over an

enormous range of brightness, there is, however, a characteristic
luminosity L* for galaxies, since the luminosity function ~(L) (such
that the number of galaxies per unit volume having luminosity in the
range L + L + dL is $(L)dL) for both field and cluster galaxies can be
fit to a function of the form proposed by Schechter (1976):

(4)

with the value of L* varying little between galaxies in the field and
those in dense clusters. We plan to show here how the characteristic
L* determines, not only the typical galaxy seen at the knee of the
luminosity function, but also the first brightest galaxy in clusters
which have undergone extensive dynamical evolution.
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 363

There are alleged to be characteristics of the observed luminosity

functio~ at the bright end, which are inconsistent with any statistical
model. The most ingenious study to date, by Tremaine and Rich-
stone (1977), compared the expected values of cr(Ml) and <6M12> = (M2 -
Ml) in a way that did not depend on the explicity form of the lumi-
nosity function. Applying their analysis to the Sandage and Hardy
(1973) cluster data, they tentatively agreed with Sandage's conclusion
that the luminosity of the brightest cluster galaxies is determined by
some special process. More explicitly, Tremaine and Richstone proved
that for any statistical luminosity function, the inequality
cr(Ml)/<~M12> =tl > 1.0 must hold. In fact, one can show that, for a
Schechter luminosity function with y = +1.25 and n* in the interval
25 - 100, tl is fairly constant and equals 1.20 ± 0.02; in contrast,
the value of tl derived from the Sandage-Hardy data, as presented in
Tremaine and Richstone, is 0.48 ± 0.10 (standard error).

Three separate processes appear to be combining to produce this

statistically quite significant effect.

1. The first brightest galaxies are somewhat better standard candles

as measured through a fixed metric diaphragm than would be ex-
pected on the basis of the slope of the luminosity function at the
bright end and the number of galaxies in the cluster (cf. Dressler
1977). The effect is most pronounced in centrally condensed spi-
ral poor clusters (type I clusters).

2. The brightness of the first brightest galaxies depends much less

on the cluster richness than would be expected with
(dMl/d log n*) = -0.16 ± 0.02 (S.E.) observed, (derived from data
presented in Richstone and Tremaine 1976) compared to
(dMl/d log n*) = -0.9 for the Schechter function with 5 < n* < 50
(cf. also Sandage 1976 and Schechter and Peebles 1976).

3. The gap between the first and second brightest galaxy is larger
than would be expected statistically, it is larger than can be
accounted for simply by saying the first galaxy is too bright,
and it is also most pronounced in centrally concentrated type I
clusters (cf. Dressler 1977). This last effect must be confirmed
using more homogeneous data than has hitherto been available.

Sandage and Hardy (1973) commented that "The brighter the dominant gal-
axy becomes, the absolutely fainter will be the second and third ranked
members. 'The rich are rich at the expense of the poor, progressively!"

I would like to suggest here that that is exactly what is happen-

ing, that in type I clusters the cannibalism described by Ostriker and
Tremaine (1975) and White (1976) will produce just these effects.

First let us treat the luminosity evolution of the first brightest

galaxy by a rough qualitative argument. Consider a simplified case of
the merging of N identical galaxies, each of mass M, core radius R(l)
364 JEREMIAH P. OSTRIKER

defined so that each has gravitational energy W = -GMZ/R(l), and total

energy E = -GMZ/ZR(l). If we combine N such galaxies, the total mass
is NM and the total energy (neglecting the contribution produced by
initial orbital binding energy) is E(N) = -NGM Z/ZR(l), which we may
identify with -G(NM)Z/ZR(N) assuming that the new galaxy differs from
its constituent parts only by scaling factors. This ~ives for the
radius and mean surface brightness within R,L(~(NM)/R )

R(N) = NR(l); L(N) = N- 1 L(l) (5)

the luminosity and radius will increase, but the mean surface brightness
will fall with increasing N (cf. Figs. Za and Zb). Suppose now that we
observe the growing galaxy through a diaphragm of radius RD which is
much larger than R(l). All the luminosity passes through the diaphragm
and, as N increases, the observed luminosity will be proportional to N
until R(N) ~. RD. Thereaftef the luminosity observed is :::: 7T~L and de-
clines proportionally to N- , the peak apparent luminosity b~ing

L
max
:::: 7T~R(l)L(l)
.DRs
:::: (~JL (6a)
s
where Ls and Rs are approximately the luminosity and characteristic
radius of the standard building-block galaxies. We can make this argu-
ment more precise if we adopt a definite intensity profile for the com-
ponent galaxies. The most accurate two-parameter fit is, according to
Kormendy (1977), the de Vaucou1eurs (1953) law which can be written
I(r) = 2141.5 10 exp[-7.669Z(r/Ra)1/4], Ltot = Z2.666 IoRoZ, where
(1 0 , Ra) are the surface brightness and radius at the cylinder contain-
ing one-half of the total light. We can now compute the luminosity ob-
served within RD from the first brightest galaxy composed of N standard
objects by integrating the intensity profile over the diaphragm area
with 10 = N- 1 Ios (or L1,tot = NLs) and Ro = NRos. The result is given
by equation (lZa). As N increases, Ll,obs(N,RD) reaches a broad flat
maximum and then slowly declines for N > Nc . The maximum value is
Lmax obs = 0.7390 (RD/Ros)Ls for the de Vaucou1eurs model. Inclusion
of t~e orbital (relative) energy affects the results only by scaling
factors: we can estimate the magnitude of the effect by comparing this
analytical treatment with the numerical results described later, and
obtain

L 1.5 (R_/R (6b)

max,o b s =
-n 0 *)L s
For the de Vaucouleurs model the critical value Nc is

Nc = 6.084 -n os )
(~/R z 14 (~/R *)
-n 0
( 7)

In Fig. (la) we show L1,obs as a function of 1n(N/N c ) since the latter

quantity is roughly proportional to the time (the accretion process be-
ing initially exponential); the analogous relation computed via the
Monte Carlo simulation is displayed in Fig. (Zc,d). We also show in
(la) the model 1n(N/N c ) dependence of the important observable parameter
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 365

1.0 a

0.9
a
0.8

FIG. 1: Ca) Observed luminosity of accreting galaxy plotted vs. number of

accreted objects N according to theory of § IV B. The ordinate Ll b /
-4 ,0 S
CLsRu/Ros) is computed parametrically as 3459.5 P(B,x)x where In(N/N c )
6.34282 - 4 In x. Also plotted is a(Ru) defined by eq. (B) and computed as
x 8e-x /[20l60 P(B,x)]. (b) The expectation value for the magnitude difference
<6M12 > for clusters of richness n* = 30,100 computed by equations (7) and (10)
- (13), assuming no luminosity evolution for any galaxy except the brightest.

a(~) = (dlnLObS) ( 8)
dlnR R=~
which is to be compared with Fig. (2e). Note that the scales in Fig.
(1) are linear in luminosity (not magnitude) and logarithmic in (NINe);
Ll,obs is a very weak function of accretion once the latter is substan-
tial. Note also that a(16), which for Oemler's sample ranges between
0.29 and 0.78, is monotonic and a good candidate for an observable to
replace the unobservable "time" or (NINe) coordinate.
366 JEREMIAH P. OSTRIKER

The value of the maximum luminosity is set partially by instru-

mental effects (RD) and partially by the intrinsic properties of the
standard galaxies contributing most to the growing supergiant. The
maximum magnitude corresponding to the numerical value in equation (3)
~epends on Ro*. From the mean of 10 Coma cluster galaxies having
1 = 1* whose properties are measured in Oemler (1976), we find Ro *
5.6 kpc. This would give a maximum luminosity within a 43 kpc dia-
phragm of MB = -23.2. This .is somewhat greater than the mean first
brightest cluster galaxy for Bautz-Morgan I (highly evolved) clusters
of -22.7 given by Sandage and Hardy (1973).

Gunn and Tinsley (1976) investigating the same problem also com-
mented that evolution could in principle lead to either an increase or
decrease in the apparent luminosity of the accreting galaxy. We see
from the above calculation (see also Hausman and Ostriker 1977) that,
for galaxies with properties like observed galaxies, (d1 0bs /dt) > 0
simply because the effective core radii are always smaller than the
radii of the diaphragms typically used by observers.

Next, let us look at the apparent luminosity evolution ~M12' the

magnitude difference between the first and second brightest galaxies.
The probability of cannibalism per unit time is proportional to the
mass of the giant times the mass of the prospective victim (neglecting
a slowly changing logarithmic factor). Thus the galaxy most likely to
be eaten is the second brightest galaxy after which the new second
brightest is the former third brightest and is consequently fainter;
it follows that M2 will become fainter andl~M121 will increase with time.
Again, a simple analysis allows us to be roughly quantitative. 1et
v(~,t)dt be the probability that a galaxy with luminosity ~ = 1/1* is
eaten in time dt. Ignoring the luminosity evolution of all except the
first brightest galaxy (to be considered elsewhere), the probability
of survival till time t is
t
p(~,t) = exp(- f v(~,t)dt) (9)
o
and the expectation value (ensemble average) of M12 is

where all quantities are explicit functions of time and Pk = p(~k,t) is

the probability of survival of the cluster member which was originally
kth rank in luminosity

~Mlk = 2.5 log (1l'ObS)= 2.5 (11)

~,obs

1
l,obs NP[8,7.6692 (R/NR
-1) os )1/4]/7! (12a)
1
s
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 367

L L
k,obs ~ = (2 + 15 - ct) (12b)
L*
w'here P(x,y) is the incomplete gamma function. Here 15 is the exponent
in the assumed mass luminosity relation (M/L) = (M*/L*)~Q and the most
probable values of ~k can be obtained from the formulae in Schechter
(1976). The total rate of accretion by the central object is given by

dN(t) = n*vn(~*,t)dt Joo~-ct+l+Oe-~d~

(13)

-(1+0)
-n*~ r(2+O-ct)d1np(~,t)

giving Pk = exp[-(N~kl+Q)/n*r(2+0-ct)1 from equation (7) since

v(~,t) = ~l+ov(~*,t). The uncertain time variable has fortunately been
bypassed; given equations (7) and (10)-(13) we can evaluate <bM12> as
a function of (N/N c ) or a for given values of (Rn,Ros) and n*. Taking
(43 kpc, 5.6 kpc) for the former we show the expected evolution of
<~M12> in Fig. (lb) for two values of n*, these curves give upper
bounds on IbM12i, since luminosity evolution of L2 (L3' etc.) has been
ignored. A direct check of the proposed correlation between ~M12 and
ct seen in Fig. (1) is possible and would test the theory presented
here.

Thus we expect that, as dynamical evolution proceeds, the first

brightest galaxy will initially gorw in core radius and luminosity;
its core radius will continue to grow, but the luminosity seen through
a metric diaphragm will level out near Lmax (equation 6) and then gradu-
ally decline; the total luminosity will steadily increase. The other
bright galaxies in the cluster core will tend to be swallowed and the
gap between the first and second brightest galaxy will grow steadily
(equation 10).

Finally, let us compare different clusters at the same epoch. For

fixed surface brightness, or distinctiveness compared to the back-
ground, the relaxation time TE, which varies as v~ms/(G2Mga1Pcl)' tends
to decrease with decreasing richness as N5/4. Thus there is relatively
more dynamical evolution in poor clusters than in rich (at the same
surface brightness) which consequently increases (-Ml) for the poor
clusters and reduces the expected amount of the Scott effect. The
Bautz-Morgan classification (1970), which essentially measures clus-
ter dynamical evolution, should be correlated with TE (and crossing
time, TDyn)' It would be interesting to test the proposed correlation
for those clusters having two of the three dynamical parameters (Rcore'
Ncore ' vrms) measured as well as Bautz-Morgan type or ~M12'

Hausman and Ostriker (1977) have numerically simulated cluster

368 JEREMIAH P. OSTRIKER

evolution. They assume that 1) All galaxies have a surface brightness

obeying the Hubble law I(r) = 10 (1 + r/B)-2 with mass density propor-
tional to light emissivity, the galaxy being truncated where its den-
sit~ reaches the cluster density. We define a core luminosity La =
loB, and assume that (M/LO) ~ Lo 0 .5 (cf. Faber and Jackson 1976) and
B ~ LOO. l (cf. Oemler 1976). The relation between (U-B), (B-V) and
MV is taken from Sandage (1972). 2) The initial distribution of galaxy
o
luminosit L is given by equation (1) with Y = +1.0 and with L* =
3 x 10 1 LO' 3) The accretion rate is given by Ostriker and Tremaine
(1975) formulae which, under these circumstances, give Ml~Ml!Mcldncl
where Ml is the growing central giant and ncl and Mel are the number
and mass of cluster galaxies of various types. 4) Collisions conserve
mass, energy (binding + orbital), and luminosity in the (U,B,U) bands.
We take an initial assumedly central first brightest galaxy with L =
3L* then, in a given time step, pick a victim galaxy from the Schechter
distribution, the probability being proportional to the mass of the
galaxy, find the new, swollen, first brightest galaxy from the conserva-
tion laws noted above, and repeat the process. Accretion is intrinsic-
ally unstable; if the first galaxy eaten happens to be particularly
large, the primary will subsequently eat at a. more rapid rate, and
vice versa. Thus a Monte Carlo stimulation is useful. A detailed
description of the numerical procedures and more extensive publication
of results is reserved for a subsequent paper (Hausman and Ostriker
1977) .

Figures (2a)-(2e) show the evolution of the first brightest gal-

axy as observed through diaphragms of radii 16 kpc and 30 kpc (for
comparison, Gunn and Oke 1975; Sandage and Hardy 1975, use 19 kpc and
43 kpc for a Hubble constant of 50 km- l Mpc- l ). Five Monte Carlo simu-
lations are shown for identical starting conditions. A major uncer-
tainty is how to treat the accreted galaxies when their mean density is
more than the central density of the growing cannibal, which decreases
continuously. Accretion will still occur and the initial relative
orbital energy is still available, but tidal forces will not neces-
sarily disrupt the accreted galaxies (unless they disrupt each other)
so the self-binding energy may not be available and parts of the
accreted galaxies may remain as intact cores to be seen as the "multi-
ple nuclei" often found in cD systems. To simplify the discussion
here, we terminated the integrations displayed when these effects be-
came Significant.
In Fig. (2c) the leveling off of the observed luminosity discussed
in §2 is seen clearly; the five runs give a mean Lmax of [4.78 ± 0.23
(standard error)] L* compared to 2.2 L* derived from equation (3), the
difference being due to the neglect of the orbital binding energy in the
analytical calculation. Figure (2a) illustrates the exponential growth
of the total luminosity (N ~ e t ), which is to be compared with observed
isophotal luminosities, and (2e) the approximately linear behavior of
a with t (or InN). In Figures (3a) and (3b) we show the expected lumi-
nosity evolution as a function of a (equation 8). Notice that the dis-
persion is smaller since a effectively measures dynamical evolution.
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 369

30r---,----,----,----,----,----,
o

.20
-l
"-
-l
10

t::t'~
J8'0~ C
'-..

14.0i....~""=f--O>-

f':~
I·te~
~t-~~, ,
0.0 1.0 2.0 3.0 4.0 5.0 6.0
TIME

FIG. 2: Five Monte Carlo runs for evolution of first brightest cluster
galaxy with initial luminosity L = 3L* (see § IV C for details). (a) Total
luminosity; each successive symbol represents an accretion of a single galaxy.
(b) Core Hubble radius vs. time; galaxy with L = L* has S = 0.67 kpc initially.
(e), (d) Luminosities observed through 16 kpc and 30 kpc diaphragms. (e) a(~),

the dimensionless measure of the core radius Ceq. 8).

Hubble radii are given in Oemler (1976) for 6 galaxies which are first
brightest in their clusters (MKW4, Virgo, A779, A1413, A2147 and A2670).
From these, estimates for (a,L) can be obtained which are accurate to
the extent that the galaxies fit Hubble laws in the range of radii
considered. For this sample the mean and dispersions in (a(16),
L(16)(a(30), L(30» were calculated and are also shown in Fig. (3). The
agreement is excellent, but of course, may be fortuitous.

Figure (4) shows core radii vs. luminosity as derived from the
Monte Carlo simulations plotted with Oemler's data. The large increase
in core radii observed for central luminous galaxies follows naturally
from the dynamical theory. Figure 5 shows the expected evolution of
370 JEREMIAH P. OSTRIKER

6.0
a
...J*
.:;; 4.0
a.
.><
<.0
S 2.0

0.3 0.4 0.5 0.6 0.7 O.B 0.9

a (l6kpc)

10.0

...J
* B.O
"-
~ 6.0
-'"
0
r0
:i 4.0

2.0

0.0
0.3 0.4 0.5 0.6 0.7 O.B 0.9
a (30 kpc)

FIG. 3: Observed luminosities plotted

versus a from Monte Carlo runs. Large cross

data point is the mean of 6 first brightest

galaxies from Oemler (1976).

the (U - B) vs. MV relation. The bright galaxies do get bluer with

time, but at only a moderate rate since they tend to preferentially
accrete rather luminous (L ~ 1.25 L*) red galaxies.

In a further numerical investigation (Hausman and Ostriker 1977)

finite clusters with various values of n* were simulated by a Monte
Carlo process and then allowed to interact according to the previously
described rules. Here it was found that the luminosity growth of the
first brightest galaxy tended to limit out at IHl - M*I = 1.5 - 2.2
mag, for n* = 30-100, rather than to continue to increase exponentially,
the reason of course being that the supergiant ultimately "uses up"
all of the easily available victims. The luminosity function change
is quite interesting. It tends to steepen at the bright end and to
develop an extra peak of super bright (cD) systems bearing some re-
semblance to the results found by Oemler (1976) and Dressler (1976).
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 371

\
\
\
\
\
\
04 \
\
\
\
\
\
\
\
\
\
02 \
u
,,
\
0. \

"-
"-
'"
0'
o
-I 00
...... --- -
..... ..............

-02

-25 -24 -23 -22

MV (16)

FIG. 4: Core Hubble radius versus magnitude from

Monte Carlo simulations (T is time coordinate): data

from Oemler (1976) lie between dashed lines with means

near T = 0, T = 18.75 lines.

In sum, it appears that many of the notable features of centrally

condensed clusters of galaxies, particularly the presence of very
luminous but low surface brightness central cD systems, having a rela-
tively small luminosity dispersion through fixed metric diaphragms
and an apparently too large gap between first and second or third
brightest galaxies, can be understood in terms of a straightforward
dynamical theory of galactic cannibalism.

Bright galaxies spiral to the center of clusters on an equiparti-

tion time scale; there they are swallowed by a central giant which be-
comes physically bigger and brighter. Many tests of the theory are
possible essentially by checking on correlations between a, ~MI2, and
relaxation or crossing times among observed clusters. In addition,
since cannibalism tends, selectively, to deplete the bright end of the
luminosity function, we can predict that the latter will become
steeper and the break (L*) displaced to lower luminosity along the
Bautz-Morgan sequence from Types III to I. To the extent that the
theory is confirmed, it should be possible to use it, with accumulated
372 JEREMIAH P. OSTRIKER

T=O
07

"
T=6 25---------... "", . . . . . ,
---
-
CD
I 06
T = 12 5
------------ .....
.... ........
.... ..........
..........
T=25 ........

05

-25 -24 -23 - 22 - 21

MV (16)

FIG. 5: Evolution of the color-magnitude relation from Hausman and Ostriker

(1977). Dashed lines represent, approximately, the envelope from Sandage's (1972)

observations.

data, to correct for the effects of dynamical evolution, thereby removing

that source of dispersion and confusion from the Hubble diagram used in
cosmological investigations. Since the Bautz-Morgan (1970) classifica-
tion is correlated with ~M12 (Dressler 1977), and ~M12 is a good measure
of dynamical evolution (for fixed n*), the Bautz-Morgan corrections cur-
rently being used by Sandage and others already compensate for dynamical
evolution, to some degree, on an empirical basis.

3. SUMMARY

The presently observed properties of clusters of galaxies have re-

cently been reviewed by Bahcall(1977). These observations, and the the-
ory of the last section, might be put together to sketch out a hypothet-
ical evolution to this state in the form of the following tale:

Once upon a time, at the relatively recent epoch of 3 < z < 10, gal-
axies formed from fluctuations of unknown origin in the expanding uni-
verse. Typical galaxies were very large (10 5 • 5 kpc) and massive (10 12 . 3
Me). Large scale fluctuations took longer to develop, and clusters of
galaxies separated out at the very recent epoch of 1 < z < 3. The inner
parts of the clusters collapsed, violently relaxed and adjusted to a
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 373

nearly isothermal state. The outer parts continue to expand but at a

decelerating rate. In the inner parts interactions with cluster gas
(largely expelled from other galaxies) strips gas from galaxies and in-
hibits the formation of secondary discs (spirals and SOs). Relatively
fast tidal interactions between galaxies strips off the dark material
from the outer parts of individual galaxies leaving systems with conven-
tional sizes and masses (10 4 . 5 pc, lOll MO) and distributing the dark
matter throughout the inner parts of the cluster. Then on a longer time
scale the giant systems tend to accumulate at the center as supergiant
low surface brightness cD systems. The type of a given cluster in, for
example, the Bautz-Morgan system is determined by the degree to which
these dynamical processes have acted. Some depend more closely on the
crossing time in a system, some on the relaxation time but in general the
most evolved clusters should be the dense but poor (small N) systems. In
extreme cases these dynamical processes may go to completion with the re-
sult that the whole cluster is transformed into one supergiant cD system
containing remnant cores of destroyed galaxies.

I would like to thank Drs. A. Dressler, J. E. Gunn, A. Sandage L.

Spitzer and S. Tremaine for useful conversations. I benefited from a
John Sherman Fairchild Fellowship at California Institute of Technology,
and the work was supported also by National Science Foundation grant
AST76-20255.

REFERENCES

Aarseth, S., Gott, R., and Turner, E.: 1977, in preparation.

Antonov, V. A.: 1962, Vestn. Leningr. Gros. Univ. Z, 135.
Bahcall, N.: 1977, Ann. Rev. Astron. & Astrophys. 12
Bautz, L. and Morgan, W. W.: 1970, Astrophys. J. Letters lQZ, L149.
Binney, J.: 1977, Monthly Notices Roy. Astron. Soc. in press.
Chandrasekhar, S.: 1943, Astrophys. J. 21, 251.
Cowie, L.: 1977, Nature 2Q~, 501.
Doroshkevich, A. G., Sunyaev, R. A., and Zeldovich, Y. B.: 1974,
Confrontation of Cosmological Theories with Observational Data,
ed. M. S. Longair, IAU Symposium 63 (Reidel: Dordrecht).
Dressler, A.: 1976, Thesis, Lick Observatory.
Dressler, A.: 1977, preprint.
Eggen, O. J., Lynden-Bell, D., and Sandage, A. R.: 1962, Astrophys. J.
Ufj, 748.
Gallagher, J. and Ostriker, J. P.: 1972, Astron. J. 77, 288.
Gisler, G. R.: 1976, Astron. & Astrophys. 21, 137. ."'-
Gunn, J. E. and Gott, J. R.: 1972, Astrophys. J. l1§, 1.
Gunn, J. E. and Oke, J. B.: Astrophys. J. ~i, 255.
Gunn, J. E. and Tinsley, B.: 1976, Astrophys. J. ~, 1.
Hausman, M. A. and Ostriker, J. P.: 1977, in preparation.
Henon, M.: 1964, Ann. Astrophys. ~, 83.
Kellogg, E. M.: 1973, lAU Symposium #55.
Kormendy, J.: 1977, preprint.
Lightman, A. and Shapiro, S.: 1977, preprint.
Lynden-Bell, D.: 1967, Monthly Notices Roy. Astron. Soc. ~'. 101.
374 JEREMIAH P. OSTRIKER

Lynden-Bell, D. and Wood, R.: 1968, Monthly Notices Roy. Astron. Soc.
],J.~, 495.
Melnick, J. and Sargent, W. L. W.: 1977, Astrophys. J. in press.
Oemler, A.: 1974, Astrophys. J. ~., 1.
Oem1er, A.: 1976, Astrophys. J. ~, 693.
Oort, J.: 1965, Stars and Stellar Systems Vol. V, ed. A. Blaauw and
M. Schmidt (Chicago: U of Chicago Press), Chapt. 21.
Ostriker, J. P. and Hausman, M.: 1977, Astrophys. J. Letters, in press.
Ostriker, J. P. and Thuan, T. X.: 1975, Astrophys. J. W6., 353.
Ostriker, J. P. and Tremaine, S.: 1975, Astrophys. J. Letters ~?, Ll13.
Peebles, P. J. E.: 1970, Astron. J. Zi, 13.
Peebles, P. J. E. and Dicke, R. H.: 1968, Astrophys. J. ~, 891.
Rees, M. J. and Ostriker, J. P.: 1977, Monthly Notices Roy. Astron. Soc.
R4... 520.
Richstone, D. 0.: 1975, Astrophys. J. ~, 535.
Richstone, D. 0.: 1976, Astrophys. J. 204, 642.
Sandage, A.: 1972, Astrophys. J. ~ 2I:
Sandage, A. and Hardy, E.: 1976, Astrophys. J. ~, 743.
Schechter, P.: 1976, Astrophys. J. ~ 297.
Schechter, P. and Peebles, P. J. E.: 1976, Astrophys. J. ~, 670.
Ser1emitsos, P. J., Smith, B. W., Boldt, E. A., Holt, S. S., and Swank,
J. H.: 1977, Astrophys. J. ~ L63.
Spinrad, H. and Ostriker, J. P.: 1978, in preparation.
Spitzer, L.: 1969, Astrophys. J. Letters 1~, L139.
Spitzer, L.: 1975, IAU Symposium #69, Dynamics of Stellar Systems,
ed. A. Hay1i (Dordrecht: Reidel), 3.
Spitzer, L. and Baade, W.: 1951, Astrophys. J. ttJ, 413.
Spitzer, L. and Thuan, T. X.: 1972, Astrophys. J. ~, 37.
Strom, S. and Strom, K.: 1977, Yale Univ. Symposium on Galactic Evolution.
Toomre, A. and Toomre, J.: 1972, Astrophys. J. ~, 623.
Tremaine, S.: 1976, Astrophys. J. ~ 72.
Tremaine, S., Ostriker, J. P., and Spizer, L.: 1975, Astrophys. J. ~,
407.
Tremaine, S. and Richstone, D.: 1977, Astrophys. J. 2~ 311.
de Vaucou1eurs, G.: 1953, Monthly Notices Roy. Astron. Soc. tll, 134.
White, S.: 1976, Monthly Notices Roy. Astron. Soc. ~ 19.
White, S.: 1977, Monthly Notices Roy. Astron. Soc. ~, 33.

DISCUSSION

Silk: Gas accretion by galaxies in rich clusters followed by subsequent

star formation also can result in the luminosity increase of dominant
cluster galaxies with time. The resulting effect on galaxy colours,
however, may be different from that predicted in the cannibalism picture.
For example, while gas accretion may initially result in a substantial
"b1uening" of the colours, this will be pronounced at redshifts 'V 1. If
star formation ceased more than 3 x 10 9 years ago, the effect of gas
accretion will be to produce net "reddening" (since intracluster gas is
found to be enriched and is inferred to have originated early in the
evolution of the cluster). My impression of the most recent data is
ON THE DYNAMICAL EVOLUTION OF CLUSTERS OF GALAXIES 375

that galaxy colours continue to redden with increasing luminosity. How

do these data compare with the predicted flattening of the colour-
magnitude relation that results from galaxy mergers?

Ostriker: This has often been raised as an objection to the theory and
I would like to show what one would expect. The observations show that
the brighter galaxies are redder and hence if massive galaxies grow by
swallowing less massive galaxies, the bright galaxies should become
bluer. However, since most of the galaxies being swallowed have lumino-
sities about L*, the massive galaxies should evolve at roughly constant
colour. If you see blue massive galaxies, they must be due to recent
star formation.

Chernin: Clusters are in equilibrium, but only dynamically, i.e. they

are gravitationally bound, not in statistical equilibrium. Relaxation
via violent interactions may be rapid enough, but it cannot lead to the
mass segregation observed in clusters.

Ostriker: I agree. The cluster centres are in dynamical equilibrium

and only for the most massive galaxies in some clusters has equipartion
begun to occur.

Tinsley: Ostriker has discussed how cannibalism of smaller elliptical

galaxies by central cluster members will cause the latter to be somewhat
too blue for their luminosity; this effect will lead to a levelling-off,
at the bright end, of the colour-magnitude relation for E galaxies. I
wish to mention that the only theoretical explanation of the colour-
magnitude relation (Larson 1974, Mon. Not. R. astr. Soc.) predicts an
intrinsic flattening of the relation at the bright end. The point is
that smaller E galaxies are bluer because they lose their gas by super-
nova-driven winds at an earlier stage of chemical enrichment; the effect
is insensitive to galaxy mass at the high-mass end because big galaxies
retain their gas for long enough to make the mean stellar metallicity
essentially the yield value (independently of the small gas fraction
lost). Therefore, any observed levelling-off of the colour-magnitude
relation for cluster galaxies may not be entirely due to cannibalism.

Ostriker: I agree.

Ozernoy: Could you talk a little more about the cD galaxies in poor
clusters and especially those in the field? What about their origin?
Should their properties differ appreciably from cD galaxies in rich
clusters?

Ostriker: Yes. They should be less luminous and have smaller internal
velocity dispersions.

Lynden-Bell: In the small clusters should the cD galaxies not be bluer

in your theory?

Ostriker: Yes.
THE EPOCH OF GALAXY FORMATION

Bernard J.T. Jones

Martin J. Rees

Institute of Astronomy,
Cambridge, England.

1. INTRODUCTION

Unfortunately, there is as yet no direct observation of a forming

galaxy, nor of a manifestly "young" galaxy. It is not even clear what
such galaxies might look like. There are, however, some lines of attack
that will yield indirect information about the environment in which
galaxies formed. There is no lack of theories of galaxy formation, many
of which provide quite different scenarios for the birth-process. The
hope is that theories of galaxy formation may indicate what young
galaxies ought to look like and thus guide us in our search for such
objects.

2. OBSERVATIONAL EVIDENCE

There are several ways in which we might hope, in the near future,
to obtain information about the evolution of galaxies, clusters of
galaxies and QSO's.

a. The correlation function of the distribution of faint samples of

galaxies may yield evidence for the evolution of clustering out to red-
shifts z ~ 0.4 or so. One of the problems with this approach is that
the all-important K-correction is poorly known. The recent work of
Phillips et al. (1977) on deep UK-Schmidt plates already suggests that
galaxies were less clustered at redshifts z ~ 0.4 than now.

b. Butcher and Oemler (1977) have recently studied over a hundred

galaxies in each of two very distant galaxy clusters (Cl 0024 + 1654 at
z = .39 and the cluster around 3C 295 at z = .46) and found that the
galaxies in these clusters are considerably bluer than the galaxies in
the Coma cluster. Not only may this be telling us about the evolution-
ary history of SO galaxies, but it may also be construed as evidence
that these clusters collapsed in the relatively recent past. The
K-correction is an underlying uncertainty in this work also.

377

M. S. Longair and 1. Einasto (eds.), The Large Scale Structure of the Universe, 377-388. All Rights Reserved
Copyrzght © 1978 by the fA U.
378 BERNARD J. T. JONES AND MARTIN J. REES

c. Dube, Wickes, and Wilkinson (1911) have reported measurements of the

extragalactic background light which serve to constrain some models of
galactic evolution in which galaxies form at redshifts z < 10 with a
bright flash of star formation. Observations of this kind in the near
infra-red may be of considerable value.

d. Quasar absorption lines can serve as an important probe of physical

conditions at different redshifts along the line of sight. B.A. Peterson
has obtained good evidence that the absorption lines are mainly due to
intervening material rather than matter being ejected relativistically
from the quasar. Interestingly, Peterson's results also suggest that
the state of the intervening material is steeply dependent upon redshift,
In the range z = 2 - 3.

e. Future X-ray telescopes (notably HEAO B) should be able to detect

individual galaxy clusters at considerable distances (z ~ 1 - 3).
Evidence for evolutionary effects in the temperature or gas density will
certainly help to discriminate between various schemes for cluster
formation.

f. Small-scale angular variations in the microwave background radiation

temperature, if detected, would provide information on the structure of
the universe at very early epochs. Such variations would arise as a
result of motions on the "last scattering surface". Distant clusters of
galaxies, if they contain hot gas, might be detected in this way as
isolated "dips".

g. The origin and evolution of quasars and related objects poses even
more problems than galaxy formation. If quasars represent activity in
the nuclei of galaxies, then the fact that we observe them out to at
least z = 3.5 sets a lower limit to the redshift when galaxies had
developed to the stage of having acquired well-defined centres. The
redshift distribution of quasars in principle provides clues to galactic
evolution, but as yet no firm conclusions can be drawn. Extended radio
sources are indirectly relevant, insofar as they serve as probes of the
external gas density at different epochs (cf Blandford and Rees 1911).

3. THE SCENARIO OF DOROSHKEVICH ET AL.

Doroshkevich et al (1914, and references cited therein) have

explored the consequences of a model in which the primordial fluctuations
were purely adiabatic or "isentropic", so that all those on mass-scales
below 10 13 - 10 14 M0 had damped out by t rec . The first bound systems to
condense would then be gas clouds with masses appropriate to clusters of
galaxies. These clouds must have turned around at a "recent" epoch
(z $ 5). They collapse in an aspherical or centrally condensed manner,
and before they have contracted by more than a factor ~ 2, shock waves
would have heated most of the gas, even if it were previously cold, to
the temperature Tvirial. Galaxies would form later in the manner discus-
sed by other speakers at this conference.
THE EPOCH OF GALAXY FORMATION 379

The most basic question one could ask about galaxies is "What deter-
mines their characteristic sizes?" The analogous question about stars
can be answered - there are good and well-known reasons why stable main-
sequence stars are r~stricted to within an order of magnitude or so of
the mass (nc/G~2)3/ ffio. An advantage of the idea that galaxies and
clusters form dlrectly from massive gas clouds is that a very simple
physical argument does yield a characteristic radius and mass in crude
but suggestive accordance with actual galaxies (Rees and Ostriker 1977,
Silk 1977, Binney 1977). The key feature of this argyment is a comparl-
son ~f th~ collapse or dyn~ical timescale (tdyn oc.p- 2 ) and ~he radiative
coollng tlmescale tcool (whlch depends on p- 1 and lS a functlon of the
gas temperature Tvirial)' Fragmentation readily occurs only if tcool ~
t dyn ' This condition is generally fulfilled only for clouds of mass
~ 10 12 M@ and/or radii ~ 75 kpc; and these are in gratifying order-of-
magnitude accordance with the properties of actual galaxies.

In this type of model, the covariance function on scales smaller

than clusters would bear only a very indirect relation to the spectrum
(op/p)rec of density inhomogeneities at recombination.

4. NON-DISSIPATIVE GRAVITATIONAL CLUSTERING

As an opposite extreme from the above, we may imagine a scenario

in which all the star formation is completed early (soon after recombina-
tion) and the subsequent evolution is stellar-dynamical (i.e. of a kind
that could in principle be simulated by N-body computations). Some key
expected features were clearly expounded by Press and Schechter (1974).
If, at some initial instant, point masses are distributed in a "random"
fashion, then at later times clustering develops on progressively larger
scales. At any epoch t, there is a characteristic mass-scale Mc(t) on
which the density fluctuations have amplitude of order unity (the corres-
ponding length-scale rc would be that for which ~(r) ~ 1). The initial
conditions satisfy

( 0 p / p) oc M-o.

where a. is in the range 1/3 - 2/3 (a. = ~ corresponding to "white noise"

or a poisson distribution of the number of particles within a comoving
cell). If the background universe is of Einstein-de Sitter type (and,
for (1 + z) ~ n- 1 , in low density models also) then

Me oc t 2 / 3o. oc (1 + z)1/o. ( 2)

On scales » Mc(t) the perturbations are of small amplitude, the kine-

matics being close to those of the background universe. But on scales
« Mc ' "virialised" systems would have condensed out.

The observed clustering of galaxies tells us that, at the present

epoch, rco ~ (5 - 10) Mpc and Mco ~ 10 14 M®, and (2) allows us to infer
380 BERNARD J. T. JONES AND MARTIN J. REES

the scale of clustering at trec (the earliest epoch at which the model
could be relevant). This scale is in the general range 10 7 - 10 10 Mo ,
depending on«: in other words, smaller than galaxies but larger than a
present-day globular cluster.

Insofar as a naive gravitational clustering model is relevant, we

expect the scales that virialise at time t (i.e. have M ~ Me(t)) to be
subsequently subsumed into larger-mass systems. The subclustering will
then eventually be erased, and its binding energy redistributed (if a <
2/3 the energy is mainly associated with the largest scales anyway);
internal dynamical relaxation processes will have time to operate on
progressively larger scales. In view of this, it should perhaps come
as no surprise that the N-body simulations by Aarseth, Gott and Turner
(these proceedings) suggest that the form of ~(r) is insensitive to the
initial conditons (i.e. to the ~alue of a in (1)) when ~(r) » 1.
Indeed, it is conceivable that some redistribution of binding energy
between different scales could occur even in the so-called "linear"
domain where ~ ~ 1.

If the clustering proceeds in the kind of "self-similar" fashion

discussed by Press and Schechter (1974), there is no natural preferred
scale for galaxies. At any time, material will be clumped into units
of typical mass Mc (equation (2)); these units will each contain a few
subunits (which themselves turned around a factor ~ 2 earlier) but any
finer substructure would have been erased.

The origin of cosmic angular momentum is a major problem in such a

scheme. The sugestion that angular momentum would be generated by tidal
torques between neighbouring protogalaxies has been a subject of consider-
able controversy. Recent numerical simulations of this theory by
Efstathiou and Jones lend' strong support to the earlier simulations by
Peebles (197la), and show that the tidally generated angular momentum
is not dynamically important unless protogalaxies undergo a significant
amount of collapse before settling down to their final configuration.
In the usual dimensionless units, the generated angular momentum is

~ = 0.06 + .03

A protogalaxy would have to collapse by almost a factor 10 in radius in

order to make this amounc of angular momentum dynamically important.
This result is a further argument that dissipation must play an important
role in the formation of galaxies. (The existence of a disk is manifest
evidence of dissipation, though it does not tell us by how much the
galaxy has collapsed.)

It might be remarked that tidal torques are not the only means of
producing galactic spin in this theory. The coalescence of substructure
at various levels of the hierarchy will lead to spinning stellar systems.
There is an upper limit on the amount of angular momentum generated in
this way because systems colliding with high relative velocity will not
THE EPOCH OF GALAXY FORMATION 381

stick together. (This is a possible explanation for the observation

that elliptical galaxies - interpreted as low-angular momentum systems -
lie predominantly in rich clusters of galaxies where such collisions
will be high velocity collisions.) A certain amount of dissipation will
be required in such a model, but possibly not as extreme as in a pure
tidal torque theory.

5. COSMIC TURBULENCE

In this theory, galaxies are supposed to grow out of primordial

turbulence thrown into a state of supersonic chaos at the epoch of
decoupling. The recombination of cosmic plasma takes a finite time
during which there is considerable damping of primordial eddy motions
on galactic scales (Chibisov, 1971). Dense lumps can form as a result
of centrifugal compression during recombination (Peebles, 1971b; Jones
1977), though the origin of cosmic angular momentum on galactic scales
becomes as serious a problem in this theory as in the gravitational
instability theory.

6. CONSEQUENCES OF EARLY (z > 100) STAR FORMATION

If the spectrum of inhomogeneities surviving the recombination

period extends to masses smaller than ~ 10 5 M@, and if the amplitude of
the spectrum increases towards smaller masses, then the first objects to
condense out would be those of the Jeans Mass at recombination

(4)

There is no reason, apparently, why such objects (incorporating most of

the primordial gas) should not quickly collapse and fragment, but we can-
not predict the characteristic masses of these "Population III" stars.
The left-over material would be re-ionised, (unless too few high mass
stars are formed), and indeed this re-ionisation may be responsible for
turning off the star formation process. There are three interesting
consequences of such early star formation. 1. Galaxies "born" in this
way would make little contribution to the background cosmic light (Dube
et al. 1977). 2. "Young" galaxies (by which in this scenario one means
stellar systems of galactic mass undergoing their violent relaxation
towards the smooth symmetric systems observed today) are not strikingly
luminous objects of the kind described by Partridge and Peebles, Tinsley,
Larson, Meier and others. 3. The small-scale angular variations of the
microwave background radiation would (as we explain below) be undetect-
able. Astronomers could hardly hope to obtain direct evidence for galaxy
formation under these circumstances!

If at epoch z a fraction f of the cosmlC matter is left over as

uncondensed gas, and has ionisation x e ' optical depth unity corresponds
to a redshift of order
382 BERNARD J. T. JONES AND MARTIN J. REES

where ~ and h describe the present universe. The uncondensed gas is

still dynamically counled to the cosmic radiation f'ield if' the "Compton
drag" timescale t drag '" (ffip/me)nH/n e t comp is shorter than the cosmic
expansion timescale. (f, the gas fraction, does not appear in this
expression.) This condition amounts to

(6 )

If star formation maintains a high level of' ionisation the cosmic radia-
tion f'ield can prevent the generation of large scale gas motions until
redshifts on the order of 100 or so. Optical depth unity corresponds to
mass scales

M
T=l
'" 10 8 [~)-6(f'X
1000 e
)-3 (~h2)-2 M
0

Combining (6) and (7) gives another expression f'or the mass-scales invol-
ved in this process:

(8)

where (6) demands the constraint xe ~ 0.05 in order that the matter-
radiation coupling should persist after recombination.

If' the f'raction of' uncondensed gas, f', were sufficiently low (say
on the order of' a few per cent), according to equation (5) astronomers
would be seeing back to redshifts where, owing to Compton drag, there
were no gas motions on scales smaller than M given by (7). A fluctuation
of' mass M subtends angular scales

Because the temperature fluctuations would be mainly due to scattering

off matter moving relative to the general cosmic background, the radia-
tion temperature would have no detectable variations on scales less than
several arc minutes.

Whether or not we can see as f'ar back as this time depends on the
details of the ionisation history and precisely how much matter is left
in gaseous f'orm. If' the ionisation is maintained near unity all the
time, and f '" 0.1, then according to (5) the last scattering "surface"
lies at a redshift zl ~ 30 (~h2)-1/3, and there has been a considerable
time since z ~ 140 in which to generate gas motions. (The gas moves
around in the fluctuating potential of the stellar matter aggregates.)
Now an additional effect comes into play due to the "fuzziness" of the
THE EPOCH OF GALAXY FORMATION 383

last scattering surface: a large number of inhomogeneities contribute

to the temperature fluctuation observed on anyone line of sight. The
number of such inhomogeneities is roughly the ratio of the horizon scale
at the redshift of last scattering to the inhomogeneity scale at that
epoch:

(10)

The angular scale is still of the order (9) provided ~z »1. The
relative temperature fluctuation oT/T produced by one fluctuation is
on the order of vic where v is the gas velocity generated. Because the
ionisation is held constant in this version of the model, the observed
oT/T is a factor N down on this. The mass scale observed depends on
the angular scale being looked at via relation (9). On galactic scales B
or smaller, where the observed fluctuation scale would be less than an
arc minute, the relevant velocities would be typical gas inflow
velocities: a few hundred kilometers per second. On larger scales like
M ~ 10 15 Mo ' we are looking at angular scales e ~ 10 1 • The clusters
will not have turned around and collapsed at the last scattering sur-
face, so the relevant gas velocities are at most several hundred kilo-
meters per second. Considering (10) then, the expected temperature
fluctuations oT/T will be at most a few times 10- 5 on angular scales
8 < 10 1 . The temperature fluctuations on angles 8 » 10 1 would be due
to mass-scales where the density enhancements are still in the linear
regime: the gravitational effects of such large perturbations on the
last scattering surface would affect the observed temperature; and the
whole "local supercluster" would acquire a detectable peculiar velocity
if it lay (off-centre) within one such irregularity.

As a last comment on these ionised remnant gas models, it should

be noted that while the matter and radiation are Compton-coupled, the
electrons remain at the temperature of the cosmic radiation field. The
gas is thus ionised without being "hot" (> 10 40 K). Consequently, the
distortions to the microwave radiation spectrum (estimated via the
famous y-parameter) due to this prolonged period of non-neutrality are
small. The coldness of the neutral component means also that much
neutral hydrogen could be in molecular form.

T. CLUSTERS AT z ~ 2.

Evidence on the gas content within clusters (or protoclusters) at

large redshifts would help to discriminate between the types of model
discussed in §3 and §4 respectively. In the Doroshkevich et al scheme,
a protocluster is predominantly gaseous when it turns around and is
shock-heated, the gas content being thereafter gradually depleted as
galaxies condense via thermal and/or gravitational instability; on the
other hand, those theories in which clusters form from pre-existing
galaxies predict that the gas content, perhaps very low at turnaround,
384 BERNARD J, T, JONES AND MARTIN J. REES

may increase with time owing to infall, ejection from individual

galaxies, etc.

Two observational handles on this question (Fabian et al 1977)

which are already feasible (or should soon become so) ar~--

(i) HEAO B observations of thermal emission from clusters at z =

(1 - 3) may be responsible for the bulk of the observed X-ray background
below 'V 10 kev. Individual clusters should be resolvable; and "counts"
in small areas of sky would indicate at least the sign of the evolution-
ary effect and also constrain the epoch at which cluster formation in
the Doroshkevich et al scenario could have occurred (cf Kellogg (these
proceedings), Cowie and Perrenod (1977), Cavaliere et al (1977)).

(ii) If clusters of galaxies at (say) z ~ 2 contained more hot gas

than present-day clusters, then each should produce a larger fractional
dip in the microwave background (by the Sunyaev-Zeldovich Compton
scattering effect) than nearby clusters in which the effect may already
have been measured (Parijskij 1973, Gull and Northover 1976, Lake and
Partridge 1977). If Les is the optical depth of a cluster to electron
scattering, then the effect
. is of order Les
. (kTvirial/me c 2 ). The
temperature perturbatlon due to the stralghtforward doppler effect on
the cluster material (even if it is undergoing systematic rotation or
collapse) could not be larger than Les (Vvirial/C). Since kTvirial ~
(Vvirial/C)2~ c 2 , the Sunyacv-Zeldovich effect must be the dominant
one if Vvirial > (me/~) c, which is c~rtainly the case for clusters.
Note, however, that thls argument applles only after the gas in the
protocluster has been shock-heated.

8. A "SYNTHETIC COMPROMISE" MODEL FOR THE FORMATION OF GALAXIES AND

THEIR HALOS.

Although all scenarios involve a systematic conversion of gas into

"stellar" material between z ~ 1000 and the present epoch, they make
different predictions about how (and exactly when) this happens, and
about the order in which the various mass-scales separate out. The
resemblance between what is seen in the real sky and the results of
N-body simulations by Aarseth and his associates and by Peebles suggests
that unadulterated gravitational clustering has basically determined the
large-scale distribution of luminous matter. On the other hand, the
basic properties of galaxies - their high-luminosity cut-off, the
flattening of disc systems, and the gradient of M/L with radius - seem
inexplicable without assigning some role to dissipative gas-dynamical
processes.

White and Rees have developed a theory for galaxy evolution which
lS in some sense a compromise or synthesis between the schemes described
In 53 and §4. The basic hypothesis is that by z = 100, about 80 per
cent of the primordial material condensed into "dark stars". These
stars would originally be grouped in units smaller than galaxies, but
THE EPOCH OF GALAXY FORMATION 385

clustering would develop on progressively larger scales. This material

eventually constitutes the halos of massive galaxies. (For the purposes
of the model, all that is required is that this material undergoes non-
dissipative gravitational clustering as in §4. More exotic possibili-
ties such as heavy neutrinos or primordial black holes would work equally
well). In this respect the theory is like the gravitational clustering
theory. The remaining gas subsequently settles into the potential wells
associated with these dark halos, and fragments into the stellar popula-
tions we now see. The characteristic upper limit to the mass and size
of galaxies is determined by the requirement that the gas must be able
to cool and fragment in the available time.

The theory is motivated by the following remarks:

(a) A galaxy formation theory should account for the existence of a

large amount of non-gaseous "dark matter" which apparently provides at
least 80 per cent of the virial mass of cluster like the Coma cluster,
and which may reside in massive halos around large galaxies. The
"luminous" material is much more "clumped" than the "dark" matter.

(b) Dissipation almost certainly played a role in the formation of disk

galaxies (and perhaps in the central luminous parts of elliptical
galaxies) .

(c) Uncondensed gas exists ln clusters of galaxies (it is also enriched

with heavy elements).

(d) The characteristic mass and size of galaxies has no natural inter-
pretation in a purely gravitational picture.

If the hierarchical clustering is described ln discrete steps, each

lasting twice as long as the preceding step, the total number of steps
since recombination is ~ 15 0- 1 / 2 . The characteristic mass which is on
the verge of collapse at each step increases by 2 2/ 3a at each step (cf
(2)). The parameter Fi denotes the initial fraction of left-over gas,
so that 1 - Fi is the fraction of "dark matter" in the universe. The
fraction of the gas available at the ith stage of the hierarchy that
forms into stars is denoted by fi. This fraction presumably depends on
the binding energy of the "dark halo" into which the gas is falling,
and the simplest assumption is
M.
f. ex: 1 (11)
1 R.
1

where Mi and Ri are the mass and maximum radius of the dark component at
that level of the hierarchy (cf Larson 1974). For different masses, f
scales as

(12)
386 BERNARD J. T. JONES AND MARTIN J. REES

so for a < 2/3 the fraction of infalling gas that gets turned into
luminous stars lS very small for the early stages of the hierarchy.

The value of fi is chosen by requiring that some 50 per cent or so

of the available gas has been turned into luminous material by the
present epoch. The 50 per cent is chosen on the basis of the observa-
tion that the mass of uncondensed gas in the Coma cluster is of the
same order as the luminous mass. The assumption that M/L for the Coma
cluster is the same as that for the whole Universe favours ~ ~ 0.2.
From the equation

gas
gas + luminous matter n
hierarchy
(1 - f.)
l

steps

we can deduce the value of f at the last step of the hierarchy, f max '
(We know the characteristic mass Mi at each step.)

The only parameter for which we have little basis for choice is
a, which is generally taken to lie in the range 1/3 - 1/2.

Several conditions must be satisfied in order that the gas falling

into a dark-halo potential well should fragment into stars. The cooling
time of the gas must be shorter than the Hubble time. It must also be
short enough so that the gas can cool before the dark matter itself is
absorbed into the next level of the hierarchy. It is also necessary
that gas should be able to cool and collapse enough for its self
gravity to dominate over that of the dark matter into which it is
falling. Otherwise it cannot fragment efficiently. The radius of the
gas cloud should therefore satisfy a condition like

R ~ (fF)1/3 R,
gas

where R is the halo radius, F is the ratio of gas mass to halo mass,
and f is the fraction of the gas that can be turned into stars. (The
exponent in this expression in fact depends on the details of the den-
sity distribution.)

It can be shown that with the preferred values of Fi, f max , ~ and
a, the conditions that the infalling gas should be able to fragment into
stars yields an acceptable upper limit to the mass and radius of the
luminous part of a galaxy.

Low-mass galaxies would tend to have formed by gas that accumulated

and fragmented into lower-mass "halos" which existed at earlier stages
in the development of the hierarchy. These "luminous cores" could pre-
serve their identity even if they are embedded in halos that coalesce
when the next stage of the hierarchy builds up. On this assumption,
one can calculate a luminosity function which turns out to be rather too
THE EPOCH OF GALAXY FORMATION 387

steep at the faint end (~ (L) varying about as L-1.8 rather than the
canonical ~ L-1.25). Mergers may however occur (large galaxies with
several smaller satellites being a common expected phenomenon on this
model). Dynamical friction in clusters, and differing MIL (arising
from variations in the star formation rate, even for a universal IMF),
are further effects which modify ~(L). There would be a general
correlation between metal abundance and galactic mass (though an impor-
tant constraint on the Population III objects is that they should not
produce too high a heavy element abundance). But the most important
feature of the model is that it suggests what physical processes distin-
guish a characteristic galactic mass from a typical cluster (or group)
mass (cf Press & Schechter 1974); and also accounts for the distinction
between the relatively diffuse "dark" material which is gravitationally
dominant and the 10 per cent of "luminous" material that arises from
"secondary" star formation in the potential wells.

REFERENCES

Blandford R.B. and Rees, M.J. 1977 Proceedings of Copenhagen conference

on Active Nuclei (in press).
Binney, J.J. 1977 Ap. J. 215, 492.
Butcher, H. and Oemler, A. 1977 Ap. J. ln press.
Cavaliere, A., Donese, L. and de Zotti, G. 1977, Astron. & Astrophys.
In press.
Chibisov, G.V. 1972, Sov. Astr. J. 16, 235.
Cowie, L. and Perrenod, S. 1977 Ap. J. in press.
Doroshkevich, A.G., Sunyaev, R.A., and Zeldovich, Ya.B. 1974 in
"Confrontation of Cosmological Theory and Observational data" ed.
M.S. Longair (Reidel, Holland).
Dube, R.R., Wickes, W.C., and Wilkinson, D.T. 1977 Ap. J. Lett. 215, L51.
Fabian, A., Nulsen, P., Rees, M.J., and Setti, G. in preparation.
Gull, S.F., and Northover, K.S.E. 1976, Nature 263, 572.
Jones, B.J.T. 1977, M.N.R.A.S. 180, 151. ---
Lake, G. and Partridge, R.B. 1977 in preparation.
Larson, R. 1974 M.N.R.A.S. 169, 229.
Parijskij, Y.N. 1973, Ap. J~ett. 180, L47.
Peebles, P.J.E. 1971a Astron. & Astrophys. 11, 377.
b Astrophys. & Sp. Sci.:ll, 443.
Phillips, S., Fong, R., Ellis, R.S., Fall, S.M~ MacGillivray, H.T.
1977 M.N.R.A.S. in press.
Press, W.H. and Schechter, P. 1974 Ap. J. 187, 425.
Rees, M.J. and Ostriker, J.P. 1977 M.N.R.A.S. 179.
Silk, J. 1977 Ap. J. 211, 638.

DISCUSSION

Ostriker: One of the principal reasons for believing that galaxy form-
ation is late is the relatively low velocity dispersion or density. The
388 BERNARD J. T. JONES AND MARTIN J. REES

halo density is even lower indicating a very recent epoch for its form-
ation. Does this present any difficulty for the theory?

Jones: It is certainly a good point. I am not sure to what extent this

rules out such scenarios.

Silk: If star formation occurs as early as z = 100, how does the pro-
posed model arrange to have just 20 per cent of the initial mass left
over in gaseous form for subsequent dissipational galaxy formation at a
much later epoch?

Jones: When stars first form the gas may be reionized and thus recouple
to the cosmic radiation field. This may prevent further star formation.
Why 80% we cannot as yet say of course: that is a difficult problem.
Observationally it has to be about that value.

Tifft: Statements such as "80% of dark matter unequivocally exists"

ignore the fact that observations, good observations, exist which under-
mine the dynamical concepts upon which the existence of the mass depends.
There are no direct observations of this mass. Redshift-magnitude bands
in clusters (5 or 6 now), the evidence for a discrete or periodic red-
shift (especially in radio redshifts), and other observations strongly
imply a part or all of the redshift is an intrinsic phenomenon. We
should at least admit of this possibility. There are no observational
inconsistencies in this type of model and we are naive to believe we
understand all of physics.

Fall: How is your calculation of small scale fluctuations in the micro-

wave radiation specific to the White-Rees model and different from other
models?

Jones: In this model, there is only a small fraction of the matter in

the Universe in the form of gas at any redshift and a low value of the
density parameter, ~ = 0.1, is adopted. Therefore, one can see very
much further back in redshift than in the conventional model so that the
peculiar velocities of the perturbations are much smaller and consequent-
ly give rise to smaller fluctuations of the microwave background radia-
tion.

Tinsley: Jones pointed out that to get enough angular momentum in

galaxies, a lot of dissipation is required. This requirement is consis-
tent with the picture I discussed yesterday, in which disk galaxies form
gradually and late. They could form from matter that was a long way out,
falling in slowly and bringing angular momentum (cf. the model of
Ostriker and Thuan, 1975, Astrophys. J., and comments by Binney, 1977,
Yale conference).

Jones: That is encouraging, but I do think we should look at Ostriker's

point more closely before being carried away.
QSO ABSORPTION LINES AND INTERGALACTIC HYDROGEN CLOUDS

Bruce A. Peterson
Anglo-Australian Observatory
Sydney, Australia

The absorption line spectra of some QSOs are similar to the absorp-
tion line spectra imposed by the interstellar medium in our own galaxy
on the spectra of hot stars (see Figure 1). The spectra of other QSOs
show P Cygni profiles similar to those of hot stars with mass outflow
(see Figure 2). The absorption lines in both of these types of spectra
are due to the resonant transitions of hydrogen and of the more abundant
elements.

The spectra of high redshift QSOs show a multitude of absorption

lines on the short wavelength side of the Ly-a emission line (see Figure
3). As almost none of these lines on the short wavelength side of the
Ly-a emission line can be identified with heavy elements in recognizable
redshift systems, these absorption lines are attributed to the neutral
hydrogen component of clouds in the general intergalactic medium and in
intervening galaxies.

In order to study the properties of the clouds in the intergalactic

medium, spectra were obtained of QSOs with redshifts in the range
2 < Z < 3.6. The QSOs were selected to have similar luminosities. The
preliminary results from a study of these spectra are reported here.
The assumptions made are that the absorption lines have similar charac-
teristics in these spectra (Figure 4) and.that the wavelength distribu-
tion of the absorption lines in QSO spectra is related to the distance
distribution along the line of sight to the QSO of the neutral component
of hydrogen clouds which are part of the intergalactic medium.

If the clouds are assumed to have a characteristic radius and

density which are constant with cosmic time, and if the co-moving
density of clouds is also contant with cosmic time, then the number of
absorption lines produced by the clouds in a given wavelength interval
(which corresponds to a unit redshift interval) will be only a slowly
varying function of redshift.

The number of clouds intercepted along the line of sight dl is given

389

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 389-392 All Rights Reserved.
COPYright © 1978 by the fA TT
390 BRUCE A. PETERSON

by dN = a P dl, where a is the cloud cross section, p is the cloud

number density and dl is the ~roper length along the line of sight.
Using the relations p = (l+z) Po and dl/dz = c/[(l+z)H(z~ , we have

(l+z)
dN/dz = a Po I (1)
(1+2qoz)~
for the number of clouds in a unit redshift interval. As can be seen
from equation (1), dN/dz varies slowly with z.

The observations have revealed a quite different result. In

Figure 4 are shown the spectra of four QSOs which have been studied
with 0.1 nm resolution. It can be seen in Figure 4 that many absorp-
tion lines lie on the short wavelength side of the Ly-a emission line,
corresponding to absorbing hydrogen at intermediate distances, while
virtually no absorption lines are seen on the long wavelength side of
the Ly-a emission line, where the wavelength region corresponds to
hydrogen at a greater redshift than the QSO, or to hydrogen falling
into the QSO. The spectra in Figure 4 are arranged in order of red-
shift with the spectrum of the QSO having the highest redshift at the
top of the figure. It can be seen in Figure 4 that the number of
absorption lines in a fixed wavelength interval is correlated with the
redshift of the QSO in the sense that the QSO with the larger redshift
has the greater number of absorption lines.

The relation between Nand z given by equation (1), for uniformly

distributed clouds with invarient cross section and density, is compar-
ed in Figure 5 with the results obtained from the QSO spectra shown in
Figure 4. The rapid increase of N with z which was found from the
observations can be understood in terms of a progressive ionization of
the intergalactic medium, starting at z ~ 4 with the turn on of QSOs'
(Bergeron and Salpeter 1970), or in terms of a change in cross section
brought about by the collapse of the clouds, perhaps to form galaxies.

The spectra discussed above cover the same intrinsic redshift

range for QSOs with different redshifts, but similar luminosity. What
remains to be done, is to obtain spectra over a large redshift range in
the same QSO so that the N of z relation obtained from spectra of a
single object can be compared with the N of z relation found from
objects at different redshifts, as discussed here. If the lines are
produced by the Ly-a transition in intergalactic clouds, the N of z
relations found from spectra covering different regions of the intrin-
sic QSO spectrum should agree.

Also, confirmation of the identification of the lines with the

Ly-a transition should be obtained by observing the region of the spec-
trum where the corresponding Ly-S lines can be seen. While the identi-
fication of individual Ly-a, Ly-S systems may not be possible, a change
in the character of the absorption spectrum due to the Ly-S lines should
be detectable.
QSO ABSORPTION LINES AND INTERGALACTIC HYDROGEN CLOUDS 391

The observations discussed here were obtained with the Image-tube

Dissector Scanner and the Image Photon Counting System on the Anglo-
Australian Telescope in collaboration with Drs D. Jauncey, A. Wright
and D. Morton.

REFERENCE

Bergeron, J., and Salpeter, E.E., 1970. Ap. Letters, L, 115.

•
290 .....

Figure 1. The spectrum of Pks Figure 2. The spectrum of Pks 1157-

0528-250, an example of absorp- 054, an example of absorption lines
tion lines produced in the metal produced in the H II region
enriched interstellar medium of surrounding the QSO.
an intervening galaxy.

Figure 3. The spectrum of Pks Figure 4. Spectra of Pks 1442+101,

2156-15, an example of mUltiple z = 3.53, Pks 0805+046, z = 2.88,
absorption line redshifts Pks 0329-255, z = 2.68, and Pks
produced by the neutral hydrogen 1448-232, z = 2.21. These spectra
component of clouds in the cover the same intrinsic wavelength
intergalactic medium. range in the spectrum of each of the
four QSOs. The Ly-a emission lines
are prominent features near the
right-hand ends of the spectra.

Z, • Figure 5. The relations between line density,

dN/dz and redshift, z. The thin curves in the
22 0·'· figure show the relation given by equation (1)
I(j 20 0-- for constant cloud properties. The diamonds
were derived from the spectra in Figure 4. The
~" ~:;.
~
,. ~.u
z extent of each diamond represents the ·~ange
in z covered while the extent in dN/dz repre-
sents an estimate of the error in determining
•
logZ
dN/dz. These are preliminary results.
392 BRUCE A. PETERSON

DISCUSSION

de Vaucouleurs: Is there any evidence which allows you to distinguish

intergalactic HI clouds from ordinary galactic discs?

Pete~son: Yes, in some cases we can see absorption lines of heavy

elements in the absorption systems as would be expected in the gaseous
discs of galaxies.

Silk: Can you comment on the metal content in any of the absorption
line systems?

Pete~son: No, not yet.

FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION

R. A. Sunyaev
Space Research Institute
USSR Academy of Sciences
Moscow, USSR

Investigations of small scale angular fluctuations and the spectrum

of the microwave background radiation is one of the main methods of
studying the large scale structure of the Universe. Figure 1 shows the
principal stages of the evolution of the Universe. Today we can direct-
ly observe galaxies, clusters of galaxies and quasars in the redshift
range z ~ 3.5 by optical, radio and X-ray astronomy. These observations
show that significant density perturbations op/p > 1 are present on mass
scales M < 10 16 M0 . The Universe is essentially uniform op/p < 1 on
large scales M » 10 16 M®.

nuclear IOniZed Ionized objects galaxies,

reactions matter gas clusters , q.uasars

'! ,. !m
10' adiabatic
sound
1500-1000 4 _ 20' 3 5 && 1 0
rf!COrnblnatl 0 n reheating y>1 M <10 16 M.
growth ot
small
perturba lions
¥<1. M>10'6 Mo

Figure 1. A schematic diagram showing the evolution of

the Universe.

We can see from Figure 1 that at z ~ 10 annihilation of electron-

positron pairs which were numerous at that epoch and nuclear reactions
associated with helium synthesis took place in the Universe, which was
filled with matter and radiation at the temperature Tr = 2.7(1+z)K. At
the epoch z ~ 1500 recombination of hydrogen took place. This is an
important moment: up till that time the optical depth of the Universe
to Thomson scattering was very large but after that the Universe became
transparent. In principle, by observing the relict radiation we can
obtain direct information about the state of the Universe at the moment
z ~ 1000. After that photons of the relic radiation do not interact
with matter.

Thus the matter in the Universe was ionized before z ~ 1500, and
afterwards became neutral. However, we do not see Lyman-a absorption
393

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe. 393-404. All Rights Reserved.
Copyright © 1978 by the [AU.
3M R.A.SUNYAEV

bands in the spectra of distant quasars. This means that the matter
must be ionized at z < 3.5. We do not know when this secondary ioniza-
tion took place. If the gas was ionized up to a redshift z ~ 10-20,
the optical depth due to the Thomson scattering was very large. These
scatterings smooth out the fluctuations by a factor e-TT that existed
up to that time and generate new ones.

According to the present view, the observed large scale structure

of the Universe was formed as a result of gravitational instability and
the growth of small density perturbations, ap/p « 1, that existed at
the early stages of the expansion z »10 10 . These perturbations had a
broad mass spectrum and somewhere at z > 1 ap/p became larger than one
which resulted in the formation of observed objects. Silk (1967) first
noted that the existence of density perturbations at the epoch of recom-
bination must lead to fluctuations of the relict radiation. The fluc-
tuations are proportional to the amplitude of perturbations, and,
therefore, by observing them it is possible to determine ap/p at the
moment of recombination (or reheating) and also the epoch of the forma-
tion of galaxies or (more accurately) of clusters of galaxies.

Three sources of the fluctuations will be considered below:

(1) Perturbations of the matter density ap/p and velocity v at
the epoch of recombination, z ~ 1000 - primaeval fluctuations.
(2) Perturbations of density ap/p and velocity v at the epoch
of reheating at z ~ 11 ~-1/3(if the intergalactic matter was
ionized at that time, of course) - secondary fluctuations.
(3) Observed sources: radio sources, clusters of galaxies,
young galaxies, non-uniformities of hot intergalactic gas (IGG),
protoc1usters of galaxies, etc.

1. PRIMAEVAL FLUCTUATIONS

There are three types of density perturbations: adiabatic,

enthropy (isothermal) and vortex perturbations. The last type will be
described in detail in the lecture of L. M. Ozernoy and therefore I will
only discuss in detail the first two types.

(a) Adiabatic perturbations of density

Before the epoch of recombination these perturbations were sound

waves in which the densities of matter and radiation increase and
decrease together. The evolution of these perturbations has been des-
cribed in a number of publications (Sunyaev and Zeldovich 1970, Peebles
and Yu 1970, Doroshkevich, Zeldovich and Sunyaev 1977).

Let us write down the perturbations of density and their velocities

in terms of a Fourier integral.
FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION 395

=
In this case
=
The mass contained within a perturbation of wavelength A 2TT/k is
equal to
m
where Ho is Hubble's constant.

If, at a sufficiently early time (z » 10 8 ), the spectrum of

fluctuations was of power law form ak ~ k n then the evolution of the
spectrum in the radiation dominated stages of expansion results in a
spectrum of fluctuations after recombination of the form shown in
Figure 2, i.e. a~ ckak
,om. kR-:r

for R < RJ and ck = 1 for R > RJ (see Doroshkevich et al. 1977). Here
RJ is the Jeans' wavelength at the moment of recombination
f11 J % Jo'~ (J/.1\.2.)-2 fflQ

and Rc is the scale on which according to Silk viscous dissipation of

the perturbations is important Mc:'ll 101Ll01S M0 •

(1): "-
"-
"-
" "-
2In-11'-kRc 2
- k • ....510 ICRJ'

..
M°.!.

Figure 2. The spectrum of adiabatic density perturbations

after recombination. The initial spectrum is assumed to be
a power law ak ~ kn. The spectrum of entropy perturbations
does not change and remains of power law form with the same
spectral index.

After the epoch of recombination fluctuations of density grow on

all scales having mass M > 105 Me according to the law
396 R. A. SUNY AEV

Jf (~) oC t~/3 oC {J+2)-1

!
until the time when op/p is of order unity and the formation of gravita-
tionally bound systems begins. We note that for cases ~ < 1 the pertur-
bations grow rapidly only up to redshift z ~ ~-1. The subsequent growth
1S very slow (see Sunyaev 1971).

Perturbations of density and velocity are related by the continuity

equation
.L(~) =
dt .f
from which it follows that fluctuations in velocity grow according to
the law v cr (1+z)-1/2 cr t 1/ 3 . This equation gives a simple relation
between ak and bk.

In the cases, which are of most interest to us, the most important
contribution to the fluctuations results from scattering of radiation
by moving electrons (resulting from the velocity field of the perturba-
tion) (see Sunyaev and Zeldovich 1970). In fact, because of the Doppler
effect, the temperature of the scattered radiation depends on the direc-
tion of motion of the electrons, T(a) = To(l + vic cosa).

For a continuous medium with variable degree of ionization we have

where
oIT =
is the optical depth to Thomson scattering; u1(z) is the projection of
the velocity vector onto the line of sight and the exponent e-T(z) takes
into account the smoothing out of the fluctuations which originate in
regions with large optical depth. The function e-TdT/dz is easy to
derive knowing the law of change of the electron density during recom-
bination. This function has a sharp maximum at z = 1100 and halfwidth
6z/z % 7%. Fluctuations originate mainly in the region of this maximum.

This function determines the scale M~ 1015 M:@

if the mass of the
perturbation is smaller, then a great number of wavelengths can be
located within this zone. The velocity of matter changes direction many
times (see Figure 3) and the effect decreases strongly due to these
anticorrelations. For perturbations on larger scales the velocity does
not change within this zone and we obtain the full effect. In this case
recombination is rather fast and by observing the fluctuations in the
radiation we can obtain an overall picture of density perturbations at
the epoch of recombination.
FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION 397

<f[ -"t d"t -"t

d'Ze dZ·

:0
~/--~----,
(0) (b)

(e)

Figure 3. Illustrating how fluctuations of the relic

radiation are formed. The bell-like figures show the
regions in which the main effects are formed. At the
bottom of top Figure 3(a) sinusoidal perturbations of
velocity with small wavelength are shown. In this case
the velocity changes sign many times in a bell zone,
and therefore the effect is reduced. In Figure 3(b)
the case of long wavelength velocity perturbations is
presented. Effect appears with full strength. In
Figure 3(c) two zones are shown where the perturbations
are formed.

Results of detailed calculations (Sunyaev and Zeldovich 1970,

Doroshkevich et al. 1977) are shown in Figure 4. The abrupt damping of
the fluctuations on scales M ~ 10 13_10 14 M~ is connected with viscous
damping of the amplitude of the fluctuations (see Figure 2). It can be
seen that the most interesting scales for observations are 2' - 20' (the
exact position of the break in the dependence ~T/T(a) depends on nh 2 ).
On small angular scales primaeval fluctuations are quite small and we
can only obtain information about the spectrum of fluctuations on scales
M p 10 13 M@.

The amplitude of intensity fluctuations depends on the amplitude of

density perturbations at the epoch of recombination. i.e. on the time of
the formation of the clusters of galaxies. Figure 4 shows the case
n = 1 and op/p(1015 M ) = 1 at Zo = 3. It is evident that at the epoch
of recombination, op/~ ~ (l+zo)/(l+zrec):::: 1/350. If clusters were
398 R. A. SUNY AEV

4.I..
T
adiabatic
perturbations

region
of
VISCOUS
diSSipation
§.,t ( 10" M.I =1
oJ
at Z =3

e
5' 25'

e
5' 25'

Figure 4. The spectrum of background fluctuations

corresponding to the power law spectrum of primaeval
perturbations in the case of adiabatic and entropy
perturbations.

formed earlier, then op/p must have been greater at the epoch of recom-
bination and, therefore, ~T/T must be greater on this scale. In the
case of the open world model n « 1, perturbations grew only at epochs
z > n- 1 and therefore the initial perturbations could be even greater in
this case. It is interesting to note that, in this case, at z ~ n- 1
the objects with op/p(z = n- 1 ) >1 are formed, but perturbations with
op/p < 1 are frozen and do not grow any further.

At present the counts of galaxies and radio sources do not give

information about the present amplitude of density perturbations on
mass scale M > 10 16 - 17 Me' Therefore observations of fluctuations of
the background radiation on corresponding angular scales may give
valuable information about their spectrum. In principle, it is possible
that the amplitude of the perturbations on this scale is small. In this
case the fluctuations must also be small on the scale e ~ 20'-30'. The
angular scale of fluctuations is related to the mass of perturbation by
the relation e = 5' M~~3 n2 / 3 .

(b) Entropy perturbations

Before recombination entropy perturbations were density perturba-

tions of protons and electrons Pm in a background of constant radiation
FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION 399

density Pro Let us note that before recombination (at z > 10 4 Qh2) the
radiation density Pr = Er/C 2 = 4 x 10-34(1+z)4 g cm- 3 exceeded the
density of matter Pm = 5 x 10-30Qh2(1+z)3 g cm 3 . While the primaeval
plasma was ionized, entropy perturbations were frozen and their spectrum
on the scale M > 1 M@ practically did not change. For this reason they
differ greatly from adiabatic and vortex perturbations, in particular
the viscous dissipation of perturbations occurs only for M < 1 M~ com-
pared to the damping scale M ~ 10 13 _10 14 M@ in the case of the adiabatic
perturbations. After recombination, the radiation is weakly coupled
with the matter. Therefore, the growth of entropy perturbations begins
on all the scales M ~ 10 5 M@ due to gravitational instability. In this
case velocity perturbations that lead to fluctuations of the relict
radiation are generated in accordance with the continuity equation as
in the case of adiabatic perturbations. For equal amplitudes at the
moment of recombination entropy perturbations lead to the fluctuations
on the scale 5' about 2-2.5 times smaller than adiabatic perturbations
(Sunyaev and Zeldovich 1970, Zentsova and Chernin 1977, Doroshkevich et
al. 1977). Essentially, this is connected with the fact that in the
first case at the epoch of recombination density perturbations that
existed before the recombination generate velocity perturbations,whereas
in the second case velocity perturbations,which were already great
before the recombination, generate density perturbations.

2. SECONDARY FLUCTUATIONS GENERATED DURING REHEATIN~ OF

THE INTERGALACTIC GAS

At z < 3.5 the IGG is highly ionized (Gunn and Peterson 1965). At
z < 5 it is easy to ionize the IGG by the UV radiation of quasars,
young galaxies and "pancakes". In this case the temperature of gas is
close to T = 10 4 K. The density of gas increases rapidly with z and
its effective emission measure increases correspondingly: to preserve
the high degree of ionization at z ~ 10-20 enormous luminosities,
L ~ 10 46 - 47 erg s-1 and high densities (of the order of the present
density of galaxies) of UV sources are necessary (Weymann 1966, Arons
and Wingert 1972, Sunyaev and Zeldovich 1972, Doroshkevich and
Shandarin 1975, Ozernoy and Chernomordik 1976, Sunyaev 1976). There are
models according to which the gas was ionized by shock waves originating
in the explosions of quasars or by subcosmic rays (Ginzburg and Ozernoy
1965) and also by shock waves generated in the formation of "pancakes" -
protoclusters of galaxies (Sunyaev and Zeldovich 1972). In these models
the gas has to be heated up to a temperature T ~ 10' K. At z > 10 these
models meet with enormous energetic difficulties. We must take the
problem of reheating very seriously because a large optical depth of the
IGG to Thomson scattering in the range z z 10-30 may lead to strong
damping of primaeval fluctuations of the relic radiation. If the bulk
of the matter in the Universe was in a gaseous state and ionized at that
time, it is easy to find the optical depth to scattering.
't' = J&rNe C ott ~ 0.03..n.
'/ZL
n
~/2
:n/NJt
400 R. A. SUNY AEV

At z ~ 11 ~-1/3 we have T ~ 1; in order to ensure strong damping of the

priiaeval fluctuations it is necessary that T ~ 3-5, i.e. zmax ~ (25-30)
~-l 3. It is exceedingly difficult to ensure such early complete
ionization of the IGG. However, let us suppose that this was the case,
and calculate the magnitude of secondary fluctuations. At the epoch
z ~ 10-20, 8p/p did not exceed unity on the scale of clusters of
galaxies (and on larger scales). These perturbations grew in accordance
with the continuity equation and there were also perturbations in the
velocity field. It is easy to calculate fluctuations arising from
scattering on moving matter, according to the same scheme as in the
previous paragraph. In the case_u~der consideration the function .
e- T dT/dz has a peak at z ~ 12 ~ 1/3 and the whole effect accumulates 1n
this zone. The scale M~ 1022~-2M0 is the characteristic mass in this
case. It is analogous to M ~ 10 15 M@ in the case of primaeval fluctua-
tions (see Figure 3). The amplitude of the effect might be very great
since the velocity after the epoch of recombination increases due to
gravitational instability, roughly v/vo·~ j(l+zr)/(l+z) ~ 10 times.
However, we do not know the amplitude of perturbations 8p/p in the
region of such large masses: they may be small and in this case the
fluctuations are small as well.

If there was a power law spectrum of perturbations after recombina-

tion, ak ~ k P the secondary fluctuations, formed due to reheating, are
equal to , p/
- '6 - '3
PI I P
bT _ 3 >< '0-":/- tf//~'1"3- /~
0(" e- _,
6T Z
- p- ~
¥XIO'f
.,
(1'/,<- oee
T '" "
corresponding to small (M < 10 22 Me) and large scales respectively
(Sunyaev 1977). The amplitude depends strongly on the spectral index P.
It is assumed that 8p/p(M = 10 15 Me) = 1 at z = 4. rhe angular dimen-
g/
sions are determined by the relation e ~ 5' M1 3 ~2'3 as before. In
the stage after recombination both adiabatic and entropy perturbations
are of potential form and do not differ except for their spectra.
Observations of fluctuations on large angular scales e ~ 30' - 10 0 may
give valuable information about the spectrum of initial large-scale
perturbations with small amplitude even at the present time which have
not led to the formation of gravitationally bound objects.

Vortex perturbations at this stage lead to

<~T>:::: 5".4 <70 >"'.2,'" ""'~ <~T > r.::

respectively, for small (M < 10 21 Me) and large masses

(Sunyaev 1977) where Vo is the present vortex velocity on scale M. At
the epoch of z ~ 10, vortex velocities were (l+z) ~ 10 times greater.
An interesting situation arises when the heating occurred at z z 12. In
that case the optical depth due to the scattering is small, T ·V 1 and
the initial fluctuations are damped only e times, whereas secondary
fluctuations are formed with full amplitude. Thus both the epoch
z "'" 1000 and z z 10 contribute to the observed fluctuations.
FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION 401

3. OBSERVABLE OBJECTS AS SOURCES OF FLUCTUATIONS

(a) Radio sources

Longair and Sunyaev (1969) have shown that radio sources observed
at long wavelengths (v = 408 MHz) and already in catalogues must lead to
noticeable fluctuations in centimetre wavelength band (oT/T ~ 10- 5-10- 6 ).
The main contribution to fluctuations is given by the sources whose
density on the sky is close to one source per beam width of the radio-
telescope. This estimate is somewhat uncertain since it requires extra-
polation of the results of the counts of sources.

(b) Clusters of galaxies

As a result of the Compton effect the presence of hot gas in a rich

cluster of galaxies must lead to a decrease in brightness temperature of
the relic radiation in the direction to the cluster (Sunyaev and
Zeldovich 1972b),
6.T = -:2 kTe o'rNee z 10-4-
T 111,,2

This effect has been observed in the last few years by Parijskij, Gull
and Northover and Partridge. It is evident that all rich clusters taken
together must be a source of fluctuations of the microwave background.
Estimates show that ~T/T z 10- 5 can be expected on an angular scale of
10'-20' .

The magnitude of the effect does not depend on the redshift at

which the object is located; it is only the angular dimensions that
change. Hence protoclusters of galaxies in which the gas was heated by
shock waves, generated at the formation of pancakes, might also make a
contribution to the fluctuations. Fluctuations may also arise as a
result of inhomogeneities in the temperature and density of the gas at
the epoch of reheating of the intergalactic gas.

(c) Young galaxies

According to present ideas (see the paper by B. Tinsley) in the

first 10 8 years of their life, the luminosity of the young galaxies was
determined by 0 and B stars. The main part of the energy was emitted in
the form of UV radiation (Weymann 1966, Partridge and Peebles 1967). If
at this stage there was a lot of gas in the galaxies, they were similar
to an aggregate of gigantic HII regions similar to Orion. Under these
conditions bremsstrahlung radio emission of gas in the HII zones and
hence the luminosity of galaxies at centimetre wavelengths were very
great. Such objects may lead to the background fluctuations of the
order of
~T ( 10 )'3
T f.t7"
402 R.A.SUNYAEV

on angular scale of the order of e ~ 10" (Sunyaev 1977). At z = 5-20

young galaxies must undergo significant clustering, since the amplitude
of fluctuations on the scale of clusters of galaxies must be large at
this time. Apparently young galaxies may ensure fluctuations of the
order of ~T/T ~ 10- 5 on the scale e ~ 10'.

4. CONCLUSION

Observations of the relic radiation impose stronger and stronger

limits on the amplitude of the fluctuations (let us note especially the
papers of Parijskij, Carpenter, Gulkis and Sato, and Partridge). The
theoretical analysis presented above shows that background fluctuations
at centimetre and millimetre wavelength must exist at a level ~T/TZ 10- 5
independent of whether there was early secondary reheating or not. At
this level, fluctuations must be observed, which are associated with
known objects - radiogalaxies etc.

REFERENCES

Arons, J. and Wingert, D., 1972. Astrophys. J., 177, 1.

Doroshkevich, A.G. and Shandarin, S.F., 1975. Astron. Zh., 52, 9.
Doroshkevich, A.G., Zeldovich, Ya. B. and Sunyaev, R.A., 197~ Preprint
of Institute of Applied Mathematics, Moscow. (Astron. Zh., in
press).
Ginzburg, V.L. and Ozernoi, L.M., 1965. Astron. Zh., ~, 943.
Gunn, J.E. and Peterson, B.A., 1965. Astrophys. J., 142, 1633.
Longair, M.S. and Sunyaev, R.A., 1969. Nature, 223, 719.
Ozernoi, L.M. and Chernomordik, V.V., 1976. Astra;. Zh., 53, 459.
Partridge, R.B. and Peebles, P.J.E., 1967. Astrophys. J.,-r47, 868.
Sunyaev, R.A., 1971. Astron. Astrophys., 12, 190. ---
Sunyaev, R.A. and Zeldovich, Ya.B., 1970. ~strophys. Sp. Sci., 7, 3.
Sunyaev, R.A. and Zeldovich, Ya.B., 1972a. Astron. Astrophys., 20, 189.
Sunyaev, R.A. and Zeldovich, Ya.B., 1972b. Comments Astrophys. Sp.
Phys., 4, 173.
Weymann, R.J~, 1966. University of Arizona, preprint.
Zentsova, A. and Chernin, A.D., 1977. Pismi Astron. Zh. (in press).

DISCUSSION

Silk: The presence of hot gas produced during the formation of galaxy
clusters suggests that limits on distortions of the Wien region of the
background radiation spectrum may be able to yield significant con-
straints on secondary reheating models.

Sunyaev: I agree with you. Observations of distortions of the

black-body spectrum in the millimeter and submillimeter band give
extremely important restrictions to the temperature of IGG and the time
FLUCTUATIONS OF THE MICROWAVE BACKGROUND RADIATION 403

of reheating (see discussion in the paper Ya B. Zeldovich and R. A.

Sunyaev, Astrophys. Sp. Sci., ~, 285, 1969).

Jones: In principle we should be able to distinguish the effects due to

cosmic fluctuations and discrete sources such as galaxy clusters and
radio sources. Can you comment on these?

Sunyaev: The primaeval and secondary fluctuations of the relic radia-

tion do not depend on the frequency. The contribution of radiosources
is strongly dependent on frequency. Therefore observations at several
frequencies might distinguish these effects. In the case of clusters of
galaxies (holes in background), the measurements at hv ~ kT r are neces-
sary, because in the Rayleigh-Jeans part of the spectrum the effect of
Comptonization does not depend on the frequency.

Ostriker: What do you believe is the origin of the fluctuations in the

distribution of matter and radiation which you assume in your discus-
sion?

Sunyaev: This question will be discussed by Zeldovich in his paper.

Parijskij: You have described many sources of fluctuations of the relic

radiation. What is the total fluctuation you expect on summing all
these effects?

Sunyaev: I do not think it is possible to obtain a good answer for the

total fluctuations because the magnitudes of all the effects I described
in my paper are very poorly known.

Chernin: Where would the energy for the reionization come from?

Sunyaev: I agree with you that there are great difficulties with the
energy sources in the case of very early reheating (z > 5). It seems to
me that heating and ionization by shock waves from "pancakes" and
quasars or by soft cosmic rays at z > 5 is very improbable. These
mechanisms give a high temperature and one needs an enormous energy
release. However, in the case of ionization of the intergalactic gas
(IGG) by U-V radiation from quasars and young galaxies, the temperature
of gas might be low, T ~ lO~ K, and the energy requirements are smaller.
Even in this case we have great trouble with the sources of the U-V-
radiation, because the effective "emission-measure" of the IGG increases
rapidly with redshift.

van der Laan: You have suggested a large, even bewildering variety of
reasons why we should expect ~T/T ~ 10- 5 • The last ten years have
demonstrated how difficult and expensive (in telescope time especially)
these measurements are. Every measurement is at a particular frequency
Vi and covers usually a small range of angular scales ~ei. To motivate
observers and to persuade committees that allocate telescope time,
theorists should specify the value or information content in the (v,e)
plane as precisely as they can.
404 R. A. SUNY AEV

Sunyaev: The best wavelength for observations of "real" cosmological

fluctuations is the band close to the maximum of the 2.7 K black body
spectrum. However, the atmospheric conditions are better in the centi-
metre band. The best angular scales are indicated in the written
version of this paper.

CONSTRAINTS ON THE MEAN DENSITY OF THE UNIVERSE WHICH FOLLOW

FROM THE THEORIES OF ADIABATIC AND WHIRL PERTURBATIONS
A. A. Kurskov and L. M. Ozernoy

The aim of this communication is to investigate what constraints

to the cosmological parameter Q = 2qo can be obtained if one assumes
that primaeval whirl motions or adiabatic density perturbations with an
appropriate initial spectrum were responsible for the formation of
large scale structure in the Universe. These constraints are readily
obtained from the two conditions: (i) an upper limit to small scale
temperature fluctuations of the microwave background radiation, and
(ii) the requirement that the primaeval perturbations should be large
enough in order to produce observed structures.

(a) Adiabatic perturbations. One usually makes the reasonable

assumption that the initial metric perturbations are scale-independent:
h = hi = constant ("white noise"). Then using Chibisov's (1972) value
for the mass Md damped at decoupling we are able to show (Kurskov and
Ozernoy 1977a) that the condition of isolation (~p/p ~ 1 at z ~ Q-1)
for inhomogeneities of maximum amplitude corresponding to masses M ~ Md
is hi > 2 x 10- 4 Q-5/4 h- 1/ 2 (1+-3Qh 2 ) ,where h = Ho/75 km s-l Mpc- 1 .
Since the r.m.s. temperature fluctuations produced at the epoch of
recombination due to the Doppler-effect on potential velocities
(Zeldovich and Sunyaev 1970) are ~T/T ~ (/3)-1(v rec /c) = 1/3(zrec/zeq)3/4
hi on characteristic mass-scale M ~ ~T(tr2c)' one finds for
0.5 < Q h 2 < 0.7 that ~T/T ? 10- 5 Q-2 h- (1+3Qh2) on angular scales
e ~ 15'-30'. Comparison with corresponding upper limit ~T/T ~ 10- 4
yields Q ~ 0.3[(~T/T)/10-4J-l~ This conclusion cannot be avoided by
introducing secondary reheating since otherwise appreciable angular
variations of ~T/T ~ 10- 3 on scales e ~ 20 0 Q2/3 will appear (Kurskov
and Ozernoy 1977a) unless the initial metric perturbation spectrum had
a cut-off on masses M « 10 22 M0 which seems to be rather artificial.

(b) Whirl perturbations. Angular variations of ~T/T produced by

the Doppler-effect on turbulent velocities (Chibisov and Ozernoy 1969)
are the sum of (i) temperature variations at t = trec weakened by
Thompson scattering due to secondary reheating of the cosmic plasma by
young galaxies and of (ii) temperature variations produced at the mom-
ent of "last scattering" when T ~ 1. The latter gives the main contri-

*This lower limit to Q is apparently even more stringent if

Qh2 » 0.1 since in that case appreciable velocities are present on
scales M > MJ(t rec ) where oscillations were always absent.
SHORT COMMUNIl AnONS 405

bution to the observed value of /',T/T at comparatively large H

(H ~ 0.24-0.31 when ~h2 = 1-0.1) which are of most importance in the
whirl cosmogony.* Comparison with the observed upper limit /',T/T ~ 10- 4
shows that the whirl theory in its present state is invalid unless
~h2 ~ 0.3-0.4.

He conclude that both primaeval whirls or "white noise" adiabatic

density perturbations responsible for galaxy formation may be compatible
with the available upper limits to /',T/T only in a dense enough Universe
(~h2 ~ 0.3-0.4).

REFERENCES

Anile, A.M., Danese, L., De Zotti, G. and Motta, S., 1976. Astrophys.J.,
205, L59.
Chibisov, G.V., 1972. Astron. Zh., 49, 74.
Chibisov, G.V. and Ozernoy, L.M., 1969. Astrophys. Lttters, 3, 189.
Kurskov, A.A. and Ozernoy, L.M., 1977a. Astrophys. Sp. Sci. (submitted).
Kurskov, A.A. and Ozernoy, L.M., 1977b. Astrophys. Sp. Sci. (submitted).
Ze1dovich, Ya. B. and Sunyaev, R.A., 1972. Astrophys. Sp. Sci., l, 3.

*It should be noted that the value of /',T/T from the last scattering
calculated by Anile et al. (1976) and used by Parijskij (this meeting)
was overestimated by about one order of magnitude (Kurskov and Ozernoy
1977b) .

RECENT ADVANCES IN MICROHAVE COSMOLOGY

Paul Boynton

In summary, this brief communication outlined first the recent

measurement of large-scale anisotropy of the relic radiation by the
Berkeley group of Smoot, Gorenstein and Muller (1977). The customary
peculiar velocity interpretation of the dipole term gives v = 390 ± 60
km/sec in the direction of a = 11~0 ± O~5 and 0 = 6 0 ± 10 0 . These
values differ from the preliminary results of Corey and Hi1kinson (1976)
by less than twice their reported errors.

Secondly, the latest observations by G. Lake and R. B. Partridge

(1977a, 1977b) of clusters of galaxies seeking evidence for inverse
Compton "cooling" of the microwave background were reviewed. Signifi-
cant effect was indicated in each of the richness class 4 clusters
observed:
Abell Cluster /',T ± 10-
1689 -1.06 ± 0.46 mK
2125 -3.10 ± 0.34 mK
2218 -2.65 ± 0.23 mK
406 SHORT COMMUNICATIONS

The rema1n1ng 8 clusters (richness class 2 and below) showed no temper-

ature deviations beyond ± 0.4 rnK (± 20). All measurements were made at
9 mm with a resolution of ~ 4'. Some of the values quoted above have
been "updated" since the symposium.

REFERENCES

Corey, B.E., and Wilkinson, D.T., 1976. Bulletin of the AAS, 8, 351.
Lake, G., and Partridge, R.B., 1977a. Nature (to be published).
Lake, G., and Partridge, R.G., 1977b. In preparation.
Smoot, G.F., Gorenstein, M.U., and Muller, R.A., 1977. Ap. J. Letters
(to be published).

DISCUSSION

Silk: There seemed to be a positive deflection in the direction of

Abell 426.

Boynton: Yes. This is the Perseus cluster in which there is known to

be an intense discrete radio source at centimetre wavelengths.

Parijskij: I mentioned on Tuesday that the new result by Partridge on

the Sunyaev-Ze1dovich effect in the Coma cluster does not contradict
our result in the sense that they do not reject the possibility that
the effect exists. The numbers quoted by Partridge and by us do dis-
agree and new observations should be made with special attention to the
problem of discrete sources.
At the same time, it seems to me that in the Coma cluster the
effect is rather small compared with ones expectation from the intensity
and spectrum of the X-ray emission. The mass of gas obtained from
observations of ~T/T seems to be less than that from the intensity of
the X-ray emission. This may be due to clumpiness in the gas. It may
well be that in the future the best way to find this out is to take the
temperature from the X-ray observations and compute the mass from the
Sunyaev-Zeldovich effect.

Boynton: Let me just comment that while your result may be consistent
with Partidge's result at your 30 level, his is not consistent with
yours at his 30 level.
v

THE FORMATION OF STRUCTURE IN THE UNIVERSE

THE THEORY OF THE LARGE SCALE STRUCTURE OF THE UNIVERSE

Ya. B. Zeldovich
Institute of Space Research
USSR Academy of Science
Moscow, USSR

INTRODUCTION

The God-father of psychoanalysis Professor Sigmund Freud taught us

that the behaviour of adults depends on their early childhood experiences.
In the same spirit, the problem of cosmological analysis is to derive the
observed present day situation and structure of the Universe from certain
plausible assumptions about its early behaviour. Perhaps the most impor-
tant single statement about the large scale structure is that there is no
structure at all on the largest scale - 1000 Mpc and more. On this scale
the Universe is rather uniform, structureless and isotropically expanding
- just according to the simplified pictures of Einstein-Friedmann .....•.•
Humason, Hubble •.... Robertson, Walker. On the other hand there is a lot
of structure on the scale of 100 or 50 Mpc and less. There are clusters
and superclusters of galaxies.

Much work has been done on the classification of these bodies into
"richness classes" and attempts have been made to deduce from observa-
tions a "mass function" giving the distribution of matter among clumps of
various sizes and masses. There is a firmly established division between
regions with enhanced, higher-than-average density of stars and radio-
sources and regions with density lower than average. In recent years
correlation functions have been used to characterize the relation between
density enhancements and the linear scales of the distribution of galax-
ies in space.

A systematic effort of measuring the redshifts (optical and 21 cm

radio) of thousands of galaxies has resulted in confirmation of Hubble's
law. Surprisingly it is approximately valid for smaller distances than
those characteristic of the density distribution. The redshift measure-
ments have opened the way for disentangling the three-dimensional struc-
ture of the universe, as opposed to the two-dimensional projection of
astronomical objects on the celestial sphere.

The present symposium has really opened up a new direction in the

409
M. s. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 409-421. All Rights Reserved.
Copyright © 1978 by the fA U
410 Y A. B. ZELDOVICH

search for the geometrical patterns governing the distribution of lumin-

ous matter in space. We heard about ribbons or filaments along which
clusters of galaxies are aligned; the model of a honeycomb was presented
with walls containing most of matter; the presence of large empty spaces
(holes - not black holes of course) was emphasized. Cosmological theory
must be aware of this information and try to use it in order to discrim-
inate between various proposed schemes. Let us briefly characterize
those schemes which seem to us most promising at the present time. There
are two extreme assumptions which can be made about the initial density
perturbations. The first concerns an ideal unperturbed metric connected
by General Relativity with ideal homogeneity of the overall density in
the early radiation dominated Universe. The perturbations consist of an
inhomogeneous distribution of "matter" - of baryon excess - on a back-
ground of homogeneous radiation. Therefore the ratio y/B (photons per
baryon) varies from place to place. But the specific entropy of matter
is proportional to this ratio and therefore those perturbations are
called "entropy perturbations".

The second type of perturbation consists of common motion of photons

and baryons. These perturbations conserve entropy and therefore they are
called "adiabatic".

A departure from the main line of this report is permissible in the

introduction. The actual value of the ratio y/B which is of the order of
lOB or 109.is most important for cosmology. The closed or open geometry
of the Universe as a whole depends on this number.

Is it possible that in due time this number will be calculated by

elementary particle theorists, taking into account the lack of exact
symmetry between particles and antiparticles as indicated by laboratory
experiments (so-called CP - violation, 1964) and also baryon non-conserv-
ation predicted by some theories? In this case it is conceivable, that
the y/B ratio is constant everywhere, just because the physical constants
are everywhere the same. But this argument is not very strong. It is
equally possible that Y/B depends on the interrelation of external physi-
cal constants and the local properties of the space-time metric; in this
case the Y/B ratio must not be a constant.

Let us return to conventional cosmology. At the present moment we

do not see any better policy than to make plausible assumptions about the
size, amplitude and character of the initial perturbations, to develop
logically and mathematically all the consequences of these assumptions
and to compare them with observations.

In this report we shall investigate adiabatic perturbations - the

second type, according to the classification given above. This investi-
gation has been carried out during approximately the last ten years by
our group, which includes Doroshkevich, Sunyaev, Novikov, Shandarin,
Sigov, Kotok and others. We use important theoretical results obtained
by Lifshitz, Bonnor, Silk, Peebles, Yu and others.
THE THEORY OF THE LARGE SCALE STRUCTURE OF THE UNIVERSE 411

We consider several phases in the development of the perturbations:

1) acoustic oscillations of the radiation-dominated plasma and

their attenuation before and during recombination;
2) the growth of small perturbations in the neutral gas;
3) the non-linear growth of perturbations leading to the formation
of compressed gas layers - pancakes;
4) the further fate of pancakes, the interaction of pancakes, their
decay into galaxies and protoclusters of galaxies.

The first two points are investigated using the "merry old" linear
theory of perturbations. In 3) and 4) an approximate nonlinear theory is
widely used and also numerical simulation. The statistical side of the
problem is considered. Radio astronomical predictions are made. The
main result is most encouraging: the adiabatic perturbation spectrum
possesses a definite cut-off wavelength as already pointed out by Silk.
We now see that this critical wavelength is also reflected in the cell
structure of the Universe.

1. THE THEORY OF PERTURBATION GROWTH

A plausible featureless initial spectrum of density fluctuations in

the radiation-dominated plasma is assumed. Due to photon viscosity and
damping during recombination the final Fourier spectrum of growing pert-
urbations in the neutral gas is given by

(~)2
S K
= -& z
k
= ~~ Kn e-I< Rc
The critical length Rc depends on 1) the radiation density during recom-
bination taking account of the specific effects of Ly-a reabsorption
through the 2s + ls+2y metastable hydrogen decay, 2) the Compton cross-
section for scattering of photons by electrons, 3) the matter density or
Y/B ratio. The best calculations give the characteristic length (multi-
plied by(l+zrec) in order to account for the expansion from recombination
to the present epoch)
Rc =8 Mpc for n =1 and Rc 40 Mpc for n = 0.1.
The wavelength AC is determined by AC

The index n and the average value of aR are adjusted to fit the
observed picture. But independent of this adjustment, due to the expon-
ential damping factor e-KRc we are sure that the surviving fluctuations
are very smooth. It is immediately clear that in the adiabatic theory
early formation of stars or globular clusters or even of galaxies is
impossible. First large-scale density enhancements must grow, and only
thereafter is their fragmentation in smaller units possible.

The second qualitative feature is gas motion under the influence of

412 YA. B. ZELDOVICH

gravitation only. The pressure forces, which depend on gradients, are

negligible on the large scales involved. In this case the growth of per-
turbations is especially simple: they grow in amplitude due to gravita-
tional instability and increase in linear dimensions, conserving their
form. The density perturbations and the peculiar velocity (the excess
over the Hubble velocity) are given by

~=
S fvrSL= I
<P(t)
re =
Obviously the density field and velocity field are connected by the con-
tinuity equation
tt (¥-) oC f oC dw- ~ oc tIIw g
and by the equation of motion in which the perturbation of the potential
by the perturbed density is included.

It is important to realize that already in the linear theory the

extra compression in places with positive and growing op/p is anisotropic:
the three components of the divergence of the peculiar velocity are not
equal

There is also shear, 3ux /3x F 0 etc. - but of course no vorticity

3ux/3y - 3uy /3x = 0 because the motion is due to potential (gravitational)
forces. The anisotropy of the deformation due to peculiar velocities is
easily understood by tidal action. The nearby density distribution dis-
torts the motion at the point under consideration.

A natural way to build an approximate theory, exact in the linear

region and also good enough in non-linear situations, is to use the
Lagrangian formulation. The position of every particle in space (i.e.
its Eulerian coordinates) f is given as a function of time t and the
initial position (i.e. Lagrangian coordinate) of the particle t.

The solution with growing perturbations only is written

;; = (II (t;) [f 4- -trlf) ifff')]
-+
The first term a~ describes the Hubble expansion a/a = H, the second
term ab~(t) describes the displacement of every particle from its legiti-
mate unperturbed position. bet) i! a growing function, bet) oc t 2 / 3 • The
perturbation due to gravitation ~(~) is of potential type ~ = grad~~.

Analytical and numerical studies confirm that this is a good approx-

imation - less than 20-30% errors occur in highly nonlinear situations;
the proofs are in our original papers.
THE THEORY OF THE LARGE SCALE STRUCTURE OF THE UNIVERSE 413

++
Given the formula for ~(~,t), it is easy to write down the velocity
of every particle
-+
u. = ~r]
"rt r -
u~ .. = =

and also the densi:; of matter( ~3r )_,

!= !(Lt) oC "i13
Here a3~/a3t is the Jacobian i.e. the determinant of the partial deriva-
tives.

Using ~ = grad~~ and choosing coordinate axes which diagonalize the

deformations, using the notation
~'Z~
= -J. ~
H2..2. = -~ ~= -'if ) "'>rs>~
Ht,
)
d~2.
:3
we obtain ,
f = f (1- .,(,-rl.){I- .fr/J)(I-.(,¥)

With ~ determined by the initial small perturbation field we find

the particles where a has local maxima ~l' am2 The condition
1 - b(ti)Om' = 0 determines the moment when infinite density is obtained
. 1. .
for the 1.-th part1.cle.

From the density formula we see that this infinity is due to the
intersection of trajectories of adjacent particles lying on the ~l coord-
inate axis. At the moment ti the perturbation along the other two axes
~2' ~3 is finite.

The approximate theory predicts the formation of thin dense gas

clouds. They grow due to fresh gas falling onto their flat boundary and
being compressed and heated by shock waves. They also spread sideways
due to new intersections of trajectories of adjacent particles.

The picture outlined above was already known at the time of the
Krakow lAU Symposium No.63, "Confrontation of cosmological theories with
observational data". Qualitatively they are described in the report by
Doroshkevich et al; formulae and detailed analyses were given in our
original papers, and also in the book by Zeldovich and Novikov "Structure
and Evolution of the Universe", published in Russian in 1975 and prepared
for publication in English by Chicago University Press. These are men-
tioned in this report for the sake of completeness and to make it poss-
ible to read this report without using references.

Now we turn to the results obtained after the Krakow Symposium,

partly published in Astronomical Journal (USSR) and partly in preprints
of the Institute of Applied Mathematics. These results are most import-
ant in connection with optical and radio astronomical observations.
414 YA. B. ZELDOVICH

2. LATE PHASES OF THE DEVELOPMENT OF PERTURBATIONS

AND CELL STRUCTURE

Numerical calculations were pushed to a late phase, when more than

half the matter is brought together into the dense phase. During the
lecture in Tallinn a movie was shown made by an electronic computer dis-
play. Here, in the written form, only a small number of selected
pictures can be shown.

There is one movie (corresponding to Fig.l) calculated using the

approximate nonlinear theory for two-dimensional perturbations. The
initial spectrum has a sharp cut-off on the short wave - and also long
wave end; it is flat (on the average) within the excited interval. The
individual Fourier coefficients in this interval are taken at random
according to a normal Gaussian distribution.

Comparing these calculations with others it must be stressed that

the potential, the velocity and density contributions are calculated for
a continuous medium, not for a finite number of discrete point masses.
The calculations are not exact and the initial conditions somewhat arti-
ficial (two dimensions, flat spectrum). But these departures from the
ideal calculation are not of the sort which arise when a finite number of
discrete masses is considered.*

The pictures in the movie contain a finite number of points. But

those points are test particles for visual ising the motion and density
distribution. The potential used in the calculation corresponds to a
continuum or,in other words, to a calculation with an infinite number of
particles with inertial and gravitational mass.

The calculation is continued to the moment after the first inter-

section of trajectories occurs. It is assumed that the particles are
non-interacting and one layer can penetrate through another. The stick-
ing together of particles, their physical, non-gravitational collisions
and the formation of shockwaves are not included. Therefore the pancakes
in this picture are somewhat thicker than would be found in a real gas-
dynamical calculation. Still they are rather thin, distinctly different
from the spherical or irregular clumps predicted in a simplified approach.

As time goes on,the pancakes spread laterally and they intersect. In

Figure I a typical net structure is seen: matter is mostly concentrated
in thin filaments, the inner regions are empty and divide up the network.

It seems plausible that a three-dimensional calculation will lead to

a cell or honeycomb structure with matter concentrated in the walls

*It is IN in two dimensions and 31N in three dimensions which are

important in incorporating shortwave perturbations involuntarily in
N-body calculations just due to the discrete character of the mass distri-
bution.
THE THEORY OF HIE LARGE SCALE STRUCTURE OF THE UNNERSE 415

Approximate theory

Approximate theory Numerical simulation

Figure 1
416 YA. B. ZELDOV1CH

surrounding large disconnected empty spaces. The intersection of walls

could give enhanced density along lines.

It is not yet clear if such types of structure, or at least its

remnants are discernible in the observational data of Joeveer, Einasto,
Tago and SeIdner, Siebers, Groth and Peebles. It is well known that the
human eye has the property of finding lines and other patterns in random
assemblies of points. One example is the Sciapparelli channels on Mars,
but even more striking are the constellations - figures of humans and
beasts found by the ancients in the distribution of stars on the sky.
Therefore one must be very cautious in interpreting the observations.
One must find some mathematical algorithm to distinguish between super-
clusters as quasispherical clumps and the honeycomb structure. It is
possible that correlation functions (two points, three points etc.) are
not the best method for this particular task.

There are obvious difficulties: 1) the walls must fragment into

separate galaxies and clusters of galaxies. The turbulence inside the
pancakes and the gravitational interaction of the fragments must partly
wash out the structure. 2) In investigations of the three-dimensional
structure we use redshift as a measure of distance. Because of peculiar
velocities, this procedure is not exact. These points need further
investigation. But with all these uncertainties, one point must be
stressed: the occurrence of cells in theoretical calculations is not an
artefact due to the use of an approximate theory.

Calculations of another type were carried out and used to make the
second part of the movie. The motion of (128)2 ~ 16000 points in two
dimensions was calculated numerically. The potential for every distribu-
tion of points was calculated using Poisson's equation ~¢ = 4TIGp with
some smoothing and interpolation on the smallest scale. Periodicity on
the largest scale was assumed: points intersecting from inside the wall
of the square reappeared on the opposite wall. The periodicity condition
was also used in the potential calculation.*

Again a flat spectrum of perturbations with cut-offs from both sides

was used in formulating the initial conditions. The results of numerical
simulations are practically indistinguishable from the results of the
approximate theory. The characteristic pattern with thin walls and dis-
connected empty spaces depends on the cut-off of the short waves - this
is our firm conclusion. It is confirmed by the fact that the average
linear dimension of the empty spaces are approximately equal to the cut-
off wavelength, 2TI/~ax'

Therefore this pattern is characteristic of the adiabatic theory

with the exponential cut-off at short wavelengths, which results from the
matter-radiation interaction.

*The force cr ~-1 and potential 00 log T is characteristic of two

dimensional gravitation; to be exact we are working with infinite bars,
not points.
THE THEORY OF HIE LARGE SCALE STRUCTURE OF THE UNIVERSF 417

For entropy perturbations, there is no cut-off except the Jeans'

mass for neutral hydrogen corresponding to masses of the order of globu-
lar clusters. Therefore probably no net or cell-structure will occur in
this case. Corresponding numerical computations with sufficient accura-
cy are still lacking. Perhaps the two dimensional case will be easier
to handle and be still meaningful.

We are optimistic about the prospects for discriminating between

entropy and adiabatic perturbations by means of investigations of the
large-scale structure of the Universe.

A three dimensional calculation was done for adiabatic perturbations

with an exponentially cut-off spectrum, but it is the visualisation of
results which is the bottleneck in this case. This report is written at
the moment when this work is still in progress.

3. STATISTICAL PROPERTIES OF THE BIRTH OF PANCAKES

AND COUNTS OF QSO

The birth of an individual pancake occurs at the moment of inter-

section of trajectories, i.e. mathematically speaking at the moment when
the smallest denominator in the density expression (l-b(t).a) vanishes
so that p + 00. Therefore we must find the local maxima of a = - a2$/d~r.
Due to the statistical character of the problem the answer is also given
in statistical terms. The function P(~) gives the density of local
maxima of given amplitude
(vm- 3 )
cl..1V =
In the case of a cut-off spectrum P is proportional to Rc -3. The depen-
dence on ~ is universal, given the normal Gaussian law of density per-
turbations. But P(Um) is not a simple Gaussian function, because in
calculations of a~we are performing the nonlinear operation of diagon-
alisation of the d $/d~idr,k matrix and we are choosing the maxima.
Doroshkevich has obtained ~
s- -nol. (of. > I/(fi)
p(.A tt ) oc oi. e
Using the connection between the amplitude of the maximum am and
the moment tm we can obtain the birth function F(t) or fez) giving the
number of pancakes born per unit comoving volume per unit time or unit
of z.

The newborn pancakes have small vorticity and low temperature. The
formation of compact objects and brightest galaxies is easiest just at
the birth-place of the pancake and even before the growth of the ends of
the pancake.

Therefore it is plausible to identify the birthrate of pancakes

with the birthrate of brightest known compact objects - quasars. If the
418 YA. B. ZELDOVICH

life time of quasars is short and independent of their absolute age,

then at every epoch the concentration of quasars is proportional to their
birthrate. The high power a 5 before the Gaussian factor leads the
(1+z)7 dependence of the density of quasars as an intermediate asymptote
in the case of a flat Universe. At high z the power law is cut off by
the Guassian exponent. By the choice of a single constant (corresponding
to the amplitude of perturbations) it is possible to obtain a good fit of
the pancake birth rate curve to Longair's results on radio source counts
and Schmidt's data on quasar evolution.

Still, the similarity between the radio source and quasar evolution
and the birthrate of pancakes should not be overestimated. The power
laws involved refer to different regions of z. The birth of cold pan-
cakes occurs from some high z (of the order of 10 or 20) up to z ~ 4 ~ 3.
It is well known that for z < 4 the gas is totally ionized; therefore
even if pancakes are formed, their physical properties are totally diff-
erent as compared with genuine pancakes formed from cold initial gas.
On the other hand, the observed counts of radio sources and quasars refer
to the range 0 < Z < 4; at z > 4 instead of evolution there is a cut-off
or stagnation. This question needs further investigation.

Another statistical test concerns the two-point correlation curve.

The adiabatic pancake theory does not contradict the most interesting
part of Peebles' correlation function 0 ~ ~-1.7 in the region near 0 ~ 1.
We refer to original papers for quantitative confirmation.

The general outlook seems to be that the adiabatic theory does not
contradict the observations.

4. THE CRUCIAL TESTS AND FURTHER PROBLEMS

Still the absence of contradiction is not positive proof. In order

to distinguish between the entropy and adiabatic theories one needs
direct observation. The observation of very early globular clusters and
galaxies at z > 30 to z ~ 100 or 200 would be strong evidence in favour
of the entropy perturbation theory with further clumping of the initial
small mass objects into clusters of galaxies. If hot gas clouds of
primordial composition (H + He) are found, identifiable with pancakes,
this would be a strong argument for the adiabatic theory. Fully ionized
very hot gas could be detected by its X-ray emission and by distortions
of the Planckian background radiation spectrum (cooling in the Rayleigh-
Jeans region). The medium-hot hydrogen gives redshifted 21 cm radiation.

In any case, the controversy with the observed limits on ~T/T of

the relic radiation fluctuations must be solved - but this is needed for
all variants. Entropy perturbations predict ~T/T only 2 or 3 times less
than adiabatic perturbations. The study of those perturbations which
are directly connected with the structure of the Universe is the most
rewarding part of the problem.
THE THEORY OF THE LARGE SCALE STRUCTURE OF THE UNIVERSE 419

Extrapolating from Krakow through Tallinn to the next symposium

somewhere in the early eighties one can be pretty sure that the question
of the formation of galaxies and clusters will be solved in the next few
years.

What remains is the wider question of the overall spectrum of pert-

urbations including the smallest scale damped in the very early radiation
dominated or hadronic era and of the longest perturbations, whose ampli-
tude remains small even now. Is the power law spectrum without any
characteristic length valid? New, indirect observational tests are need-
ed. Still the major theoretical questions remain unsolved: what is the
fundamental theory of the initial perturbations? And what is the
ultimate reason for the homogeneous and isotropic expansion from the
singularity which is the background for the perturbations?

REFERENCES

Doroshkevich, A.G., Sunyaev, R.A. and Zeldovich, Ya.B., 1974. "Confront-

ation of Cosmological Theories 'vi th Observational Data", M. S.
Longair (ed.), Dordercht,Holland/Boston, USA.
Zeldovich, Ya.B. and Novikov, I.D., 1975. "Structure and Evolution of
the Universe", Nauka, Moscow.
Zeldovich, Ya.B., 1970. Astron. & Astrophys., 5, 84.
Doroshkevich, A.G., 1970. Astrophysica, 6,581-
Sunyaev, R.A. and Zeldovich, Ya.B., 1972.- Astron. & Astrophys., 20, 189.
Doroshkevich, A.G., Ryabenkyi, V.S., Shandarin, S.F., Astrophysica;
9, 257.
Doroshkevich, A.G., Shandarin, S.F., 1973. Astrophysica,~, 549.
Doroshkevich, A.G., 1973. Astrophys. Lett., 14, 11.
Doroshkevich, A.G., Shandarin, S.F., 1974. Sov. Astron., ~, 24.
Shandarin, S.F., 1974. Astron. Zh. USSR, 51, 667.
Doroshkevich, A.G., Shandarin, S.F., 1975. Sov. Astron., 19, 4.
Sunyaev, R.A., Zeldovich, Ya.B., 1975. Mon. Not. Roy. Astron. Soc.,
171, 375.
Doroshkevich, A.G., Shandarin, S.F., 1976. Hon. Not. Roy. Astron. Soc.,
175, l5p.
Doroshkevich, A.G., Zeldovich, Ya.B., Sunyaev, R.A., 1976. In "Formation
and Evolution of Galaxies and Stars", S.B. Pikelner (ed.), Nauka,
Moscow.
Doroshkevich, A.G., Shandarin, S.F., 1976. Preprint lAM No.3.
Doroshkevich, A.G., Shandarin, S.F., 1977. Mon. Not. Roy. Astron. Soc.,
179, 95p.
Doroshkevich, A.G., Shandarin, S.F., 1978. Mon. Not. Roy. Soc. Astron.,
182 (in press).
Doroshkevich, A.G., Saar, E.M., Shandarin, S.F.,1977. Preprint lAM
No.72.
Doroshkevich, A.G., Shandarin, S.F., 1977. Preprint lAM No.73.
Doroshkevich, A.G., Shandarin, S.F., 1977. Astron. Zh. USSR, 54, 734.
Doroshkevich, A.G., Shandarin, S.F., 1977. Preprint lAM No.84:-
420 Y A. B. ZELDOVICH

Doroshkevich, A.G., Saar, E.M., Shandarin, S.F., 1978. This volume.

SeIdner, M., Siebers, B., Groth, E.J., Peebles, P.J.E., 1977. Astron. J.,
82, 249.
Joeveer, M., Einasto, J., Tago, E., 1977. Preprint A-I, Struve Astro-
physical Observatory Tartu.

DISCUSSION

Suchkov: There are quite distinct knots in your array of pancakes. Now,
if the pancakes are destined to be superclusters or clusters of galaxies,
what kind of future do you foresee for these knots?

Zeldovich: The numerical calculations need to be pushed further in order

to obtain unambiguous answers. Possibly the filaments along which
clusters of galaxies are aligned (if this effect is statistically veri-
fied) will be identified with intersection lines but it is not yet clear
theoretically.

Chernin: Hhat kind of relaxation could lead to the evolution of a flat

pancake into a cluster like Coma with more or less spherical form?

Zeldovich: Pancake formation is due to compression on one axis, but this

does not exclude less dramatic compression (without intersection) in one
or two other directions. Therefore at least a part of pancakes can
transform into rather dense clumps. Turbulence inside the pancake and
also its curvature tend to make the clump thick. The last effect tending
to make the cluster spherical is gravitational interaction.
On the other hand, there must also be pancakes which are expanding
in the two directions tangential to the pancake surface and in this case
one should observe Hubble's law in a region with strongly enhanced dens-
ity. Of course, the Hubble constant for this region is different from
the genuine long-range H; the local H is subject to quadrupole perturba-
tions. One should ask Prof. de Vaucouleurs and Profs Sandage and Tammann
if perhaps we are living in such a region.

Binney: One cannot but be impressed that Dr Zeldovich's beautiful film

gives a better representation of the sky as published recently by Dr
Peebles and collaborators than does that shown earlier by Dr Aarseth
(Peebles et al. 1977). Further strong evidence in support of the picture,
based on a spectrum biassed towards large masses, are the facts that both
most rich clusters of galaxies and elliptical galaxies are as often as
not nearly as aspherical as a slowly rotating body can be (Klingworth
1977, Rood and Chincarini 1974, Macgillivray 1976, Schipper and King
1977) .
I should like to ask Dr Zeldovich, however, whether he believes
large-scale shock formation is a necessary part of this picture. I ask
this because I have difficulty in believing that the cold cosmic gas will
fail to fragment soon after it starts to contract in one dimension. This
THE THEORY OF TilE LARGE SCALE STRUCTURE OF THE UNIVERSE 421

will destroy the pressure-balance required across the centre of the pan-
cake. My belief is that one may retain the cellular structure and the
aspherical cluster formation even without large-scale shock formation.
Certainly one cannot overemphasize the importance of anisotropic collapse
on a large scale.

Zeldovich: Dr Binney is making a statement rather than a question. I

should point out that the film was made by Doroshkevich, Shandarin, Sigov
and Kotok; I would also add Einasto and Joeveer to the list of people
observing large scale structure.
As to the origin of the structure: it is the cut-off of short wave
perturbations which is most important. The cell structure remains
(perhaps somewhat weaker, with thicker walls) in the collisionless case
with trajectories continuing without break after intersection, i.e. in
the absence of the shock. Concerning fragmentation, when the perturba-
tions are small (linear regime) the exponent of the gravitational insta-
bility has no maximum; it is an increasing function of wavelength. The
cut-off short wave perturbations do not outgrow those of long wavelength.
The compression time before pancake formation is so short that it does
not compensate the handicap due to short wave damping. We feel that the
overwhelming part of fragmentation occurs after shock wave compression -
if there are no primordial short wave entropy perturbations of course.
THE INTERNAL STRUCTURE OF PROTOCLUSTERS AND THE FORMATION OF GALAXIES

A. G. Doroshkevich, E. M. Saar and S. F. Shandarin

Institute of Applied Mathematics
USSR Academy of Sciences
Moscow, USSR

We give a short review of the general picture and main features of

the formation of galaxies and clusters of galaxies on the basis of the
adiabatic theory. Detailed discussions of these questions are published
in references 1-6. Some related problems of the formation and develop-
ment of large scale structure in the Universe according to the same
theory are considered in Zeldovich's report.

The general picture of the theory under consideration is as

follows:
(i) For redshifts z > 10 5 , there are small density and velocity
perturbations having a monotonic (e.g. power-law) spectrum.
(ii) For redshifts 10 5 > Z > zrec ~ 10 3 , large scale perturbations
grow whereas small scale perturbations are dissipated; a characteristic
length arises and this determines the future processes of galaxy forma-
tion.
(iii) For redshifts 10 3 > Z ~ 15 following recombination, the per-
turbations grow, preserving their form.
(iv) For redshifts 15 > Z > 10, the non-linear stages begin; the
first "pancakes" are formed although the mean square density perturba-
tions «op/p)2)~ are less than 1. At the same time protoclusters of
rich clusters like Coma, Perseus, and Virgo begin to form.
(v) For redshifts 10 > Z > 1 to 2, the bulk of the matter (~ 70%)
turns into "pancakes". The central part of a "pancake" cools and
fragments into small clouds while the "pancake" grows as a whole. At the
same time these small clouds cluster forming larger complexes, which
then develop into galaxies.

In this report we concentrate on the late stages of evolution of

"pancakes". We recall briefly the general features of the formation and
evolution of pancakes. Zeldovich has shown that the growth of perturba-
tions results in the origin of one-dimensional gas structures which are
thin and very dense; we call them pancakes. Gas layers fall onto these
423

M. S. LongOlr and J. Einasto reds.), The Large Scale Structure of the Universe, 423-425. All Rights Reserved.
Copynght © 1978 by the IAU.
424 A. G. DOROSHKEVICH ET AL.

pancakes and lose their velocity in shock waves. Kinetic energy is

transformed into heat and the gas acquires a high temperature. The
dense central part of the pancake cools rapidly due to radiation, but
the out-lying layers remain hot up to the present epoch.

In spite of the fact that the motion outside pancakes was without
vorticity, inside pancakes the motion acquires a component of vorticity.
We believe that thermal instabilities in the cooling medium moving with
vorticity results in the formation of cool gas glouds moving in the hot
gas. In this case the local density is much greater than the mean
density. "Mean temperature" means the temperature of the hot gas or the
temperature associated with the motion of cool clouds. The masses and
sizes of the clouds are determined by their thermal conductivity and by
the turbulence in the hot gas.

It is very important that the pressure in the shock front is great-

er than the self-gravitation of the "pancake". The pressure decreases
with time and determines the evolution of cool matter.

It should also be remembered that the thickness of the pancake, its

temperature, density and so on depend on radial coordinates and change
with time. We would like to point out that the final results which may
be compared with observation need an analysis of the problem as a whole.

Galaxies are formed by the clustering of cool clouds which formed

due to thermal instability and turbulence. A galaxy was never at any
stage a homogeneous gas cloud; the first stars can form in the cool
clouds before the origin of the galaxy. But it is necessary to empha-
size that both turbulence and cool clouds are a secondary phenomenon;
they arise inside the pancakes and their parameters are closely related
to the parameters of the pancakes.

We would like to give a short summary of the results we have

obtained within the framework of the "pancake theory" of galaxy forma-
tion which can be compared with observation.

(1) The masses of galaxies formed in pancakes depend on their

radial coordinates. We can derive an approximate mass function of
galaxies

M -\) . \)=1-l-1.5,
0:
g ,

(2) We can derive an approximate mass-angular momentum relation

~g ~ 10 30 (Mg/l0 11 M@)2/3,
where ~g is angular momentum per unit mass.

(3) We can obtain the mass function of clusters of galaxies

INTERNAL STRUCTURE OF PROTOCLUSTERS AND THE FORMATION OF GALAXIES 425

dN cl
dMCl '"

M > Me, and }flo = l015Me,.

(4) In this picture we can also explain the existence of hot gas
in rich clusters as well as HI clouds in superclusters of galaxies, for
example in the Local Supercluster.

REFERENCES

1. Doroshkevich, A.G., Sunyaev, R.A. and Zeldovich, Ya.B., 1974.

"Confrontation of cosmological theories with observational
data", ed. M. S. Longair, Reidel, Holland.
2. Doroshkevich, A.G. and Shandarin, S.F., 1977. Astron. Zh., 54, 734.
3. Doroshkevich, A.G. and Shandarin, S.F., 1977. Astron. Zh. (in
press).
4. Doroshkevich, A.G., Saar, E.M. and Shandarin, S.F., 1977. Mon. Not.
R. astr. Soc. (in press).
5. Doroshkevich, A.G. and Shandarin, S.F., 1977. Preprint No.8 of
Institute of Applied Mathematics.
6. Doroshkevich, A.G. and Klipin, A.A., 1977. Preprint No.ll of
Institute of Applied Mathematics.

DISCUSSION

Efstathiou: I would like to make a comment concerning the pancake

theory. This morning Dr Jones presented the results of our work on the
tidal torque theory. The value of the parameter A obtained is a scale-
independent quantity and so should apply to pancakes as well as proto-
galaxies. If these pancakes then dissipate a lot of energy during
their collapse, one might be in danger of producing a rotation-dominated
pancake contrary to observations.

Silk: How did you obtain your mass function of galaxies and of galaxy
clusters?

Doroshkevich: The mass function is a result of numerical calculations.

THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNIVERSE

L. M. Ozernoy
Lebedev Physical Institute
Moscow, USSR

1. INTRODUCTION

There are at least two reasons for examining different theories of

galaxy formation:
(i) We do not know the initial conditions in the early Universe;
(ii) We do not know which forces were most important for the origin
and evolution of initial perturbations.
The first reason forces us to deal with perturbations of different
types, i.e. adiabatic, turbulent and entropy perturbations. The second
makes it necessary to examine the influence of non-gravitational forces
- for instance, local vortices. The whirl theory of the formation of
structure in the Universe is an example of a consistent theory which
introduces - by means of initial conditions - non-potential vortex
perturbations.

One point should be emphasized even in these introductory remarks.

One of the first questions which often arises in whirl cosmogony is:
why should one consider the rotation of galaxies to be primordial if it
may be obtained as a result of the evolution of entropy or adiabatic
perturbations? These arguments are however far from being completely
convincing, at least in a quantitative respect. In fact, the nature of
the perturbations has not yet been established. At the same time, the
assumption that the rotation of galaxies is of primordial origin does
not imply the introduction of any additional parameters. Moreover, the
initial whirl velocity is the only essential parameter of the whirl
theory. In this theory, as distinct from both adiabatic and entropy
perturbation theories, there is no free choice of the velocity spectrum
since it will develop into the standard Kolmogorov form in the course of
the evolution of cosmological turbulence. Hence what, at first glance,
seems to be a weakness of the theory is, in fact, one of its attractive
features.

The whirl theory, l.n its present state, explains, of course, much
427

M. S. Longair and J. Einasto (eds.). The Large Scale Structure of the Universe. 427-438. All Rights Reserved.
Copyright © f 9 78 by the fA U.
428 L. M. OZERNOY

more than the rotation of galaxies alone. This theory enabled us to

find the spectrum of cosmological turbulence and the picture of its
decay, to explain the origin and magnitude of rotational velocities, as
well as the main dynamical parameters of both galaxies and systems of
galaxies, gave estimates for the redshifts at which the birth of
galaxies and systems of galaxies occurred - and all this by means of
substantially only one single parameter which characterizes the ampli-
tude of primaeval whirls. Most of these results have been obtained
during the last few years jointly with A. A. Kurskov (see Kurskov and
Ozernoy 1974a,b,c, 1975), and I would like to review them briefly (for
a more detailed presentation, see Ozernoy 1976). Previous reviews
(Ozernoy 1974, Jones 1976) deal with the theory in an earlier and less
well developed stage.

2. THE EVOLUTION OF WHIRLS PRIOR TO RECOMBINATION

According to the main assumption of the theory, large-scale whirl

motions with subsonic velocity Vo existed on all scales during the
radiation-dominated era. This means that the dimensionless amplitude of
the whirls was W = vo/c~ 1/13. In the following we shall use a time-
independent distance scale R which is related to physical size r by
R = r(l+z) (z is the redshift). We restrict ourselves to those stages
of evolution when the maximum size of a whirl does not exceed the size
of the horizon. Then we have a simple picture for the dynamical evolu-
tion which is determined by three processes: cosmological expansion,
hydrodynamical (inertial) readjustment of motions and, finally, viscous
dissipation. Characteristic times are different for different scales,
and so they evole in a different manner. It is convenient to introduce
a characteristic scale Rh = vtz for which the hydrodynamical time
Th = r/v is equal to that of cosmological expansion, Texp = r/r.

On large scales (R » Rh) cosmological expansion is dominant, and

the velocity of whirl motions, despite the expansion, preserves its
initial value because of conservation of angular momentum, up to a red-
shift, Zeq = 1.B x 10~ nh 2*,corresponding to the epoch of equality of
matter and radiation densities, after which v ~ (l+z).

On intermediate scales (Rd « R « Rh, where Rd is the scale of

viscous dissipation) the primaeval whirl spectrum undergoes readjustment
due to energy flow from large scales into smaller ones. A universal
(Kolmogorov) spectrum is established on these scales as in normal labor-
atory turbulence. The mass corresponding to the upper limit of the
Kolmogorov spectrum, attains its maximum at a redshift z = Zeq and equals
5 x l015W3 n- 2Me • For reasonable values of W ~ 0.2-0.5 and n ~ 0.5 this
mass corresponds to rich clusters and even superclusters. A detailed
investigation, both analytical and numerical, of the pre-recombination
evolution of whirl motions showed (Kurskov and Ozernoy 1974a) that the
THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNNERSE 429

resulting turbulent spectrum becomes of universal form and its amplitude

does not depend in any important way on the details of the initial
motions.

The inertial readjustment of an initial whirl spectrum into a

Kolmogorov spectrum proceeds only down to some minimum scale. On
smaller scales the motions dissipate due to radiative viscosity, and the
most important damping occurs both immediately and in the process of
decoupling. The resulting scale Rd of damped motions corresponds to a
mass Md ~ 4.7 x lOll(nh2)-7/2M@.

The relation between Rd and Rh at the instant of recombination is

of great importance with respect to the post-recombination evolution of
cosmological turbulence. At z = zrec the sound velocity drops, and all
the motions become supersonic. The problem is whether at the instant of
recombination the scale of hydrodynamical "freezing" of motions Ii is
larger or smaller than Rd' In an earlier rather crude sketch of this
theory (Ozernoy and Chibisov 1970) it was assumed that ~ > Rd' On that
basis the adversaries of the whirl theory came to the conclusion that
immediately after recombination large density jumps produced by shock
waves would appear on scales RQ < R < R.
However, this conclusion has
not been confirmed by the deta1led calculations made by Kurskov and
Ozernoy (1974c) which, though model-dependent, were performed using
rather reasonable assumptions about the influence of dissipation on
ve10~ity gradients. The calculations showed that not only on scales
R > R but on all scales up to R ~ Rd the motions by the epoch of recomb-
ination become supersonic and at the same time "frozen-in". This means
that the generation of shock waves by cosmological turbulence is impos-
sible. Hence, the corresponding evolution of cosmological turbulence is
"silent" rather than "rumbling".

It should be emphasized that the conclusion concerning the silent

character of the evolution follows only from taking into account consis-
tently the damping of turbulence, and not by the choice of a lower value
of W which remains the same. The evolution of cosmological turbulence
is "silent" up to the maximum value of W = 1/13.

3. THE SPECTRUM OF DENSITY INHOMOGENEITIES PRODUCED BY COSMOLOGICAL

STUDIES

Knowledge of the evolution of cosmological turbulence allows one to

calculate the spectrum of small density inhomogeneities produced by
turbulence at the epoch of decoupling of matter and radiation after
which the inhomogeneities start to grow due to gravitational instability.
The resulting spectrum of these density perturbations is shown in
Figure 1. The spectrum is a sum of undamped inhomogeneities generated
before recombination and those produced by undamped motions after recom-
bination (marked A). Concerning the latter, they are generated on large
scales where the hydrodynamical time is much larger than the expansion
time. As a result, these inhomogeneities are rather small and are given
430 L. M. OZERNOY

Figure 1
by a perturbation theory. Taking account of velocity damping, it
follows that their amplitude attains its maximum value 0max ~ 0.3 W4 / 3
(~h2)-1/3 on the scale R ~ Rd and diminishes sharply at R < Rd. On
scales R > Rd "frozen-in" motions generate inertially inhomogeneities
with the amplitude 0 ~ R-4/3 ~ M-4/9. Their amplitudes exceed those of
local density inhomogeneities produced by turbulence before recombina-
tion (they are marked C in Figure 1). It is interesting that on scales
R < Rd post-recombination inhomogeneities (marked as B) are much smaller
than those produced before recombination in the course of velocity
damping. In the two-component cosmological substratum (matter +
radiation), velocity damping due to radiative viscosity and thermal con-
ductivity is accompanied by the generation of composition inhomogenei-
ties, i.e. specific entropy perturbations. After decoupling, the
radiation density becomes homogeneous and perturbations remain only in
the matter. The amplitude of these entropy inhomogeneities attains a
maximum value 0 ~ 10- 2 W4/3(~h2)-1/6 on a scale which contains the mass
M ~ 3 x 1010(~h2)-1114 Me.

4. FORMATION OF GALAXIES AND SYSTEMS OF GALAXIES

Having derived the spectrum of density perturbations produced by

cosmological turbulence, it is easy to calculate their growth during
the linear stage of gravitational instability. One can obtain the gross
parameters of galaxies and systems of galaxies at the stage of their
isolation from the background before the transition into the non-linear
regime. If we do not require high precision, these estimates may be
used for comparison with the observational data.

First, let us consider the parameters of objects which correspond

to the maximum of the spectrum of density perturbations, i.e. to objects
with mass M ~ Md. As can be seen from Figure 1, the spectrum near the
maximum is rather flat. This means that at the instant of isolation of
the mass Md hierarchical fragmentation will take place. This process
will continue up to the mass at which the spectrum diminishes by a
factor e, i.e. of a mass approximately 10 2 times smaller than Md.
THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNIVERSE 431

A flat maximum near M ~ ~ indicates that in addition to fragmenta-

tion of this mass into smaller pieces the opposite process of clustering
will also proceed. In other words, from objects of mass M ~ Md both
massive galaxies and agglomerates of galaxies with a mass M ~ Md and
M p Md may be formed. It is reasonable to assume that in this statisti-
cal process the maximum mass which a galaxy may attain coincides with
~.

The principal parameters of the most massive galaxies (or of

complexes of massive galaxies), which are first formed with spectrum of
the form shown in Figure 1, are as follows:

Redshift of isolation

Mass M 5 x

Mean density at isolation

Virial radius*

Vi rial density*

Specific angular momentum

*Dissipationless collapse is assumed.

The last expression requires some comment. Since inhomogeneities

with mass M~ Md , where internal motions are damped, participate in the
rotation on larger scales, the galaxies formed will possess angular
momentum. In these regions of turbulence where the specific angular
momentum is less than the mean value elliptical galaxies will form and
where it is larger than the mean spiral galaxies will originate. The
dependence of specific angular momentum on the mass for spiral galaxies
is the relation shown abqve and this is similar to the observed relation
which is described by M2/3 (Ozernoy 1967, Nordsieck 1973) or M3/4
(Freeman 1970). It is tempting to note that a spatial velocity corre-
lation in cosmological turbulence is able to explain a correlation of
the morphological type of a galaxy with the type of cluster to which it
belongs: ellipticals occur mostly in rich clusters, and spirals in
irregular clusters.
432 L. M. OZERNOY

As can be seen from these expressions, the above results resemble

typical gross galactic parameters if we adopt W ~ 0.2 - 0.4, Q ~ 0.5 - 1.
The correspondence between calculated and observed parameters is
surprisingly good if one remembers the simplicity of the model.

Now, let us turn to the parameters expected for systems of galaxies

(Ozernoy 1971, Kurskov and Ozernoy 1975). As mentioned above, after
recombination when the pressure drops, turbulent motions with a
Kolmogorov spectrum generate small inhomogeneities with amplitude which
is smaller the larger the scale. Their subsequent growth and isolation
through gravitational clustering lead to the formation of groups and
clusters of galaxies. The isolation of a system occurs the later the
larger its mass:

Assuming as above that the collapse of a system of galaxies into the

vi rial state is dissipationless, one obtains the vi rial density of a
system
l:l,l~ -2 -12/'f

Pvir ~ 4 x 10- 26 (O~2) (~~~) C~pJ g cm- 3

where r is its radius. This relation is shown in Figure 2 together with

the corresponding observational data summarized by Karachentsev (1967).

-25
-26

-27

-28
SUPERCLUSTERS
-29
19 R, em
23 24 25 26

Figure 2

De Vaucouleurs (1971) gives an analogous relation for the virial mean

density of systems of galaxies as a function of their size.

<p> " 10-26. It ( __ r_)-1·1

1 Mpc
g cm-3.

From extensive statistical analyses of catalogues of galaxies, Peebles

(1974) obtained a similar value (-1.77) for the exponent of the spect-
rum. All these estimates are in rather good accord with the theoretical
value 12/7 = 1.71 produced by the Kolmogorov velocity spectrum of
THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNIVERSE 433

cosmological turbulence. According to the theory the maximum scale of

motions is of the order of 100 Mpc, and on larger scales the density
contrast has a cut-off. Evidently this may be compared with the dimen-
sions of superclusters as the largest scale of inhomogeneities in the
Universe.

The theory predicts that the velocities of galaxies in groups and

clusters are a mixture of primordial rotation and motions produced by
subsequent gravitational instability. The ratio of chaotic and rota-
tional velocities is expected to be vchaot/vrot ~ (M/lOI2~)4/9 for
gravitationally bound systems. For rich clusters of galaxies
(M ~ 10 15 MQ) one obtains v rot ~ 10 2 km/s. Applying the same formula at
the limit of its applicability to superclusters, we have Vrot ~ 10- 2
vchaot. Thus, for the Local Super cluster one expects Vrot -$ 10-10 2 km/s,
in qualitative agreement with the recent revision by de Vaucouleurs
(1976) of previous cruder estimates.

5. COSMOLOGICAL ASPECTS OF WHIRL COSMOGONY

First of all, let us consider what observational data limit the

amplitude W of the initial whirl velocity on the largest scales of
motion (this scale is naturally identified with the horizon scale at
t = t eq ). The most interesting of these limits are shown in Figure 3.

"'-"'--r-T"""T -1

'~ D

~ 1

..
10 '--::_~~~~~-'-'-,----~--'~~~~,---J
10·& aka 10- 4

Figure 3

A lower limit to W (marked A) follows from the natural requirement that

the scale containing most of the energy must not dissipate by the end
of the recombination era. An upper limit to W (marked B) is provided
by the fact that dissipation of cosmological turbulence must produce
very small distortions (if any) of the Rayleigh-Jeans part of the
spectrum of the relic radiation.
434 L. M. OZERNOY

Recently Kurskov and Ozernoy (1977), using the available observa-

tional limits to the small-scale anisotropy of the relict radiation,
improved considerably the constraints on W. A more stringent lower
limit to W (marked D) follows from the temperature variations produced
at recombination and then weakened by Thompson scattering as a result
of secondary ionization of the metagalctic gas during the initial bright
phase of galactic evolution (Ozernoy and Chernomordik 1976). A more
stringent upper limit to W (marked C) follows from the temperature
variations produced in the last scattering of the microwave background
when the optical depth was T ~ 1 at redshift Zl ~ 10. From Figure 1 it
follows that values W ~ 0.3 - 0.4, which provide a reasonable explana-
tion of the parameters of both galaxies and systems of galaxies, do not
contradict these more severe observational constraints. However, it
should be stressed that a further improvement of the upper limit to
angular variations of the relict radiation may be very dangerous for the
present theory.

The whirl theory raises some problems not only for observational
but also for theoretical cosmology. At early epochs, generally speaking,
whirls make the cosmological expansion highly anisotropic, and we have
some kind of "space-time curvature turbulence" (Tomita 1972). The
transition from a Friedmann universe to an anisotropic one takes place
at t < tF ~ W4 teq ~ 2 x 10 7 (W/D.2)4(~h2)-2sec (Ozernoy 1971), and this
means that helium and other light element production may be different
from the standard picture. There is some controversy concerning esti-
mates of the helium abundance produced in the anisotropic stage of cos-
mological turbulence (see, e.g., Tomita 1973, Barrow 1977), and further
analyses are necessary. It is interesting to mention in this connection
that the hypothesis of Chibisov (1976), who proposed that near the
singularity the relativistic motions of both plasma and radiation were
compensated for by oppositely directed vortex motions of free particles
(for instance, of gravitons). Since the resulting vortex is zero, the
metric is Friedmannian and such "zero vortices" do not influence element
production at all; at the same time the whole picture of galaxy forma-
tion remains unchanged. "Zero vortices" do not lead to the problem of
particle creation which may be serious for the usual whirl concept
(Lukash et al. 1975). However, very special initial conditions are
needed to have ab initio a total compensation of whirl motions for the
normal and free-Particles.

6. DISCUSSION. COMPARISON OF THE WHIRL AND POTENTIAL CONCEPTS

OF GALAXY FORMATION

Although a number of problems in the whirl theory remain unsolved,

a test of the credibility of the theory in its present state consists in
its ability to explain rather satisfactorily the main characteristics of
galaxies and systems of galaxies. However, many of these features may
be explained just as well according to the entropy or adiabatic theories.
Evidently, a choice between different concepts should be made from some
THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNIVERSE 435

fundamental corollaries of these theories such as the mechanism for the

formation of structure. The adiabatic concept is associated very
closely with the fragmentation hypothesis, the entropy concept is based
on the clustering hypothesis, whereas the whirl theory includes both.
The situation is presented by the following scheme (cf Rees 1977):

FRAGMENTATION HYPOTHESIS
(associated with adiabatic perturbations)

collapse,
cooling and gas clouds hierarchical stars
protoc1usters ~ of galactic ;0-
inside
fragmenta- fragmentation
mass galaxies
tion

B
CLUSTERING HYPOTHESIS
(associated with entropy perturbations)

gravitational gravitational systems of

stars ~------~.~--~~~ galaxies ----------~----~~~
c1uster1ng clustering galaxies

FRAGMENTATION-CLUSTERING HYPOTHESIS
(as given by the whirl theory)
,
collapse,
gas clouds galaxies
cooling and gravitational systems of
of galactic consistinr )
galaxies
mass fragmenta- of stars clustering
tion

Of course, the sequence of the formation of galaxies and clusters

of galaxies depends also on the initial perturbation spectrum. In the
adiabatic theory the natural choice of an initial white noise spectrum
appears to be in contradiction with the observational data (Peebles
1974). The entropy theory is much better in this respect: it may be
brought into agreement with the observed correlation function for a
large interval of masses. One should recall that the whirl theory
explains these data as well but with the principal difference that the
spectrum of inhomogeneities is not designed initially to produce the
observed result: it is calculated consistently and is independent
(within broad limits) of the initial velocity spectrum. One can hope
that further analysis, both observational and theoretical, of the frag-
mentation and clustering hypotheses may provide reliable tests for the
choice between these different concepts.
436 L.M.OZERNOY

7. CONCLUSIONS

During the last few years many theoretical problems which seemed to
present difficulties for the whirl concept have been resolved. I mean,
first of all, the evolution of cosmological turbulence immediately after
recombination, which according to detailed calculations and contrary to
simplified estimates turned out to be shockless and, at the same time,
to be able to explain the main parameters of galaxies, not only their
rotation. There are now two difficult problems for the theory: (i)
creation of particles near the singularity and their influence on whirl
motions, and (ii) the production of light elements in the whirl model
with appropriate parameters W andn. Until much theoretical work is
completed, these problems will remain unresolved as well as the main
problem of the origin of the whirls themselves. By the way, the origin
of primaeval perturbations is a problem common to all theories of galaxy
formation. Some observational aspects of the theory seem to be more
important at the moment. New measurements of the small-scale isotropy
of the blackbody radiation leave only a narrow margin for the main para-
meter W of the whirl theory. At the same time it is worth noting that
for values of W which do not contradict the observational constraints,
the theory explains quite reasonably the main parameters of galaxies and
systems of galaxies. Further ob~ervations and more detailed models
will establish the plausibility of the whirl theory.

REFERENCES

Barrow, J.D., 1977. Monthly Notices Roy. Astron. Soc., 178, 625.
Chibisov, G.V., 1976. Pis'ma Astron. Zh., 2, 131. ---
de Vaucouleurs, G., 1971. Publ. Astr. Soc.-Pacific, 83, 113.
de Vaucouleurs, G., 1976. Report at the IAU Meeting,lGrenoble.
Freeman, K.C., 1970. Astrophys. J., 160,811.
Jones, B.T.J., 1976. Rev. Mod. Phys.:-48, 107.
Karachentsev, 1.0., 1967. Soob. Byurakan Obs.,39, 76.
Kurskov, A.A. and Ozernoy, L.M., 1974a. Astron:-Zh.,~, 270 (Engl.
transl. Soviet Astron.-A.J., 18, 157).
Kurskov, A.A. and Ozernoy, L.M., 1974b. Astron. Zh., ~, 508 (Engl.
trans1. Soviet Astron. - A.J., 18, 300).
Kurskov, A.A. and Ozernoy, L.M., 1974c. Astron. Zh.,~, 1177 (Engl.
transl. Soviet Astron. - A.J., 18, 1975, 700).
Kurskov, A.A. and Ozernoy, L.M., 197~ Astron. Zh., 52, 937 (Engl.
transl. Soviet Astron. - A.J., 19, 1976, 569). --
Kurskov, A.A. and Ozernoy, L.M., 197~ Astrophys. Space Sci. (in press).
Lukash, V.N., Novikov, 1.0., Starobinskii, A.A., 1975. Zh. Exp. Teor.
Fiz., 69, 1484 (Soviet Phys. - JETP, 42, 1976, 757).
Nordsieck, K.H., 1973. Astrophys. J., 184:-735.
Ozernoy, L.M., 1967. Astron. Tsirk. of Acad. Sci. of USSR, No.407, 1.
Ozernoy, L.M., 1971. Astron. Zh., 48, 1160.
Ozernoy, L.M., 1974. In "Confrontation of Cosmological Theories with
Observational Data" (Ed. by M. S. Longair). Dordrecht-Holland,
p.227.
THE WHIRL THEORY OF THE ORIGIN OF STRUCTURE IN THE UNIVERSE 437

Ozernoy, L.M., 1967. In "Origin and Evolution of Galaxies and Stars"

(Ed. by S.B.Pikel'ner) "Nauka", p.l05 (in Russian).
Ozernoy, L.M. and Chernomordik, V.V., 1976. Astron. Zh., 53, 459.
Ozernoy, L.M. and Chibisov, G.V., 1970. Astron. Zh., 47,769 (Engl.
transl. Soviet Astron.- A.J., 14, 1971, 615). --
Peebles, P.J.E., 1974. Astrophys.~, 189, LSI.
Rees, M., 1977. Preprint. ---
Tomita, K., 1972. Progr. Theor. Phys., 48, 1503.
Tomita, K., 1973. Progr. Theor. Phys., 50, 1285.

DISCUSSION

Audouze: Could you comment in more detail on the nucleosynthetic impli-

cations of the whirl theory of the origin of the Universe? For instance,
is helium produced in too large or in too small quantities compared to
the canonical big-bang calculations?

Ozernoy: The problem is to produce a helium abundance which does not

exceed the observed one. In order to calculate the expected abundance
correctly it is necessary to construct reasonable anisotropic cosmolog-
ical modeis containing whirls at stages when the size of whirls exceeds
that of the horizon.

Shandarin: I should like to stress that according to the adiabatic

theory stars do not form in galaxies. Stars can form in gas glouds
before galaxies have been formed.

Ozernoy: In most versions of the adiabatic theory (including the earlier

version of the "pancake" concept) galaxies are formed in the course of
fragmentation of protoclusters. It is true that another choice of free
parameters makes it possible to obtain the picture you mentioned.

Silk: How do you account for the large number of dwarf galaxies in your
model?

Ozernoy: I do not see any major difficulties with dwarf galaxies, which
may be formed as a result of the fragmentation process on mass-scales
M < Md' where appreciable "entropy" inhomogeneities are present.

de Vaucouleurs: The low velocities you quoted for superclusters were

for spherical systems. Do you have any estimate for the case of strong-
ly flattened ("pancake") systems?

Ozernoy: If you consider a one-dimensional rather than a spherical

character of the collapse for large scale systems you will obtain a
smaller mean density for a system and, consequently, a smaller velocity
for its rotation.

Jones: Before commenting I should stress that calculations in this kind

of theory are extremely difficult and perhaps it is not surprising that
438 L.M.OZERNOY

various theoretical calculations disagree. However, two things are

clear:
(a) the calculations of Barrow are better than those of Tomita and
the nucleosynthesis problem must be taken seriously;
(b) turbulence, by its very nature, dissipates and one has to
fight hard to overcome this natural decay.
We can only continue our attempts to evaluate this theory - it is
very important to resolve the existing conflicts.
ORIGIN OF PROTOGALACTIC EDDIES

A. D. Chernin
A. F. Ioffe Physico-Technical Institute
Leningrad, USSR

A highly excited state of the cosmic medium might develop in the

isotropic Big Bang starting with weak irrotational perturbations that
were likely to be of entropy type. Supersonic matter currents are an
indispensable feature of this state. Initial irrotational gas motions
become vortical when they become supersonic. One can expect that this
may provide a clue in galactic cosmogony (1970, Nature, 226, 440).

The most effective mechanisms of vorticity generation are related

to violent interactions of currents of strongly supersonic matter
involving intersections of shocks and so-called non-evolutionary proces-
ses. It is of particular interest that both of them are generally
accompanied by tangential discontinuities.

When two gaseous masses, "clouds", come into contact with a super-
sonic velocity, a component of their initial relative velocity along the
line of the centres of the masses leads to two shocks propagating in
opposite directions from the surface of contact of the masses. A trans-
verse component of the initial velocity makes the material of the clouds
move with different tangential velocities on the two sides of the sur-
face. Such a tangential discontinuity is absolutely unstable, and this
"vortex sheet" transforms eventually into a turbulent layer by the two
shock fronts.

It seems reasonable to suppose that gaseous proto-clusters which

originated as turbulent layers of large mass evolved into clusters like
Virgo with many rapidly rotating galaxies. The encouraging resemblance
between rough estimates of the characteristics of the layers and the
major parameters of rotating galaxies and clusters of galaxies appears
naturally as a definite consequence of gas-dynamical phenomena in the
proto-galactic medium.
439

M S. Longalr and J Einasto (eds), The Large Scale Structure of tlze Universe, 439-440. All Rights Reserved.
Copyright © 1978 by the fA U.
440 A.D.CHERNIN

DISCUSSION

Peebles: D. Hawley and I found that the observed orientation of the

long axes of galaxies are remarkably close to random. Is this difficult
to reconcile with the idea that the momenta of galaxies were produced in
shocks continuous over large scales?

Chernin: I do not believe the statistics are yet large enough and close
pairs are excluded from these samples.

Jones: Are you sure that the tidal torques theory predicts antipara11e1
spins? I can certainly imagine contrary situations.

Chernin: I agree these situations can occur but only rarely.

EVIDENCE FOR THE GRAVITATIONAL INSTABILITY PICTURE IN A DENSE UNIVERSE

Marc Davis
Center for Astrophysics
Cambridge, Mass. U.S.A.

The statistical nature of the galaxy distribution is in a sense

remarkably simple. The two-point correlation function ~(r), which meas-
ures the count of galaxies at separation f in excess of that expected
for a random distribution, varies as ~crr- .8 for ~>1(r~15 Mpc). At
larger separations ~ apparently decreases more rapidly. The power law
behavior is observed in different galaxy catalogs of varying depth and
positions in the sky. What is the explanation of this universal behav-
ior of ~(r), and what do correlation functions tell us about the ini-
tial conditions at the recombination epoch and/or the value of ~?

A power law shape for ~(r) is a natural expectation of gravita-

tional instability in a Universe with no fixed scales. There exists a
similarity solution of the BBGKY hierarchy equations describing the time
evolution of ~(r) in the limit of: 1) an Einstein-de Sitter cosmology
(~=l); 2) a power law initial spectrum of small perturbations; 3) ab-
sence of non-gravitational forces; 4) no effects caused by the discrete-
ness of the particles. If boundary conditions are chosen to match the
growth rate of linear perturbations on large scales, and to form stable
(non-collapsing) clusters on small scales, then the observed slope of
1.8 is expected for a white noise initial spectrum.

A detailed calculation based on an observed model of the three

point correlation function has been performed by Davis & Peebles (1977)
and compares favorably to the available data, suggesting ~~.3. Analysis
of galaxy catalogs complete with redshift for each member will yield
consistency checks on our model. A study of the Shapley-Ames catalog
(Davis, Geller, and Huchra, 1977) again suggests rather high values of
~, but this sample is too biased by the Virgo supercluster to be a
fair test.

Davis, M., and Peebles, P.J.E., Ap.J. Supp., li, 4

Davis, M., Geller, M.J., and Huchra, J., 1977, Preprint

441

M. S. Longair and J. Einasto (eds.) , The Large Scale Structure of the Universe, 441-443. All RIghts Reserved.
Copyright © 1978 by the fA U
442 MARC DAVIS

DISCUSSION

Zeldovich: I would like to understand in simple physical terms the

influence of~. If you change ~, small perturbations grow at the same
rate but they do so at a different epoch. Bound systems, once they have
formed are independent of~. This is where the difference in the shapes
of the correlation functions comes from.

Davis: This is correct.

Turner: Those of us (Drs Aarseth, Gott, and myself) who have been
analyzing the N-body simulations of galaxy clustering do not feel that
the discrepancy between the calculated BBCKY ~(r) and the measured N-
body ~(r) is necessarily due to the introduction of a mean initial
interparticle separation in the simulations. Indeed, it seems implaus-
ible that the presence of this characteristic scale of which there is
no sign in the N-body ~(r) could cause a (BBCKY predicted) break in the
~(r) power law to disappear and leave a pure power law with no prefer-
red scales. We feel that there is some evidence that ~(r) is determined
by relaxation processes and is relatively independent of ~ and the
initial conditions.

Davis: Indeed, relaxation may occur, but it is caused by the discrete-

ness of particles in the N-body calculation and cannot be included
within a similar solution. The relationship between the spectral index
and the slope y follows from the BBCKY equations in the similar solution
with no approximations, if boundary conditions on small scales are
chosen to ensure that virialized clusters are stable against collapse.
In addition the predicted break in ~(r) occurs on scale lengths
unattainable in the N-body simulations.

Silk: Is it fair to say that your conclusion of Q : 0.3 is dominated

by Local Supercluster galaxies, and that the "background" value of Jl..
could therefore be somewhat lower?

Davis: We evaluated ~ separately in the northern and southern galactic

hemispheres. In the south, where it was argued earlier in this meeting
that the sample of galaxies to m : 13 is a fair sample of the Universe,
we found ~ : 0.26 from the cosmic virial theorem. In the north, even
excluding the Virgo cluster, we found 0.46 which is probably not repre-
sentative.

Tinsley: Can anyone explain why the cosmic virial theorem results in
values of ~ three times that found from the analyses of groups and
knowledge of the mean luminosity density?

Davis: There is a trend towards larger values of MIL as one proceeds

to larger and larger scales and these cosmic virial theorem estimates
refer to the largest possible scales.
THE GRAVITATIONAL INSTABILITY PICTURE IN A DENSE UNIVERSE 443

Gatt: As I said in my talk I would correct Davis' value of ~ = 0.3

downward by a factor of 3/2 to give ~ = 0.2. This is because statisti-
cal vi rial theorem methods always weight large clusters with large
velocity dispersions more heavily. Davis' estimate is also larger than
some previous ones because Davis, Geller and Huchra found an amplitude
of the covariance function that is lower than previous estimates.

Davis: The downward correction of 3/2 is model dependent on the N-body

simulations. According to model dependent theory of the BBGKY equations
my estimate of ~ should be increased by a factor of as much as 1.4. I
have chosen a middle ground and have made no further model dependent
corrections.

Fessenka: How do you account for the influence of observational selec-

tion due to the irregular absorption of light in our Galaxy?

Davis: We restricted our sample to galactic latitudes greater than 40 0

where the effects of absorption can be neglected.

Ostriker: How do you find a break on a characteristic scale in the co-

variance function in a closed Universe?

Davis: The break corresponds to the scale on which the perturbations

become non-linear. The division between the linear and non-linear
regimes is determined by the amplitude of the initial perturbation
spectrum.

Tammann: There must be an observational error of the velocities o(v)

which makes your result insignificant. For what value o(v) does this
become true?

Davis: If the RMS velocity error of a single galaxy o(v) were ~ 225 km
s-l, our data would be consistent with no peculiar motions.
PRIMEVAL CLUSTERS OF GALAXIES AND THE X-RAY BACKGROUND

Edwin M. Kellogg
Center for Astrophysics
Harvard College Observatory/Smithsonian Astrophysical Observatory
Cambridge, Massachusetts

The X-ray telescope for HEAO-B, due for launch next year, will extend our
observing power in X-ray astronomy by a very large factor. For example, its
sensitivity to point sources will be 103 greater than the limit of existing X-ray
sky surveys. Even more intriguing will be its capability to image the all-sky
X-ray background radiation on an arc minute or sub arc minute scale. The
origin of this background is still a mystery. The simplest hypothesis, that it
is the integral of radiation from the more distant members of the classes of
discrete X-ray sources such as clusters of galaxies, Seyferts, QSO's and other
active galaxies, can only explain part of the background.
The investigation I report here begins an attempt to explain the origin of
the X-ray sources in clusters of galaxies as due to primeval gas clouds associ-
ated with density perturbations in the early universe. The clouds become the
sites for formation of clusters of galaxies. Sufficient heating of the gas in the
protocluster occurs to prevent further collapse, but not enough to cause evapo-
ration of the gas away from the cluster. I call this maximal heating. The
temperature is related to the size and mass of the protocluster. One result is
the prediction that there were more protocluster X-ray sources at z '" 1 than
there are now. These are the primeval gas clouds that were maximally heated
but had a high enough density so their bremsstrahlung cooling times were short
compared with their present age. Therefore, they are not now X-ray sources,
but the larger, less dense clouds as in the Coma cluster have not cooled signifi-
cantlyand are still luminous in X-rays.

I have applied this model to predict the appearance of the X-ray sky on a
scale of 'V 10 arc minutes in the HEAO-B telescope. Free parameters are:
qo, density perturbation spectrum at the point of separation, range of cloud sizes,
epoch at which heating begins, and total space density of maximally heated clouds.

Preliminary conclusions are: 1) such primeval gas clouds are observable,

2) angular sizes of the clouds can be measured from 'V 10" to'" 10', and 3) it
should be possible to analyze the distribution of angular sizes and apparent
brightnesses to test the model and perhaps to obtain information on %, zo and
the density perturbation spectrum in the early universe.

445

M. S. Longair and J. Einasto (eds.), The Large Scale Structure of the Universe, 445-446. All Rights Reserved.
Copyright © 1978 by the [AU.
446 EDWIN M. KELLOGG

DISCUSSION

Ozernoy: How does the X-ray luminosity per cluster within the superclus-
ter X-ray sources compare with that of cluster sources such as Coma?

Kellogg: My rough estimates suggest there is not much difference.

Zeldovich: Can HEAO-B distinguish between power-law and bremsstrahlung

spectra of such weak sources as a distant protocluster of galaxies?

Kellogg: The HEAO-B observatory is equipped with four spectrophoto-

meters: a monitor proportional counter of ~ 10 3 cm 2 and 10 field of
view and three others. The objective grating spectrometer disperses
the image of an X-ray source in the focal plane to give a resolution of
A/oA ~ (A/4~) x 40. Its efficiency is ~ 10% in first order. The solid
state spectrometer has E/oE ~ 10 and 100% efficiency above 1 keV. The
focal plane crystal spectrometer can provide A/oA up to 10 3 at much
lower efficiency. Both the grating and the solid state spectrometers
should be capable of detecting iron lines in cluster sources at z = 1-2.
It will be more difficult to obtain an accurate measurement of the shape
of the continuum for these sources, due to the background counts in the
monitor counter.
OBSERVATIONAL LIMITS ON NEUTRAL HYDROGEN IN PRIMORDIAL PROTOCLUSTERS

R. D. Davies
University of Manchester
Nuffield Radio Astronomy Laboratories
Jodrell Bank
Macclesfield
Cheshire, SKII 9DL

After recombination in the early Universe, gas will pass through a

neutral phase on its way to condensation into galaxies and stars. Radio
emission in the reshifted 21 cm line of neutral hydrogen can be used to
detect condensations in the early Universe. Sunyaev and Zeldovich
(1972,1975) have suggested that primordial protoclusters may be predom-
inantly neutral at z = 3 to 10 and may have neutral hydrogen masses
~ 10 14 M@. Two observational approaches have been made to the search
for this emission.

The first was to choose a direction in which intergalactic gas is

known to lie and to make a deep search for redshifted hydrogen in an
associated protocluster. Such a feature is the low excitation gaseous
cloud seen at z = 2.3099 in the absorption spectrum of the radio-quiet
quasar PHL 957. Observations have been made of the redshifted hydrogen
line radio emission at 429.123 MHz. If this cloud were of the Sunyaev
and Zeldovich type then its neutral hydrogen mass would be less than
3 x 10 13 M@ (Davies, Booth and Pedlar 1977).

The alternative approach is to use a frequency in the range of

emission expected and search a number of regions of the sky for the
characteristic 1000 km s-1 wide emission spectrum of a protocluster.
Observations of this type have been made at 328 MHz (z = 3.33) and 240
MHz (z = 4.92) in 20 fields at each frequency (Davies, Pedlar and
Mirabel 1977). The upper limits set by these observations show that the
neutral hydrogen masses are < 10 15 Mo and that the number of protoclus-
ters in the early Universe is < 5 x 10 5 ; the time spent by such proto-
clusters in the neutral form is assumed to be 10 percent of their age.
These results present significant limits on the properties of the neut-
ral gas in the early Universe.

REFERENCES

Davies, R.D., Booth, R.S. and Pedlar, A., 1977. Mon. Not. R. astr. Soc.,
in press.
447
M. S. Longair and J. Einasto (eds.J, The Large Scale Structure of the Universe, 447-448. All Rights Reserved.
Copyright © 1978 by the 1AU.
448 R. D. DAVIES

Davies, R.D., Pedlar, A. and Mirabel, F., 1977. Mon. Not. R. astr. Soc.,
in press.
Sunyaev, R.A. and Zeldovich, Ya. B., 1972. Astr. Astrophys., 20, 189;
1975. Mon. Not. R. astr. Soc., 171, 375.

DISCUSSION

Sunyaev: This is a very important observation, because this is a poss-

ible way to observe protoclusters of galaxies at the stage when there
were no galaxies and there was only gas. The mass of cold gas may be
10 to 100 times greater than today, because the gas was not condensed
into stars at that time. I want to mention that the limits obtained by
Dr Davies and his colleagues are very useful and give an upper limit on
the lifetime of a protocluster at this stage. They are also important
when we choose the epoch of cluster formation. I must mention also that
in the open Universe (n « 1) it is much more difficult to find these
objects because their angular dimensions are smaller. Therefore the
restrictions on M, V and 6t depend strongly on the accepted value of n.
We are very grateful to Dr Davies for obtaining data so important for
our theory.
There is an interesting possibility of finding the redshifts of
distant radio sources at radio wavelengths. If there are protoclusters
or galaxies or clouds of cold matter on the line of sight between the
source and the observer, absorption radiolines might appear in the radio
spectrum. They must be redshifted. For example, the detection of an
absorption line in the metre waveband might be interpreted as a strongly
redshifted 2l(1+z) cm line. In this case we can measure the redshift of
absorbing matter and find a lower limit to the redshift of the source.
It is important that there are no well-known radio lines in the metre
waveband.
It is possible also to predict other strong absorption lines in the
spectra of distant radio sources. Among these are A 2.6 mm CO line
(redshifted to the 1 cm band) and A 6 cm line of formaldehyde (redshift-
ed to the decimetre waveband).

ORIENTATION OF SPIRAL GALAXIES AS A TEST OF

THEORIES OF GALAXY FORMATION
J. Jaaniste and E. Saar

The cosmological turbulence theory and the theory of adiabatic

fluctuations predict different orientations of galaxies in clusters and
superclusters. The first theory favours the alignment of the planes of
galaxies with the supergalactic plane, whereas the planes of spiral
galaxies formed according to the second theory are perpendicular to the
plane of a supergalaxy.

We compared these alternative predictions with the observed distri-

SHORT COMMUNICATIONS 449

bution of orientations of spiral galaxies in the Local Supercluster

within the volume of a radius of 50 Hpc (Jaaniste and Saar 1977). As a
result, the picture expected in the turbulence theory (alignment of the
planes) is excluded at a high confidence level. This is due to the
observed concentration of face-on galaxies towards the supergalactic
equator, a fact that cannot be explained by the turbulence theory. As
only nearly edge-on galaxies have been used in most statistical studies
of orientation, the effect of face-on galaxies has been ignored so far.

The distribution of poles of spiral galaxies is significantly con-

centrated towards the supergalactic plane in agreement with the theory
of adiabatic fluctuations.

REFERENCE

Jaaniste, J. and Saar, E., 1977. Tartu Astrophys. Obs. Preprint A-2.

DISCUSSION

de Vaucouleurs: Do you conclude that there is an excess of galactic

planes parallel or perpendicular to the supergalactic planes?

Saar: There is a definite excess of galaxies with planes perpendicular

to the supergalactic plane.

de Vaucouleurs: Do your results agree with those of Roberts and

Reinhardt or not?

Saar: No, they disagree, and this is probably due to the difference of
our samples. Roberts and Reinhardt used all the galaxies of the
Reference Catalogue, while we considered only those galaxies that
undoubtedly belong to the Local Supercluster.

Reinhardt: You are aware that your investigation contradicts the result
of Reinhardt and Roberts, Nilson and de Vaucouleurs, all derived from
slightly different data samples and giving a weak parallelism of planes
of galaxies to the equatorial plane of the Local Supercluster. Of
course, your sample is smaller and different from all of those used in
the previous investigations. Also your method is different from those
of your precursors. Thus without reanalyzing your data with another
method, it is difficult to say how your results compare to the previous
ones.
450
SHORT COMMUNICATIONS

NON-LINEAR EFFECTS IN FLAT GRAVITATING SYSTEMS

A. M. Fridman

It is shown that in flat gravitating systems, such as spiral gala-

xies, groups of galaxies and some superclusters (which have an axis
ratio of, say, 1/5) there are nonlinear effects such as the formation of
collapsars in disks which are stable according to Toomre's criterion;
the formation of solitons (spiral and ring-shaped), which can move with
supersonic as well as subsonic velocities depending on the characteris-
tic parameters of the disk. In an isothermal gas, shock waves can only
originate in spirals with two or more arms. In the background of two
bright spiral arms one may be able to trace a weaker ring structure. In
contrast to supersonic spirals, the velocity of the circular structure
is subsonic. A bright ring structure can in principle indicate that in
such systems for some reason the adiabatic index y is close to that of
an ideal gas y ~ 5/3. This last point can be of interest in the future
if non-isothermal perturbations will be found as seems possible in some
systems. A break-up instability will lead to practically uniform
enhanced star formation allover the disk; it will have a tendency to
simulate narrow rings and tightly wound spirals with one arm according
to the law (oa/ao)~(t-to)-l.

The condition for applicability of the model of interacting galax-

ies is (M/m) ~ 2R/h, where R, H, m are radius, thickness and mass of the
disk and M is the mass of the galaxy centre (the bulge component). For
spiral galaxies, (M/m)max ~ 2, 2R/h ~ 20 to 30, therefore condition of
applicability of the model of interacting galaxies is not fulfilled.
PERSONAL VIEW - THE LARGE SCALE STRUCTURE OF THE UNIVERSE

M. S. Longair
Mullard Radio Astronomy Observatory
Cavendish Laboratory, Cambridge

I should explain at once that I very much regret that it is the

Chairman of the Scientific Organising Committee who has been press-
ganged by his Committee into giving the concluding remarks at this
symposium. I obviously cannot take an unbiassed view of the proceed-
ings. I have therefore refrained from calling this a "survey" or
"concluding remarks" - it is a personal view of those aspects of this
symposium which have struck me as particularly interesting - I will
make no attempt to provide a complete survey of the last five days.

It is the privilege of those landed with the task of attempting to

bring a symposium to a rousing conclusion to offer comments which may
be termed "philosophical", "sociological" or "banal",depending on your
point of view, and I will not be the exception to this rule. First of
all, I often think conferences are not particularly successful at
communicating information or ideas, in the sense of what people really
think about different problems. Partly this is because the symposium
format does not allow enough time for proper discussion of particularly
thorny points, partly it is because people are too modest in expressing
their worries (or perhaps their ignorance) and partly it is because many
people who have serious criticisms of a particular piece of work do not
express their views, perhaps because they think the work under discus-
sion is obviously wrong, perhaps because they cannot be bothered or
perhaps because they have said too much already. I don't know how to
overcome these problems. What I will try to do is to expose some of
the questions and worries which were running through my mind during the
symposium.

A second point which struck me is that there are basically two

types of astrophysicist - those who like simple things and those who
love complicated things. This applies to both observers and theoreti-
cians. This dichotomy is particularly apparent in a subject like the
large-scale structure of the Universe. We are all aware of the tremen-
dous complexity of the Universe and sooner or later we will have to
account for every little bit of it. However, this should not prevent
us seeking the overall regularities in the Universe. What many of us
451

M. s. LongQlr and J. Einasto feds.). The Large Scale Structure of the Universe. 451-461. All Rights Reserved.
Copyright © 1978 by the IA U.
452 M. S. LONGAIR

want to find out are the basic principles according to which the large
scale structures form and evolve and by which the complexities which we
observe today came about. You will gather from this remark that I am
a proponent of the "simplistic" school of astrophysics.

Finally, before tackling the science, it is important to remember

the methodology involved in studies of the large scale structure of the
Universe. It is not simply a question of observations and theory but
rather a complex interaction of observations, their interpretation and
theory. Theories which are too strongly dependent on interpretation of
observations must live dangerously. The oftener the theories confront
the observations directly, the more secure the validation or otherwise
of the theory.

The Scientific Organising Committee had well-defined objectives in

designing the programme. In the first three parts, we considered large
scale structures in the Universe in order of increasing dimension and
discussed the astrophysical problems which they pose in the context of
the observations:
1. GALAXIES IN SMALL GROUPS.
2. CLUSTERS OF GALAXIES.
3. LARGE SCALE SYSTEMS.
Naturally, these discussions involved consideration of the evolution of
these systems over cosmological time-scales but, as we heard, there is
little direct observational evidence for evolution in these systems.
Therefore, part 4 was devoted to related fields in which there may be
direct evidence for the effects of cosmological evolution
4. OBSERVATIONAL EVIDENCE FOR COSMOLOGICAL EVOLUTION.
Having then completed the description of the large scale properties of
the Universe as we know it, we then grasped the thorny problem of the
origin of these structures
5. THE FORMATION OF STRUCTURE IN THE UNIVERSE.

I will devote most attention to optical observations of the large

scale structure of the Universe in Sections 1, 2 and 3 before consider-
ing the theory of such systems in Section 4. In the remaining sections,
I will consider evidence for cosmological evolution and the origin of
large scale systems.

1. GALAXIE S IN SMALL GROUP S

De Vaucouleurs' classical work on groups of galaxies in the
Reference Catalogue of Bright Galaxies was referred to many times. The
problems of establishing group membership and their reality as bound
self-gravitating systems came through clearly as the most serious
problems, the resolution of which has repercussions throughout many
different aspects of cosmology, for example, the mean density of matter
THE LARGE SCALE STRUCTURE OF THE UNIVERSE 453

in the Universe and the nature of the hidden mass which must be binding
relaxed groups and clusters.

Despite the fact that many different workers were using basically
the same material, there were two distinct schools of thought about what
the mean mass to luminosity ratio, MIL, of groups of galaxies within
about 20 Mpc really is. One school believes <MIL; ~ 5 - 10, the other
<MIL) '\, 100. First of all, it should be emphasised that this is only
a factor of 10 which is not too bad by some astronomical standards but
most people feel the discrepancy is too big. Fortunately, the methodo-
logy used by each observer was described clearly and from this we (or
rather they) should be able to decide whether there is a real discre-
pancy or not. Many times we heard of the importance of including
properly the high velocity members of the group since they make a large
contribution to its kinetic energy and hence require large masses to
bind them to the group. I very much hope this symposium will help bring
together these workers so that they can decide if there is a real
discrepancy and whether or not it is entirely a question of group
membership.

It did strike me that there is no very good reason why all groups
should have the same MIL ratio and partly the discrepancy might be due
to the selection of different types of groups by different observers.
It was also not clear whether or not the discrepancy, if real, can be
reconciled within a single picture. In view of the uncertainty about
what form the binding mass of groups and clusters takes, I would be most
surprised if one could not reconcile the different values within a
single picture.

Having raised the question of the form of the binding mass of

clusters and groups, I was somewhat surprised that we heard relatively
little about its nature. Is this because everyone agrees it must be
there but they have given up for the present trying to choose among the
various possibilities? Hegyi presented evidence for an almost spheri-
cal halo around NGG 4565 with radius'\, 40 kpc and Lynden-Bell, in his
talk in which within 25 minutes he changed the Hubble constant and the
size and mass of the Galaxy, suggested that there may be a massive halo
around our Galaxy. Against this view, Karachentsev showed that the MIL
ratio for widely separated double galaxies is "normal", i.e. '\, 8-10.
This last approach would appear to be a most promising method for
obtaining further real data about massive haloes. Another manifestation
of massive haloes may be Einasto's hypergalaxies. The velocities of
satellite galaxies in hypergalaxies must give dynamical information
about the total mass of these systems and it is to be hoped the objects
in his first catalogue of hypergalaxies will be the subject of intensive
study.

I was particularly impressed by the vast amount of high quality

redshift data which is now becoming available for nearby groups, both
from optical studies and from HI velocity data. I feel that so much
454 M. S. LONGAIR

data is now being accumulated that many of the problems of the nearby
groups must soon be clarified.

2. CLUSTERS OF GALAXIES

The optical properties of the great clusters received relatively

little attention, perhaps because many of their properties are now well
estbalished. One point which strikes me on reading the literature on
the great clusters is that there are many classification schemes but all
of them seem to boil down to one continuous classification in which
there are three basic types, Regular, Intermediate and Irregular
clusters. This point is made very clearly in Neta Bahcall's review of
clusters of galaxies to appear in Annual Reviews of Astronomy and Astro-
physics 1977. An abbreviated version of her table is given in Table 1
for reference. I believe that most of the interesting astrophysical
aspects of clusters are contained within this simple scheme.

Table 1
Classification of clusters of galaxies (after N. Bahcall)
Classification
or Property REGULAR INTERMEDIATE IRREGULAR

Zwicky Compact Medium-Compact Open

Bautz-Morgan I, .. I-II, II (II), II-III (II-III), III
Rood-Sastry cD, B, (L,C) (L), (F), (C) (F), I

Content Elliptical rich Spiral poor Spiral rich

E : SO : S ratio 3 : 4 : 2 1 : 4 : 2 1 : 2 : 3
Symmetry Spherical Intermediate Irregular Shape
Central High Moderate Very little
Concentration
Central Profile Steep gradient Intermediate Flat gradient
Mass segregation Marginal Marginal No segregation

3. LARGE SCALE SYSTEMS

To me, some of the most exc~t~ng results presented at this sympos-

ium concerned the structure of the Universe on the largest scales.
Everyone seemed to agree about the existence of superclusters - de
Vaucouleurs' description of the local supercluster, Tully's film of
local supercluster and Peeble's 10' x 10' plot of the Shane-Wirtanen
counts are convincing direct evidence of systems on scales ~ 30-100 Mpc.
THE LARGE SCALE STRUCTURE OF THE UNIVERSE 455

But perhaps even more surprising are the great holes in the Universe.
Peeble's picture, Einasto's analysis of the velocity distribution of
galaxies which suggests a "cell-structure" and Tifft's similar analysis
argue that galaxies are found in interlocking chains over scales ~ 50-
100 Mpc forming a pattern similar to a lace-tablecloth. The holes are
particularly interesting since they might appear to be at variance with
the idea of continuous clustering on all scales which we have been
educated into understanding from the original work of Kiang up to the
more recent analyses of Peebles and Abell. I do not believe there is
any basic contradiction here - one cannot expect a covariance function
approach to reproduce sharp features like holes. I am still a firm
believer in the basic correctness of the results of the covariance
analysis - more detailed astrophysical arguments will have to account
for details such as holes which are about 10 Mpc in size and void of
bright galaxies.

On the very largest angular scales, patchy galactic obscuration

becomes a problem but we heard from Kalinkov that there remains
structure on these scales, "third-order clustering", when the analysis
is restricted to high galactic latitudes. His beautiful pictures in
which the distribution of clusters was convolved to a very large
angular scale suggested two main groupings of clusters. If these
structures are real, they must be the largest associations of optical
galaxies known in the Universe. One wonders whether their existence is
consistent with the isotropy of the distribution of extragalactic radio
sources and of the microwave background radiation.

4. THEORETICAL STUDIES

Turning now to theoretical studies of these structures, I must

begin with Peebles' covariance function. I have always regarded this
as one of the-most beautiful pieces of analysis of the distribution of
objects in the Universe and, although the procedure is fairly complex,
the result is amazingly simple. The covariance function on all scales
from ~ 30 kpc to ~ 30 Mpc has a simple power-law form ~(r) ~ r- 1 • 77 .
Within this picture one can account for structures observed on different
scales - de Vaucouleurs' groups and local supercluster, Zwicky's super-
clusters, Abell's associations of Abell clusters, etc. There are of
course further complications which result from more detailed analyses
(such as chains of galaxies and clusters and the "holes" in the
Universe), but it would seem that any satisfactory theory of the origin
of clustering in the Universe should account for this.

This is what is so impressive about the work of Aarseth and his

colleagues in their simultations of galaxy clustering. The model is of
great basic simplicity. Galaxies are regarded as "fuzzy particles" and
the particles interact only through their mutual gravitational attrac-
tion. The amazing result of these computations is a more-or-less
perfect power-law with slope close to that found by Peebles which
456 M. S. LONGAIR

extends over 4 orders of magnitude in physical scale. Granted the

simplicity of the basic model, I find it staggering that the end point
of these calculations should end up giving so close agreement with the
observed covariance function for galaxies in the real Universe. As I
understand the results of this work, the predicted power-law covariance
function is insensitive to many of the initial assumptions. It should
also be noted that this power-law results from a myriad of non-linear
interactions between particles and that there is as yet no simple way
of deriving this result by physical arguments.

Peebles mentioned some of his worries about the way in which the
problem was set up, in particular that on small enough scales the
density fluctuations must have op/p ~ I and hence non-linear effects
are already important. I would tend to argue the other way round. One
of the great strengths of the approach of Aarseth and his colleagues is
that they set about their simulations using sufficient particles (4000)
to mimic our region of the actual Universe at the present day. Thus,
if the general picture of collapse is correct, they are mimicking
exactly what must have happened to the actual Universe and it too must
have had to worry about the fact that the fluctuations were already
large on a small enough scale. I interpret the success of the computa-
tions as telling us something about the initial conditions from which
large scale structures evolved.

Having been elated by these results, which, incidentally, I classi-

fy as "simple", I then went through my low point at the conference when
de Vaucouleurs said in discussion that these simulations looked nothing
like the real Universe!! I have been privately canvassing opinions
about this since then and most people seem to think the simulations do
not look too unlike the real Universe. It is true that we do not see
all the complexity of the real Universe but I feel we at least have the
beginnings of the picture. We do not perhaps see the well developed
"cell-structure" in these simulations (although I believe I can see
things like them) but it must be remembered that the computations were
only made with 4000 "soft" point masses. One can envisage in addition
many astrophysical processes which would change the details of the
simulations and perhaps make them look more like che real Universe but
which would not change the overall result of the present simulations.
I interpret de Vaucouleurs' remark as encouragement for us all to work
harder! !

Ostriker's presentation of the dynamical evolution of clusters

using only Newtonian gravity was, to me, wholly convincing and need not
be amplifed here. The scheme can account naturally for the various
types of cluster listed in Table I as well as phenomena such as the
formation of cD galaxies at the centres of clusters, possibly with
mUltiple nuclei. In addition, we can all see reasonable ways in which
hypergalaxies and even Vorontsov-Velyaminov's "nests" of galaxies could
come about.
THE LARGE SCALE STRUCTURE OF THE UNNERSF 457

The one thing which I found missing in this presentation was the
problem of the MIL ratio for clusters. Ostriker made the convincing
remark that regular clusters,such as the Coma cluster in which the
velocity distribution is Maxwellian, lS just as much in equilibrium as
the inside of a star and hence there is no question of the mass being
missing - it is merely not visible. We heard the important result that
all the hidden mass could not be in the galaxies because dynamical
friction would then be so important that the clusters would today be in
a state of extreme "stripping" and "cannibalism" which apparently has
not yet happened. One guesses that the binding mass is some form of
"stellar" distribution which follows roughly the galaxy distribution -
maybe it is in the form of very faint stars, possibly neutron starts or
blackholes. One wonders exactly what the nature of the diffuse light
in the Coma cluster is which was described by Melnick and his colleagues.
A conservative intuition suggests that the binding mass may be no more
than low-mass stars which are easily stripped from galaxies and now
belong to the cluster as a whole.

I have omitted so far all mention of the radio and X-ray properties
of clusters. This is because at the present day I feel they are perhaps
diagnostic tools for studying more detailed aspects of cluster evolution
rather than the basic problem of cluster formation. However, I must
immediately emphasise their central importance in the study of the
astrophysics of the evolution of gas in clusters and of extragalactic
radio sources. In addition, studies of extragalactic radio sources
provide a powerful method of detecting distant rich clusters of galax-
ies. The X-ray satellite HEAO-B will be able to observe X-ray cluster
sources at cosmological distances and these will be of central import-
ance in studying the early evolution of clusters. The reviews by
Perola and Ekers and by Culhane of radio and X-ray observations of
clusters respectively indicate clearly the wide range of exciting
astrophysics which these disciplines have contributed.

At this point in the symposium, we discussed explicitly the ques-

tion of the mean density of matter in the Universe, a problem which had
been casting its shadow over a number of the discussions. The present
situation is admirably summarised in the discussion chaired by Tammann.
The range of mass-to-light ratios considered ranges from about 5 to 200
and the corresponding values of the density parameter ~ from about 0.02
to 0.3. There were a few suggestions of higher values, ~ ~ 0.5-0.B.
What was particularly interesting about the discussion are the "new"
methods of estimating ~ - from perturbations of the Hubble flow, from
the "statistical virial theorem", from the correlation of the peculiar
motions of galaxies with density perturbations. All of these methods
require large bodies of high quality redshift data but one should be
reasonably optimistic that these observations will become available in
the foreseeable future. The consensus view appeared to be that values
of ~ ~ 0.1 are favoured by I do not think anyone would want to exclude
values of ~ ~ 1 at present.
458 M. S. LONGAIR

5. EVIDENCE OF COSMOLOGICAL EVOLUTION

None of the presentations in Section 4 of the programme actually

provided observational evidence for the evolution of large scale systems
but rather evidence for the evolution of very rare, super-luminous
objects - quasars and radio galaxies. The V/Vmax test for quasars and
the counts of radio sources are still by far the most significant pieces
of evidence for the cosmological evolution of any class of object in the
Universe. Broadly interpreted, they suggest that the Universe went
through a period of violent activity when it was about 0.1-0.2 of its
present age. Quasars and radio galaxies were much more common events
at these epochs than they are now by a factor of about 10 3 • It is my
belief that we will eventually learn a great deal about when galaxies
and clusters formed and also about the evolution of the environments of
radio sources from these studies as I outlined in my lecture.

In comparison, the evidence for the evolution of galaxies and

clusters is sparse. Tinsley revieved the few available pieces of
evidence and the problems of interpretation. However, one can confid-
ently expect the amount of data on these topics to increase dramatically
over the next few years as much more data is accumulated by the increas-
ing number of optical telescopes in the 4-6 m class. In the mid 1980's
we have the prospect of the NASA Space Telescope which everyone believes
will revolutionise extragalactic studies just as the construction of the
Hale 200-inch telescope opened up a new era of extragalactic studies in
the 1940's.

Two topics struck me as being ripe for immediate attention. First,

Karachentsev and Kopylov presented counts of galaxies to 24th magnitude
made with the SAO 6-m telescope which appeared to be in remarkable
agreement with the predictions of uniform world models. Quite different
results were reported in discussion by Abell, de Vaucouleurs and
Tinsley. The problem of making self consistent counts of galaxies to
faint magnitudes are well known, in particular the problem of knowing
the magnitude scale at these faint magnitudes. Tinsley has emphasised
how sensitive the counts of galaxies are to evolutionary changes with
cosmological epoch. An interesting comparison is with the counts of
radio sources where strong evolutionary changes are observed. The
problem in interpreting the radio source counts is that the theory of
the origin and evolution of individual radio sources is at too primitive
a stage to provide astrophysical foundations for the interpretation of
this phenomenon. On the other hand, the interpretation of the optical
spectra of galaxies is much more secure and thus may provide more
definite evidence on the evolutionary history of galaxies as a whole.

The second point relates to the first - namely, young galaxies, if

they formed at redshifts Z ~ 5-10, may well be detectable as non-
variable stellar objects similar to quasars. Sunyaev expressed this
view (which I share) in discussion. It is particularly intriguing to
speculate what the nature of Bolton's non-variable radio-quiet quasars
THE LARGE SCALE STRUCTURE OF THE UNIVERSE 459

is. These are exactly the types of object which Sunyaev and I believe
may be strong candidates for young galaxies with redshifts Z ~ 3-5. The
discovery of such objects would put the whole subject of the formation
and early evolution of galaxies on a firm observational foundation and
hence programmes to search for them are well worth an intensive effort.

6. THE ORIGIN OF STRUCTURE IN THE UNIVERSE

Finally, we corne to the problems of the origin of structure in the

Universe, probably the most fundamental question of all. I and many of
my colleagues have been deeply impressed by the work of the Moscow
groups in this field. I regard their work as among the most exciting
and imaginative in modern astrophysics and their efforts have made major
contributions to our understanding of these basic problems. However, as
we have heard, widely differing theories can account for the observed
structure of the Universe - specifically, these models are the adiabatic
and isothermal models developed by Zeldovich and his colleagues and the
whirl theory described by Ozernoi. The basic problem is that the end
product of the models is determined to a large extent by the initial
conditions. In all models, there is no convincing physical explanation
of the initial spectrum of perturbations which eventually leads to the
formation of galaxies, clusters, superclusters, etc. Zeldovich express-
ed the belief that eventually we will be able to understand this once we
have a better understanding of the physics of the very earliest stages,
including the quantum epochs, of the evolution of the Universe. This is
an exciting prospect but I wonder how many astrophysicists believe this
is really attainable in the foreseeable future.

According to Zeldovich and Ozernoi all three theories mentioned

above can account for observed structures in the Universe. I confess
that I was disappointed not to hear critiques of these models by the
proponents of the different theories. Many of us are aware of the hot
disputes between the Moscow schools which have stimulated splendid
pieces of astrophysics. Personally, I would have liked to hear more of
the physical problems of these theories rather than of their successes.
This is not just a question of theoretical interest but one of immediate
importance for observation which I will corne to below.

All of us must have been impressed by the film of the development

of "pancakes" by Doroshkevich and his colleagues and by the remarkable
resemblance to the cell-structure of the Universe described by Einasto,
Tifft and others. Intuitively, I am attracted to the adiabatic picture
of the origin of galaxies, principally because of the simplificity of
the picture. There is a marked contrast in complexity between this
adiabatic picture and the whirl picture. To oversimplify, in the former,
the non-linear stages of collapse happen once when large scale struc-
tures collapse at Z ~ 5-10 whereas in the latter the theory is non-
linear at all epochs. In the adiabatic model, the Universe may be
considered to be effectively isotropic and homogeneous up to the epoch
460 M. S. LONGAIR

op/p ~ 1 and hence all the attractions of the canonical hot big-bang
model are also valid in this model. This is not necessarily so in the
whirl model. The development of the turbulent spectrum in the radiation
dominated phase is the result of non-linear interactions and galaxy
formation results from the collapse of eddies which have a supersonic
turbulent spectrum after recombination. As mentioned by Jones, the
theory is constantly fighting the problem of the dissipation of turbu-
lence which must be continuously regenerated by the (non-linear) decay
of larger-scale eddies corning through the horizon. The very early
stages of evolution of this model are also much more complex than in the
adiabatic model and it is not clear that it is consistent with the
isotropy of the microwave background radiation and with cosmological
synthesis of the light elements. However, just because the whirl theory
is more complicated, it does not mean that it is wrong and I heard no
piece of evidence at the symposium which showed that the whirl theory
was in immediate danger of suffering a mortal blow.

The problem with all these theories is one of observational valida-

tion. The methodological problem is that all models are designed to
produce the observed Universe at the present epoch. There are very few
observational tools by which the early development of these models can
be tested, the only viable one at present being observations of the
microwave background radiation. The most promising test is the obser-
vation of fluctuations in the intensity of the microwave background
radiation which, according to most workers, gives us information about
the amplitudes and velocities associated with density perturbations at
or immediately after the epoch of recombination. We heard of the very
low limits to intensity fluctuations now being obtained with instruments
such as the Ratan-600 radio telescope by Parijskij, 6T/T ~ 10- 4 or
better. The problem as I see it is that, although all theories are
still consistent with these upper limits, as described by Sunyaev and by
Ozernoi, it is not at all clear how the theories could be distinguished
even if fluctuations are observed. In addition, as I have emphasised,
there is no independent method of obtaining information about these
early epochs. In this situation we are particularly in need of clarifi-
cation of the theoretical difficulties of the models so that we can
judge for ourselves how we are to interpret the data.

A final important point about these observations is the following.

The sensitivities and angular scales which are most important for
studying these problems are such as to require either very large amounts
of observing time on existing radio telescopes or the construction of
space experiments with total cost of ~ 50-100 x 10 6 $. The enthusiasm
of radio astronomers (and more important, the grant-awarding agencies)
is considerably dampened if they learn that following the experiment we
will have got no further forward in the resolution of which theory is
the best description of how the Universe has evolved. Sunyaev
described very clearly the many possible sources of fluctuations in the
microwave background radiation which are all expected to become import-
ant at about the same intensity level as those due to collapsing proto-
THE LARGE SCALE STRUCTURE OF THE UNNERSE 461

structures in the Universe ~T/T ~ 10- 5 • Everyone would agree that the
measurement of fluctuations in the microwave background radiation
would tell us something important about the Universe but I'm not sure
that we will be able to agree what that is.

These remarks are not intended to discourage theorists or

observers. They are intended to spur theorists to great efforts to
eliminate at least some of the many possible theories of the origin of
fluctuations in the microwave background radiation and possibly of
rival theories of galaxy and cluster formation.

I hope you have enjoyed the symposium as much as I have. On

behalf of the Scientific Organising Committee, I thank you all for
coming and making it such a memorable occasion.
 AUTHOR INDEX

Aarseth, S. J. 189 KaIinkov, M. 276

Abell, G. O. 253 Kaneva, K. 276
Arakelian, M. A. 274 Karachentsev, I. D. 11,339
Kellogg, E. M. 445
Baier, F. W. 198 Khachikian, E. 105
Biermann, P. 121 Kopylov, A. I. 339
Bolton, J. G. 295 Kraan, R. 71
Boynton, P. E. 317,405 Kunkel, W. E. 130
Burczyk, J. 199 Kurskov, A. A. 404

Casini, C. 100 Lari, C. 137

Chernin, A. D. 439 Longair, M. S. 305,451
Chincarini, G. 263 Lorenz, H. 303
Culhane, J. L. 165 Lynden-Bell, D. 123

Davies, R. D. 447 Machalski, J. 200

Davis, A. G. 131 Materne, J. 93
Davis, M. 441 Melnick, J. 135
Dermendjiev, V. 276 Metik, L. 103
De Vaucouleurs, G. 205 Metlov, V. 117
Doroshkevich, A. G. 423 Michalec, A. 200
Dostal, V. 117
Ostriker, J. P. 357
Einasto, J. 51,96,241 Ozernoy, L. M. 98,404,427
Ekers, R. D. 149
Paal, G. 197
Fessenko, B. I. 279 Parijskij, Yu. N. 315
Fisher, J. R. 31,214 Peebles, P. J. E. 217
Flin,P.275 Perala, G. c., 137
Fridman, A. M. 450 Peterson, B. A. 389
Philip, A. G. D. 131
Ginzburg, V. L. 161 Phillipps, S. 280
Gott, J. R. 63 Pronik, I. 103
Gregory, S. A. 267
Grishchuk, L. 226 Rees, M. J. 377
Gursky, H. 327 Reinhardt, M. 98
Richter, N. 105
Hegyi, D. J. 95 Rood, H. J. 263
Heidmann, J. 100 Rudnicki, K. 229
Hoessel, J. 135
Hagner, W. 105 Saar, E. M. 423, 448
Huchra, J. 271 Schmidt, K.-H. 101
Huchtmeier, W. K. 49 Schmidt, M. 289
Shandarin, S. F. 423
Jaaniste, J. 448 Silk, J. 179
J6eveer, M. 241 Staikov, B. 276
Jones, B. J. T. 377 Stavrev, K. 276
Sunyaev, R. A. 393

463
 AUTHOR INDEX 464

Tammann, G. A. 71 Valentijn, E. A. 153

Tarenghi, M. 263 Van der Laan, H. 153
Thompson, L. A. 263 VorontsoY-Velyaminoy, B. A. 3, 117
Tifft, W. G. 263, 267
Tinsley, B. M. 343 White, S. 135
Tomoy, A. 102 Wielebinski, R. 157
Tomoy, B. 276 Wielen, R. 121
Toomre, A. 109
Tully, R. B. 31, 214 ZeldoYich, Ya. B. 409
Turner, E. L. 21 Zieba, A. 199
Zieba, S. 229
Urbanik, M. 275